Cell Cycle Deregulation and TP53 and RAS Mutations Are Major Events in Poorly Differentiated and Undifferentiated Thyroid Carcinomas

Background:

Anaplastic thyroid carcinomas (ATCs) are among the most lethal malignancies, for which there is no effective treatment.

Objective:

In the present study, we aimed to elucidate the molecular alterations contributing to ATC development and to identify novel therapeutic targets.

Design:

We profiled the global gene expression of five ATCs and validated differentially expressed genes by quantitative RT-PCR in an independent set of tumors. In a series of 26 ATCs, we searched for pathogenic alterations in genes involved in the most deregulated cellular processes, including the hot spot regions of RAS, BRAF, TP53, CTNNB1 (β-catenin), and PIK3CA genes, and, for the first time, a comprehensive analysis of components involved in the cell cycle [cyclin-dependent kinase (CDK) inhibitors (CDKI): CDKN1A (p21^CIP1); CDKN1B (p27^KIP1); CDKN2A (p14^ARF, p16^INK4A); CDKN2B (p15^INK4B); CDKN2C (p18^INK4C)], cell adhesion (AXIN1), and proliferation (PTEN). Mutational analysis was also performed in 22 poorly differentiated thyroid carcinomas (PDTCs).

Results:

Expression profiling revealed that ATCs were characterized by the underexpression of epithelial components and the up regulation of mesenchymal markers and genes from TGF-β pathway, as well as, the overexpression of cell cycle-related genes. In accordance, the up regulation of the SNAI2 gene, a TGF-β-responsive mesenchymal factor, was validated. CDKN3, which prevents the G₁/S transition, was significantly up regulated in ATCs and PDTCs and aberrantly spliced in ATCs. Mutational analysis showed that most mutations were present in TP53 (42% of ATCs; 27% of PDTCs) or RAS (31% of ATCs; 18% of PDTCs). TP53 and RAS alterations showed evidence of mutual exclusivity (P = .0354). PIK3CA, PTEN, and CDKI mutations were present in 14%–20% of PDTCs, and in 10%–14% of ATCs. BRAF, CTNNB1, and AXIN1 mutations were rarely detected.

Conclusion:

Overall, this study identified crucial roles for TP53, RAS, CDKI, and TGF-β pathway, which may represent feasible therapeutic targets for ATC and PDTC treatment.

Most thyroid neoplasias, arising from follicular cells, are well-differentiated tumors (WDTCs), namely papillary (PTC) and follicular (FTC) carcinomas, which can be successfully treated with surgery and radioiodine. On the other hand, poorly differentiated (PDTC) and anaplastic thyroid carcinomas (ATC) are dedifferentiated tumors and present a more aggressive behavior. In particular, ATC rapid onset, extensive invasion, and distant metastases contribute to be one of the most lethal malignancies, with a median survival of 3–4 months (1). In addition, most current therapeutic options have been ineffective.

PDTCs and ATCs have been identified in coexistence with well-differentiated areas, suggesting that these tumors can arise from preexisting PTC and/or FTC cases (2–4). This process of dedifferentiation is supported by the detection, in PDTCs and ATCs, of BRAF and RAS gene mutations (commonly found in WDTCs) in association with later acquired alterations in TP53, CTNNB1 (β-catenin), and PIK3CA genes (2–6).

By global gene expression profiling, authors have described that compared with WDTC cases, PDTCs and ATCs exhibit deregulation of different cellular events, eg, focal adhesion, cell motility, TGF-β signaling, chromosome segregation, cell cycle, and proliferation (7–9). Previously we have compared the transcriptional profiles of PDTC, WDTC, and normal thyroid samples and found that PDTCs presented molecular signatures mainly related to cell proliferation, poor prognosis, spindle assembly checkpoint, and cell adhesion (10).

