DNA Methylation in Thyroid Cancer

26 7

Endocrine-Related C Zafon et al. DNA methylation in thyroid 26:7 R415–R439 Cancer cancer

REVIEW DNA methylation in thyroid cancer

Carles Zafon1,2, Joan Gil3, Beatriz Pérez-González3 and Mireia Jordà2,3

1Diabetes and Metabolism Research Unit (VHIR) and Department of Endocrinology, University Hospital Vall d’Hebron and Autonomous University of Barcelona, Barcelona, Spain 2Consortium for the Study of Thyroid Cancer (CECaT), Catalonia, Spain 3Program of Predictive and Personalized Medicine of Cancer, Germans Trias i Pujol Research Institute (PMPPC-IGTP), Barcelona, Spain

Correspondence should be addressed to M Jordà: [email protected]

Abstract

In recent years, cancer genomics has provided new insights into genetic alterations and Key Words signaling pathways involved in thyroid cancer. However, the picture of the molecular ff thyroid cancer landscape is not yet complete. DNA methylation, the most widely studied epigenetic ff DNA methylation mechanism, is altered in thyroid cancer. Recent technological advances have allowed ff global hypomethylation the identification of novel differentially methylated regions, methylation signatures ff biomarkers and potential biomarkers. However, despite recent progress in cataloging methylation ff demethylating drugs alterations in thyroid cancer, many questions remain unanswered. The aim of this review is to comprehensively examine the current knowledge on DNA methylation in thyroid cancer and discuss its potential clinical applications. After providing a general overview of DNA methylation and its dysregulation in cancer, we carefully describe the aberrant methylation changes in thyroid cancer and relate them to methylation patterns, global hypomethylation and gene-specific alterations. We hope this review helps to accelerate the use of the diagnostic, prognostic and therapeutic potential of DNA methylation for the benefit of thyroid cancer patients. Endocrine-Related Cancer (2019) 26, R415–R439

Introduction

Thyroid cancer, the most prevalent endocrine malignancy, and more extensive use of imaging techniques such as covers the full range of phenotypes from indolent to neck ultrasound (Kitahara et al. 2017). the worst forms of human cancer. It is categorized into Currently, total thyroidectomy, the removal of the differentiated thyroid cancer (DTC), poorly differentiated affected neck lymph nodes of the central compartment, thyroid cancer (PDTC) and undifferentiated or anaplastic radioiodine (RAI) therapy for the ablation of thyroid thyroid cancer (ATC), all of which are derived from thyroid remnants or metastases and TSH suppression with follicular cells, and into medullary thyroid cancer (MTC), l-thyroxin are the treatment schedule for a large portion of which is derived from parafollicular cells. Moreover, DTC DTC patients (Haugen et al. 2016). With these therapeutic has three basic subtypes: papillary thyroid cancer (PTC), approaches, the majority of DTC patients exhibit good follicular thyroid cancer (FTC) and Hurthle cell thyroid prognosis with a >98% 5-year survival rate. However, a cancer (HCTC). Globally, DTC accounts for 95% of all subset of tumors progress to display more aggressive thyroid carcinomas. During the last few decades, several behavior, and some of these undergo a progressive epidemiologic studies have reported that DTC incidence process of dedifferentiation that makes them less capable has increased worldwide (Lim et al. 2017). The reasons for of producing thyroglobulin and concentrating iodine, this are not clear; it has been attributed both to a true producing a poor response to RAI. To identify patients increase in the incidence of DTC and to the improvement with a progressive course of the disease, a number of

-19-0093 Endocrine-Related C Zafon et al. DNA methylation in thyroid 26:7 R416 Cancer cancer prognostic factors and clinical scores have been proposed a methyl group to the 5-carbon of the cytosine, giving that are mainly age, the histological variant, the initial rise to 5-methylcytosine (5mC) (reviewed in Portela & extent of the disease and the size of the primary tumor Esteller 2010). In humans, DNA methylation occurs (Asa 2017). However, these prognostic factors have almost exclusively within CpG dinucleotides, which are some limitations. underrepresented (i.e., found in a lower than expected In recent years, the management of thyroid cancer in proportion based on the G/C content) and not evenly patients is shifting toward more personalized medicine to distributed throughout the genome (Bird 1980). Most avoid the overdiagnosis and overtreatment of tumors with human genome (approximately 60–80% of CpG sites) an indolent course and, at the same time, to identify those is methylated, except for some CpG-rich regions called tumors that will progress (Dralle et al. 2015). The final goal is CpG islands (CGIs), which are often unmethylated to deliver the most effective but least aggressive treatment. and encompass the promoters of approximately 60% A better understanding of the molecular mechanisms of protein-coding genes (Ehrlich et al. 1982, Bird 1986, underlying thyroid cancer progression may be key to tailor Lister et al. 2009). the management of this disease. In this regard, significant DNA methylation is frequently described as a progress has been made in the last 20 years (Riesco-Eizaguirre repressive epigenetic mark. However, DNA methylation & Santisteban 2016). The major event in PTC carcinogenesis function varies depending on the genomic context is the constitutive activation of mitogen-activated protein (reviewed in Jones 2012, Baubec & Schubeler 2014) kinase (MAPK), whereas the PI3K/AKT pathway is involved (Supplementary Fig. 1, see section on supplementary in the progression of FTC. Recently, the genetic landscape of data given at the end of this article). DNA methylation in some thyroid cancer histotypes has been largely deciphered proximal and distal regulatory elements (i.e., promoters (Cancer Genome Atlas Research Network 2014, Kunstman and enhancers, respectively) represses transcription et al. 2015), and some of these genetic alterations have by affecting the binding of transcription factors been used both as diagnostic tools (in the study of thyroid and/or recruiting enzymes that modify chromatin nodules) and as prognostic tools. Importantly, for the first structure. Conversely, DNA methylation of the gene time, the last set of American Thyroid Association (ATA) body may enhance transcriptional elongation and affect guidelines recommended the use of mutations in the BRAF splicing. In the case of repetitive elements, which are gene and the TERT promoter as prognostic factors in PTC densely methylated, DNA methylation is the major (Haugen et al. 2016). repression mechanism. However, cancer is not only caused by genetic Therefore, DNA methylation is a key player in abnormalities but also by epigenetic alterations (reviewed the regulation of gene expression and is implicated in in Jones & Baylin 2007). The most widely used definition many cellular processes such as imprinting (Reik et al. for epigenetics is ‘the study of changes in gene function 1987, Swain et al. 1987), X-chromosome inactivation that are mitotically and/or meiotically heritable and that (Mohandas et al. 1981) and the establishment and do not entail a change in DNA sequence’ (Wu & Morris maintenance of cell type-specific expression programs 2001, Bird 2007). Epigenetics can explain how two (reviewed in Suelves et al. 2016). DNA methylation identical genotypes can lead to different phenotypes. There is also essential for the maintenance of genome are several epigenetic mechanisms: DNA methylation, stability by modeling chromatin structure (reviewed posttranslational modifications of histones, chromatin in Madakashira & Sadler 2017) as well as by silencing remodeling or non-coding RNAs. These mechanisms have repetitive sequences to prevent chromosomal been reviewed elsewhere (Bannister & Kouzarides 2011, rearrangements (Gaudet et al. 2003) and the expression Holoch & Moazed 2015, Längst & Manelyte 2015, Allis and expansion of transposable elements (reviewed in & Jenuwein 2016, Feinberg et al. 2016), and here we will Belancio et al. 2010). focus on DNA methylation. Furthermore, it is noteworthy that DNA methylation is an important source of promising cancer biomarkers for many reasons: DNA methylation is stable even What is DNA methylation? in fixed samples over time, easily detected by well- DNA methylation was the first discovered epigenetic established techniques (Fig. 1 and Supplementary modification Hotchkiss( 1948, Holliday & Pugh 1975, Table 1), present in various bodily fluids, and cell type Riggs 1975) and consists of the covalent addition of specific Koch( et al. 2018).

Writers, erasers and readers of DNA methylation called DNA methyltransferases (DNMTs), transfer a methyl group to cytosine residues (reviewed in DNA methylation does not function alone but is Goll & Bestor 2005) (Fig. 2). involved in a complex crosstalk with many other players Despite the high stability of DNA methylation, 5mC to reinforce specific regulatory programs. Although can be demethylated by passive or active mechanisms, the how DNA methylation is interpreted in the context of latter mediated by erasers that generate DNA demethylation genome regulation is not completely understood, there intermediates, such as 5-hydroxymethylcytosine (5hmC), are some proteins known to modulate DNA methylation. 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) They are classified as writers, erasers and readers of (reviewed in Wu & Zhang 2017) (Fig. 2). Interestingly, DNA methylation. although the levels of 5hmC in human cells are very low DNA methylation writers are proteins that establish (in general, 14-fold lower than the levels of 5mC) and and maintain DNA methylation patterns through vary greatly between different tissues, 5hmC is present at development and differentiation. These proteins, a relatively stable abundance, suggesting that it is not just

How to measure DNA methylation?

Global Sequence-Specific

Compartment of Candidate Entire Genome Genome-wide the Genome approach

Physico-chemical Repetitive elements RE sites Amplicon sequencing methods •QUAlu •AIMS •BS-Sanger sequencing •RP-HPLC •LINE-1 •NSUMA •BS-Pyrosequencing •HPCE pyrosequencing •MS-AFLP •BS-MALDI-TOF-MS •LINE-1 ELISA • MCA •LC-MS Discriminative amplicon •AuNPs RE sites Antibody affinity •PCR + MS-RE Antibody affinity •LUMA •MeDIP-array •MSP •ELISA •MeDIP-Seq •COBRA •IHC Bisulfite qPCR based •BeadArray •MethyLight •RRBS •MethylQuant •WGBS •In-tube melting curves analysis

Figure 1 Main DNA methylation techniques according to the type of DNA methylation measured (global or sequence-specific) and the principle of DNA methylation discrimination (physicochemical properties, 5mC antibody affinity, methylation-sensitive restriction enzymes and bisulfite conversion) (more detailed information in Supplementary Table 1). In recent decades, a large number of techniques to measure DNA methylation have been developed that can be classified into two main groups: (i) those providing unique value as a global measure of DNA methylation and (ii) those that are sequence-specific and measure the levels of DNA methylation in particular regions at single CpG resolution. Moreover, global methods can be subdivided into those measuring the DNA methylation of the entire genome and those measuring the DNA methylation of a compartment of the genome used as surrogate reporter of the genome (e.g., repeat sequences such as LINE-1 and Alu elements, which comprise 20% and 10% of the human genome, respectively). Sequence-specific methods can also be subdivided into those that are genome-wide (mostly based on bead arrays or NGS) and those measuring specific regions of interest (mostly based on PCR). A more comprehensive list of available techniques can be found elsewhere (Jordà et al. 2009, Toraño et al. 2012). Importantly, techniques to analyze DNA methylation do not differentiate 5mC from 5hmC; thus, in recent years, some approaches to specifically measure 5hmC have been developed Skvortsova( et al. 2017). AIMS, analysis of DNA methylation by amplification of intermethylated sites; AuNPs, Au nanoparticles; BS, bisulfite; COBRA, combined bisulfite restriction analysis; ELISA, enzyme-linked immunosorbent assay; HPCE, high-performance capillary electrophoresis; IHC, immunohistochemistry; LC-MS, liquid chromatography coupled with mass spectrometry; LUMA, luminometric methylation assay; MALDI-TOF-MS, matrix-assisted laser desorption/ionization time-of-flight mass spectrometry; MCA, methylated CpG island amplification; MeDIP, methylated DNA immunoprecipitation; MSP, methyl-sensitive PCR; MS-AFLP, methylation-sensitive amplification length polymorphism; NGS, next-generation sequencing; NSUMA, next-generation sequencing of unmethylated Alu; RE, restriction enzyme; RP-HPLC, reversed-phase high-performance liquid chromatography; RRBS, reduced representation bisulfite sequencing; QUAlu, quantification of unmethylated Alu; WGBS, whole genome bisulfite sequencing.

