Gene expression and functional evidence of epithelial-to-mesenchymal transition in papillary thyroid carcinoma invasion

Abstract

Papillary thyroid carcinomas (PTCs) that invade into local structures are associated with a poor prognosis, but the mechanisms for PTC invasion are incompletely defined, limiting the development of new therapies. To characterize biological processes involved in PTC invasion, we analyzed the gene expression profiles of microscopically dissected intratumoral samples from central and invasive regions of seven widely invasive PTCs and normal thyroid tissue by oligonucleotide microarray and performed confirmatory expression and functional studies. In comparison with the central regions of primary PTCs, the invasive fronts overexpressed TGF β, NFκB and integrin pathway members, and regulators of small G proteins and CDC42. Moreover, reduced levels of mRNAs encoding proteins involved in cell–cell adhesion and communication were identified, consistent with epithelial-to-mesenchymal transition (EMT). To confirm that aggressive PTCs were characterized by EMT, 34 additional PTCs were examined for expression of vimentin, a hallmark of EMT. Overexpression of vimentin was associated with PTC invasion and nodal metastasis. Functional, in vitro studies demonstrated that vimentin was required both for the development and maintenance of a mesenchymal morphology and invasiveness in thyroid cancer cells. We conclude that EMT is common in PTC invasion and that vimentin regulates thyroid cancer EMT in vitro.

Thyroid carcinoma is the most common classical endocrine malignancy, and its incidence is rising rapidly, due almost entirely to an increase in papillary thyroid carcinoma (PTC) diagnoses (1). Patients diagnosed with PTC at an early stage have an excellent prognosis; however, individuals with large, invasive tumors and/or distant metastases have a 5-year survival rate of ≈40% (2, 3). Thus, there is a need to better understand the molecular causes of thyroid cancer progression to develop new treatment options.

The genetic defects believed to be responsible for PTC initiation have been identified in the majority of cases; these include genetic rearrangements involving the tyrosine kinase domain of RET and activating mutations of BRAF and RAS (3–5). Although some correlation studies support an association between specific genetic alterations and aggressive cancer behavior (6–9), there are a number of events that are found nearly exclusively in aggressive PTCs, including mutations of P53 (10, 11), dysregulated β-catenin signaling (12), up-regulation of cyclin D1 (13), and overexpression of metastasis-promoting, angiogenic, and/or cell adhesion-related genes (14–20). We have determined that invasive regions of primary PTCs are frequently characterized by enhanced Akt activity and cytosolic p27 localization (21, 22). We, and others, have also demonstrated functional roles for PI3 kinase, Akt, and p27 in PTC cell invasion in vitro (16, 23, 24). However, the correlation between increased Akt activity and invasion was not found for PTCs with activating BRAF mutations. Most importantly, these focused studies do not address the more global question of which biological functions and signaling pathways are altered in invasive PTC cells.

The process of epithelial-to-mesenchymal transition (EMT), whereby epithelial cells lose cell–cell contact, undergo remodeling of the cytoskeleton, and manifest a migratory phenotype has been implicated in the conversion of early-stage tumors to invasive malignancy (25). Signaling pathways that directly regulate EMT include Integrin, Notch, MET, TGFβ, NF-κB, PI3K, and p21-activated kinase (Pak) (reviewed in ref. 26), some of which have been implicated in PTC progression. A defining feature of EMT is the loss of E-cadherin expression and gain of fibronectin and vimentin expression (27, 28).

To identify pathways involved in PTC invasion, we performed oligonucleotide microarrays of total RNA isolated from intratumoral invasive and central regions of widely invasive PTCs and performed confirmatory expression, clinical correlation, and in vitro studies. These studies together demonstrate that EMT is associated with PTC invasion in vivo and is mechanistically involved in thyroid cancer cell morphology and aggressive behavior in vitro. Disruption of pathways involved in EMT may represent a therapeutic approach to invasive papillary thyroid carcinoma.