In the present study, to further elucidate the main molecular pathways and alterations contributing to PDTCs and ATCs, we profiled the ATC gene expression and analyzed the mutational status of N-, H-, K-RAS, BRAF, TP53, CTNNB1, and PIK3CA genes in a series of 48 tumors (26 ATCs and 22 PDTCs), four ATC-, and two PDTC-derived cell lines. Additionally, in accordance with the deregulated pathways in ATCs and PDTCs, we sequenced the coding regions and splice sites of five cyclin-dependent kinase inhibitor (CDKI)-encoding genes [CDKN1A (p21^CIP1), CDKN1B (p27^KIP1), CDKN2A (p14^ARF, p16^INK4A), CDKN2B (p15^INK4B), CDKN2C (p18^INK4C)] and the PTEN gene, which are critical regulators of the cell cycle/proliferation. We also analyzed the AXIN1 gene, which encodes a negative regulator of WNT signaling pathway, and was found to be mutated in 81.8% of ATCs from a Japanese series (11).

Materials and Methods

Tissue samples

A total of 11 classical variants of PTC (cPTC), 12 follicular variants of PTC (fvPTC), 11 FTCs, and seven normal thyroid tissues samples (all taken from the opposite lobe of thyroid tumors) were used in the study. All these samples, 12 PDTCs and 15 ATCs were obtained at the time of surgery and were immediately frozen in liquid nitrogen. Eight PDTCs and five ATCs were preserved as formalin-fixed paraffin-embedded samples. Two PDTCs and seven ATCs were collected during fine-needle aspiration biopsies (FNABs), being conserved in RLT buffer (RNeasy mini kit; QIAGEN) with 1% (vol/vol) 2-mercaptoethanol and maintained at −70°C. Histological classifications followed previously described criteria (12). A pool of human thyroid total RNA (BD Bioscience), two PDTC-derived (T243 and T351) and four ATC-derived (T235, T238, T241, and C643) cell lines were also used.

All samples were obtained with permission, and the project was approved by our institutional ethical committee.

Array hybridization and data analysis

RNA integrity was assessed by microcapillary electrophoresis (Agilent 2100 Bioanalyzer). Samples were processed following the whole transcript sense target labeling assay from Affymetrix and were hybridized in a GeneChip Gene 1.0 ST array (Affymetrix).

Microarray data analysis

Partek Genomics Suite Software (Partek Inc) was used for the unsupervised hierarchical clustering of the samples, applying Pearson's dissimilarity and Ward's clustering method. A robust multiarray average method (13) was first used for array data normalization and expression levels determination.

DNA-Chip Analyzer (dChip) 2010.01 software (14) was used to obtain differentially expressed genes between ATC and normal thyroid samples. Arrays were normalized with the invariant set normalization method and gene-level expression was determined by summarizing the multiple probes across the gene (median of 26 probes per gene) into a single gene level-probe set expression, using model-based expression analysis with a perfect match-only model. Gene level-probe sets that were absent in all samples or those that did not change across samples (coefficient of variation lower than 0.2 and higher than 10) were eliminated from further analysis. Gene level-probe sets were considered to be differentially expressed, with a lower 90% limit of the confidence interval of the fold change (ratio of the expression level in the two groups), equal or higher than 2-fold, and with an unpaired t test considered significant at P ≤ .001. Onto-Express (15) and Pathway-Express (16) from the Onto-Tools package were used for functional profiling according to cellular components and for pathway impact analysis of differentially expressed genes, respectively.

Gene set enrichment analysis software (17), using the GenePattern platform (18), was applied to the complete list of 33 252 probe sets to determine Gene Ontology-defined gene sets (groups of genes, which share common features) associated with increased or decreased expression in ATCs. Statistical significance was estimated by a nominal P value obtained by gene set permutation (70 000 permutations performed). P values were corrected for multiple hypothesis testing, using a false discovery rate and family-wise error rate. Gene sets were considered significant at P ≤ .05 and a false discovery rate of 0.25 or less.

A Venn diagram analysis was performed using the GenePattern platform (18).

The expression data set has been deposited in the National Center for Biotechnology Information's Gene Expression Omnibus (GEO) and is accessible through GEO series accession number GSE53072.

