The Long Non-Coding RNA Prader Willi/Angelman Region RNA5 (PAR5) Is Downregulated in Anaplastic Thyroid Carcinomas Where It Acts as a Tumor Suppressor by Reducing EZH2 Activity

Anaplastic thyroid carcinoma (ATC) represents one the most aggressive neoplasias in humans, and, nowadays, limited advances have been made to extend the survival and reduce the mortality of ATC. Thus, the identification of molecular mechanism underlying its progression is needed. Here, we evaluated the long non-coding RNA (lncRNA) expression profile of nine ATC in comparison with five normal thyroid tissues by a lncRNA microarray. By this analysis, we identified 19 upregulated and 28 downregulated lncRNAs with a fold change >1.1 or <−1.1 and p-value < 0.05, in ATC samples. Some of them were subsequently validated by qRT-PCR. Then, we investigated the role of the lncRNA Prader Willi/Angelman region RNA5 (PAR5), drastically and specifically downregulated in ATC. The restoration of PAR5 reduces proliferation and migration rates of ATC-derived cell lines indicating that its downregulation contributes to thyroid cancer progression. Our results suggest that PAR5 exerts its anti-oncogenic role by impairing Enhancer of Zeste Homolog 2 (EZH2) oncogenic activity since we demonstrated that PAR5 interacts with it in thyroid cancer cell lines, reducing EZH2 protein levels and its binding on the E-cadherin promoter, relieving E-cadherin from the negative regulation by EZH2. Consistently, EZH2 is overexpressed in ATC, but not in differentiated thyroid carcinomas. The results reported here define a tumor suppressor role for PAR5 in undifferentiated thyroid neoplasias, further highlighting the pivotal role of lncRNAs in thyroid carcinogenesis.


Introduction
Thyroid cancer (TC) is the most common endocrine malignancy, accounting for 1% of all human cancers, and its incidence has been significantly increased worldwide in the last decades [1]. Based on the histological features, thyroid carcinomas have been classified into well-differentiated papillary (PTC) and follicular thyroid carcinomas (FTC), poorly differentiated (PDC), and undifferentiated anaplastic thyroid carcinomas (ATC) [2]. Among them, ATC is less common but represents the most aggressive and lethal thyroid neoplasia with an overall survival rate of 3-5 months after diagnosis, being resistant to chemo-and radio-therapy [3].
It has been previously reported that p53 impairment represents a feature of ATC [4,5]. More recently, B-RAF mutations have been detected in about 30% of ATC [6,7]. Additionally, it is proven that ATC aggressiveness is further due to mutations of human telomerase reverse transcriptase (TERT) [8]. As far as epigenetic events in ATC are concerned, the Enhancer of Zeste Homolog 2 (EZH2), a member of the polycomb group (PcG) proteins, has been found overexpressed in ATC, but not PTC [9]. Moreover, a significant downregulation of miR-30d, miR-125b, miR-26a, and miR-30a-5p was detected in ATC samples in comparison to normal thyroid (NT) tissues [10].
However, the molecular bases of ATC development are still far from being understood, and, therefore, the unveiling of novel molecular mechanisms involved in ATC carcinogenesis is required to propose an efficient treatment for ATC.
Long non-coding RNAs (lncRNAs) are a class of heterogeneous non-coding transcripts longer than 200 nucleotides [11]. Recently, it has been shown that lncRNAs play a pivotal role in several cellular processes such as differentiation, development, and cellular senescence [12,13]. Accumulating evidence implies that lncRNAs exert regulatory functions at both transcriptional and post-transcriptional levels [14][15][16]. Indeed, several studies have reported that lncRNAs interact with DNA, RNA, and protein to regulate protein function, induce chromatin remodeling and histone modification, acting as a scaffold, and modulate protein stability and DNA methylation [17][18][19]. The deregulation of lncRNAs has been frequently found in many pathophysiological processes, including cancer, underlying their crucial role in human diseases and considering them as potential diagnostic biomarkers for cancer [20,21].
We, recently, demonstrated the deregulation of lncRNA expression in PTC, and that the restoration of the downregulated MPPED2-AS1 (RP5-1024C24.1) reduced cell proliferation and migration in thyroid carcinoma cell lines by modulating the PTEN/Akt pathway [22]. Then, we aimed to analyze the role of lncRNAs also in ATC.
In this study, the analysis of the lncRNAs expression in ATC and NT tissues has allowed the identification of up-and downregulated lncRNAs in ATC samples. We focused our attention on the lncRNA Prader Willi/Angelman region RNA5 (PAR5) significantly and constantly decreased in ATC, but not in PTC tissues. Interestingly, the restoration of PAR5 reduces proliferation and migration rates of two ATC-derived cell lines. Finally, we reported that PAR5 directly interacts with EZH2 in thyroid cancer cells, reducing its binding on the E-cadherin promoter, then relieving E-cadherin from the negative control by EZH2 in PAR5-overexpressing ATC cell lines. These results suggest that PAR5 exerts its tumor suppressor action by impairing the oncogenic activity of EZH2.

