The authors have declared that no competing interests exist.
Conceived and designed the experiments: AH JED CM. Performed the experiments: AH GD MD. Analyzed the data: AH. Contributed reagents/materials/analysis tools: GA CT EL. Wrote the paper: AH.
Anaplastic thyroid carcinoma (ATC) is the most lethal form of thyroid neoplasia and represents the end stage of thyroid tumor progression. No effective treatment exists so far. ATC frequently derive from papillary thyroid carcinomas (PTC), which have a good prognosis. In this study, we analyzed the mRNA expression profiles of 59 thyroid tumors (11 ATC and 48 PTC) by microarrays. ATC and PTC showed largely overlapping mRNA expression profiles with most genes regulated in all ATC being also regulated in several PTC. 43% of the probes regulated in all the PTC are similarly regulated in all ATC. Many genes modulations observed in PTC are amplified in ATC. This illustrates the fact that ATC mostly derived from PTC. A molecular signature of aggressiveness composed of 9 genes clearly separates the two tumors. Moreover, this study demonstrates gene regulations corresponding to the ATC or PTC phenotypes like inflammatory reaction, epithelial to mesenchymal transition (EMT) and invasion, high proliferation rate, dedifferentiation, calcification and fibrosis processes, high glucose metabolism and glycolysis, lactate generation and chemoresistance. The main qualitative differences between the two tumor types bear on the much stronger EMT, dedifferentiation and glycolytic phenotypes showed by the ATC.
Thyroid tumors are divided into encapsulated benign tumors (autonomous and follicular adenomas) and carcinomas. These carcinomas are themselves subdivided into differentiated carcinomas (follicular carcinomas (FTC) or papillary carcinomas (PTC)) which may evolve into the very aggressive and dedifferentiated anaplastic carcinomas (ATC)
Despite its low frequency (<5% of all thyroid carcinomas), ATC is responsible for more than half of thyroid carcinoma deaths, with a mean survival of 6 months after diagnosis
mRNA expression analysis based on microarray technology has been largely used to characterize human cancers. This approach allows the identification of genes important in the tumorigenesis process, and the definition of diagnosis and prognosis signatures.
Until now, only a limited number of ATC have been investigated for mRNA expression with incomplete and sometimes not very sensitive microarray sets
To identify the molecular mechanisms involved in tumor evolution, we analyzed the mRNA expression profiles of 59 thyroid tumors (11 ATC and 48 PTC) using the Affymetrix microarray technology and real-time qRT-PCR and the mutational status of 11 ATC.
The analysis of the genes regulated in ATC revealed several very interesting known and unknown features: a strong similarity with PTC, a signature of 9 genes discriminating ATC and PTC which may be related to clinical prognosis, and biological signatures which suggest new therapeutic approaches. The study defines the molecular phenotypes corresponding to the qualitatively described pathological features of these cancers.
16 ATC and 53 PTC were obtained from different hospitals: Regional Reference Cancer Center of Lille (Lille, France), Pitié-Salpêtrière (Paris, France), Jules Bordet Institute (Brussels, Belgium), Cliniques Universitaires Saint-Luc (Brussels, Belgium), Katholieke Universiteit Leuven (Leuven, Belgium) and from the Chernobyl Tissue Bank (
Total RNA was extracted from thyroid tissues using Trizol reagent (Invitrogen), followed by purification on RNeasy columns (Qiagen). The RNA concentration was spectrophotometrically quantified, and its integrity was verified using an automated gel electrophoresis system (Experion, Biorad).
In order to determine the mutational status for TP53, BRAF, H-RAS, N-RAS, K-RAS, PI3KCA and β-catenin in the 11 ATC samples, the sequences containing the most frequent mutations were amplified by PCR using appropriate primer pairs (primer sequences and PCR conditions provided in
Two µg of total RNA from 11 ATC and 48 PTC were engaged for cDNA synthesis. Labeled cRNA was synthesized, purified and hybridized on Affymetrix HU 133 Plus 2.0 arrays, following the Affymetrix Protocol.