Figure 2 Process of DNA methylation and demethylation. DNA methyltransferases (DNMTs) catalyze the methylation of cytosine by adding a methyl group to C5 position. DNMT3A and DNMT3B are responsible for de novo methylation, while DNMT1 maintains DNA methylation patterns by copying the 5mC pattern on the newly synthesized strand after DNA replication. DNA demethylation can occur through different mechanisms: passive demethylation due to the impairment of the DNA methylation maintenance machinery that results in the dilution of DNA methylation after multiple rounds of replication, and active demethylation involving several proteins considered to be erasers of DNA DNMT1 methylation. Specifically, active demethylation occurs through the iterative oxidation of 5mC to 5-hydroxymethylcytosine (5hmC), Maintenance 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) mediated by ten-eleven-translocation (TET) proteins. Then, these oxidized forms can be subsequently diluted during DNA replication, or 5fC and 5caC can be excised by thymine DNA glycosylases (TDG) coupled with base excision repair (BER) (reviewed in Wu & Zhang 2017). Even though other mechanisms of active demethylation have been proposed, the TDG-BER passive demethylation mechanism has gained the most support (Wu & Zhang 2010, Bochtler et al. 2017). a DNA demethylation intermediate. 5hmC is associated been ignored for decades mainly due to the technical with gene transcription, although this relationship complexity of its analysis. Conversely, research has focused is not fully understood (reviewed in Shi et al. 2017) on hypermethylation (i.e., the increase in the methylation (Supplementary Fig. 1). of CpG sites compared to normal cells), which often but Finally, DNA methylation readers are proteins that not exclusively occurs in localized sequences within specifically bind to methylated CpGs and coordinate the regulatory elements associated with CGIs (Herman et al. crosstalk between DNA methylation, histone modifications 1995, Melki et al. 1999, Esteller et al. 2000, Nguyen and chromatin organization to reinforce downstream et al. 2001). Both global hypomethylation and focal regulatory programs. The paradigm of these proteins are hypermethylations are constant features of the cancer the methyl-CpG-binding domain (MBD) family proteins, genome and often coexist in tumoral cells, but although which have the ability to recruit chromatin remodelers, there is interplay between them, their underlying histone deacetylases (HDAC) and DNMTs to methylated mechanisms seem to be independent. CpGs associated with gene repression (reviewed in The main consequence of the hypermethylation Du et al. 2015) (Supplementary Fig. 1). of promoters and enhancers is the repression of the expression of genes functionally important in the neoplastic process, whose silencing may have a tumor- Aberrant DNA methylation is a hallmark of cancer promoting effect. The hypermethylation profile is tumor Disruption of DNA methylation is a common feature specific and affects all cellular pathways (reviewed in human disease, both in noncancerous diseases and in Esteller 2007). Some genes such as p16INK4A and in cancer (Fernandez et al. 2012). Although the first MLH1 are frequently hypermethylated in many cancers discovered alteration of DNA methylation in cancer including thyroid cancer (Schagdarsurengin et al. 2006, was an overall reduction in 5mC levels, that is, a global Guan et al. 2008) while others are tumor specific (e.g., hypomethylation in tumoral cells compared to the the sodium iodide symporter gene – SLC5A5 or NIS – in methylation in normal cells (Feinberg & Vogelstein 1983, thyroid cancer) (Neumann et al. 2004, Galrão et al. 2014). Gama-Sosa et al. 1983), this epigenetic alteration has In contrast to global hypomethylation, an overall increase

https://erc.bioscientifica.com © 2019 Society for Endocrinology https://doi.org/10.1530/ERC-19-0093 Published by Bioscientifica Ltd. Printed in Great Britain Downloaded from Bioscientifica.com at 09/28/2021 06:06:52AM via free access Endocrine-Related C Zafon et al. DNA methylation in thyroid 26:7 R419 Cancer cancer in 5mC levels in tumors compared to its levels in normal in thyroid cancer, although further studies should be tissues is much less common, despite the high number of performed to understand the underlying relationship. local hypermethylations (Ehrlich 2002). On the other hand, MBD proteins are mutated in several For several decades, it has been widely accepted cancers, including PTC (Du et al. 2015). Although they that the hypomethylation of repetitive sequences is represent less than 5% of patients, the study of these responsible for global hypomethylation (Ehrlich 2009). proteins in thyroid cancer could be a promising field. However, recent approaches enabling the mapping of DNA methylation at the genome scale have shown that global hypomethylation affects large genome domains DNA methylation changes in thyroid cancer: including both repetitive and unique sequences (Berman drivers of disease progression et al. 2011, Hansen et al. 2011, Timp et al. 2014). In this and biomarkers context, hypomethylation can encompass regulatory elements and affect gene expression (reviewed in DNA methylation has been extensively studied in many Wilson et al. 2007). Nevertheless, hypomethylation- cancers, such as colorectal and breast cancer, and due dependent transcriptional activation is less frequent to technological advances, enormous progress has been than hypermethylation-dependent transcriptional made in the understanding of the epigenetic landscape of silencing. In contrast, numerous studies indicate that these tumors. Conversely, the role of DNA methylation global hypomethylation is associated with chromosomal in thyroid cancer has received comparatively less instability and the reactivation of transposable elements attention (Fig. 3). The first DNA methylation studies in (Gaudet et al. 2003). thyroid cancer were based on candidate gene approaches 5hmC is also perturbed in cancer in a similar way as assessing the DNA methylation levels of specific gene 5mC, i.e., there is a strong global loss of this epigenetic promoters (Supplementary Table 2). It was not until mark in tumors. However, its involvement in thyroid 2011 that Hou et al. performed the first array-based, cancer is completely unknown. A recent study that genome-wide DNA methylation study using two PTC analyzed 5hmC in circulating cell-free DNA and in cell lines to analyze the effect of the BRAF(V600E) tumoral and normal tissues from different cancer types, mutation on DNA methylation (Hou et al. 2011). To including thyroid cancer, found that 5hmC was mainly our knowledge, 11 more array-based studies using distributed in transcriptionally active regions (Li et al. different platforms (Goldengate, 27K or 450K) to profile 2017). Importantly, they identified cancer-specific 5hmC the methylomes of tissue samples from patients with signatures. To the best of our knowledge, this is the only thyroid cancer have been published since then (Table 1). study of 5hmC in thyroid cancer; thus, 5hmC has opened All these studies showed that thyroid cancer is not an a new field to explore in this disease. exception and exhibits DNA methylation alterations. However, different pan-cancer analyses based on data from the Cancer Genome Atlas Research Network Disruption of epigenetic pathways in cancer revealed that PTC has one of the lowest frequency of The recent whole exome sequencing of thousands of DNA methylation alterations. Specifically, Yanget al. tumoral samples of different cancer types has revealed performed a differential DNA methylation analysis that many genes controlling the epigenome are mutated, between normal and tumoral samples (n = 5480) for which can lead to epigenetic aberrations (reviewed in You 15 cancer types showing high variability in the numbers & Jones 2012). This is the case for mutations in the writers, of differentially methylated CpGs, which ranged from erasers and readers of DNA methylation. Mutations in the 3722 in PTC to 57,290 in uterine corpus endometrial DNMT and TET genes have been identified in different carcinoma (Yang et al. 2016a). Accordingly, another DNA cancers; for example, DNMT3A and TET2 are frequently methylation pan-cancer study focused on promoters mutated in hematologic malignancies. In thyroid cancer, found that PTC exhibited one of the lowest frequencies mutations in these genes are rare (<1.5% in PTC and in both hypomethylation and hypermethylation events <3% in ATC and PDTC) (data from The Cancer Genome (Saghafiniaet al. 2018) (Fig. 4). In addition, these authors Atlas Research Network; http://www.cbioportal.org/) introduced the concept of DNA methylation instability, (Cerami et al. 2012). However, the expression of some of which was found to be very low in PTC. In contrast, ATC these genes is altered (Supplementary Fig. 2) and could exhibits a high frequency of DNA methylation alterations contribute to the dysregulation of DNA methylation (10-fold higher than PTC; Bisarro dos Reis et al. 2017).

600 CRC BRCA LUAD PRAD THCA 500 s Figure 3 n o

i Comparison of the overall number of science t

a 400 citation indexed publications in the field of DNA c i l methylation in different cancer types over the last b

u 300 25 years (1990 to 2017). A search was carried out p

f on the Web of Science (Thompson Reuters) on o

r Nov 2018 using a date-restricted search

e 200

b (1990–2017) and ‘DNA methylation’ AND ‘colorectal

m cancer’ OR ‘colon cancer’, ‘DNA methylation’ AND u

N 100 ‘breast cancer’, ‘DNA methylation’ AND ‘lung cancer’, ‘DNA methylation’ AND ‘prostate cancer’ or ‘DNA methylation’ AND ‘thyroid cancer’ as search 0 terms. BRCA, breast cancer; CRC, colorectal 1990 1995 2000 200520102015 2020 cancer; LUAD, lung cancer; PRAD, prostate cancer Publication year PRAD; THCA, thyroid cancer.

Interestingly, these epigenetic differences between origin in carcinomas of unknown primary origin (CUPs) PTC and ATC also can be found at the level of genetic (Moran et al. 2016). alterations. A recent pan-cancer analysis on whole exome sequencing revealed that the mutation frequency in PTC Association between methylomes and histology in was one of the lowest (approximately 1 change/Mb across thyroid cancer the entire exome) among solid tumors (Lawrence et al. Genome-wide studies to profile thyroid cancer 2013) (Fig. 4), while the mutation frequency in ATC was at methylomes, most of which used BeadArrays (Table 1), the opposite extreme and was closer to that in melanoma revealed histology-associated DNA methylation profiles. and lung cancer, exceeding 100 changes/Mb (Kunstman Specifically, PTC is characterized by a higher number et al. 2015, Riesco-Eizaguirre & Santisteban 2016). As of hypomethylations (most of them outside promoter mutations are largely caused by errors in DNA replication regions) than hypermethylations in comparison to (Tomasetti et al. 2017), some researchers propose that the normal thyroid tissues (Ellis et al. 2014, Mancikova et al. cell division rate also participates in shaping the cancer 2014, White et al. 2016, Beltrami et al. 2017, Bisarro DNA methylation landscape (Yang et al. 2016b). Thus, dos Reis et al. 2017) (Supplementary Fig. 3). Only the different proliferation rate between PTC (low) and the study by Rodríguez-Rodero et al. identified more ATC (high) could explain, at least in part, the different hypermethylations than hypomethylations, which frequencies of their DNA methylation alterations. could be explained by the low number of analyzed PTCs (Rodríguez-Rodero et al. 2013). In contrast to PTC, FTC exhibits more hypermethylations than Thyroid cancer DNA methylation patterns hypomethylations (most of them outside promoter Since the initial studies assessing the DNA methylation of regions) (Rodríguez-Rodero et al. 2013, Mancikova a single locus, there has been accelerating technological et al. 2014, Bisarro dos Reis et al. 2017, Affinito et al. progress providing a plethora of DNA methylation 2019) as well as follicular adenomas (FA), although the techniques that have allowed the generation of single number of DNA alterations in these benign tumors CpG resolution maps (Fig. 1 and Supplementary Table 1). is low, thus resembling normal thyroid methylomes These maps have improved our understanding of DNA (Supplementary Fig. 3). There is debate within the methylation and have shown that DNA methylation field about whether FA and FTC are distinct molecular patterns, the so-called methylomes, are tissue specific, entities or represent a biological continuum (Arora et al. allowing us to distinguish different normal tissues 2008, Krause et al. 2011, Yoo et al. 2016). Interestingly, from each other (Hansen et al. 2011, Fernandez et al. Mancikova et al. showed that most of the FA-associated 2012). Moreover, methylomes differ largely between promoter hypermethylations that they identified normal and tumoral cells and between different types were also found in FTC, suggesting a progressive gain of tumors, which is key from a translational point of of hypermethylations along the tumorigenic process view. An example of the clinical use of this specificity is from adenomas to carcinomas, thereby reinforcing the that methylomes allow the identification of the tissue of hypothesis that some FAs have the malignant potential

Table 1 Summary of studies in thyroid cancer using methods for genome-wide analysis of DNA methylation.