Deregulation of lncRNAs Expression in ATC
We analyzed the lncRNA expression profile of nine human ATC tissues with respect to five NT tissues through a single channel Agilent array. This analysis revealed 47 lncRNAs aberrantly expressed in ATC (fold change >1.1 and <−1.1, p < 0.05, FDR ≤ 0.1), including 19 up-and 28 downregulated lncRNAs (Table S1) in comparison with NT tissues. The 10 most up-and downregulated lncRNAs, in terms of fold change are given in Table 1. The fold change and p-value presented for each lncRNA are the ones of the probe with the highest absolute fold change. Then, to validate the microarray results, the expression of three upregulated (MIAT, BCYRN1, BIC) and three downregulated (RMST, PAR5, IPW) lncRNAs was evaluated by qRT-PCR in the same ATC samples used for the microarray analysis. As shown in Figure 1A, the lncRNAs MIAT, BCYRN1, and BIC were found overexpressed whereas RMST, PAR5, and IPW were found significantly downregulated (p = 0.001) in almost all analyzed ATC tissues in comparison with the mean of five NT ( Figure 1B).

The lncRNA PAR5 Was Drastically Downregulated in ATC, but Not in PTC
Subsequently, our attention was focused on the lncRNA PAR5 since it showed a drastic downregulation in all the analyzed ATC samples. Moreover, it has been reported that PAR5 was able to directly bind EZH2 [23] that, interestingly, was found overexpressed in ATC samples, but not in the differentiated thyroid carcinomas [9]. Consequently, PAR5 expression levels were further assessed by qRT-PCR in 11 differentiated PTC and additional 11 undifferentiated ATC samples. Intriguingly, no significant expression changes were detected in the PTC samples when compared to NT tissues (Figure 2A), while PAR5 expression levels were drastically and significantly downregulated in all ATC samples in comparison with NT ones (p = 0.0005) ( Figure 2B), clearly indicating that PAR5 downregulation is strictly related to ATC progression.

PAR5 Overexpression Reduces Cell Proliferation of Thyroid Cancer Cell Lines
To better define the role of PAR5 in thyroid cancer progression we set up functional assays to evaluate its effects on cell proliferation. First, PAR5 levels were evaluated in four ATC-derived cell lines (FB-1, FRO, 8505c, ACT1) and, as expected, its expression was found to be extremely lower with respect to five NT used as control ( Figure 3A). Then, PAR5 expression was restored in FRO and 8505c cells by transfecting them with a PAR5-expressing vector (pCMV-PAR5). As shown in Figure 3B,C, a great increase of PAR5 was detected in both FRO and 8505c cells transfected with pCMV-PAR5, by qRT-PCR, but not in those transfected with the empty vector (EV).
Then, the proliferation rate of PAR5-overexpressing ATC cells was evaluated. A significant delay of growth in the FRO-PAR5 (PAR5 vs. EV at 72 h and 96 h, p < 0.001) and 8505c-PAR5 (PAR5 vs. EV at 96 h, p < 0.001) cells was observed when compared with the corresponding EV-transfected cells ( Figure 3D,E). Subsequently, to corroborate this result a colony formation assay was performed on FRO and 8505c cells transfected with pCMV-PAR5 or EV used as control. As shown in Figure 3F, the colony number of both PAR5-transfected ATC cells was significantly lower in comparison with the EV control-transfected cells, confirming that PAR5 slows down cell growth.