CEL file data were normalized by GCRMA. For each spot, data were expressed as the log2 ratio of fluorescence intensities of the tumor tissue and the reference normal tissues pool. All gene expression data are released on GEO under the accession number GSE33630.
The subset of probes that varied by at least 2-fold compared to the normal pool (ratios tumor/normal) in all the 11 ATC samples (without any opposite regulation) was selected and called the ATC list. Similarly, those that varied by at least 2-fold compared to the normal pool in all the 48 PTC samples (without any opposite regulation) were selected and called the PTC list.
Probes which were upregulated more than 1.5-fold in all the ATC and downregulated or not modulated in PTC, and probes which were downregulated more than 1.5-fold in all the ATC and upregulated or not modulated in PTC were selected for a potential signature of aggressiveness of ATC versus PTC. Similarly, probes which were upregulated more than 1.5-fold in all the PTC and downregulated or not modulated in ATC and probes which were downregulated more than 1.5-fold in all the PTC and upregulated or not modulated in ATC were also selected for this aggressiveness signature.
Nonsupervised analyses were performed on the basis of between-sample correlation distances. Multidimensional scaling (MDS, as implemented by R isoMDS function) was performed considering all the probes present on the microarray with the 11 ATC and the 48 PTC samples.
A number of modulated genes on the microarray slides were validated using qRT-PCR (SYBR Green, Eurogentec, Liege, Belgium). The following mRNA expressions were evaluated using, when possible, transexonic primers, designed with Primer Express software (Applied Biosystems): NELL2, SPINT2, MARVELD2, DUOXA1, RPH3AL, TBX3, PCYOX1, c5orf41, PKP4 (primers sequences provided in
On the 11 ATC, p53 mutation was found in 4 (36%), BRAF mutation in 2 (18%), PI3KCA mutation in 1 (10%). One sample showed both BRAF and p53 mutations (ATC1). No mutation was found for RAS (H-RAS, K-RAS and N-RAS) nor for β-catenin.
Differences in the molecular phenotypes of ATC and PTC can best be demonstrated by a comprehensive microarray analysis of gene expressions in the two types of tissues. Because of the absence of normal tissue counterparts for ATC, gene expression profiles were compared with a common reference pool of 23 normal, non-neoplastic, tissues from the contra-lateral lobe with respect to the thyroid carcinomas. Overall gene expressions from the 11 ATC and the 48 PTC were analyzed using multidimensional scaling (MDS) (
The ATC are labelled according to their mutation as depicted in the figure.
ATC showed 1051 commonly upregulated and 1113 commonly downregulated probes using the criteria of 2 fold described in Material and methods (ATC list (
Using stringent selection criteria (1.5 fold regulation), described in Material and Methods, we identified a minimal signature of 9 genes called signature of aggressiveness (
The differential expression of several genes in the ATC and PTC lists confirmed the previously published data on ATC by two other groups
To further validate our data, the modulation of the 9 genes composing the signature discriminating ATC from PTC was investigated by real-time qRT-PCR. Experiments were performed on 5 new ATC and on 5 new PTC. Similar modulation patterns were found for the expression of all the genes comparing microarray analysis with qRT-PCR, thus validating the microarray data in independant set of tumors (
The microarray expressions is also represented. Log2 ratios represent the expression ratios of the genes in the tumors versus a pool of 23 normal thyroid tissues.
First, several immediate early genes (IEG) transcription factors are strongly downregulated in ATC (cJun (Jun), JunD, JunB (9/11), FOS (9/11), FOSB, EGR1, EGR2) attesting a complete program switch between the ATC and the normal tissues. These downregulations were also generally observed in PTC.
Second, there is a switch of gene expression in ATC within functional categories of genes from one set to another: for example, the switch of ion channel genes (e.g. the SLC family with 18 upregulated and 20 downregulated genes), of structural proteins (e.g. the cadherin (CDH family), heatshock proteins (HSP family)), organelle proteins (e.g. ribosomal proteins (RPL family), metabolic enzymes (e.g. aldehyde dehydrogenase (ALDH family)). Most of these switches are common to a great majority of PTC and ATC. The biological meaning of these switches must be analyzed in each case; however collectively they testify to a whole different program of the tumor compared to the normal tissue.