Study Integration Identified genes No. Method Discovery series (n) Hypera Hypob expression of interest References 1 MCA/CGI array Cell lines (2) NA NA RT-qPCR HMGB2, FGD1 Hou et al. (2011) 2 GoldenGate NT (25); TC (36) NA NA – Hansen et al. (2011) 3 27K NT (10); PTC (14) NA NA RT-qPCR HIST1H3J, POU4F2, SHOX2, Kikuchi et al. (2013) PHKG2, TLX3, HOXA7 4 27K NT (2) GEO array ADAMTS8, HOXB4, ZIC1, Rodríguez-Rodero data KISS1R, INSL4, DPPA2, et al. (2013) PTC (2) 309 14 TCL1B, NOTCH4, MAP17 FTC (2) 408 24 ATC (2) 114 174 cell lines (4) 5 27K NT (8) GEO array COL4A2, DLEC1, KLK10, Mancikova et al. (2014) FA (18) 89 9 data EI24, WT1 PTC (42), fvPTC (5) 39 53 FTC (18) 460 83 6 450K NT (8) – Ellis et al. (2014) PTC (29) 255 2582 fvPTC (15) 164 405 recurrent PTC (7) 1023 2796 7 450K NT (56); cPTC NA NA RNA-Seq; Cancer Genome Atlas (324); fvPTC (99); miRNA-Seq Research Network tcPTC (35); Other (2014) PTC (38) 8 450K NT (12) – Timp et al. (2014) DTC (24) 24130 19857 9 450K NT (16) – CDKN1B, mir-146, PDGF, White et al. (2016) PTC (13) 165 1061 SERPINA1, TGFB1, TPO, DUSP5, ERBB3, FGF1, FGFR2, GABRB2, HMGA2 10 450K NT (41) GEO array ERBB3, FGF1, FGFR2, Beltrami et al. (2017) PTC (41) 645 5425 data GABRB2, HMGA2, RDH5 11 450K NT (50) – PFKFB2, LAIR2, THSD7B, Bisarro dos Reis et al. OR52B2, OR2T6, FFAR2, (2017) BTLs (17) 1531 222 RTN3, DCD, ADGRE2, PTC (60) 242 2773 HRH1, GPR21, MBP, FTC (8); HCTC (2) 4100 1475 YPEL4, ATP6V0C PDTC (1); ATC (3) 6195 28252 Affinitoet al. (2019) 12 450K NT (7) RNA-Seq FA (10) 0 0 FTC (11) 2979 585 13 RRBS NT (3) RNA-Seq CXCL12, FBLN7, FAM3B, Zhang et al. (2017) PTC (3) 639 907 PROX1, COL23A1, GJB3, LAD1 14 RRBS NT (39) – Diagnostic DNA Yim et al. (2019) BTLs (28) NA NA methylation signature (DDMS) PTC (39) NA NA aHyper, number of hypermethylation events compared to normal tissue; bHypo, number of hypomethylation events compared to normal tissue. 27K, Infinium HumanMethylation27 BeadChip; 450K, Infinium HumanMethylation450 BeadChip; ATC, anaplastic TC; BTLs, benign thyroid lesions; cPTC, classical PTC; DTC, differentiated TC; FA, follicular adenoma; FTC, follicular TC; fvPTC, follicular variant of PTC; GEO array; Gene Expression Omnibus; MCA/CGI, methylated CpG island amplification/CpG island; NA, not available; HCTC, Hurthle cell TC; NT, normal tissue; PDTC, poorly differentiated TC; PTC, papillary TC; RRBS, reduced representation bisulfite sequencing; TC, thyroid cancer, tcPTC, tall cell PTC. to give rise to FTC (Mancikova et al. 2014). Accordingly, and normal thyroid tissues (Bisarro dos Reis et al. 2017, unsupervised clustering analysis in the study by Bisarro Affinito et al. 2019). dos Reis et al. showed that FAs clustered with FTCs, and The majority of genome-wide studies are focused on a recent study by Affinito et al. found that FAs displayed PTC. Interestingly, some of them specify the PTC variants an intermediate DNA methylation profile between FTCs used, revealing differential DNA methylation profiles.

Figure 4 Epigenetic and genetic alterations by tumor type. (A) Hypermethylation and (B) hypomethylation event frequencies in 6010 human tumors across 24 cancer types. Frequencies are estimated as the percentage of probes found hypermethylated (out of 64,414) or hypomethylated (out of 3423) Epigeneticalterations compared to normal tissues. (C) Somatic mutation A Hypermethylation frequency frequencies in 3025 tumor-normal sample pairs

2 across 27 cancer types. Tumor types are sorted by their median hypermethylation, hypomethylation 1 and somatic mutation frequencies. Papillary

0 thyroid tumors (red bars) are among tumors with the lowest frequency of epigenetic and genetic -1

log(Frequency) alterations, mostly leukemias and pediatric -2 cancers. The x-axis gives the number of samples 502 276 66 89 600 381 334 57 304468 151282 308350 80 192 225 257374 186 323 134379 292 -3 for each tumor type. Data from Lawrence et al. PTCKIRPKICHBRCA.B OV UCECPRAD UCS LUAD BRCA.L BLCAKIRC HNSCCRC ACCAML LUSC CESC SKCM ESCASTADGBM LIHC LGG (2013) and Saghafiniaet al. (2018). ACC, adrenocortical carcinoma; AML, acute myeloid B Hypomethylation frequency leukemia; BLCA, bladder carcinoma; BRCA.B, basal

2 breast invasive carcinoma; BRCA.L, luminal breast invasive carcinoma; Car, carcinoid tumors; CESC, 1 cervix squamous cell carcinoma; CLL, chronic 0 lymphocytic leukemia; CRC, colon and rectum

-1 carcinoma; DLBCL, diffuse large B-cell lymphoma; log(Frequency) ESCA, esophageal carcinoma; EwngSRC, Ewing -2 276 334 502 282 192 350 186257 292 304 308468 89 66 381323 225 134 80 600 151 57 374 379 sarcoma; GBM, glioblastoma multiforme; HNSC,

KIRP PRAD PTCKIRC AML CRC ESCACESCLGG LUAD HNSCBRCA.LBRCA.BKICH UCECSTAD LUSC GBMACC OV BLCA UCSSKCMLIHC head and neck squamous cell carcinoma; KICH, kidney chromophobe carcinoma; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LGG, low grade glioma; Genetic alterations LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell C Somatic mutation frequency 5 carcinoma; MD, medulloblastoma; NB, neuroblastoma; MM, multiple myeloma; OV, 4 ovarian carcinoma; PAAD, pancreatic 3 adenocarcinoma; PRAD, prostate 2 adenocarcinoma; PTC, papillary thyroid cancer; 1 RhD, Rhabdoid tumor; SKCM, skin cutaneous log(Frequency/Mb) 0 melanoma; STAD, stomach adenocarcinoma;

-1 22 20 52 134262381 227 91 57 121 13 63 21411 335219 20 49 181231 76 88 35 336 179 121 UCEC, uterine corpus endometrial carcinoma; RhD EwngSRC PTC AML MD CarNBPRADCLL LGGBRCA PAAD MM KIRC KIRP OV GBMCESC DLBCL HNSCCRC ESCASTAD BLCA LUAD LUSC SKCM UCS, uterine carcinosarcoma.

Ellis et al. found that classical PTC (cPTC) displayed which was characterized by the hypermethylation of a high number of DNA alterations, most of which numerous CGIs) and two groups enriched by cPTC and were hypomethylations, whereas follicular variant tcPTC (Meth-classical 1 and Meth-classical 2, which were of PTC (fvPTC) exhibited a smaller proportion of characterized by hypomethylations outside of CGIs). hypomethylations (Ellis et al. 2014). In this regard, Interestingly, a small subset of fvPTCs resembled tcPTC this study, as well as those by Mancikova et al. and the and cPTC. Mancikova et al., who also included FA and Cancer Genome Atlas Research Network, found that FTC in the study, showed that fvPTC methylomes fvPTC exhibited a methylome that was not as different were more similar to follicular tumors than to cPTC from that of normal thyroid tissue (Mancikova et al. (Mancikova et al. 2014). These results indicate that PTCs 2014, Cancer Genome Atlas Research Network 2014). with follicular architecture are different from PTCs with Apart from cPTC and fvPTC, the Cancer Genome Atlas papillary architecture. In the future, it will be of great Research Network also analyzed tall cell PTC (tcPTC). interest to investigate whether the new noninvasive Based on an unsupervised clustering analysis, they follicular neoplasm with papillary-like nuclear features classified tumors into four groups: two groups enriched by (NIFTP) entity shows a specific methylation profile. fvPTC (Meth-follicular, which exhibited few methylation Two of the array-based studies included several PDTC changes compared to normal tissue and Meth-CGI, and ATC samples (Table 1 and Supplementary Fig. 3).

While the study by Rodríguez-Rodero et al. identified cells in tumors may affect DNA methylation patterns fewer DNA methylation alterations in ATC than in (Bisarro dos Reis et al. 2017). DTC, Bisarro dos Reis et al. identified a drastically higher Globally, most DNA methylation alterations in number of DNA methylation alterations in PDTC and thyroid cancer occur outside promoter regions and are ATC (six-fold higher than in FTC and ten-fold higher than specifically associated with histology. in PTC) (Rodríguez-Rodero et al. 2013, Bisarro dos Reis et al. 2017). The contradictory results of these two studies are probably due to the low number of analyzed samples Association between DNA methylation and genetic and the use of different arrays (27K vs 450K platforms) drivers in thyroid cancer that cover different regions of the genome (Rodríguez- Another important finding derived from these studies Rodero et al. 2013, Bisarro dos Reis et al. 2017). However, is the relationship between DNA methylation profiles both studies concluded that PDTC and ATC exhibited and mutations in BRAF and RAS genes; BRAF-mutated more hypomethylation than hypermethylation events, tumors harbor more hypomethylations (which is suggesting the association of hypomethylation with expected since this mutation is almost exclusively dedifferentiation. The similarity between the methylomes detected in PTC), while RAS-mutated tumors harbor more of ATC, PDTC, extensively invasive FTC and lymphocytic hypermethylations (which is expected since this mutation thyroiditis found by Bisarro dos Reis et al. is noteworthy mostly occurs in fvPTC and FTC). These observations (Bisarro dos Reis et al. 2017); as these aggressive tumors are were confirmed by the pan-cancer study by Saghafinia characterized by a high level of immune cell infiltration et al. who found a significant association between NRAS (Ryder et al. 2008), these results suggest that part of the mutation and hypermethylation events and between DNA methylation alterations in these tumors may come BRAF mutation and hypomethylation events in thyroid from infiltrating immune cells. cancer (Saghafiniaet al. 2018). Interestingly, although RAS BeadArrays are widely used in genome-wide DNA mutations are common in many types of tumors such as methylation studies due to their low cost, but they are lung adenocarcinoma and prostate cancer this genetic– limited to the CpGs covered by the array. However, reduced epigenetic relationship was not detected in other cancers representation bisulfite sequencing (RRBS), which is based or even in melanomas that mostly harbor NRAS mutations on next-generation sequencing, has a higher sensitivity, such as thyroid cancer. Therefore, an association between resolution and coverage than BeadArrays. There are DNA hypermethylation and a specific RAS isoform could two studies that investigated PTC methylomes using be discarded. The BRAF(V600E) mutation is also frequent RRBS (Table 1). Zhang et al. focused on the methylation in colorectal cancer and melanoma, but there is no of mRNA and lncRNA promoters and confirmed association between this mutation and hypomethylation previous results showing more hypomethylation than events. Conversely, BRAF mutation is strongly associated hypermethylation events in PTC (Zhang et al. 2017). with hypermethylation events in colorectal cancer However, when DNA methylation and expression data (Weisenberger et al. 2006, Cancer Genome Atlas Research from the same samples were crossed, only 19 mRNAs Network 2012, Saghafinia et al. 2018). Specifically, in were upregulated/hypomethylated, and 26 mRNAs colorectal cancer, BRAF(V600E) has been associated with and 3 lncRNAs were downregulated/hypermethylated. the so-called ‘CpG island methylator phenotype’ (CIMP) These findings, which are consistent with results from (including tumors that exhibit an exceptionally high Mancikova et al. and Affinito et al. (Mancikova et al. 2014, frequency of the hypermethylation of some CGIs) (Toyota Affinito et al. 2019), suggested that DNA methylation in et al. 1999). From studies done in melanoma, there are promoters does not have a widespread role in controlling some controversial results, but most studies do not find gene expression in PTC. On the other hand, by using any significant association between BRAF mutation and RRBS, Yim et al. identified a unique DNA methylation hyper- or hypomethylations (Lauss et al. 2015, Saghafinia signature of 4,575 CpGs specific to benign nodules, most et al. 2018). of which were hypermethylated compared to adjacent Altogether, these findings show a cancer type-specific normal tissues and malignant nodules, that may have relationship between BRAF or RAS mutations and aberrant important diagnostic applications (Yim et al. 2019).They DNA methylation. However, whether these events are also analyzed specimens with lymphocytic thyroiditis dependent on one another requires further studies in and, in accordance with the results from Bisarro dos Reis controlled experimental systems (e.g., cell lines, mouse et al., suggested that the presence of immune-infiltrating models). In this regard, in BRAF-mutated colorectal tumors,

HighlyHighly frequentfrequent geneticgenetic alaltterationerations1 Lowly frequent genetic alterations2 DNA methylation alterations3

Histology

cPTC, tcPTC BRAF(V600E),BRRAF(V600E), RETRET ffusions,usions, BRBRAFRAF ffusiusiononss N/H/K-RAS Hypo

BRAF other, BRAF fusions, RET fusions, N/H/K-RAS,N/H/K-RAS, BRABRAF(V600E)F(V600E) Hyper, hypo fvPTC EIF1AX

4 FTC N/H/K-RASN/H/K-RAS, EIEIF1AX,F1AX, PPAX8AX8-PPARG-PPPARG Hyper

BRAF/RAS mutational state cPTC,cPTC, tcPTC,tcPTC, fvPTfvPTC fvPTC,fvPTC, FTFTC cPTC, tcPTC fvPTC, FTC cPTC,cPTC, tcPTC,tcPTC, ffvPTvPTC FTFTC

BRAF BRAF(V600E)BRRAF(V600E) Hypo

RAS N/H/K-RASN/H/K-RAS N/H/K-RAS Hyper

BRAF other, EIF1AX 1 , PAX8-PPARG, RET BRAF/RAS wt BRAFBRRAF ffusions,usions, RRETET ffusionusionss Hypo, hyper fusions

cPTC, tcPTC, fvPTC FTC cPTC, tcPTC, fvPTC FTC cPTC, tcPTC, fvPTC FTC BRAF/RAS-like phenotype

BRAF(V600E),BRRAF(V600E), RETRET BRAF fusions Hypo BRAF-like fusionfusionss

RAS-like N/H/K-RASN/H/K-RAS BRAF other, PAX8-PPARG, EIF1AX Hyper

Figure 5 Genetic and epigenetic alterations associated with DTC according to histology, BRAF and RAS mutational state and BRAF-like and RAS-like phenotypes. 1Genetic alterations were considered highly recurrent when present in >5% tumors based on Cancer Genome Atlas Research Network data. 2Genetic alterations were considered lowly recurrent when present in <5% tumors based on Cancer Genome Atlas Research Network data. 3Hyper, more hypermethylations than hypomethylations; hypo, more hypomethylations than hypermethylations. 4FTC data from Yoo et al. (2016). cPTC, classical PTC; FTC, follicular thyroid cancer; tcPTC, tall cell PTC fvPTC, follicular variant of PTC.