Restoration of PAR5 Expression Decreases Migratory Ability of Thyroid Cancer Cells
To further evaluate the consequence of PAR5 expression on cancer progression we analyzed the migratory ability of the PAR5-overexpressing ATC cells. Transwell migration assays revealed that the restored PAR5 expression significantly reduced migration ability of about 50% (p < 0.01) and 30% (p < 0.0001) in FRO and 8505c cancer cells, respectively ( Figure 4A). To deeper investigate the lncRNA influence on cell migration ability, scratch wound-healing assay was carried out in PAR5-overexpressing ATC cells. As shown in Figure 4B, FRO-EV cells were able to close the wound after 24 h, but this effect was not achieved by FRO-PAR5 cells at the same time. Similar results were obtained in 8505c-PAR5 cells, in which a significant inhibition of migration ability was detected after 24 h ( Figure 4C). These results clearly evidence the role played by PAR5 reduction in the contribution to ATC development.

PAR5 Negatively Modulates EZH2 in Thyroid Cancer Cells
Since it has been reported that PAR5 suppresses cell growth and migration by direct binding EZH2 in glioma cells, regulating the expression of its targets [23], we performed RNA immunoprecipitation (RIP) assays using specific EZH2 antibody in FRO and 8505c cells in order to confirm that the same mechanism occurs also in thyroid cancer system. As expected, we found an enrichment of PAR5 in EZH2 immunoprecipitated lysates when compared to the IgG controls in both cell lines ( Figure 5A,B), confirming that the interaction between PAR5 and EZH2 also occurs in ATC cells. To further examine this mechanism, we assessed EZH2 levels after PAR5 overexpression in FRO and 8505c cells by Western blot and qRT-PCR analyses. The obtained results showed that PAR5 was able to reduce EZH2 protein levels in both PAR5-overexpressing ATC cells ( Figure 5C), but no decrease of EZH2 mRNA levels were found by qRT-PCR in FRO-PAR5 and 8505-PAR5 cells ( Figure 5E), meaning that PAR5 does not affect EZH2 expression at transcriptional level. In fact, despite the higher EZH2 mRNA levels found in the same ATC tissues by qRT-PCR, no correlation between EZH2 mRNA and PAR5 levels was detected (Supplementary Materials, Figure S1A). Subsequently, increased EZH2 protein levels in ATC tissues, that inversely correlate with PAR5 RNA expression ( Figure 5E), were found by Western blot analysis (Supplementary Materials, Figure S1B), suggesting that PAR5 influences EZH2 protein expression. Additionally, the decrease of H3K27me3 levels were detected in FRO-PAR5 and 8505-PAR5 cells by Western blot technique ( Figure 5F). Intriguingly, the reduction of EZH2 protein levels was associated with the upregulation of both mRNA and protein levels of its target gene E-cadherin, a key marker of epithelial-mesenchymal transition (EMT) ( Figure 5G,H).
Subsequently, EZH2 and E-cadherin expression levels were assessed in human thyroid cancers of different histotypes by qRT-PCR. The results confirmed the specific overexpression of EZH2 in the undifferentiated ATC tissues, while no increase was found in PTC samples ( Figure 6A). Conversely a marked downregulation of E-cadherin was found in ATC, whereas no changes in its expression were detected in PTC samples ( Figure 6A). Additionally, as shown in Figure 6B, a significant negative correlation was observed between EZH2 and E-cadherin levels (r = −0.7214, p < 0.0001), meaning that EZH2 negatively regulates the E-cadherin gene. Then, chromatin immunoprecipitation (ChIP) assay was performed to investigate whether PAR5 expression could modify the binding of EZH2 on E-cadherin promoter. To this aim, the crosslinked DNA-protein complexes were immunoprecipitated with a specific