Results and Discussion

Initial analysis was performed comparing expression differences between normal and central tumor mRNA isolated from PTCs from the Ukraine. Results from this analysis were compared with those from a second group of paired normal/central PTC samples from the Ohio State University tissue bank that were simultaneously analyzed. Overall, the results were similar between the two groups (gene expression data available at www.ncbi.nlm.nih.gov/projects/geo; accession no. GSE6004). Functional analysis by using Expression Analysis Systemic Explorer (EASE) demonstrated a marked reduction in heavy metal-interacting genes in the central tumors, particularly due to down-regulation of genes in the metallothionein family. An increase in expression of genes relating to cell–cell adhesion and cell–cell communication was also identified [EASE analysis summary of Ukraine series in supporting information (SI) Table 1]. In addition, a group of genes shown to have altered expression levels in PTC versus normal tissue (29) were similarly regulated in both groups. In the U.S. group, five of nine had activating mutations in BRAF and none had RET/PTC translocations or RAS mutations, whereas three of seven Ukraine samples had RET/PTC rearrangements and one had a BRAF V600E mutation.

A unique feature of our data set is the comparison between the central and invasive regions of PTCs. To be certain we included only tissues that were intratumoral in which the cellular component comprised mostly thyroid cancer cells, tumor dissection on frozen tissue samples and subsequent study of slices from adjacent paraffin-embedded samples was performed by two experienced thyroid pathologists (V.V. and V.S.) for all cases. For a sample to be included in this study, the pathologists had to confirm that >90% of the cells in the tissue sample were thyroid cancer cells and that there was no evidence of peritumoral immune cell invasion or chronic lymphocytic thyroiditis. In support of the overall similarity in cell types between the intratumoral samples, the differences in gene expression levels between the paired samples (invasive and central regions) were much smaller than those between normal and either tumor tissue sample.

To identify genes uniquely up- or down-regulated in the invasive fronts of PTCs, ratios between gene expression levels in the central regions of the tumor and normal tissue and between the invasive and central regions of the tumors were created for each patient and analyzed. The unsupervised clusters of gene ratios of paired center/normal and invasion/center samples are shown in Fig. 1A.

Fig. 1.

Analysis and confirmation of oligonucleotide expression microarrays. (A) Unsupervised cluster analysis of gene arrays comparing the ratios of center/normal and invasion/center was performed. Red indicates overexpression, and green indicates underexpression as determined by the ratio of center versus normal or invasion versus center for individual pairs. The total number of genes per cluster is denoted in parentheses. Columns are labeled as tumor number. (B) RUNX2, fibronectin 1 and vimentin immunohistochemistry of PTCs. Immunoactive RUNX2, fibronectin, and vimentin were low or absent in normal tissue and were up-regulated in tumors for each protein. Immunoactive RUNX2 and fibronectin 1 proteins were qualitatively increased the invasive regions of PTCs, whereas vimentin expression was similar in central and invasive areas.

Because the expression differences were smaller between the central and invasive intratumoral samples, and similar numbers of gene expression ratios are required in each group for functional analysis, we needed to reduce the stringency of ratio of gene expression required for inclusion to 1.2-fold between the central and invasive regions. In accordance with minimizing the inappropriate inclusion of genes with minimal expression change without excluding commonly over- or underexpressed genes in the invasive fronts, >99% of genes showed at least 1.2-fold increased or 0.8-fold reduced expression in at least five of seven (usually six or seven of seven) paired samples. Confirmatory gene and protein expression studies were performed for genes of functional interest.

By using EASE analysis (SI Table 1), the invasive regions of the PTCs, in comparison with the central regions, overexpressed genes related to (i) transcription; (ii) cell signaling, including genes in the integrin and TGFβ pathways; (iii) genes regulating nuclear localization of proteins; and (iv) small G proteins and guanine nucleotide exchange factors. In addition, a reduction in expression of genes related to cell–cell processes related to adhesion and communication was noted in the invasive regions. It was of interest that the normal/central comparison revealed an increase in expression of genes involved in cell–cell adhesion and communication, suggesting that the loss of cell–cell interactions might be particularly unique to the invasive regions.