Third, another important level at which changes in protein expression can be regulated is the mRNA stability, as controlled by uridylate rich elements (ARE) in the mRNAs. The proportion of deregulated genes in ATC containing ARE was evaluated by using ARE-mRNA database (ARED:
The ATC gene list was analyzed by gene ontology (GO) using Database for Annotation, Visualization and Integrated Discovery (DAVID) software
Analysis of
One of the most striking clinical and pathological feature of ATC is EMT and invasion
In the following paragraphs, gene regulations which are not common to all the ATC are indicated into brackets by the fraction of positive tumors (no bracket for those regulated in all ATC).
Biochemically, EMT is characterized by a downregulation of E cadherin (CDH1) (10/11 ATC) and an upregulation of N cadherin (CDH2) (9/11 ATC) and vimentin (VIM). Presumably the decreased expression of CDH16 and the increased expression of CDH11 also reflect this program. Genes of the TGFβ signalling pathway inducing EMT are upregulated (eg. TGFβ1, TGFβ2 (9/11 ATC), TGFβ3 (9/11 ATC), TGFβI and the TGFβR1 receptor) and the expression of transcription factors involved in EMT is also increased (snail (SNAI2), sprouty (SPRY4) (10/11 ATC), zinc finger E-Box binding homeobox (ZEB2 and ZEB1 (8/11 ATC)), twist (TWIST1 and TWIST2 (7/11 ATC)))
ATC is a highly proliferative tumor. It is therefore not surprising that gene expression analysis (
As expected from their fully dedifferentiated status and as illustrated in
A downregulation of enzymes involved in H2O2 metabolism in thyroid (catalase (CAT), metallothioneins (MT1F and MT1G) and superoxide dismutase (SOD3)) was also observed, consonant with the decreased expression of H2O2 generation genes, DUOXs and DUOXAs
One interesting cluster of
Anaplastic carcinomas are often characterized by high fluorodeoxyglucose uptake allowing their visualization by positron emission tomography (PET). This feature illustrates the metabolic shift from an oxidative to a pseudoanaerobic glycolytic metabolism: the Warburg effect. A lot of mRNA encoding enzymes involved in this metabolism were upregulated in ATC (
The expression of angiogenesis factor genes does not give a clear indication of angiogenesis. Indeed, on the one hand, we observed a downregulation of some mRNA encoding angiogenesis factors such as angiopoietin (ANGPTL1) and an upregulation of thrombospondins (THBS1, THBS2)
A deregulation of the tree main biomarkers
The known resistance of ATC to classical chemotherapies could be explained by the presence of a large proportion of CSC-TPC like cells in the tumor and by the increased expression in our data of mRNAs encoding multidrug resistance proteins such as some ATP-binding cassette family members (ABCA8, ABCB10, ABCC5, ABCC10, ABCG1). The acquisition of such tumor resistance mechanisms has also been associated with EMT transition
The postulated role of p53 inactivation in the important program switch from PTC to ATC
In this study, the molecular phenotypes of ATC and PTC have been correlated with their biological phenotypes. The major shift from normal tissue to ATC and PTC is illustrated by the alteration of more than one third of the genes and by the switches of gene expression within same gene families. This is further multiplied by additional changes in the nature of the mRNA expressed as suggested by the increased expression of mRNA containing the regulatory elements ARE, controlling gene expression.