BRAF or RAS exhibiting more hypomethylations than cancer types, the degree of global DNA hypomethylation hypermethylations are BRAF-like while those exhibiting is strongly associated with the tumor grade and stage, more hypermethylations than hypomethylations are which has attracted great interest for its potential clinical RAS-like. This is in agreement with the study from Chen use. Nevertheless, little is known about global DNA et al., although they do not specifically use the BRAF- hypomethylation in thyroid cancer. As shown in Table 2, like and RAS-like terms (Chen et al. 2017). These findings as far as we know, a total of nine studies have assessed suggest that hypo- and hypermethylation events may global DNA methylation levels in thyroid tumors and be downstream of the pathways overactivated in the report conflicting results that may be partly explained by BRAF-like and RAS-like phenotypes, respectively. Further the low number of samples included in some studies and investigation should be conducted in this area. the different methods used. Five of the studies used techniques based on repetitive sequences. They revealed that LINE-1 elements of normal Global DNA hypomethylation in thyroid cancer and tumoral samples did not show different levels of A wide range of techniques to analyze global DNA DNA methylation (Chalitchagorn et al. 2004, Lee et al. methylation have been developed (Fig. 1 and 2008, Keelawat et al. 2015), while Alu elements were Supplementary Table 1) (reviewed in Jordà & Peinado slightly hypomethylated in PTC and FTC (Buj et al. 2010, Toraño et al. 2012), but there are some key points 2016). A deeper study of the hypomethylation of Alu to take into account when interpreting results that are elements showed that it occurred in distant metastatic summarized in the Supplementary material. As explained DTC, PDTC and ATC but not in low-risk DTC and above, global DNA hypomethylation is a common pediatric PTC (Klein Hesselink et al. 2018), suggesting the epigenetic feature of cancer. Interestingly, in many involvement of global hypomethylation of Alu elements

Table 2 Summary of studies in thyroid cancer analyzing global DNA methylation.

Study No. Method Global methylationa Discovery series (n)b Result References 1 COBRA-LINE1 Entire genome NT (7); PTC (7) No differences Chalitchagorn et al. (2004) 2 IHC with 5mC antibody Entire genome NT (9);NG (1); Global hypomethylation in de Capoa et al. PTC (3); fvPTC (1); tumors (2004) FTC (2) 3 IHC with 5mC antibody Entire genome NT (17);NG (19); Global hypomethylation in Galusca et al. HCA (10); FA (16); PTC and FTC (2005) PTC (17); FTC (6) 4 LINE-1 pyrosequencing/ Compartment NT (21); FTC (21) No differences Lee et al. (2008) LUMA 5 ELISA with 5mC antibody Entire genome NT (10); NG (24) No differences Brown et al. (2014) 6 COBRA-LINE1/IHC with 5mC Compartment/ NT (50); FA (15); No differences in LINE-1 Keelawat et al. antibody entire genome PTC (17); FTC (18) methylation but global (2015) hypermethylation in tumors 7 QUAlu Compartment NT (9); cPTC (31); Global Alu Buj et al. (2016) FTC (14) hypomethylation in PTC and FTC 8 QUAlu Compartment NT (20); PTC (40); Global Alu Klein Hesselink FTC (21); PDTC hypomethylation in et al. (2018) (7); ATC (9); M1 distant metastatic PTC (24); pediatric and distant metastatic PTC (13) FTC as well as the paired M1, and in PDTC and ATC 9 ELISA with 5mC antibody Entire genome Blood: controls (6); No differences Ceolin et al. PTC (12) (2018) aPart of the genome in which global DNA methylation has been analyzed. bDNA methylation was assessed in postsurgical tissue unless otherwise stated. 5mC, 5-methylcytosine; ATC, anaplastic thyroid cancer; COBRA, combined bisulfite restriction analysis; cPTC, classical PTC; FA, follicular adenoma; fvPTC, follicular variant of PTC; FTC, follicular thyroid cancer; IHC, immunohistochemistry; HC, Hurthle cell adenoma; LUMA, luminometric methylation assay; M1, distant metastasis; NG, nodular goiter; NT, normal tissue; PDTC, poorly differentiated thyroid cancer; PTC, papillary thyroid cancer; QUAlu, quantification of unmethylated Alu.

https://erc.bioscientifica.com © 2019 Society for Endocrinology https://doi.org/10.1530/ERC-19-0093 Published by Bioscientifica Ltd. Printed in Great Britain Downloaded from Bioscientifica.com at 09/28/2021 06:06:52AM via free access Endocrine-Related C Zafon et al. DNA methylation in thyroid 26:7 R426 Cancer cancer in thyroid cancer progression and dedifferentiation. Keelawat et al. did not find global hypomethylation but This is in agreement with studies in other cancer types, rather found global hypermethylation (Keelawat et al. such as hepatocellular carcinoma or cervical cancer, in 2015). These inconsistent results are probably due to which global hypomethylation correlates with disease technical issues that mainly include the use of different progression (Lin et al. 2001, Yegnasubramanian et al. 2008). antibodies with different sensitivities. In this regard, These findings in thyroid cancer may have important the two studies showing global hypomethylation used prognostic applications, especially in preoperative fine- the same antibody. Although further investigation is needle aspiration biopsies (FNAB), which would help in needed, these results suggest the diagnostic potential of treatment planning. The differences between the results global DNA methylation measured by specific antibodies. about LINE-1 and Alu elements could be explained by the Moreover, the fact that benign lesions do not display fact that the studies analyzing LINE-1 elements included global alterations is very promising, especially for thyroid FA, PTC and FTC but did not include aggressive tumors. nodules with indeterminate cytology. We cannot discard the use of different techniques with Only one of the studies analyzed the relationship different sensitivity and accuracy as being responsible for between global hypomethylation and mutations in BRAF the different results. On the other hand, as explained in and RAS, and it found an association in distant metastatic the Supplementary material, these apparently opposite DTC but not in low-risk DTC (Klein Hesselink et al. findings are biologically plausible. In this regard, 2018). Specifically, BRAF-mutated distant metastatic DTC the different methylation between LINE-1 and Alu showed hypomethylation of the Alu elements, but RAS- elements also occurs in other cancers (Benard et al. 2013, mutated tumors did not. Interestingly, distant metastatic Park et al. 2014). DTC harboring no mutations in BRAF or RAS showed Another interesting result from these studies was notable variability, which could reflect the BRAF-like and that while global methylation of LINE-1 elements varied RAS-like phenotypes. between different normal tissues, especially in the normal Several studies indicate that the global DNA thyroid, all normal tissues displayed similar levels of methylation in peripheral blood leukocytes differs unmethylated Alu elements (Chalitchagorn et al. 2004, Buj significantly between healthy individuals and patients et al. 2016). Conversely, tumors showed a broad variation with different cancer types (Li et al. 2012). There is of the DNA methylation of both LINE-1 and Alu elements. only one study analyzing global DNA methylation in For example, colon and lung cancer exhibited 2- to 3-fold blood from PTC patients and control individuals (using higher levels of unmethylated Alu elements than thyroid a 5mC DNA ELISA), and they did not find differences cancer. On the other hand, the Alu hypomethylation was (Ceolin et al. 2018). similar between distant metastases and matched primary Altogether, these findings indicate that global tumors, suggesting that Alu methylation remained hypomethylation plays a role in thyroid tumorigenesis. stable during metastatic spread in thyroid cancer (Klein However, further investigation is required. The analysis Hesselink et al. 2018). of global hypomethylation in different compartments of Four more studies used antibodies that recognize 5mC the genome using different techniques in a large cohort to evaluate global hypomethylation in thyroid cancer, of thyroid samples would provide enlightening insight. and three of them were based on immunohistochemistry while one used ELISA. The analyses of benign lesions Focal DNA methylation alterations in thyroid cancer: (hyperplasia, FA and Hurthle adenoma) did not find gene-specific studies differences between benign lesions and normal thyroid tissues (Galusca et al. 2005, Brown et al. 2014, Keelawat Gene-specific DNA methylation has been broadly studied. et al. 2015), except for de Capoa et al., but they only Accordingly, many investigators in thyroid cancer have analyzed one sample (de Capoa et al. 2004). Thus, focused on the hypermethylation of specific tumor global DNA hypomethylation in thyroid cancer does suppressor genes (TSGs) as an alternative to mutational not seem to be an early event as described in other inactivation (Supplementary Table 2). Some of the most cancer types (Ehrlich 2009). The results on global DNA recurrently hypermethylated TSGs in thyroid cancer are hypomethylation in DTC were more variable. de Capoa Ras association domain family 1, isoform A (RASSF1A), et al. and Galusca et al. showed global hypomethylation cyclin-dependent kinase inhibitor 2A (CDKN2A or in malignant tumors compared to normal tissues P16INK4A) and death-associated protein kinase1 (DAPK) (de Capoa et al. 2004, Galusca et al. 2005). In contrast, (Table 3). The RASSF1A gene encodes a signaling protein

et al. et al. et al. (2005) (2011) (2013) Continued ( (2011) (2015) (2018) (2005) (2014) et al. et al. et al. (2012) (2004) (2008) et al. et al. et al. (2008) et al. et al. et al. et al. et al. et al. (2002) (2006) (2011) Reference Schagdarsurengin Xing Hoque Nakamura Schagdarsurengin Lee Hou Mohammadi-asl Czarnecka Stephen Brait Kunstman Brown Stephen Stephen

mutation mutation

BRAF

BRAF BRAF

mutation tumorigenesis. No relationship with mutation in older patients tumorigenesis tumorigenesis tumorigenesis exclusive with aggressiveness tumorigenesis compared to FTC NT and tumors frequency in aggressive variants tumorigenesis. Hypermethylation mutually exclusive with BRAF exclusive with Early event in thyroid Higher methylation frequencies – – Early event in thyroid Early event in thyroid Early event in thyroid Hypermethylation mutually Related to Early event in thyroid Hypermethylation in HCTC Good discrimination between Higher hypermethylation Early event in thyroid Hypermethylation mutually Potential clinical value a

PDTC/ATC M – – M M – M – – – – – – – –

DTC M hyper hyper M M hyper M hyper M M hyper hyper hyper M hyper

Methylation status

BTLs U hyper hyper M M – M M M M M – hyper – hyper

U U M U – U – – M M M U U – U NT

) n

(

FTC (10), PDTC (1), ATC (9), CL (9) (12), PTC (30) (24), PTC (23), fvPTC (10), CL (5) (23),PTC (42), FTC (4), ATC (12), CL (3) (13), FTC (10), ATC (9), CL (9) (36), CL (5) fvPTC (15), tcPTC (3) FTC (2) cPTC (17), fvPTC (10), FTC (7), HCTC (2) (10), FTC (10) HCTC (44), FTC (46), fvPTC (42), cPTC (53) NT (4), NG (1), PTC (13), PTC (1), NG (4), NT NT (14), BTLs (9), FTC NT (15), NG (20), FA NT (42), FA (3), HTT NG (12), FA (10), PTC NT (21), FTC (21) FA(42), FTC (65), ATC BTLs (25), PTC (25) NG (20), cPTC (27), (11), PTC (3), NG (5), NT NT (15), BTLs (44), NT (18), PTC (41) NT (29), BTLs (23), FA HCTC (26), FTC (27) NT (71), FA (83), Discovery series

MSP qMSP MSP, qMSP MSP MSP Pyrosequencing qMSP COBRA MSP MS-MLPA qMSP PCR + MS-RE MS-RE + qPCR qMSP qMSP Method

1 2 3 4 5 6 7 8 9 15 10 11 12 13 14 Study No. Summary of candidate approach DNA methylation studies on classical tumor suppressor genes.