antibody raised again EZH2 or IgG. Immunoprecipitation of chromatin was then analyzed by qRT-PCR examining a region spanning nucleotides from −300 to +40 related to the transcription start site (TSS) of the E-cadherin promoter. As shown in Figure 6C,D, anti-EZH2 antibodies precipitated human E-cadherin promoter in both EV-transfected cell lines but not in those overexpressing PAR5. Next, the effects of EZH2-PAR5 interaction on the modulation of the E-cadherin promoter were evaluated by performing luciferase assays. To this aim, ATC cell lines were transfected with a vector encoding the luciferase gene under the control of the E-cadherin promoter. Then, FRO and 8505 cells were transfected with single or both vectors encoding for EZH2 protein or PAR5 gene. As shown in Figure 6E,F, PAR5 and EZH2 behaved in an opposite way on the modulation of E-cadherin promoter activity. In fact, EZH2 clearly displayed a negative transcriptional regulation on E-cadherin promoter. On the other hand, when both genes were co-transfected, a partial rescue of the transcriptional E-cadherin activity was obtained, showing comparable levels to those observed in ATC cells expressing only the PAR5 gene. These results indicate that PAR5 not only reduces the EZH2 protein levels, but also decreases EZH2 binding to the E-cadherin promoter, thereby, affecting the expression of its target genes. (A,B) RNA immunoprecipitation (RIP) assay was performed on extracts obtained from FRO and 8505c cells transfected with pCMV-PAR5 or the EV, using an anti-EZH2 antibody or a pre-immune (IgG) serum, as control. qRT-PCR was performed on the samples to assess fold enrichment of PAR5 precipitated by the anti-EZH2 antibody relative to IgG control, set equal to 1. (C) Western blot analysis of EZH2 expression in FRO and 8505c stably expressing PAR5 or carrying the corresponding EV. β-actin was used to normalize the amount of loaded protein. Densitometric analysis was performed by using ImageJ software to analyse protein expression compared to the EV, set equal to 1. (D) qRT-PCR analysis to evaluate the expression of EZH2 after PAR5 transfection. Data were reported as relative expression ± SD and were compared to the EV, set equal to 1. (E) Correlation scatter plot (Spearman's Rank) between qRT-PCR levels of PAR5 and Western blot levels of EZH2 analyzed in six ATC samples (r = −0.3143; p = 0.5639). (F,G) Western blot analysis of H3K27me3 and E-cadherin expression in FRO and 8505c stably expressing PAR5 or carrying the corresponding EV. H3 and β-actin were used to normalize the amount of loaded protein. Densitometric analysis was performed by using ImageJ software to analyze protein expression compared to the EV, set equal to 1. (H) qRT-PCR analysis to evaluate the expression of E-cadherin after PAR5 transfection. Data were reported as relative expression ± SD and compared to the EV, set equal to 1. Chromatin immunoprecipitation (ChIP) assay was carried out in FRO and 8505c cells transiently transfected with PAR5 or the corresponding EV. ATC cell lines were then crosslinked, sonicated, and subjected to pre-clearing. The chromatin was immunoprecipitated using antibodies against EZH2. IgG were used as negative control. The immunoprecipitated chromatin was analyzed by qPCR assay with primers specific for the E-cadherin promoter. (E,F) Luciferase assays performed in ATC cells transfected with the E-cadherin-luc vector and with single or both vectors expressing EZH2 or PAR5. The total amount of the transfected DNA was balanced with the EV. Relative E-cadherin luciferase activity was compared to that observed in cells transfected with the only EV, assuming that the control is equal to 1. Values are the mean of three independent experiments ± SD.