Mutational analysis

Primers were designed to amplify hot spot regions of BRAF (exon 15), NRAS, KRAS (exons 2 and 3), HRAS (exons 1 and 2), PIK3CA (exons 1, 9, and 20), CTNNB1 (exon 3), and TP53 (exons 5–9) genes. Analysis of these genes was performed in the 22 PDTCs, 26 ATCs, and six cell lines. The entire coding sequence and exon-intron boundaries were sequenced for CDKN2A (both p14^ARF and p16^INK4A transcripts), CDKN2B, CDKN2C, CDKN1A, CDKN1B, AXIN1, and PTEN genes. All genes but AXIN1 were analyzed in 20 PDTCs, 22 ATCs, and the six cell lines. Due to the high guanine and cytosine (GC) content, exon 1 of the PTEN gene could not be assessed in the formalin-fixed paraffin-embedded samples. Details of the mutational analysis and nucleic acid extraction are described in the Supplemental Methods, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org. The primer sequences and assay conditions are available on request.

Splice variants analysis and quantitative RT-PCR

The presence of the CDKN3 splice variants was searched in RNA from five normal thyroid tissues, commercial pool of human thyroid, three fresh-frozen ATCs, and six cell lines. Splice variants were sequenced after amplification and separation by subcloning (details are supplied as Supplemental Methods).

CDKN3 quantitative RT-PCR analysis was performed in eight cPTCs, eight fvPTCs, eight FTCs, seven fresh-frozen PDTCs, nine ATCs (five FNABs and four fresh frozen), six cell lines, six normal thyroid tissues, and a commercial pool of human thyroid RNA. The sequences of the primers used are supplied in Supplemental Table 1. Seven cPTCs, seven fvPTCs, seven FTCs, 12 PDTCs (one FNAB and 11 fresh frozen), 19 ATCs (three FNABs and 16 fresh frozen), six cell lines, seven normal samples, and a commercial RNA pool were used for SNAI2 quantitative analysis. Details of the quantitative RT-PCR analyses are supplied as Supplemental Methods.

Statistical analysis

GraphPad Prism version 4.00 (GraphPad Software, Inc) was implemented for statistical analysis and for comparison of survival distributions by log rank test. Differences were considered significant at P ≤ .05.

Results

ATC transcriptome analyses

We determined the ATC global gene expression profile, using RNA extracted from three fresh-frozen tumors and two FNABs, and compared it with three normal thyroid tissues and a commercial pool of normal human thyroid RNA. Unsupervised analysis using hierarchical clustering showed that all five ATC samples clustered together, separated from the normal samples (Figure 1). Gene set enrichment analysis identified sets of genes with increased expression in ATCs, markedly related to cell cycle and division processes, checkpoints, and chromosome segregation (Supplemental Table 2). Underexpressed genes in ATCs were mainly related to tight junctions, intercellular junctions, and oxidoreductase activity (Supplemental Table 3).

Gene expression comparison analysis showed that ATC and normal thyroid samples had 1333 differentially expressed genes, defined as those with expression level (set at the lower limit of 90% confidence) equal to, or higher, than 2-fold between the defined two groups, with a statistical significance of P ≤ .001. Among these genes, 983 (74%) were down regulated and 350 (26%) were up regulated in ATCs. Pathway impact analysis pointed out several deregulated pathways in ATCs (Table 1). Adherens junction (Supplemental Figure 1), tight junction (Supplemental Figure 2), and focal adhesion (Supplemental Figure 3) pathways, comprised mainly down regulated genes, contrary to cell cycle, which was completely associated with overexpressed genes (Supplemental Figure 4). The TGF-β signaling pathway (Supplemental Figure 5) was associated with an increased expression of genes involved in TGF-β ligand reception but to decreased expression of components from bone morphogenetic protein signaling. The functional profiling of the differentially expressed genes revealed the involvement of different cell components, such as the cytoskeleton, spindle, microtubules, chromosomes, and the aforementioned cell junctions (Supplemental Table 4).