In addition to the unsupervised cluster and EASE analyses, we also analyzed the gene list focusing on pathways known to be involved in cancer progression. It was notable that, in addition to TGFβ and integrin pathway genes, other genes known to be involved in EMT were also overexpressed in the invasive regions of PTCs (SI Table 2). This finding, in combination with the EASE analysis, suggested that EMT might be common in invasive PTCs.

To confirm the oligonucleotide microarray findings, we selected a group of genes known to be functionally involved in EMT and performed quantitative RT-PCR (qRT-PCR) from RNA used in the microarrays. Expression levels of FN1, RUNX2, PRICKLE1, CCND1, KFL6, and SLIT1 were determined. In all cases, the qRT-PCR confirmed an equal or more significant change than the microarrays (data not shown). To confirm that the proteins were overexpressed in thyroid cells rather than stroma, and to confirm that the mRNA expression profiles correlated with protein expression, we performed immunohistochemistry (IHC) for four of the confirmed genes, Runx2, fibronectin1, Prickle1 (data not shown), and cyclin D1 (data not shown). Vimentin, a key regulator of EMT whose gene expression level was not altered in the invasive PTC fronts, was also assessed by IHC. Representative data for Runx2, fibronectin1, and vimentin are shown in Fig. 1B. Expression of each protein was increased in the central regions of PTCs versus normal tissue. IHC staining for RUNX2 and fibronectin was even more intense in the invasive regions, whereas staining for vimentin was similarly intense in the invasive and central areas, consistent with the mRNA data.

Because the functional analysis of gene expression suggested EMT, we next determined whether there was an association between vimentin overexpression and tumor invasion or metastasis in a second larger series of tumors. Vimentin was chosen because of its functional role in EMT and its regulation by EMT-related pathways identified in the microarray, thereby serving to validate the functional analysis of the array data. RUNX2 protein expression was also analyzed in the entire validation series.

Thirty-four PTCs, normal histological tissues from the opposite lobes, and nodal metastases from the 25 tumors that had nodal spread were examined for expression of vimentin mRNA and protein. Twenty three of the tumors were invasive, including 19 with microscopic (T3) and 4 (T4) with gross invasion. Vimentin mRNA was detected in all PTCs and nodes and was nearly absent in normal tissues (data not shown). By IHC (Fig. 2A), vimentin expression was absent in the thyroid cells in the normal tissue. Expression was detected in vascular endothelial cells in these samples, thereby serving as a positive control. In contrast, vimentin was expressed in PTC cells in both the primary tumors and in metastatic nodes (Fig. 2A). Increasing vimentin staining intensity was associated with invasion (P = 0.005), nodal metastasis (P = 0.004), and mutifocality (P = 0.013). There was no association between vimentin staining and tumor size, patient gender, or age. Histopathology and correlation data are summarized in SI Tables 3 and 4.

Fig. 2.

Expression of mesenchymal proteins in PTC. (A) Immunostaining using antibodies against vimentin and osteopontin (OPN) was performed on thyroid samples from normal tissue, primary tumor, and nodal metastasis from individual patients. In comparison with normal tissue, vimentin and osteopontin were overexpressed in primary PTCs and in lymph node metastases in all examined cases. Representative data are shown. (B) Vimentin, OPN, and RUNX2 immunostaining was performed in PTCs that express RET/PTC or mutant BRAF as well as in patients whose tumors did not express either oncogene. Vimentin was overexpressed in all cases, whereas osteopontin and RUNX2 were overexpressed more dramatically in tumors with RET/PTC or a BRAF mutation.

Similar to vimentin, osteopontin has been shown to be functionally important for cancer cell invasion, and its expression is regulated in part by RUNX2, which was overexpressed in the invasive fronts in the microarrays (30). The same PTC samples from 34 patients were evaluated and scored for RUNX2 expression by IHC. Osteopontin expression was analyzed in a subset of the samples. RUNX2 and osteopontin overexpression were detected in the PTCs and metastases (Fig. 2), and RUNX2 expression scores correlated with vimentin scores (P = 0.0025) and nodal metastasis (P = 0.01).