Discussion
In this study we evaluated the contribution of lncRNAs to thyroid cancer progression by examining their expression profile in nine ATC and five NT samples through microarray analysis. By this approach, we discovered that several lncRNAs were differently expressed in ATC tissues in comparison with NT, underlying once more their crucial role in thyroid carcinogenesis. We, then, confirmed the microarray results by examining the expression levels of three up-and three downregulated lncRNAs by qRT-PCR. Among these lncRNAs, we focused our attention on the downregulated PAR5, previously proposed as a novel tumor suppressor lncRNA. Indeed, recent studies reported a drastic PAR5 downregulation in a specific hepatitis C virus-related hepatocellular carcinoma (HCC) and gliomas [23,24]. Interestingly, a significant inverse correlation between PAR5 levels and the malignant phenotype of glioblastoma multiforme (GBM) was described [25] underlying the critical role of PAR5 reduction in cancer progression. Noteworthy, it has been also reported that PAR5 can delay malignant transformation by interacting with EZH2 protein, a catalytic subunit of PRC2, that triggers H3K27 trimethylation, thereby inducing transcriptional repression of target genes implicated in important biological processes, such as cell proliferation, stem cell pluripotency, and oncogenic transformation [26][27][28][29]. Intriguingly, a significant number of studies evidenced a strict correlation of EZH2 overexpression with patient poor outcome and high aggressiveness in several cancer types [28,30,31]. Supported by this evidence, we investigated the possible functional role of PAR5 in thyroid carcinogenesis.
To achieve this, we first analyzed PAR5 expression levels by qRT-PCR in a set of differentiated and undifferentiated thyroid carcinoma samples. The results of this analysis showed a strong and constant PAR5 downregulation in ATC tissues, whereas no significant changes were observed in PTC samples, meaning that PAR5 downregulation is associated with the highly malignant phenotype also in thyroid cancer. Next, we examined the effects of PAR5 overexpression on cellular growth and migration capability in two ATC-derived cell lines. We found that PAR5 overexpression was able to reduce cell proliferation and cell migration ability, thus supporting the anti-oncogenic role played by this lncRNA and, thereby, the contribution of its downregulation to ATC development.
To investigate the molecular mechanisms by which the lncRNA acts in thyroid carcinogenesis we examined the effects of PAR5 on EZH2 expression and activity since recent studies reported EZH2 overexpression in undifferentiated thyroid carcinomas, and its capability to inhibit thyroid cell differentiation [9]. Therefore, by performing RIP assays we demonstrated that PAR5 directly binds EZH2. Moreover, we observed that PAR5 restoration reduced EZH2 protein and H3K27me3 levels in absence of any change of EZH2 mRNA abundance, suggesting that this lncRNA may affect EZH2 stability by a regulatory mechanism already reported for other lncRNAs [32,33]. Consistently with these findings, we found an increased EZH2 expression in ATC tissues and, interestingly, we observed a negative correlation between EZH2 protein and PAR5 RNA levels, strongly supporting the effects of PAR5 on EZH2.
Since it is well known that E-cadherin, whose lack is a hallmark of EMT [34][35][36][37], is repressed by EZH2 through H3K27 trimethylation at its promoter [29], we analyzed E-cadherin expression in the ATC-PAR5 overexpressing cells, and increased E-cadherin protein levels were observed. Accordingly, PAR5 restoration reduces the EZH2 binding to E-cadherin promoter thus preventing the EZH2 inhibitory activity on E-cadherin expression. Therefore, these results indicate that PAR5 exerts its tumor suppressor activity by impairing the EZH2 oncogenic role. However, further investigations are needed to better understand how PAR5 can regulate EZH2 activity. Equally, this study does not exclude that PAR5 exerts its tumor suppressor activity interacting with other proteins involved in cancer progression.
Moreover, our study demonstrates that PAR5 modulates EZH2 expression. It is worth to note that other non-coding RNA, including microRNA and circular RNA, can modulate the expression of EZH2. Indeed, it has been reported that miR-25, miR-30d, and miR26a (downregulated in ATC) target EZH2 [38,39]. Additionally, circ-PRMT5 can upregulate EZH2 via sponging miR-377-382-498 in non-small cell lung cancer (NSCLC) [40]. Intriguingly, also many lncRNAs have been described to participate in the EZH2 oncogenic regulatory network [41,42], further highlighting the complex regulation of EZH2 activity in human cancer, and reinforcing its critical role in cancer progression.
Noteworthy, other significant deregulated lncRNAs came out from the microarray analysis. In fact, it would be worthwhile to better investigate the role in ATC development of the downregulated rhabdomyosarcoma 2-associated transcript (RMST), a lncRNA able to regulate neurogenesis through the interaction with SOX2 (a critical player in pluripotent stem cell) [43]. RMST also exerted anti-oncogenic role in triple-negative breast cancer (TNBC) since it was found downregulated in TNBC tissues, where its restoration inhibits cell proliferation, cell cycle, invasion, and migration in TNBC cells [44]. Additionally, it has been found that this lncRNA is even able to enhance DNMT3 expression by interacting with HuR, thus leading to an aberrant DNA methylation pattern [45]. Interestingly, our preliminary results show that also RMST expression was decreased in ATC samples, without any change in PTC tissues. It is quite interesting to observe that lncRNAs, similarly to miRNAs [10,46], show an opposite behavior in the different subtype of thyroid carcinomas, in which their expression was found upregulated or unchanged in PTC and downregulated in the most aggressive undifferentiated subtypes, highlighting the critical role of lncRNA deregulation in the process that leads to ATC development.
In conclusion, our findings demonstrate that PAR5 downregulation was strictly related to the undifferentiated thyroid carcinomas and that PAR5 expression affected cell growth and migration rate by interacting with EZH2 and inhibiting its oncogenic activity.