Previously we analyzed the gene expression of cPTCs, fvPTCs, FTCs, and PDTCs (10) (microarrays data accessible through GEO series accession number GSE53157). To compare the ATC gene expression with our previous work, we searched for differentially expressed genes relative to normal thyroid that ATCs had in common with each other tumor type (Figure 2). Of the 1333 genes in ATCs, 71 were common to cPTCs, 31 to fvPTCs, 30 to FTCs, and 34 to PDTCs. Six genes were shared between all tumors (Supplemental File 1). Exclusion of all these common genes left 1216 genes that are specifically deregulated only in ATCs (Supplemental File 2).

Validation of gene expression results

Cyclin-dependent kinase inhibitor 3 (CDKN3) was the most overexpressed gene in ATCs. We searched for the presence of CDKN3 abnormal splice variants, associated with this up regulation. The RT-PCR amplification of CDKN3 and the separation of each of the amplicons by subcloning revealed two expression patterns: in six normal thyroid tissues, only CDKN3 full-length transcripts were present, whereas in three ATC samples, two PDTC cell lines, and four ATC cell lines, we detected full-length transcripts in combination with splice variants (Figure 3A). These splice variants were characterized by the skipping of the entire exon 2. Quantitative RT-PCR assays allowed the differential analysis of the expression levels due to splice variants or due to full-length transcript in 46 thyroid tumors and seven normal samples (Figure 3B). Despite a higher total CDKN3 (full length transcript plus splice variant) expression in some WDTCs, and in PDTCs (P < .05) relative to normal samples, most of the tumors expressed only the full-length transcript. In ATC tumors and cell lines, we also found a statistically significant overexpression of total CDKN3 transcription, relative to normal tissue (P < .01 and P < .001, respectively), but in these groups, total CDKN3 quantification was significantly higher than the correspondent quantification of the sole full-length transcript (P = .0122 and P = .0111, respectively). So, contrary to WDTCs and PDTCs, the CDKN3 up regulation in the ATCs and cell lines was associated with an expression of abnormal splice variants.

We used TaqMan RT-PCR to further validate our microarray results for SNAI2 gene in a set of thyroid tumor samples and cell lines (Figure 4). In agreement with the global gene expression, SNAI2 was significantly more expressed in ATCs than in normal tissues (P < .01). A subgroup of eight ATCs was clearly distinguished, with relative expression levels higher than 3-fold. The ATC cell line (C643) and a minimally invasive FTC also had more than 3-fold expression. SNAI2 expression was not correlated to age at diagnosis or tumor size in ATCs (data not shown).

ATC and PDTC mutational analysis

A total of 22 PDTCs, 26 ATCs, two PDTCs cell lines, and four ATC cell lines were screened for sequence variations in 14 genes. Overall, 59 different alterations, totaling 72 mutated sequences, were identified (Supplemental Table 5) and were present in 13 PDTCs (59%) and 20 ATCs (77%). Eight of the alterations were predicted in silico to be benign and were not taken into account. Three variants identified in the AXIN1 gene in tumor samples were also present in the corresponding normal tissues. RAS, TP53, PTEN, CDKN1B, and PIK3CA genes were mutated at similar frequencies (14%–27%) in PDTCs (Table 2 and Supplemental Table 6). By contrast, TP53 and RAS were the most frequently mutated genes (42% and 31%, respectively) in ATCs (Table 2 and Supplemental Table 7), and alterations in other genes (PIK3CA, BRAF, PTEN, CDKN2A, CDKN2C, and AXIN1) were less frequent (4%–10%). None of the mutations had a significant association with ATCs or PDTCs (Table 2). In 48 screened tumors, only one ATC, of 17 TP53 mutated samples and 12 RAS mutated samples, had both mutations in coexistence, suggesting that these events were mutually exclusive (P = .0354, Fisher's exact test).

Correlation of molecular data with clinical-pathological features and patients survival

Comparisons between PDTC and ATC groups showed that ATC patients had a higher age at presentation (P = .0009) and a considerably lower median survival (P = .0001) (Supplemental Table 8). We attempted to correlate the clinical-pathological characteristics (Supplemental Table 9) with the expressional and mutational results. No significant associations were found between the clinical-pathological features (gender, age at diagnosis, size of the tumor, and presence of metastasis) and SNAI2 overexpression, presence of TP53 or RAS mutations, or number of mutated genes (data not shown). It should be noted that the tumor size information was very limited and available for only 8 of the 26 ATC samples.