PTCs with RET/PTC or mutant BRAF expression displayed EMT-like mRNA expression profiles in regions of invasion. To determine whether mesenchymal protein expression patterns are associated with RET/PTC expression or NRAS or BRAF mutations, 20 invasive PTCs from the validation series were evaluated for BRAF V600E (five positive) and NRAS (two positive) activating mutations (PCR and sequencing) and for RET/PTC1 (three positive) or RET/PTC3 (five positive) expression (RT-PCR), and immunostained for vimentin, osteopontin, and RUNX2 by using antibodies specific for their targets on Western blot (data not shown). Five tumors did not express any of the four tested oncogenes. All invasive tumors expressed all three proteins regardless of tumor genotype (Fig. 2B); suggesting that expression of EMT-related proteins is common to invasive PTC regardless of tumor genotype.

To determine the functional relevance of EMT in thyroid cancer cells, we next evaluated three poorly differentiated human thyroid cancer cell lines (ARO, WRO, and NPA) for vimentin expression and mesenchymal cell behavior in vitro. In our prior work (22) we noted that the NPA cell line displays a more mesenchymal phenotype than WRO and ARO. Consistent with their ability to invade in a matrigel invasion Boyden chamber assay and their morphology, only the NPA cells expressed vimentin mRNA and protein (Fig. 3A and B). Similarly, NPA cells expressed higher levels of RUNX2 and osteopontin compared with the other cell lines (Fig. 3C).

Fig. 3.

EMT-related mRNA and protein expression in NPA, WRO, and ARO thyroid cancer cells. (A and B) Qualitative RT-PCR (A) and Western blots (B) demonstrate that only NPA cells express vimentin mRNA or protein. (C) Western blot reveals higher levels of osteopontin (OPN) and RUNX2 in the NPA cells. The arrow denotes the size of RUNX2.

We next assessed whether vimentin expression was required for NPA cell invasion by reducing vimentin expression levels using siRNA. In comparison to cells transfected with scrambled control siRNA, cells transfected with vimentin siRNA had a markedly reduced ability to invade (Fig. 4 A–D). The mean (SD) percent reduction of invasion induced by the vimentin siRNA versus control was 74.57% (± 9.8%, P < 0.01). Despite their poorly differentiated nature and ability to form tumors in vivo, ARO cells move in groups, maintain an epithelioid appearance in cell culture, and do not invade efficiently in Boyden chamber assays (Fig. 4D). To determine whether vimentin expression altered the morphology of the ARO cells and/or increased their ability to invade, ARO cells were transiently transfected with vimentin or vector alone. Transfection efficiency and expression of vimentin was monitored by Western blot and IHC by using an anti-vimentin antibody (Fig. 4D). Nuclear morphology was evaluated by using an anti-lamin A antibody (Fig. 5). ARO cells expressing exogenous vimentin developed a fibroblastoid shape associated with “smoothing” of the nuclear envelope contour, similar to wild-type NPA cells with endogenous vimentin (Fig. 4D). Moreover, vimentin expression was sufficient to increase ARO cell invasion in Boyden chambers (Fig. 4E). We therefore conclude that vimentin expression is required for NPA invasion, sufficient for induction of invasion in Boyden chambers in ARO cells, and associated with EMT-like morphology in NPA and ARO cells.

Fig. 4.

Vimentin expression is required for thyroid cancer cell invasion. (A and B) Reduction of vimentin expression using siRNA versus scrambled sequence control is shown by immunofluorescent staining (A) and by Western blot analysis (B) using an anti-vimentin antibody. (C) NPA cell invasion is reduced by vimentin siRNA in comparison to control (scrambled sequence) transfectant. (D) Expression of vimentin in vimentin-null ARO cells induces a mesenchymal morphology (brown stained cells), similar to nontransfected NPA cells. (E) Vimentin expression in ARO cells increases their ability to invade in Boyden Chamber Assays.