Table 3 Gene RASSF1A

et al. et al. et al. (2011) (2018) (2015) (2018) (2005) (2005) (2003) (2007) et al. (2013) (1998) (2012) (2012) et al. et al. et al. (2006) et al. et al. et al. et al. et al. et al. et al. et al. et al. (2006) (2011) (2006) Reference Elisei Boltze Hoque Schagdarsurengin Ishida Mohammadi-asl Czarnecka Brait Wang Stephen Hoque Hu Schagdarsurengin Brait Stephen Stephen

– – – – – – – frequency in aggressive variants. Related to N1 and M1 parameters frequency in aggressive variants parameters parameters frequency in cPTC and tcPTC. Related with multifocality frequency in aggressive variants Higher hypermethylation No clorrelation with clinical any Higher hypermethylation Related to progression No clorrelation with clinical any No clorrelation with clinical any Related to aggressiveness Higher hypermethylation Higher hypermethylation Potential clinical value a

PDTC/ATC – M – M – – – – – – – M – –

DTC M M U M M hyper M U M U M M M M U U Methylation status

BTLs M M U U – U M U – U M – U M – U

– U U – – – – U U U M – – M – U NT ) n (

(16), FTC (18), PDTC (12), ATC (13) (24), PTC (23), fvPTC (10), CL (5) (13), FTC (10), ATC (9), CL (9) fvPTC (15), tcPTC (3) cPTC (17), fvPTC (10), FTC (7), HCTC (2) (44), FTC (46), fvPTC (42), cPTC (53) (24), PTC (23), fvPTC (10), CL (5) tcPTC (22) (13), FTC (10), ATC (9), CL (9) cPTC (17), fvPTC (10), FTC (7), HCTC (2) (44), FTC (46), fvPTC (42), cPTC (53) FA (8), PTC (12), CL (4) NT (15), FA (18) PTC NT (15), NG (20), FA NG (12), FA (10), PTC PTC (39) BTLs (25), PTC (25) NG (20), cPTC (27), NT (15), BTLs (44), NT (21), PTC (74) NT (71), FA (83), HCTC NT (15), NG (20), FA cPTC (127), fvPTC (82), NG (12), FA (10), PTC NT (15), BTLs (44), HCTC (26), FTC (27) NT (71), FA (83), HCTC Discovery series

MSP MSP MSP, qMSP MSP MSP COBRA MSP qMSP MSP qMSP MSP, qMSP qMSP MSP qMSP qMSP qMSP Method

1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 10

Study No. Continued

(CDKN2A) Methylation status of gene promoter: M, methylated; U, unmethylated; hyper, hypermethylated compared to normal tissue. Table 3 Gene P16INK4A DAPK1 a ATC, anaplastic thyroid cancer; BTLs, benign lesions; CL, cell lines; COBRA, combined bisulfite restriction analysis; cPTC, classical PTC; FA, follicular adenoma; FTC, fvPTC, follicular variant of PTC; HCTC, Hurthle cell thyroid cancer; HTT, hyalinizing trabecular tumor; MSP, methyl-sensitive PCR; MS-RE, methylation-sensitive restriction enzyme; NG, nodular goiter; NT, normal tissue; PDTC, poorly differentiated thyroid cancer; PTC, papillary qMSP, quantitative MSP.

https://erc.bioscientifica.com © 2019 Society for Endocrinology https://doi.org/10.1530/ERC-19-0093 Published by Bioscientifica Ltd. Printed in Great Britain Downloaded from Bioscientifica.com at 09/28/2021 06:06:52AM via free access Endocrine-Related C Zafon et al. DNA methylation in thyroid 26:7 R429 Cancer cancer containing a Ras association domain and is involved in In addition, the methylation of thyroid-specific multiple apoptotic and cell cycle checkpoint pathways. genes, such as thyroid-stimulating hormone receptor The main mechanism of RASSF1A inactivation, a (TSHR) or NIS, has also been extensively investigated in frequent event in many cancers, appears to be through thyroid cancer (Table 4 and Supplementary Table 2). TSHR promoter methylation rather than mutational events plays a central role in the regulation of thyroid growth (Dammann et al. 2000, Agathanggelou et al. 2005). and function. Somatic TSHR mutations have been found Schagdarsurengin et al. found for the first time that in both benign and malignant thyroid neoplasms (Davies RASSF1A was hypermethylated in thyroid cancer with a et al. 2010), but the involvement of TSHR genetic status in slightly higher frequency in more aggressive hystotypes thyroid carcinogenesis is not clear. A recent study showed (Schagdarsurengin et al. 2002). Many other studies that TSHR mutations may be associated with an increased confirmedRASSF1A hypermethylation in thyroid tumors, cancer risk when present at high allelic frequency (Mon including a recent meta-analysis, although most results et al. 2018). In thyroid cancer, TSHR expression is also revealed a considerable overlap in methylation levels repressed by aberrantly methylation of gene promoter between benign and malignant tumors (Nakamura (Table 4). However, several studies found TSRH methylated et al. 2005, Hou et al. 2008, Mohammadi-asl et al. 2011, in bening lesions (Hoque et al. 2005, Schagdarsurengin Stephen et al. 2011, Niu et al. 2017) (Table 3), suggesting et al. 2006, Brait et al. 2012, Kartal et al. 2015, Stephen that RASSF1A hypermethylation may be an early et al. 2018), which limits the discriminatory power for epigenetic event in thyroid carcinogenesis (Xing et al. diagnostic purposes. Interestingly, TSHR methylation has 2004, Brown et al. 2014). Different studies have shown the been found inversely associated with tumor recurrence potential of RASSF1A hypermethylation as a biomarker of (Smith et al. 2007). NIS is a transmembrane glycoprotein aggressive tumors, while other authors have failed to find that mediates the active transport of iodide from the any relationship between RASSF1A hypermethylation bloodstream into the follicular thyroid cells and is and prognostic factors (Schagdarsurengin et al. 2006, mainly regulated by the thyroid-stimulating hormone Mohammadi-asl et al. 2011, Brait et al. 2012, Niu et al. (TSH). The role of NIS is key for effective diagnosis and 2017). Thus, further studies are required in larger series treatment of thyroid cancer since RAI accumulation is of samples using more quantitative techniques. P16INK4A primarily mediated by NIS. Accordingly, the decreased is a cell cycle regulator that induces G1 phase arrest, NIS expression and/or impairment in NIS plasma whose functional loss is frequent in cancer. Mutations in membrane trafficking De( la Vieja & Santisteban 2018) this gene are rarely observed in primary thyroid tumors are well demonstrated factors showing poor prognosis in (Calabrò et al. 1996, Yane et al. 1996), whereas promoter thyroid cancer. However, the relationship between NIS hypermethylation, which causes gene silencing, is quite and thyroid cancer is complex and not well understood common (Table 3). Many studies, including a meta- (de Morais et al. 2018). Mutations in the NIS gene do not analysis based on 17 case–control studies (804 thyroid appear to be a major cause for reduced NIS expression/ cancer patients, 487 controls), confirmed the significantly function in thyroid cancer (Russo et al. 2001). In contrast, higher frequency of P16INK4A hypermethylation in many studies have reported the methylation of NIS thyroid cancer than in normal samples (Boltze et al. 2003, promoter although results are controversial (Table 4). Schagdarsurengin et al. 2006, Melck et al. 2007, Zafon Interestingly, Galrao et al. identified a distal enhancer that et al. 2008, Wu et al. 2015). However, its usefulness as a was hypermethylated in DTC regulating NIS expression prognostic marker remains questionable, as summarized (Galrão et al. 2014). This new finding provides a basis for in Table 3. DAPK1, which belongs to the DAPK family of further investigation in the epigenetic regulation of NIS. calcium/calmodulin-dependent kinases, participates in many cellular processes such as apoptosis, autophagy, and Focal DNA methylation alterations in thyroid cancer: cell survival, and is also involved in cancer (reviewed in genome-wide studies Farag & Roh 2019). DAPK1 promoter hypermethylation has been associated with an increased risk of developing In addition to validating results from candidate approach cancer and poor prognosis in several cancer types studies, genome-wide DNA methylation studies have (Dai et al. 2016, Qi & Xiong 2018, Yang et al. 2018). As allowed the identification of novel differentially shown in Table 3, DAPK1 is hypermethylated in benign methylated sequences that may regulate the expression and malignant thyroid tumors but the clinical value of of genes involved in thyroid cancer tumorigenesis DAPK1 hypermethylation is not well established. (Table 1). In this regard, Rodríguez-Rodero et al. showed

https://erc.bioscientifica.com © 2019 Society for Endocrinology https://doi.org/10.1530/ERC-19-0093 Published by Bioscientifica Ltd. Printed in Great Britain Downloaded from Bioscientifica.com at 09/28/2021 06:06:52AM via free access Endocrine-Related C Zafon et al. DNA methylation in thyroid 26:7 R430 Cancer cancer that in ATC, the membrane-associated protein 17 a new diagnostic method, the so-called diagnostic (MAP17) gene was hypomethylated in its promoter region DNA methylation signature (DDMS) approach, which and overexpressed compared to normal tissues. They is based on 373 differentially methylated regions with showed that overexpression of MAP17 induced tumor tissue-specific DNA methylation patterns in benign growth in vitro and in vivo (Rodríguez-Rodero et al. 2013). and malignant nodules (Yim et al. 2019). A notable Zhang et al. identified 14 novel genes regulated by DNA proportion of these markers were associated with active methylation in PTC (Table 1) that were used to construct enhancers and cancer-related genes. Importantly, the a core cofunction network that revealed the potential DDMS approach distinguishes benign from malignant of the C-X-C motif chemokine ligand (CXCL12), a nodules with high sensitivity and specificity and thus has chemokine involved in the immune response, as a key the potential to provide outstanding diagnostic accuracy player in thyroid tumorigenesis (Zhang et al. 2017). for thyroid nodules, which may decrease overdiagnosis Moreover, the expression levels of these 14 genes gave and unnecessary thyroidectomies. the ability to discriminate between PTC patients and healthy individuals. By integrating DNA methylation and transcriptomic data, Beltrami et al. found 185 DNA methylation as a therapeutic target in genes with a negative correlation between methylation thyroid cancer and expression that mostly affected fibroblast growth factor (FGF) and retinoic acid (RA) signaling pathways As explained, the aberrant methylation of DNA can play a (Beltrami et al. 2017). Other interesting hypomethylated key role in tumorigenesis. In addition, DNA methylation genes that were identified in genome-wide DNA is inherently reversible, which makes targeted therapies methylation analyses were high-mobility group box 2 against it very attractive for cancer treatment. Therefore, (HMGB2), which may play a role in PTC cell proliferation, much effort has been made to study the potential of and FYVE, RhoGEF and PH domain-containing 1 drugs that inhibit this type of epigenetic modification (FDG1), which may be involved in cell invasion (Hou to induce the re-expression of silenced genes in different et al. 2011). Additionally, Lin et al. identified HORMA malignancies. Over the past few decades, different domain-containing 2 (HORMAD2) and showed that demethylating drugs have been developed and tested in its hypermethylation and repression induced the different human neoplasms. There are two different classes progression of thyroid cancer, while its hypomethylation of demethylating agents: nucleoside DNMT inhibitors and and overexpression retarded cell growth and mobility and non-nucleoside DNMT inhibitors. Treatment with these facilitated apoptosis (Lin et al. 2018). agents causes a reduction in global DNA methylation From a translational point of view, genome-wide rather than demethylation in specific regions (reviewed DNA methylation studies are an important source of in Mani & Herceg 2010). new biomarkers to develop algorithms and tools with The most commonly used demethylating agents are diagnostic and prognostic value. For example, Mancikova the first ones described: 5-azacytidine (azacitidine, AZA) et al. identified two putative biomarkers associated with (Sorm et al. 1964) and 5-aza-2′-deoxycytidine (decitabine, recurrence-free survival, etoposide-induced 2.4 (EI24) DAC), both of which are nucleoside DNMT inhibitors. and Wilms’ tumor 1(WT1) (Mancikova et al. 2014). They Around 1970, clinical trials in Europe and the United also found kallikrein 10 (KLK10) to be hypomethylated States using AZA began focusing on the treatment of both and overexpressed in BRAF-mutated tumors. Further solid and blood neoplasms (Sorm & Vesely 1968). The analyses based on KLK10 allowed the development of results showed the effectiveness of treating patients with an algorithm related to BRAF- and RAS-like phenotypes acute myeloid leukemia (AML) resistant to conventional with prognostic implications in thyroid cancer (Buj treatment and/or with relapse with AZA and DAC. In et al. 2018). On the other hand, Bisarro dos Reis et al. contrast, no significant responses were found in other developed a prognostic algorithm based on 21 CpGs able types of blood cancers or in solid tumors to those drugs. to predict recurrence in DTC with high specificity but At that time, the US Food and Drug Administration (FDA) low sensitivity (Bisarro dos Reis et al. 2017). However, did not approve AZA due to its high levels of toxicity the series of samples used contained a low number of relative to its antitumor efficacy. Nearly 40 years later, in recurrent cases; thus, further analyses are required to 2004, after adjusting the dosage to reduce toxicity and validate its potential for prognostic use. Finally, Yim et al., increase efficiency, it was approved for clinical use to treat who profiled PTC DNA methylation by RRBS, developed myelodysplastic syndromes (MDS) (Kaminskas et al. 2005).