No differences in survival were detected in SNAI2 overexpressed ATC patients (data not shown). Survival analysis of patients according to the mutational data was not performed due to the limited number of ATC and PDTC.

Discussion

Although representing a minority of thyroid malignancies, PDTCs and particularly ATCs, may contribute for more than half of the deaths attributable to thyroid cancer. ATCs are associated with elderly patients, presenting as a rapidly growing mass with widespread invasion of soft tissues and extensive hemorrhagic and necrotic areas (12). Contrariwise, PDTC definition has not been clear, and a definite designation as a separate entity, intermediate between WDTCs and ATCs (12), was obtained only in 2004. Subsequent studies demonstrated that PDTC patients had distinctive older age, more aggressive features, and poorer prognosis relatively to WDTC cases (19). In agreement with others (20, 21), in this study we found ATC patients had statistically significant older age and lower survival than PDTC patients.

Due to the high risks of recurrence and metastasis, unfeasible surgical resection, and failure of conventional chemotherapy and radiotherapy, novel treatment strategies against PDTCs and ATCs are particularly needed. Therefore, complete molecular profiling of these tumors is considerably important. Previously we analyzed the genome-wide expression of PDTCs. In the present study, we expanded our analysis to ATCs and, in addition, searched for genetic alterations potentially driving the aberrant expression and aggressiveness of PDTCs and ATCs.

Three previously published works have assessed the gene expression of primary ATC tumors. Salvatore et al (7) have analyzed the transcriptional profiles of five ATCs, 10 PTCs, and four normal thyroid samples. Among the 914 differentially expressed genes between ATCs and normal thyroid tissues, found in Salvatore et al, 76 of these genes were also present in our data. In agreement with that study, most of the common genes were up regulated and associated with proliferation, cell cycle, and chromosome segregation. Montero-Conde et al (8) have compared a group of seven ATCs and six PDTCs with 31 WDTCs (seven FTCs and 24 PTCs). A comparison of our results with the data from Montero-Conde et al was not possible because the complete list of differentially expressed genes was not available. Nevertheless, these authors have described deregulated molecular profiles (proliferation, chromosome segregation, TGF-β pathway, and thyroid functions), which we have also identified in our cases. Hébrant et al (22) studied 11 ATCs, 48 PTCs, and 23 normal thyroid tissues. A total of 2164 genes distinguished ATCs from normal thyroid tissues, and among these, 391 genes were common to our analysis. These genes were, in agreement with the other studies, indicative of enhanced proliferation, TGF-β pathway up regulation, and repression of thyroid-related functions.

Analyzing ATC differentially expressed genes relative to normal thyroid tissue, we found, in accordance with others (7, 23, 24), that the majority were underexpressed, a pattern also present in PDTCs and FTCs (10, 23). From the comparison of differentially expressed genes relative to normal tissues, between ATCs and our previous studied tumors (10), we observed 1216 genes that were associated only with ATCs and thus represent specific pathways involved in ATC progression.

As previously observed (8, 22–24), loss of thyroid follicular cell identity in ATCs was notably represented by the down regulation of genes encoding critical thyroid transcription factors and proteins important for thyroid hormones metabolism. Furthermore, deregulated gene sets were related to different cell junctions, cell-cell adhesion, and actin cytoskeleton, reflecting the loss of the epithelial morphology and transition to a mesenchymal state. Loss of epithelial phenotype is a key process in the activation of epithelial-to-mesenchymal transition (EMT), whereby epithelial cells lose contact, undergo cytoskeleton remodeling, and manifest a higher migratory phenotype (25). In our gene expression analysis, ATCs presented defining features of EMT, such as the down-regulation of the CDH1 gene (E-cadherin) (8, 22, 26), and gain of mesenchymal markers, like fibronectin (22, 27) and WNT5A. Moreover, as described before (8, 22), we detected the overexpression of TGF-β signaling components (TGFBI, TGFB1, LTBP1, TGFBR1), which suggests an autocrine pathway, potentially relevant due to its role in the promotion of EMT (25). A recent report described the activation of the EMT program in mammary cells (28) by combined autocrine loops of TGF-β signaling and WNT signaling (through WNT5A). Most interesting, the induction of EMT had to be accompanied by reduced levels of inhibitors of these pathways, such as bone morphogenetic protein, which we also found to be down regulated in ATCs. In PTC samples, increased TGF-β1 staining was present at the invasive fronts (29), in which an EMT-like expression profile was found (30).