Array Design and Data Analysis
The arrays utilize nucleic acid hybridization of a 52 nt biotin-labeled cDNA target with DNA oligonucleotide probes attached to a gel matrix. The biotin-labeled cDNA targets were prepared by a reverse transcription into first strand cDNA. Total RNA extracted from nine ATC and five NT was primed for reverse transcription by a random octamer conjugated with two biotins and a 52 nt long poly-A tail. This procedure results in an equal copy number of biotin cDNA targets to the ncRNA templates [47]. The array is a customized single-channel Agilent array (Agilent Technologies, Santa Clara, CA, USA). It contains a collection of probes (sense and antisense) for various types of non-coding RNAs (18009 probes corresponding to 1271 human pre-miRNAs, 8660 probes corresponding to 626 mouse pre-miRNAs (miRBase 21), 2745 probes corresponding to 479 ultra-conserved elements, 16,314 probes corresponding to 1283 pyknon, and 2197 probes corresponding to 97 long non-coding RNAs). Data preprocessing steps of background-correction, normalization, and summarization were performed in R version 3.5.1 (https://www.r-project.org/) using functions in Limma library (http://www.bioconductor.org/packages/devel/bioc/vignettes/limma/inst/doc/usersguide.pdf).
A threshold for positive spot selection for microarray data was calculated as the mean value of all the dark corner spots plus twice the standard deviation [48]. For class comparison we employed again Lima library, a linear model was fitted to each gene, and empirical Bayes methods were used to obtain the statistics. The statistical significance was defined as a p-value < 0.05 and we imposed a cut-off of functional relevance on the fold change in absolute value of 1.1.

Cell Culture and Transfections
Human thyroid cancer cell lines FRO and 8505c were grown in DMEM (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% fetal bovine serum (FBS) (Euroclone, Milan, Italy), 1% L-glutamine, 1% penicillin/streptomycin (Sigma-Aldrich). Cells were maintained at 37 • C under 5% CO 2 atmosphere. Lipofectamine 2000 (Life Technologies, Grand Island, NY, USA) reagent was used to transfect the cells according to the manufacturer's instructions. The stable-transfected FRO and 8505c cells were selected in a medium containing 1200 and 1000 µg/mL of G418 (Life Technologies), respectively.

RNA Extraction and Quantitative Real-Time PCR (qRT-PCR)
Total RNA from thyroid cancer tissues and cell lines were extracted using Trizol reagent (Life Technologies). A total of 1 µg of total RNA from each sample was used to obtain double strand cDNA with the QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany). Quantitative Real-Time PCR (qRT-PCR) was performed with the CFX96 thermocycler (Bio-Rad, Hercules, CA, USA) in 96-well plates. For each PCR reaction, 10 µL of 2× Sybr Green (Bio-Rad), 200 nM of each primer, and 20 ng of the cDNA, previously generated, were used. The oligonucleotides for qRT-PCR, encompassing exon-exon junctions, were purchased from Integrated DNA Technologies (San Diego, CA, USA) and designed with Primer-BLAST software. Relative gene expression was determined using comparative C(T) method, as described elsewhere. Peptidylprolyl isomerase A (PPIA) was used as housekeeping gene [49,50]. Detailed primer sequences are available as Supplementary Materials, Table S2.