Fig. 5.

Vimentin phosphorylation and perinuclear localization is associated with thyroid cancer cell invasion. Immunohistochemical staining was performed for ser-56 phospho–vimentin in NPA cells with endogenous vimentin expression (A) and in ARO cells transiently transfected with vimentin cDNA (B) before and after invasion. Immunofluorescense staining with nuclear envelope marker lamin A/C was performed for colocalization. In the overlap images, there is enhanced yellow staining, representing the colocalization of p-vimentin with lamin A/C in migrated cells.

Vimentin phosphorylation at Ser-55 occurs via PAK and is functionally required for localization into organized intermediate filaments in the perinuclear region and for motility (31). Functional analysis of the microarrays suggested activation of CDC42/PAK signaling in the invasive fronts. To determine whether PAK-modified vimentin was detectable in thyroid cancer cells and was associated with perinuclear vimentin localization, we performed immunofluorescense (IF) on nonmigrated and postmigrated NPA cells and in vimentin-transfected motile ARO cells (Fig. 5). Perinuclear localization was assessed by IF for the nuclear envelope protein lamin, and overlap images for lamin and Ser-55 phospho-vimentin were performed to visualize colocalization. Ser-55 phospho-vimentin colocalized with lamin in migrated NPA and vimentin-transfected ARO cells (Fig. 5). Thus, PAK-modified vimentin appears to be associated with organization of vimentin intermediate filaments in thyroid cancer cell invasion in vitro.

We have shown that an expression profile consistent with EMT is common in PTC invasion. Because overexpression of vimentin is associated with local invasion and metastasis in PTC, and because it plays an important functional role in initiating and maintaining EMT in the three thyroid cancer cell lines studied in vitro, identifying unique pathways that regulate EMT in thyroid cancer may have potential therapeutic implications. The role oncogenes play in initiation of EMT-related changes and the roles of other EMT-related proteins, including osteopontin, RUNX2, and fibronectin 1 remain to be defined; however, recent evidence from in vitro models demonstrate that RET/PTC and BRAF influence the invasive capacity of benign rat thyroid cells through both PI3 kinase and ERK signaling, in part, through regulation of matrix metalloproteinases and osteopontin (16, 32). Whereas we have studied a small panel of thyroid cancer cell lines, studies from other thyroid (33, 34) and nonthyroid cancer cell systems support a role for incomplete or complete EMT as common endpoints in tumor progression (35). Additional studies are needed to determine the role of specific EMT regulators, such as vimentin, in a larger number of cancer cell lines. Our data support a potentially important role for PAK signaling in EMT in PTCs; however, the precise pathways responsible for PAK activation in PTCs remain to be defined. Based on our data, it is likely that pathways regulated by integrin, TGFβ, Notch, NFκb, and small G proteins are all important for PTC invasion. This study focuses on alterations in thyroid cancer cells in two carefully selected regions of the primary tumors. It is likely that invasion represents a continuum between the central portion of the tumor and the invasive front. Finally, there are important interactions among tumor, stromal, and immune cells in thyroid cancer invasion not addressed in the current study that likely influence PTC progression. Characterization of cellular interactions among tumor cells, and among tumor, immune, and stromal cells will be critical to explore in future studies.

Subjects and Methods

Subjects.

Patients were selected who were undergoing thyroidectomy as part of their clinical management for PTC. A series of seven PTCs that were widely invasive through the thyroid capsule into local soft tissue were included (stage T4). All samples were microscopically dissected at the time of surgery by two experienced thyroid pathologist (V.V. and V.S.) after confirmation of tissue diagnosis from frozen section. The final diagnosis was confirmed by hematoxylin–eosin staining after the operation. Samples were taken from the central part of the cancer, from an intratumoral part of the invasive region close to the leading edge, and from histologically normal tissue. On histological examination, >90% of the cells were thyroid cancer epithelial cells. Total RNA was isolated, and quality and quantity were assessed by Agilent Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA). Adequate total RNA was obtained from all seven central and invasion regions, and for four of seven normal tissues. Because this set of samples was from Ukraine, a region that received radiation exposure from Chernobyl (36), the comparison of central vs. normal tissues were compared with nine paired central and normal samples from Ohio State University simultaneously analyzed by using the same reagents and chips. The protocol was approved by the Human Use Committees at The Hospital for Endocrine Surgery in Ukraine and the Uniformed Service University of the Health Sciences and the Institutional Review Board at Ohio State University and conformed to the standards of human research from the Helsinki Mandate.