Interestingly, despite presenting mostly underexpressed genes (similarly to PDTCs and FTCs), ATCs shared more common differentially expressed genes (relative to normal thyroid) with cPTC. This observation is in agreement with recent work (22), reporting that 43% of the genes deregulated in PTCs were similarly regulated in 11 ATCs (two with mutated BRAF) and suggests that ATCs could derive from PTC through a progressive transition to an undifferentiated and more mesenchymal state. For PDTCs, similarity with cPTC was not evident and genes related to EMT were not found (10), probably because our series did not include PDTCs with associated classical papillary components. Nevertheless, analysis of paired PDTC-PTC foci from BRAF mutant mice did reveal a characteristically EMT gene profile in PDTCs, which was driven by concomitant MAPK and TGF-β signaling (31). So upon MAPK constitutive activation, TGF-β may be the key promoter of invasion and metastasis in ATCs and papillary-associated PDTCs.

In our search for ATC therapeutic targets, we observed the specifically overexpression of the SNAI2 gene in our ATC series (2.24-fold relative to normal), which in combination with activation of the EMT and TGF-β signaling, prompted us to validate this gene by quantitative RT-PCR. The SNAI2 gene encodes a zinc-finger transcription factor that represses the expression of the epithelial marker, E-cadherin, as well as other epithelial components like occludin, claudins, and cytokeratins (32). In addition, SNAI2 is one effector of the EMT, which is induced in response to TGF-β (32). Several data indicate that SNAI2 has a crucial role, by integrating proliferation, apoptosis, and differentiation signals (32, 33) and by participating in tumorigenesis induced by RAS (34) and TP53 mutants (35). Accordingly, among the subgroup of eight ATCs, which clearly expressed higher SNAI2 levels (representing 42% of the ATC tumors), three had RAS mutations, three had TP53 mutations, and one was mutated for both genes. However, we were unable to identify other clinical or histological parameters that correlated with the SNAI2 overexpression in the subgroup of eight ATCs.

Another pathway contributing to EMT (25), and shown to be involved in thyroid cancer, is WNT/β-catenin signaling. Prior results showed that decreased β-catenin expression (26) or aberrant nuclear expression (42%–48%) and CTNNB1 exon 3 mutations (61%–65%) were involved in ATC development (5, 36). However, Kurihara et al (11) found CTNNB1 mutations in only 4.5% and APC mutations in 9% of ATCs, pointing out the AXIN1 gene (mutated in 82%) as the main altered component of the WNT pathway. Hébrant et al (22) also found no CTNNB1 mutations in 11 ATCs. In PDTCs, β-catenin mutations and nuclear expression were present in 25% and 21% of cases, respectively (5), whereas in other cases, no alterations were found (37). Unexpectedly, our mutational analysis revealed only three different germline AXIN1 variants (two predicted in silico to be pathogenic and one to be benign) and a single CTNNB1 mutation in a PDTC that also harbored mutations in other genes, undermining a major role for this pathway. Mutational analyses are required to elucidate the observed discrepancies and to clarify the importance of this pathway in thyroid progression.