Table 4 Summary of candidate approach DNA methylation studies on thyroid-specific genes.

Study Discovery series Methylation statusb Potential clinical Gene No. Method (n)a NT BTLs DTC ATC value Reference TSHR 1 MSP, qMSP FA (8), PTC (39), – U hyper hyper Diagnostic marker Xing et al. (2003) FTC (15), ATC (11), CL (6) 2 MSP NG (12), FA – M M M – Schagdarsurengin (10), PTC (13), et al. (2006) FTC (10), ATC (9), CL (9) 3 MSP NT(2), NG (15), U M hyper – Inverse Smith et al. (2007) FA (10), PTC correlation with (30) recurrence 4 MSP FNAB: BTLs – M M – More frequent Kartal et al. (2015) (35), FA (4), methylation in PTC (28), PTC FTC (6) 5 qMSP NT (71), FA (83) M M M – – Stephen et al. (2018) cPTC (53), fvPTC (42), HCTC (44), FTC (46) 6 qMSP NT (15), NG M M M – – Hoque et al. (2005) (20), FA (24), PTC (23), fvPTC (10), CL (5) 7 qMSP NT (15), BTLs M M hyper – – Brait et al. (2012) (44), cPTC (17), fvPTC (10), FTC (7), HCTC (2) NIS (SLC5A5) 1 MSP NT(2), NG (15), U U hyper – No prognostic Smith et al. (2007) FA (10),PTC factor (30) 2 MS-MLPA NT (5), NG (3), M M M – Early event Stephen et al. (2011) PTC (11), FTC (2) 3 MSP NT (30), BTLs M M M – – Galrão et al. (2013) (10), PTC (18), FTC (2) 4 BS-sequencing NT (30), BTLs M M hyper – Correlation with Galrão et al. (2014) (10), PTC (18), expression FTC (2) 5c MSP, NT (24), PTC M – hyper – Related to Choi et al. (2014) BS-sequencing (24) BRAF(V600E) 6 qMSP HCTC (26), FTC – – M – No differences Stephen et al. (2015) (27) between HCTC and FTC aDNA methylation was assessed in postsurgical tissue unless otherwise stated. bMethylation status of gene promoter: hyper, hypermethylated compared to normal tissue; M, methylated; U, unmethylated. cMethylation status of NIS enhancer. ATC, anaplastic thyroid cancer; BS-sequencing, bisulfite sequencing; BTLs, benign thyroid lesions; CL, cell lines; cPTC, classical PTC; FA, follicular adenoma; FNAB, fine-needle aspiration biopsy; FTC, follicular thyroid cancer; fvPTC, follicular variant of PTC; HCTC, Hurthle cell thyroid cancer; MSP, methyl-sensitive PCR; NG, nodular goiter; PTC, papillary thyroid cancer; qMSP, quantitative MSP.

In 2006, DAC was also approved for the treatment of MDS Curiously, demethylating agents are more effective (Kantarjlan et al. 2006). More recently, other agents have in treating hematologic cancers than solid tumors been identified such as zebularine or procaine, and their despite the large amount of evidence showing that potential use in demethylating therapy is being tested aberrant DNA methylation is a trait common to all (Villar-Garea et al. 2003, Marquez et al. 2005, Mani & tumorigenic processes (Sharma et al. 2010). Such trouble Herceg 2010). in accomplishing therapeutic effectiveness could be due

https://erc.bioscientifica.com © 2019 Society for Endocrinology https://doi.org/10.1530/ERC-19-0093 Published by Bioscientifica Ltd. Printed in Great Britain Downloaded from Bioscientifica.com at 09/28/2021 06:06:52AM via free access Endocrine-Related C Zafon et al. DNA methylation in thyroid 26:7 R432 Cancer cancer to a variety of reasons, such as lower DNMT activity in caused serious side effects. Therefore, as with other types solid tumors (Lin et al. 2009) or that the starting level of of malignancies, efforts have also been made toward aberrant methylation in hematological malignancies is exploring the potential of these agents in combination higher than that in solid tumors (Issa et al. 1997). Another with other treatments. Significant redifferentiation limitation is that these agents need actively dividing accompanied by growth inhibition and cell apoptosis cells to take action. Therefore, slow-growing tumors was observed when cells were treated with DAC and RA. might require longer dosing schedules due to their short However, no increase in RAI uptake was observed due to half-life or improvement in drug delivery and plasma the cytoplasmic localization of the NIS protein (Vivaldi stability (Howell et al. 2010). et al. 2009). Other studies use demethylating drugs to It may seem that the encouraging effects seen increase the sensitivity of thyroid cancer cells to other in trials to treat hematological malignancies and agents such as TNF-related apoptosis-inducing ligand preclinical data on solid tumors will never reach a (TRAIL), which induces apoptosis (Siraj et al. 2011) or clinical application. However, several studies have to upregulate immune-related genes in cancer cells to shown the association between the overexpression of enhance their response to cancer immunotherapies DNMTs and chemoresistance (Wang et al. 2001, Qiu et al. (Gunda et al. 2013, 2014). A strong synergistic effect was 2002, Segura-Pacheco et al. 2006), and the treatment of also seen when DAC was combined with everolimus (an cancer cell lines with DNMT inhibitors can revert this mTOR inhibitor) to treat thyroid cancer cells, opening a resistance to therapy (Qiu et al. 2005). In light of these promising scenario to overcome drug resistance (Vitale findings, clinical trials with promising results have et al. 2017). However, the most explored and promising been conducted to test whether demethylating drugs option, not just in thyroid cancer, is the combination can enhance susceptibility to other therapies when of demethylating agents with HDAC inhibitors such as administered in combination, especially in resistant trichostatin A (TSA), sodium butyrate or valproic acid. tumors (Linnekamp et al. 2017). In vitro, this combination can restore NIS transcription Thyroid cancer is not an exception to all the details to levels approaching those present in RAI-responder explained above. Initially, in vitro studies focused on the tumors (Li et al. 2007) and even increase RAI uptake ability of demethylating drugs to restore the expression (Provenzano et al. 2007, Massimino et al. 2018). In of different genes to sensitize thyroid cancer cells to RAI addition, they can also inhibit cell growth and invasion treatment. Venkataraman et al. reported an increase in (Mitmaker et al. 2011). the mRNA and gene expression of NIS in thyroid cancer Finally, it is important to mention the role that cell lines treated with AZA compared to those observed demethylating agents have been playing through the without AZA treatment. The increased expression of the years as important tools to discover and study new NIS gene was correlated with an increase in RAI uptake prognostic and diagnostic biomarkers (Murgo 2005, in some of the cell lines (Venkataraman et al. 1999). Zuo et al. 2010, Latini et al. 2011, Moraes et al. 2016, Nevertheless, other similar studies could not show a Wu et al. 2016, Cao et al. 2018). significant increase in RAI uptake when different thyroid In conclusion, there is little evidence of the cell lines were treated with AZA or DAC, highlighting effectiveness of demethylating agents in thyroid cancer. that the mechanism may depend on the methylation and Most studies have tested these drugs in a variety of cancer differentiation status of the cell (Tuncel et al. 2007, Miasaki cell lines obtaining promising results that have not been et al. 2008). Although significant cell redifferentiation is translated into clinical practice. However, despite the not achieved, DAC and zebularine are able to inhibit cell unsuccessful results in clinical trials with their use as solo proliferation and migration in thyroid cancer cell lines agents, they may be a potentially useful therapy when (Miasaki et al. 2008, Kim et al. 2013). combined with other drugs. There have only been two clinical trials focusing on the treatment of thyroid cancer patients with demethylating agents to sensitize tumors to RAI (ClinicalTrials.gov DNA methylation in thyroid cancer cell lines Identifier: NCT00085293 and NCT00004062). Both included patients with recurrent and/or metastatic DTC Established human thyroid cancer cell lines are the most that were resistant to RAI. Unfortunately, no partial widely used models to study thyroid tumorigenesis, or complete responses were observed, while treatment including studies aimed at understanding the DNA

https://erc.bioscientifica.com © 2019 Society for Endocrinology https://doi.org/10.1530/ERC-19-0093 Published by Bioscientifica Ltd. Printed in Great Britain Downloaded from Bioscientifica.com at 09/28/2021 06:06:52AM via free access Endocrine-Related C Zafon et al. DNA methylation in thyroid 26:7 R433 Cancer cancer methylation landscape. However, it has been shown that Concluding remarks and perspectives cell lines derived from DTC, both PTC and FTC, display Numerous studies on DNA methylation in thyroid mRNA expression profiles closer to dedifferentiated cancer have improved our understanding of thyroid in vivo thyroid tumors (i.e., ATC) than to differentiated carcinogenesis. However, we still do not have a complete ones (van Staveren et al. 2007, Saiselet et al. 2012). picture of the methylation landscape, especially for This can be explained by the prior selection of initiating histological subtypes other than PTC. The huge catalog cells and the in vitro evolution of the cell lines. of DNA methylation alterations, the association of Interestingly, some of the genes commonly upregulated DNA hypomethylation with cancer progression and in ATC and thyroid cancer cell lines are related to DNA dedifferentiation, the existence of different methylomes replication, which is in accordance with their high related to different clinical and molecular phenotypes and proliferation rate. the influence of immune-infiltrating cells in tumor DNA Considering that DNA methylation is involved in methylation patterns are some of the recent findings that the regulation of gene expression, how does cellular will most likely define the direction of future research in the immortalization affect DNA methylation in thyroid field of DNA methylation in thyroid cancer. In addition, cancer cell lines? Although there are few studies profiling numerous studies confirm the importance of DNA DNA methylation in thyroid cancer cell lines, they methylation as a source of novel biomarkers in thyroid note that DNA methylation follows a similar pattern cancer. Indeed, some studies propose potential diagnostic as gene expression. Rodero-Rodríguez et al. analyzed and prognostic markers, although the combination four cell lines (one derived from PTC, one from FTC, of DNA methylation alterations with other epigenetic one from ATC and one from MTC), and all of them and/or genetic alterations may improve their clinical exhibited methylomes that more closely resembled value. Finally, DNA methylation is also a fundamental undifferentiated tumors than differentiated ones area of interest from a therapeutic perspective. Therefore, (Rodríguez-Rodero et al. 2013). Typically, immortalized further in vitro and in vivo functional experiments to better cell lines exhibit hypermethylation (Smiraglia et al. 2001). understand the implications and underlying mechanisms However, this is not the case for thyroid cancer cell lines, of DNA methylation alterations in thyroid cancer as which, based on the study by Rodero-Rodríguez et al., well as the evaluation of candidate biomarkers through show more hypomethylation than hypermethylation case–control studies and prospective trials are warranted. events (Rodríguez-Rodero et al. 2013). This result is in agreement with Klein Hesselink et al. who found that the global hypomethylation level of Alu elements in PTC- and FTC-derived cell lines was similar to that observed in Supplementary data This is linked to the online version of the paper at https://doi.org/10.1530/ ATC-derived cell lines and in vivo PDTC and ATC samples ERC-19-0093. (Klein Hesselink et al. 2018). However, some gene-specific DNA methylation studies found a good agreement between in vivo DTC tumors and Declaration of interest cell lines. Therefore, despite the limitations of the use The authors declare that there is no conflict of interest that could be of cell lines, they provide a good model for controlled perceived as prejudicing the impartiality of this review. experiments, for example, to study the DNA methylation- mediated regulation of candidate genes identified by genome-wide studies or to investigate the effect of specific Funding This work was supported by a grant from the Instituto de Salud Carlos III, treatments, such as the ability of demethylating drugs co-funded by ERDF/ESF, ‘Investing in your future’ (FIS PI18/00654 to M J). to restore the expression of different genes to sensitize thyroid cancer cells to RAI. Altogether, these studies indicate that thyroid cancer cell lines are an important tool for thyroid cancer References research, but the differences in gene expression and DNA Affinito O, Salerno P, D’Alessio AM, Cuomo M, Florio E, Carlomagno F, methylation compared to in vivo tumors should be taken Proietti A, Giannini R, Basolo F, Chiariotti L, et al. 2019 Association between DNA methylation profile and malignancy in follicular- into account when extrapolating results obtained from patterned thyroid neoplasms. Endocrine-Related Cancer 26 451–462. these cells. (https://doi.org/10.1530/ERC-18-0308)