Plasmids
The expression vector encoding human PAR5 gene was generated by cloning cDNA sequence in the pCMV6-AC-GFP vector (Origene Technologies, Rockville, MD, USA) by using In-Fusion cloning method (Takara, Beijing, China) and HindIII restriction site. After cloning, the plasmid was subjected to sequencing (Eurofins Genomics, Vimodrone, Italy) and PAR5 expression was validated by qRT-PCR analysis.

RNA Immunoprecipitation (RIP) Assay
RIP experiments were performed using the Magna RIP™ RNA-Binding Protein Immunoprecipitation Kit (Millipore, Billerica, MA, USA) according to the manufacturer's protocol. Briefly, FRO and 8505c cells were lysed in RIP lysis buffer. First, 5 µg of human anti-EZH2 antibody (#4905, Cell signaling, Danvers, MA, USA) and normal rabbit IgG (Millipore) used as negative control, were incubated with magnetic beads for 30 min. Then, 100 µL of whole lysates were incubated overnight on a rocking platform at 4 • C. The next day, samples were incubated with Proteinase K buffer at 65 • C for 30 min and then immunoprecipitated RNA was purified. Purified RNA was reverse transcribed into cDNA by using random primer with the QuantiTect Reverse Transcription Kit (Qiagen), and the presence of PAR5 transcripts was detected by qRT-PCR. Enrichment of PAR5 transcripts in the EZH2 immunoprecipitated lysate was calculated relative to the levels of PAR5 transcript in the rabbit IgG control sample, set equal to 1 [51,52].

Chromatin Immunoprecipitation (ChIP) Assay
ChIP experiments were performed in ATC cell lines transiently transfected with PAR5-overexpressing vector. Briefly, 48 h after transfection, 5 × 10 6 FRO and 8505c cells were cross-linked to fix the DNA/protein complexes using 1% formaldehyde at RT for 10 min and the reaction was then stopped by adding glycine at a final concentration of 0.125 M. Cells were lysed in 300 µL of buffer containing 10 mM EDTA, 50 mM Tris-HCl pH 8.0, 1% SDS and protease inhibitors and then sonicated three times for 30 cycles (30 sec On, 30 sec Off) at maximum settings (Bioruptor TM Next Gen, Diagenode Inc., Denville, NJ, USA), obtaining fragments between 0.3 and 1.0 kb. After centrifuging samples at 14,000 rpm for 15 min at 4 • C, 3% of supernatant amount was used as control of the total chromatin obtained (input), and the remaining part of the sample was diluted 2.5-fold in IP buffer (100 mM NaCl, 2 mM EDTA pH 8.0, 20 mM Tris-HCl pH 8.0, 0.5% Triton X-100 and protease inhibitors). After 3 h of pre-clearing at 4 • C with protein A-Sepharose saturated with salmon sperm (Millipore), samples were mixed overnight at 4 • C with the EZH2 antibodies (#4905, Cell signaling) and rabbit normal IgG (sc-2027, Santa Cruz Biotechnology, Santa Cruz, CA, USA). Subsequently, the DNA-protein-antibodies complexes were immunoprecipitated with the protein A previously used and then the chromatin was released from the beads through 30 min incubation with 250 µL of 1% SDS, 0.1 M NaHCO 3 at 37 • C and finally with 200 nM NaCl at 65 • C overnight. Subsequently, 10 µL of 0.5 mM EDTA, 20 µL of 1 M Tris-HCl pH 6.5 and 20 µg of Proteinase K were added to the reaction tube and then the complexes were incubated for 1 h at 45 • C. DNA from chromatin immunoprecipitation was purified by phenol/chloroform extraction (Life Technologies) and precipitated by adding two volumes of ethanol and 0.1 M CH 3 COONa. IgG were used as non-specific control and input DNA values were used to normalize the values from ChIP samples. The percentage of IP chromatin was calculated as 2 −∆Ct × 3, where ∆Ct is the difference between Ct input and Ct IPsample, and 3 is the percentage of total sample used for the input. The relative abundance of immunoprecipitated chromatin was expressed as percentage of binding of interested promoter compared to the input. Detailed primer sequences are available as Supplementary Materials, Table S2.