The present study showed that ATC gene expression was deregulated for gene sets related to cell cycle regulation and checkpoints, chromosome segregation, and spindle structure. Thus, similarly to PDTCs (10), ATCs have up regulation of genes related to proliferation, cell cycle (22), and chromosomal instability (7). Progression through the cell cycle may be controlled by the interaction of cyclin-dependent kinases (CDKs) and cyclins complexes with CDKIs. Based on sequence homology and CDK specificity, CDKIs are divided into two distinct families, inhibitors of cyclin-dependent kinase 4 (INK4) members (p16^INK4A, p15^INK4B, p18^INK4C, and p19^INK4D) and CDK interacting protein (CIP)/kinase inhibitory protein (KIP) members (p21^CIP1, p27^KIP1, and p57^KIP2). In addition, all CIP/KIP members may be involved in actin dynamics and cell migration. Upon cytoplasmic mislocalization, CIP/KIP proteins regulate, at distinct levels, the Rho pathway, promoting cell motility and invasion (38). This suggests that rather than being inactivated, CIP/KIP members are subverted for increased tumorigenesis, justifying the lower frequency of CDKN1A (p21^CIP1) and CDKN1B (p27^KIP1) mutations in ATCs.

Mutations in INK4 members were found in 4 PDTCs [one with CDKN2A (p16^INK4A), two with CDKN2B (p15^INK4B), and one with CDKN2C (p18^INK4C) mutations] and in three ATCs (one with CDKN2A and two with CDKN2C mutations). To our knowledge, only one report (39) found CDKN2B mRNA expression to be normal in five ATCs, and except for the role of the p18^INK4C in medullary thyroid carcinoma, CDKN2C (or p18^INK4C protein) has never been studied in thyroid tumors. CDKN2A decreased expression and/or hypermethylation have been frequently reported in ATCs and PDTCs (40, 41), but, in agreement with our results, mutations were uncommon in ATCs (39, 42). To our knowledge, we demonstrated for the first time that mutations in the CDKI family might be involved in up to 20% of PDTCs and in up to 14% of ATCs.

Interestingly, the most overexpressed gene found in ATCs (CDKN3) is also a CDKI. This gene encodes a dual-specificity phosphatase that blocks G₁/S phase progression through CDK2 dephosphorylation. However, CDKN3 was found to be overexpressed, and to increase neoplastic transformation, in breast and prostate cancer (43). These contradictory observations were clarified with the findings in glioblastoma, in which up regulation was associated with increased aberrant splicing and loss of full-length protein (44). Upon the finding that CDKN3 was the most overexpressed gene in ATCs, we investigated the presence of aberrant splicing, similarly to that described for glioblastoma. We showed that CDKN3 is aberrantly spliced, specifically in ATCs, but did not observe the spliced form described for glioblastoma, in which exon 3 is lost. The splice variants found lacked exon 2, which is predicted to originate a truncated protein of only 23 amino acids. It remains to be clarified whether these variants could inhibit the full-length CDKN3 transcription or whether an abnormal peptide may be translated and affect the CDKN3 functions.

We found that PIK3CA mutations were more prevalent in PDTCs (3 of 22) than in ATC (1 of 26) and coexisted with other alterations. Indeed, prior studies in ATCs found PIK3CA mutations/copy number gains to be overlapped with BRAF, RAS mutations, or p53 increased expression (3, 6, 45), suggesting that PIK3CA alterations often cooperate with other oncogenic events in these types of tumors. Unexpectedly, the frequency of PIK3CA mutations in our ATC samples (4%) is lower than the reported in the literature (12%–23%). The phosphatidylinositol 3-kinase/AKT signaling pathway can also be aberrantly activated through the inactivation of phosphatase and tensin homolog (PTEN), a negative regulator of the pathway. Previous studies have not found PTEN mutations in ATCs (46, 47), or alterations were present in only 6%–16% of the samples (3, 45). In our analysis, PTEN alterations were identified in two ATCs (10%) and were also detected in three PDTCs (20%).

Our results corroborated p53 inactivation, rather than CTNNB1 and PIK3CA mutations, as the main event in ATC and PDTC progression. Prior results (20) have shown that PDTC and ATC patients presenting TP53 mutation had decreased survival, stressing its major influence in tumorigenesis. In our study, RAS was the second most mutated gene, whereas BRAF mutations were present in only three samples, as expected for a series with no histological evidence of PTC derivation (48, 49). Our detection of mutated RAS is in agreement with other PDTC (50) and ATC (2, 51) studies. From a progression point of view (Supplemental Figure 6), it may be suggested that these dedifferentiated tumors derive from RAS-mutated WDTCs, that is, FTCs and/or fvPTCs. In our previous work (10), based on the similarity of gene expression profile, we suggested fvPTCs as plausible precursors of RAS-mutated PDTCs. In contrast, our and others' (22) expression analysis suggested that ATCs were molecularly more similar to cPTCs. Rather than a common origin, this may suggest a molecular mechanism affecting both cPTCs and ATCs (for example TGF-β activation). In either case, our work suggests that RAS and TP53 mutations are alternative events, rather than progressively accumulated events, during PDTC and ATC progression (2).