RNA Isolation and Oligonucleotide Microarrays.

Human Genome U133 Plus 2.0 Arrays (Affymetrix, Santa Clara, CA) were used. Eight micrograms of total RNA per sample was processed according to manufacturer's methods (Affymetrix) and 20 μg of cRNA was hybridized per chip. CEL files were generated from DAT files by using GCOS software (Affymetrix). Gene expression levels were estimated from probe intensities by using RMA, including quantile normalization and background correction (37, 38). Complete microarray data are available at www.ncbi.nlm.nih.gov/projects/geo (accession no. GSE6004).

Statistical Analysis.

Comparisons between the paired normal–central tumor groups and central–invasive groups were performed with BRB Array Tools software (http://linus.nci.nih.gov/BRB-ArrayTools.html) consisting of a modified paired t test with random variance model. For the Ukrainian thyroid samples, probe sets with two-sided P < 0.025 (permutation tests were not applied because of low sample numbers in Ukrainian data) and with at least a median fold change of 1.5 between four normal–central tumor pairs or a median fold change of 1.2 between seven central tumor–invasive pairs were considered statistically significant in each comparison. The lower median fold change for the central tumor–invasive pairs was chosen because of an overall small differential gene regulation within the tumors in comparison with that between normal tissue and the central tumor to allow for similarly sized groups for functional comparison (see below). For the U.S. thyroid samples, permutation tests were conducted in addition to paired t test with random variance model. Probe sets with a two-sided t test P value <0.005 and permutation P value <0.01, having at least a median fold change of 1.75 between pairs, were considered differentially expressed. The median fold change cut off >1.75 between normal and PTC expression in these samples was chosen to allow for sample sizes of comparable length for functional analysis.

Biological interpretation of gene lists was performed by using EASE software (39). For measurement of statistical significance of overrepresented gene categories, “EASE score” <0.05 was chosen. Based on this analysis, functional groups of genes that were differentially expressed were identified. Initial comparisons were made to determine biological themes that might be involved in PTC progression. An unsupervised cluster analysis was performed comparing the ratios of center/normal and invasion/center for all mRNAs to establish genes that were differentially up- or down-regulated. This was performed by grouping of the probe sets that show differences between the normal–center or center–invasion (median fold change >1.5 in normal–center and median fold >1.2 in center–invasive) by using the EPCLUST program (http://ep.ebi.ac.uk) with k-means clustering method to identify five clusters.

Because of the nonnormal distribution of the data, the correlation between vimentin IHC score (see IHC methods) and clinicopathological characteristics (age, tumor diameter, extra-thyroidal extension, number of metastases, and RUNX2 score) in the validation series was assessed by using Spearman correlation (P < 0.05 considered significant). To test the trend of increasing frequency of invasion, metastasis, or multifocality with increasing vimentin score, Cochrane–Armitage trend tests were performed. Invasion and metastasis were the primary endpoints tested. To adjust for two tests, an α = 0.025 was considered significant. All statistical tests were performed with the Biostatistical Shared Resource at Ohio State University Comprehensive Cancer Center.

Tumor Genotype.

Expression of RET/PTC1 or -3 was determined by RT-PCR of the RNA samples from the central part of each tumor as described (40). Total RNA isolated from HEK293 cells transfected with RET/PTC1 or -3 plasmids was used as a positive control. Absence of genomic DNA was proven by analysis of samples in which reverse transcriptase was omitted. BRAF mutation status was evaluated in all tumor samples by analyzing genomic DNA isolated from paraffin blocks from each primary tumor by PCR, followed by direct sequencing of the region including the V600E mutation and the adjacent amino acids that account for all known BRAF mutations in PTC (41). N-RAS mutation status was analyzed by PCR-sequencing as described (42).

qRT-PCR.