Agathanggelou A, Cooper WN & Latif F 2005 Role of the Ras-association susceptibility event in thyroid carcinogenesis. JAMA Surgery 149 domain family 1 tumor suppressor gene in human cancers. Cancer 1146–1152. (https://doi.org/10.1001/jamasurg.2014.1694) Research 65 3497–3508. (https://doi.org/10.1158/0008-5472.CAN-04- Buj R, Mallona I, Díez-Villanueva A, Barrera V, Mauricio D, Puig- 4088) Domingo M, Reverter JL, Matias-Guiu X, Azuara D, Ramírez JL, et al. Allis CD & Jenuwein T 2016 The molecular hallmarks of epigenetic 2016 Quantification of unmethylated Alu (QUAlu): a tool to assess control. Nature Reviews: Genetics 17 487–500. (https://doi. global hypomethylation in routine clinical samples. Oncotarget 7 org/10.1038/nrg.2016.59) 10536–10546. (https://doi.org/10.18632/oncotarget.7233) Arora N, Scognamiglio T, Zhu B & Fahey III TJ 2008 Do benign thyroid Buj R, Mallona I, Díez-Villanueva A, Zafón C, Mate JL, Roca M, Puig- nodules have malignant potential? An evidence-based review. World Domingo M, Reverter JL, Mauricio D, Peinado MA, et al. 2018 Journal of Surgery 32 1237–1246. (https://doi.org/10.1007/s00268- Kallikreins stepwise scoring reveals three subtypes of papillary 008-9484-1) thyroid cancer with prognostic applications. Thyroid 28 601–612. Asa SL 2017 The evolution of differentiated thyroid cancer. Pathology 49 (https://doi.org/10.1089/thy.2017.0501) 229–237. (https://doi.org/10.1016/j.pathol.2017.01.001) Calabrò V, Strazzullo M, La Mantia G, Fedele M, Paulin C, Fusco A & Bannister AJ & Kouzarides T 2011 Regulation of chromatin by histone Lania L 1996 Status and expression of the p16INK4 gene in human modifications. Cell Research 21 381–395. (https://doi.org/10.1038/ thyroid tumors and thyroid-tumor cell lines. International Journal of cr.2011.22) Cancer 67 29–34. (https://doi.org/10.1002/(SICI)1097- Baubec T & Schubeler D 2014 Genomic patterns and context specific 0215(19960703)67:1<29::AID-IJC7>3.0.CO;2-1) interpretation of DNA methylation. Current Opinion in Genetics and Cancer Genome Atlas Research Network 2012 Comprehensive molecular Development 25 85–92. (https://doi.org/10.1016/j.gde.2013.11.015) characterization of human colon and rectal cancer. Nature 487 Belancio VP, Roy-Engel AM & Deininger PL 2010 All y’all need to know 330–337. (https://doi.org/10.1038/nature11252) ‘bout retroelements in cancer’. Seminars in Cancer Biology 20 Cancer Genome Atlas Research Network 2014 Integrated genomic 200–210. (https://doi.org/10.1016/J.SEMCANCER.2010.06.001) characterization of papillary thyroid carcinoma. Cell 159 676–690. Beltrami CM, dos Reis MB, Barros-Filho MC, Marchi FA, Kuasne H, (https://doi.org/10.1016/j.cell.2014.09.050) Pinto CAL, Ambatipudi S, Herceg Z, Kowalski LP & Rogatto SR 2017 Cao YM, Gu J, Zhang YS, Wei WJ, Qu N, Wen D, Liao T, Shi RL, Integrated data analysis reveals potential drivers and pathways Zhang L, Ji QH, et al. 2018 Aberrant hypermethylation of the disrupted by DNA methylation in papillary thyroid carcinomas. HOXD10 gene in papillary thyroid cancer with BRAFV600E Clinical Epigenetics 9 45. (https://doi.org/10.1186/s13148-017-0346-2) mutation. Oncology Reports 39 338–348. (https://doi.org/10.3892/ Benard A, van de Velde CJH, Lessard L, Putter H, Takeshima L, or.2017.6058) Kuppen PJK & Hoon DSB 2013 Epigenetic status of LINE-1 predicts Ceolin L, Goularte APP, Ferreira CV, Romitti M & Maia AL 2018 Global clinical outcome in early-stage rectal cancer. British Journal of Cancer DNA methylation profile in medullary thyroid cancer patients. 109 3073–3083. (https://doi.org/10.1038/bjc.2013.654) Experimental and Molecular Pathology 105 110–114. (https://doi. Berman BP, Weisenberger DJ, Aman JF, Hinoue T, Ramjan Z, Liu Y, org/10.1016/j.yexmp.2018.06.003) Noushmehr H, Lange CP, van Dijk CM, Tollenaar RA, et al. 2011 Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, Regions of focal DNA hypermethylation and long-range Jacobsen A, Byrne CJ, Heuer ML, Larsson E, et al. 2012 The cBio hypomethylation in colorectal cancer coincide with nuclear lamina- cancer genomics portal: an open platform for exploring associated domains. Nature Genetics 44 40–46. (https://doi. multidimensional cancer genomics data. Cancer Discovery 2 401–404. org/10.1038/ng.969) (https://doi.org/10.1158/2159-8290.CD-12-0095) Bird AP 1980 DNA methylation and the frequency of CpG in animal Chalitchagorn K, Shuangshoti S, Hourpai N, Kongruttanachok N, DNA. Nucleic Acids Research 8 1499–1504. (https://doi.org/10.1093/ Tangkijvanich P, Thong-ngam D, Voravud N, Sriuranpong V & nar/8.7.1499) Mutirangura A 2004 Distinctive pattern of LINE-1 methylation level Bird AP 1986 CpG-rich islands and the function of DNA methylation. in normal tissues and the association with carcinogenesis. Oncogene Nature 321 209–213. (https://doi.org/10.1038/321209a0) 23 8841–8846. (https://doi.org/10.1038/sj.onc.1208137) Bird A 2007 Perceptions of epigenetics. Nature 447 396–398. (https:// Chen YC, Gotea V, Margolin G & Elnitski L 2017 Significant associations doi.org/10.1038/nature05913) between driver gene mutations and DNA methylation alterations Bisarro dos Reis M, Barros-Filho MC, Marchi FA, Beltrami CM, Kuasne H, across many cancer types. PLoS Computational Biology 13 e1005840. Pinto CAL, Ambatipudi S, Herceg Z, Kowalski LP & Rogatto SR 2017 (https://doi.org/10.1371/journal.pcbi.1005840) Prognostic classifier based on genome-wide DNA methylation Choi YW, Kim H-J, Kim YH, Park SH, Chwae YJ, Lee J, Soh EY, Kim J-H profiling in well-differentiated thyroid tumors. Journal of Clinical & Park TJ 2014 B-RafV600E inhibits sodium iodide symporter Endocrinology and Metabolism 102 4089–4099. (https://doi. expression via regulation of DNA methyltransferase. Experimental and org/10.1210/jc.2017-00881) Molecular Medicine 1 46.e120. (https://doi.org/10.1038/emm.2014.68) Bochtler M, Kolano A & Xu GL 2017 DNA demethylation pathways: Czarnecka K, Pastuszak-Lewandoska D, Migdalska-Sek M, Nawrot E, additional players and regulators. BioEssays 39 1–13. (https://doi. Brzezinski J, Dedecjus M, Pomorski L & Brzezianska E 2011 Aberrant org/10.1002/bies.201600178) methylation as a main mechanism of TSGs silencing in PTC. Boltze C, Zack S, Quednow C, Bettge S, Roessner A & Schneider-Stock R Frontiers in Bioscience 3 137–157. (https://doi.org/10.2741/e228) 2003 Hypermethylation of the CDKN2/p16INK4A promotor in Dai L, Ma C, Zhang Z, Zeng S, Liu A, Tang S, Ren Q, Sun Y & Xu C thyroid carcinogenesis. Pathology, Research and Practice 199 399–404. 2016 DAPK promoter methylation and bladder cancer risk: a (https://doi.org/10.1078/0344-0338-00436) systematic review and meta-analysis. PLoS ONE 11 e0167228. Brait M, Loyo M, Rosenbaum E, Ostrow KL, Markova A, Papagerakis S, (https://doi.org/10.1371/journal.pone.0167228) Zahurak M, Goodman SM, Zeiger M, Sidransky D, et al. 2012 Dammann R, Li C, Yoon JH, Chin PL, Bates S & Pfeifer GP 2000 Correlation between BRAF mutation and promoter methylation of Epigenetic inactivation of a RAS association domain family protein TIMP3, RARβ2 and RASSF1A in thyroid cancer. Epigenetics 7 710–719. from the lung tumour suppressor locus 3p21.3. Nature Genetics 25 (https://doi.org/10.4161/epi.20524) 315–319. (https://doi.org/10.1038/77083) Brown TC, Juhlin CC, Healy JM, Prasad ML, Korah R & Carling T 2014 Davies TF, Yin X & Latif R 2010 The genetics of the thyroid stimulating Frequent silencing of RASSF1A via promoter methylation in hormone receptor: history and relevance. Thyroid 20 727–736. follicular thyroid hyperplasia: a potential early epigenetic (https://doi.org/10.1089/thy.2010.1638)