Luciferase Assay
First, 3. × 10 5 ATC cells were seeded in a 6-well plate, transfected with 100 ng of the E-cadherin-luciferase reporter gene [53] and expression vectors encoding the EZH2 and PAR5. Then, 24 h after transfection, cell extracts were prepared and the luciferase activity was measured by using a Lumat LB9507 luminometer (Berthold Technologies, Bad Wildbad, Germany) and the Dual-Luciferase Reporter System kit (Promega, Fitchburg, WI, USA). A vector expressing Renilla gene under the control of the cytomegalovirus (CMV) promoter was used to normalize transfection efficiency. For all transfections, the total amount of the transfected DNA was balanced with the empty vector.

Cell Migration Assay
Transwell migration assays were performed using transwell chamber (8-µm pores). Briefly, thyroid cancer cells (3 × 10 4 ) were plated in the upper transwell chamber in serum-free medium. Then, 0.3 mL complete medium was added in the lower chamber as chemoattractant. After 24 h of incubation, migrated cells on the membrane of the chambers were fixed and stained with crystal violet solution (crystal violet 0.05%, methanol 20%). FRO and 8505c cells were also seeded in a 96-well plate to normalize the number of the cells. After 2 h, the absorbance at 490 nm was read using cell titer (Promega) in a microplate reader (LX800, Universal Microplate Reader, BioTek Instruments, Inc., Winooski, VT, USA). Then, crystal violet in the chamber was de-stained with PBS-0.1% SDS solution and was read at 590 nm. Results were obtained by normalizing the crystal violet values to cell titer ones.

Wound-Healing Assay
Anaplastic thyroid cancer cells overexpressing PAR5 or carrying EV (as control) were grown to confluence in 60 mm plates in DMEM medium containing 10 µg/mL Mitomycin C (#M4287, Sigma-Aldrich) for 2 h to completely inhibit cell proliferation. A linear scratch was made on the monolayer using a P200 pipette tip. The cells were then washed with PBS three times, and further cultured in DMEM medium. After 24 h of incubation, the gap size of scratch was measured and recorded, and then compared with the initial gap size at 0 h. Using the ImageJ 1.43 image processing program (NIH), the wound closed area was determined at each time point from the digital images.

Colony Formation Assay
FRO and 8505c cells at 80% of confluency in 6-well plates were transfected with pCMV6-AC-GFP-EV and pCMV6-AC-GFP-PAR5. After 48 h of transfection, cells were treated with 1200 µg/mL and 1000 µg/mL G418, respectively. Medium containing G418 was refreshed every two days and, after 3 weeks of antibody selection, cells were fixed and stained with a solution containing crystal violet. Then, the stained crystal violet was resolved with PBS-0.1% SDS and absorbance at 590 nm was determined. The data were performed in triplicate and showed as mean ± SD.

Statistical Analysis
All data were reported as mean ± standard deviation (SD). Results were analyzed using Student's t-test, Mann-Whitney's test and ANOVA test, when required. The correlations were evaluated through non-parametric Spearman's Rank correlation coefficient with 95% confidence interval. Statistical analyses were performed using GraphPad Prism software 6.0 and the difference was considered significant when p < 0.05.

Conclusions
In conclusion, the results reported here strongly support the key role played by the lncRNA PAR5 during thyroid carcinogenesis. Indeed, we showed that the reduction of PAR5 was strictly related to the most aggressive ATC phenotype, where its restoration is important to inhibit the main feature of thyroid cancer cell lines. Additionally, we showed that PAR5 acted by impairing the activity of EZH2, relieving the downstream target gene E-cadherin from EZH2 suppression. Our findings thus demonstrated the anti-oncogenic role of PAR5 in thyroid carcinogenesis, suggesting a new mechanism of action by which the lncRNAs impair cancer progression.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/1/235/s1, Figure S1: Analysis of EZH2 expression in ATC samples, Figure S2: Originals blots for Western blot relative to β-actin, H3K27me3, H3, E-cadherin, and EZH2 in FRO and 8505c cell lines, Figure S3: Originals blots for Western blot relative to β-actin and EZH2 in ATC and NT samples, Table S1: Whole list of lncRNAs deregulated in ATC vs. NT, Table S2: List of primers used for the following analysis.