The downregulation of several immediate early genes (IEG) transcription factors is another major shift and a counter intuitive result. It illustrates, as previously shown by our group
The PTC and FTC origin of the majority of ATC is demonstrated by pathology, mutation analysis and here by gene expression. Pathological examination shows that about 25% of the ATC appear in a differentiated carcinoma (mostly PTC) background
These different elements suggest that both PTC and ATC exhibit a fundamental program switch compared to the normal tissues. ATC thus results from two major transitions: the first one, a complete gene switch leading to differentiated cancer resulting from the altered expression of ∼ 40% of the genes
Four candidate mechanisms which may induce a program switch from PTC to ATC have been proposed: the suppression of the p53 pathway, the expression of activated β-catenin and PIK3CA and the acquisition of chromosomal instability. Inactivating mutations of p53 are found in some ATC (36% in our work and >50% in the literature) and overexpression of presumably inactivated p53 protein was detected in an even larger population of one set of ATC (50%) but never in the neighbouring PTC cells
Secondly, activation of β-catenin could represent another mechanism leading to the ATC phenotype. Indeed in one study, such mutations were found in 61% of the lesions but not in the precursor lesions
Thirdly, the increase in PIK3CA copy number or the presence of activated PI3KCA following mutations, much more prevalent in ATC than in differentiated carcinomas, could also reflect a progression in the cancer phenotype
Finally, some CGH gains and losses specific for ATC have been reported, in addition to common DNA copy number changes described in the precursor differentiated thyroid carcinomas. This may suggest that the development of chromosomal instability underlies tumor progression
The gene expression phenotype of ATC corresponds very well to and explains different clinical and pathological features: hypoxia, glycolysis and fluorodeoxyglucose uptake
When comparing the molecular profiles of PTC and ATC, the two main distinctive gene expression characteristics concern EMT and the repression of differentiation. EMT markers are expressed by both ATC and PTC. However, only ATC present a marked induction of the fundamental transcription factors causing EMT (ZEB, TWIST and SNAIL). This again corresponds very well to the much more invasive phenotype of ATC.
The dedifferentiation phenotype of the ATC cells is well known and is much more severe than the one of the differentiated cancers as attested by the downregulation of thyroid specific differentiation genes and by the total downregulation of the three fundamental thyroid determination factors (TTF1, TTF2 and Pax8): the cells seem to have lost all traces of their thyroid origin. This is also illustrated by the complete absence of expression of Tg and TSHR but also of DUOX1, DUOX2, DUOXA1, DUOXA2, and NKX2-1. These correspond to the
As the same tissue displays a high expression of proliferation, EMT and CSC-TPC markers, one could, as is often done, assume that the same cells exhibit these three properties. However, this would run contrary to the often assumed concept that the CSC-TPC would be a slowly proliferating cells
The fibrosis of ATC is already well known to the pathologist. Whether this results from or induces a strong inflammation is so far unknown. The extensive fibrosis may explain the molecular phenotype of anoxia with the induction of glycolytic enzymes as consequence or as a cause
Beside giving a molecular understanding of the pathways involved in tumorigenesis, the identification of proteins corresponding to pathological features allows to propose new putative targets for diagnosis and treatment.
Several characteristics of ATC, demonstrated or confirmed in this work, offer clues about possible therapies to target this so far untreatable disease. The downstream effects of the activated proliferation cascade (eg cell cycle and cell division) would suggest to target common steps of these cascades (i.e. mTOR, cyclin dependent kinases (CDK)) rather than diverse upstream activators. The high proliferation rate should make the tumor sensitive to chemotherapies or radiation therapies, but the low accessibility of the inside of the tumor to blood flow, evidenced by its anoxia, would make it insensitive to these therapy and should be overcome. Therefore pretreatment with capillary blood flow enhancers or treatment in conjunction with high O2 supply could improve these classical therapies. Our results also lead to propose other possible therapies: anti-inflammatory treatment to decrease inflammation, a combination of epigenetic and cAMP enhancing treatments to re-establish the differentiation and temporarily the radioiodide uptake, antiglycolytic and antilactate transporter treatments to target the high glycolytic metabolism and, perhaps, the reinduction of expression of the 9 downregulated genes discriminating PTC from ATC. As no single treatment is successful yet, multitargeted approaches should be investigated.
(XLS)
(DOC)
(XLS)
(XLS)
(XLS)
We thank Chantal Degraef for excellent technical assistance, and Eric Raspe and Véronique Kruys for helpful and constructive discussions.