Real-time qRT-PCR was performed by using RNA samples isolated for the microarrays for several genes that were of functional interest as described (21). Primers and sequence-specific probes for FN1, RUNX2, PRICKLE1, KFL6, and SLIT1 were designed by using Primer Express ver 1.5 (Applied Biosystems, Burlingame, CA) or purchased for cyclin D1 (Applied Biosystems).

IHC.

IHC using specific antibodies for RUNX2 (AB/RN20; Cemines, Golden, CO), fibronectin (sc-6952; Santa Cruz Biotechnology, Santa Cruz, CA), osteopontin (cat. no. 915-034; Assay Designs, Ann Arbor, MI), and Vimentin (B-6634; Sigma) was performed by using methods as described (22). Negative control experiments omitting primary antibodies were included for each experiment. Antibody specificity was confirmed by Western blot (data not shown). Vimentin and RUNX2 IHC in the entire validation series was scored blinded by two investigators on a 0–3 scale based on intensity because nearly all thyroid cells expressed both proteins when the sample was positive.

Cell Culture and Transfection.

NPA and ARO cells were the generous gifts of G. Julliard (University of California, Los Angeles, CA) (10, 43). Cells were propagated in RPMI medium 1640 supplemented with 5% FCS. Human full-length vimentin cDNA in pCMV-Sport6 expression vector was obtained from American Type Culture Collection (Manassas, VA) (cat. no. 7489939). Cells were incubated with LipoFectamine Plus reagent before incubation with plasmids as described (44). For knockdown experiments, siRNA specific for vimentin and scrambled control was obtained from Santa Cruz Biotechnology. Cells were incubated with Lipofectamine 2000 Reagent (Invitrogen) as per the manufacturer's recommendations (22). Transfection efficiency was monitored in all cases by cotransfection with GFP-labeled siRNA and was assessed by direct visualization. The knockdown experiments were also performed by using the silentGene-2 U6 Hairpin Cloning system (Promega, Madison, WI) as per the manufacturer's recommendations. Vimentin siRNA and a scrambled sequence for vimentin are shown in SI Text.

Protein Isolation and Western Blotting.

Whole-cell proteins were collected, and Western blotting was performed by using 20 μg of protein as described (22, 45). Relative quantitation of proteins was determined by scanning and band densitometry using ImageGauge software (Fuji, Tokyo, Japan). Immunoblots of the same protein samples was performed for β-Actin or GAPDH.

Cell Migration Experiments.

Cell migration was examined by using a Boyden chamber (8-μm pore size), and cells were stained by using a Diff-Quick staining kit (Dade Behring, Newark, DE). Percent migration of total cells was calculated as described (46). All migration experiments were performed on at least three occasions in duplicate.

Immunofluorescense.

Cells were grown on well slides (Labtek, Campbell, CA), fixed with 3.7% formaldehyde, permeabilized with 0.1% Triton X-100, and incubated for 15 min in Vectastain blocking solution. Primary anti-P-Ser⁵⁵-Vimentin (MBL, Nagoya, Japan) and anti-lamin A (Santa Cruz Biotechnology) antibodies were applied for 1 h. Incubation with Alexa Fluor 488 and Texas red-conjugated secondary antibody was for 1 h in the dark. Cover slips were mounted onto the slides with Slow Fade Antifade reagent with DAPI (Molecular Probes, Eugene, OR), dried in the dark for 2 h, and then stored at 4°C. Images were collected with a Zeiss LSM 510 confocal microscope with 488-nm excitation.

Abbreviations

EASE

Expression Analysis Systemic Explorer

EMT

epithelial-to-mesenchymal transition

IHC

immunohistochemistry

PTC

papillary thyroid carcinoma

qRT-PCR

quantitative RT-PCR.