Overall, nine PDTCs (41%) and six ATCs (23%) harbored no mutations in the 14 genes profiled. Other mechanisms may be driving the progression in these tumors and further molecular characterization is required, especially for PDTCs. The application of next-generation sequencing, which has a lower limit of detection when compared with the Sanger sequencing, could reveal new ATC and PDTC driver mutations occurring in samples with a low percentage of tumor cells and the evaluation of the intratumor genetic heterogeneity. For Sanger sequencing, the limit of detection by visual inspection of the electropherogram has been described to be approximately 15%–20% of mutant alleles (52–54), which corresponds to approximately 30%–40% of mutant cells (for a heterozygous alteration). This may represent a drawback in the technique because PDTCs and ATCs have been found to be densely infiltrated with tumor-associated macrophages (55), representing 50%–60% of the nucleated cells in ATCs (56). Nevertheless, it is interesting to note that in our work, we have actually detected more mutated ATC samples (77%) than PDTC mutated samples (59%), despite the fact that macrophage infiltration occurs more frequently in ATCs.

For the mutated cases found, different therapeutic approaches are available, depending on the alteration uncovered. The present work points out additional targets that may be used, alone or in combination. The detection of TP53 mutations in 27% of PDTCs and 42% of ATCs allows the use of molecules for the reactivation of the p53 functions (57). The deregulated profile related to proliferation, CDKI mutations (in 20% of PDTCs and 14% of ATCs) and CDKN3 abnormal splicing in ATC suggest that inhibitors of CDK, for example roscovitine and flavopiridol, may be efficient in such cases (58). SNAI2 overexpression in 42% of cases and the likely influence of TGF-β signaling in ATC suggests that TGF-β inhibitors, of which some are already in clinical trials (59), also represent a reasonable therapeutic option.

Acknowledgments

We gratefully acknowledge Dr Rita Santos (Departamento de Endocrinologia, Instituto Português de Oncologia de Lisboa) for the identification of PDTC and ATC cases. We are also grateful to the Serviço de Cirurgia de Cabeça e Pescoço and Serviço de Anatomia Patológica (Instituto Português de Oncologia de Lisboa) for the supply and histological analysis of thyroid tumors.

This work was supported by the Sociedade Portuguesa de Endocrinologia, Diabetes, e Metabolismo and received the award “Prémio Nacional de Endocrinologia SPEDM/Novartis Oncology 2012.” J.M.P. was the recipient of a Portuguese government (Fundação para a Ciência e Tecnologia) PhD fellowship (SFRH/BD/46096/2008).

Disclosure Summary: The authors have nothing to disclose

Abbreviations

 
• ATC

  anaplastic thyroid carcinoma

 
• CDK

  cyclin-dependent kinase

 
• CDKI

  CDK inhibitor

 
• CIP

  CDK interacting protein

 
• cPTC

  classical variant of PTC

 
• EMT

  epithelial-to-mesenchymal transition

 
• FNAB

  fine-needle aspiration biopsy

 
• FTC

  follicular thyroid carcinoma

 
• fvPTC

  follicular variant of PTC

 
• GC

  guanine and cytosine

 
• GEO

  Gene Expression Omnibus

 
• INK4

  inhibitors of cyclin-dependent kinase 4

 
• KIP

  kinase inhibitory protein

 
• PDTC

  poorly differentiated thyroid carcinoma

 
• PTC

  papillary thyroid carcinoma

 
• PTEN

  phosphatase and tensin homolog

 
• WDTC

  well-differentiated thyroid carcinoma.

References