de Capoa A, Grappelli C, Volpino P, Bononi M, Musolino A, Ciardi A, Galrão AL, Sodré AK, Camargo RY, Friguglietti CU, Kulcsar MA, Cavallaro A & Cangemi V 2004 Nuclear methylation levels in Lima EU, Medeiros-Neto G & Rubio IGS 2013 Methylation levels of normal and cancerous thyroid cells. Anticancer Research 24 sodium-iodide symporter (NIS) promoter in benign and malignant 1495–1500. thyroid tumors with reduced NIS expression. Endocrine 43 225–229. De la Vieja A & Santisteban P 2018 Role of iodide metabolism in (https://doi.org/10.1007/s12020-012-9779-8) physiology and cancer. Endocrine-Related Cancer 25 R225–R245. Galrão AL, Camargo RY, Friguglietti CU, Moraes L, Cerutti JM, Serrano- (https://doi.org/10.1530/ERC-17-0515) Nascimento C, Suzuki MF, Medeiros-Neto G & Rubio IGS 2014 de Morais RM, Sobrinho AB, de Souza Silva CM, de Oliveira JR, da Hypermethylation of a new distal sodium/iodide symporter (NIS) Silva ICR & de Toledo Nóbrega O 2018 The role of the NIS (SLC5A5) enhancer (NDE) is associated with reduced NIS expression in thyroid gene in papillary thyroid cancer: a systematic review. International tumors. Journal of Clinical Endocrinology and Metabolism 99 Journal of Endocrinology 2018 9128754. (https://doi. E944–E952. (https://doi.org/10.1210/jc.2013-1450) org/10.1155/2018/9128754) Galusca B, Dumollard JM, Lassandre S, Niveleau A, Prades JM, Estour B Dralle H, Machens A, Basa J, Fatourechi V, Franceschi S, Hay ID, & Peoc’h M 2005 Global DNA methylation evaluation: potential Nikiforov YE, Pacini F, Pasieka JL & Sherman SI 2015 Follicular cell- complementary marker in differential diagnosis of thyroid neoplasia. derived thyroid cancer. Nature Reviews: Disease Primers 1 15077. Virchows Archiv 447 18–23. (https://doi.org/10.1007/s00428-005- (https://doi.org/10.1038/nrdp.2015.77) 1268-5) Du Q, Luu PL, Stirzaker C & Clark SJ 2015 Methyl-CpG-binding domain Gama-Sosa MA, Slagel VA, Trewyn RW, Oxenhandler R, Kuo KC, proteins: readers of the epigenome. Epigenomics 7 1051–1073. Gehrke CW & Ehrlich M 1983 The 5-methylcytosine content of (https://doi.org/10.2217/epi.15.39) DNA from human tumors. Nucleic Acids Research 11 6883–6894. Ehrlich M 2002 DNA methylation in cancer: too much, but also too (https://doi.org/10.1093/nar/11.19.6883) little. Oncogene 21 5400–5413. (https://doi.org/10.1038/sj. Gaudet F, Hodgson JG, Eden A, Jackson-Grusby L, Dausman J, Gray JW, onc.1205651) Leonhardt H & Jaenisch R 2003 Induction of tumors in mice by Ehrlich M 2009 DNA hypomethylation in cancer cells. Epigenomics 1 genomic hypomethylation. Science 300 489–492. (https://doi. 239–259. (https://doi.org/10.2217/epi.09.33) org/10.1126/science.1083558) Ehrlich M, Gama-Sosa MA, Huang LH, Midgett RM, Kuo KC, Goll MG & Bestor TH 2005 Eukaryotic cytosine methyltransferases. McCune RA & Gehrke C 1982 Amount and distribution of Annual Review of Biochemistry 74 481–514. (https://doi.org/10.1146/ 5-methylcytosine in human DNA from different types of tissues of annurev.biochem.74.010904.153721) cells. Nucleic Acids Research 10 2709–2721. (https://doi.org/10.1093/ Guan H, Ji M, Hou P, Liu Z, Wang C, Shan Z, Teng W & Xing M 2008 nar/10.8.2709) Hypermethylation of the DNA mismatch repair gene hMLH1 and Its Elisei R, Shiohara M, Koeffler HP & Fagin JA 1998 Genetic and association with lymph node metastasis and T1799A BRAF mutation epigenetic alterations of the cyclin-dependent kinase inhibitors in patients with papillary thyroid cancer. Cancer 113 247–255. p15INK4b and p16INK4a in human thyroid carcinoma cell lines and (https://doi.org/10.1002/cncr.23548) primary thyroid carcinomas. Cancer 83 2185–2193. (https://doi. Gunda V, Cogdill AP, Bernasconi MJ, Wargo JA & Parangi S 2013 org/10.1002/(SICI)1097-0142(19981115)83:10<2185::AID- Potential role of 5-aza-2′-deoxycytidine induced MAGE-A4 CNCR18>3.0.CO;2-4) expression in immunotherapy for anaplastic thyroid cancer. Surgery Ellis RJ, Wang Y, Stevenson HS, Boufraqech M, Patel D, Nilubol N, 154 1456–1462; discussion 1462. (https://doi.org/10.1016/j. Davis S, Edelman DC, Merino MJ, He M, et al. 2014 Genome-wide surg.2013.07.009) methylation patterns in papillary thyroid cancer are distinct based Gunda V, Frederick DT, Bernasconi MJ, Wargo JA & Parangi S 2014 A on histological subtype and tumor genotype. Journal of Clinical potential role for immunotherapy in thyroid cancer by enhancing Endocrinology and Metabolism 99 E329–E337. (https://doi. NY-ESO-1 cancer antigen expression. Thyroid 24 1241–1250. (https:// org/10.1210/jc.2013-2749) doi.org/10.1089/thy.2013.0680) Esteller M 2007 Epigenetic gene silencing in cancer: the DNA Hansen KD, Timp W, Bravo HC, Sabunciyan S, Langmead B, hypermethylome. Human Molecular Genetics 16 R50–R59. (https:// McDonald OG, Wen B, Wu H, Liu Y, Diep D, et al. 2011 Increased doi.org/10.1093/hmg/ddm018) methylation variation in epigenetic domains across cancer types. Esteller M, Silva JM, Dominguez G, Bonilla F, Matias-Guiu X, Lerma E, Nature Genetics 43 768–775. (https://doi.org/10.1038/ng.865) Bussaglia E, Prat J, Harkes IC, Repasky EA, et al. 2000 Promoter Haugen BR, Alexander EK, Bible KC, Doherty GM, Mandel SJ, hypermethylation and BRCA1 inactivation in sporadic breast and Nikiforov YE, Pacini F, Randolph GW, Sawka AM, Schlumberger M, ovarian tumors. Journal of the National Cancer Institute 92 564–569. et al. 2016 2015 American Thyroid Association management (https://doi.org/10.1093/jnci/92.7.564) guidelines for adult patients with thyroid nodules and Fang M, Ou J, Hutchinson L & Green MR 2014 The BRAF oncoprotein differentiated thyroid cancer: the American Thyroid Association functions through the transcriptional repressor MAFG to mediate Guidelines Task Force on thyroid nodules and differentiated the CpG island methylator phenotype. Molecular Cell 55 904–915. thyroid cancer. Thyroid 26 1–133. (https://doi.org/10.1089/ (https://doi.org/10.1016/J.MOLCEL.2014.08.010) thy.2015.0020) Farag AK & Roh EJ 2019 Death-associated protein kinase (DAPK) family Herman JG, Merlo A, Mao L, Lapidus RG, Issa JP, Davidson NE, modulators: current and future therapeutic outcomes. Medicinal Sidransky D & Baylin SB 1995 Inactivation of the CDKN2/p16/MTS1 Research Reviews 39 349–385. (https://doi.org/10.1002/med.21518) gene is frequently associated with aberrant DNA methylation in all Feinberg AP & Vogelstein B 1983 Hypomethylation distinguishes genes common human cancers. Cancer Research 55 4525–4530. of some human cancers from their normal counterparts. Nature 301 Holliday R & Pugh JE 1975 DNA modification mechanisms and gene 89–92. (https://doi.org/10.1038/301089a0) activity during development. Science 187 226–232. (https://doi. Feinberg AP, Koldobskiy MA & Göndör A 2016 Epigenetic modulators, org/10.1126/science.1111098) modifiers and mediators in cancer aetiology and progression. Nature Holoch D & Moazed D 2015 RNA-mediated epigenetic regulation of Reviews: Genetics 17 284–299. (https://doi.org/10.1038/nrg.2016.13) gene expression. Nature Reviews: Genetics 16 71–84. (https://doi. Fernandez AF, Assenov Y, Martin-Subero JI, Balint B, Siebert R, org/10.1038/nrg3863) Taniguchi H, Yamamoto H, Hidalgo M, Tan AC, Galm O, et al. 2012 Hoque MO, Rosenbaum E, Westra WH, Xing M, Ladenson P, Zeiger MA, A DNA methylation fingerprint of 1628 human samples. Genome Sidransky D & Umbricht CB 2005 Quantitative assessment of Research 22 407–419. (https://doi.org/10.1101/gr.119867.110) promoter methylation profiles in thyroid neoplasms. Journal of

Clinical Endocrinology and Metabolism 90 4011–4018. (https://doi. Kim WG, Zhu X, Kim DW, Zhang L, Kebebew E & Cheng SY 2013 org/10.1210/jc.2005-0313) Reactivation of the silenced thyroid hormone receptor beta gene Hotchkiss RD 1948 The quantitative separation of purines, pyrimidines, expression delays thyroid tumor progression. Endocrinology 154 and nucleosides by paper chromatography. Journal of Biological 25–35. (https://doi.org/10.1210/en.2012-1728) Chemistry 175 315–332. Kitahara CM, Devesa SS & Sosa JA 2017 Increases in thyroid cancer Hou P, Ji M & Xing M 2008 Association of PTEN gene methylation with incidence and mortality-reply. JAMA 318 390–391. (https://doi. genetic alterations in the phosphatidylinositol 3-kinase/AKT org/10.1001/jama.2017.7910) signaling pathway in thyroid tumors. Cancer 113 2440–2447. Klein Hesselink EN, Zafon C, Villalmanzo N, Iglesias C, van Hemel BM, (https://doi.org/10.1002/cncr.23869) Klein Hesselink MS, Montero-Conde C, Buj R, Mauricio D, Hou P, Liu D & Xing M 2011 Genome-wide alterations in gene Peinado MA, et al. 2018 Increased global DNA hypomethylation in methylation by the BRAF V600E mutation in papillary thyroid distant metastatic and dedifferentiated thyroid cancer. Journal of cancer cells. Endocrine-Related Cancer 18 687–697. (https://doi. Clinical Endocrinology and Metabolism 103 397–406. (https://doi. org/10.1530/ERC-11-0212) org/10.1210/jc.2017-01613) Howell PM, Liu Z & Khong HT 2010 Demethylating agents in the Koch A, Joosten SC, Feng Z, de Ruijter TC, Draht MX, Melotte V, treatment of cancer. Pharmaceuticals 3 2022–2044. (https://doi. Smits KM, Veeck J, Herman JG, Van Neste L, et al. 2018 Analysis of org/10.3390/ph3072022) DNA methylation in cancer: location revisited. Nature Reviews: Hu S, Liu D, Tufano RP, Carson KA, Rosenbaum E, Cohen Y, Holt EH, Clinical Oncology 15 459–466. (https://doi.org/10.1038/s41571-018- Kiseljak-vassiliades K, Rhoden KJ, Tolaney S, et al. 2006 Association 0004-4) of aberrant methylation of tumor suppressor genes with tumor Krause K, Prawitt S, Eszlinger M, Ihling C, Sinz A, Schierle K, Gimm O, aggressiveness and BRAF mutation in papillary thyroid cancer. Dralle H, Steinert F, Sheu SY, et al. 2011 Dissecting molecular events International Journal of Cancer 119 2322–2329. (https://doi. in thyroid neoplasia provides evidence for distinct evolution of org/10.1002/ijc.22110) follicular thyroid adenoma and carcinoma. American Journal of Ishida E, Nakamura M, Shimada K, Higuchi T, Takatsu K, Yane K & Pathology 179 3066–3074. (https://doi.org/10.1016/J. Konishi N 2007 DNA hypermethylation status of multiple genes in AJPATH.2011.08.033) papillary thyroid carcinomas. Pathobiology 74 344–352. (https://doi. Kunstman JW, Korah R, Healy JM, Prasad M & Carling T 2013 org/10.1159/000110028) Quantitative assessment of RASSF1A methylation as a putative Issa JP, Baylin SB & Herman JG 1997 DNA methylation changes in molecular marker in papillary thyroid carcinoma. Surgery 154 hematologic malignancies: biologic and clinical implications. 1255–1261; discussion 1261–1262. (https://doi.org/10.1016/j. Leukemia 11 S7–S11. surg.2013.06.025) Jones PA 2012 Functions of DNA methylation: islands, start sites, gene Kunstman JW, Juhlin CC, Goh G, Brown TC, Stenman A, Healy JM, bodies and beyond. Nature Reviews: Genetics 13 484–492. (https://doi. Rubinstein JC, Choi M, Kiss N, Nelson-Williams C, et al. 2015 org/10.1038/nrg3230) Characterization of the mutational landscape of anaplastic thyroid Jones PA & Baylin SB 2007 The epigenomics of cancer. Cell 128 cancer via whole-exome sequencing. Human Molecular Genetics 24 683–692. (https://doi.org/10.1016/j.cell.2007.01.029) 2318–2329. (https://doi.org/10.1093/hmg/ddu749) Jordà M & Peinado MA 2010 Methods for DNA methylation analysis Längst G & Manelyte L 2015 Chromatin remodelers: from function to and applications in colon cancer. Mutation Research 693 84–93. dysfunction. Genes 6 299–324. (https://doi.org/10.3390/ (https://doi.org/10.1016/j.mrfmmm.2010.06.010) genes6020299) Jordà M, Rodríguez J, Frigola J & Peinado MA 2009 Analysis of DNA Latini FRM, Hemerly JP, Freitas BCG, Oler G, Riggins GJ & Cerutti JM methylation by amplification of intermethylated sites (AIMS). 2011 ABI3 ectopic expression reduces in vitro and in vivo cell Methods in Molecular Biology 507 107–116. (https://doi. growth properties while inducing senescence. BMC Cancer 11 11. org/10.1007/978-1-59745-522-0_9) (https://doi.org/10.1186/1471-2407-11-11) Kaminskas E, Farrell A, Abraham S, Baird A, Hsieh LS, Lee SL, Lauss M, Ringnér M, Karlsson A, Harbst K, Busch C, Geisler J, Leighton JK, Patel H, Rahman A, Sridhara R, et al. 2005 Report from Lønning PE, Staaf J & Jönsson G 2015 DNA methylation subgroups the FDA approval summary: azacitidine for treatment of in melanoma are associated with proliferative and immunological myelodysplastic syndrome subtypes. Clinical Cancer Research 11 processes. BMC Medical Genomics 8 73. (https://doi.org/10.1186/ 3604–3608. (https://doi.org/10.1158/1078-0432.CCR-04-2135) s12920-015-0147-4) Kantarjlan H, Issa JPJ, Rosenfeld CS, Bennett JM, Albitar M, DiPersio J, Lawrence MS, Stojanov P, Polak P, Kryukov GV, Cibulskis K, Klimek V, Slack J, De Castro C, Ravandi F, et al. 2006 Decitabine Sivachenko A, Carter SL, Stewart C, Mermel CH, Roberts SA, et al. improves patient outcomes in myelodysplastic syndromes: results of 2013 Mutational heterogeneity in cancer and the search for new a phase III randomized study. Cancer 106 1794–1803. (https://doi. cancer-associated genes. Nature 499 214–218. (https://doi. org/10.1002/cncr.21792) org/10.1038/nature12213) Kartal K, Onder S, Kosemehmetoglu K, Kilickap S, Tezel YG & Lee JJ, Geli J, Larsson C, Wallin G, Karimi M, Zedenius J, Höög A & Kaynaroglu V 2015 Methylation status of TSHr in well-differentiated Foukakis T 2008 Gene-specific promoter hypermethylation without thyroid cancer by using cytologic material. BMC Cancer 15 824. global hypomethylation in follicular thyroid cancer. International (https://doi.org/10.1186/s12885-015-1861-1) Journal of Oncology 33 861–869. (https://doi.org/10.3892/ Keelawat S, Thorner PS, Shuangshoti S, Bychkov A, Kitkumthorn N, ijo_00000074) Rattanatanyong P, Boonyayothin W, Poumsuk U, Ruangvejvorachai P Li W, Venkataraman GM & Ain KB 2007 Protein synthesis inhibitors, in & Mutirangura A 2015 Detection of global hypermethylation in synergy with 5-azacytidine, restore sodium/iodide symporter gene well-differentiated thyroid neoplasms by immunohistochemical expression in human thyroid adenoma cell line, KAK-1, suggesting (5-methylcytidine) analysis. Journal of Endocrinological Investigation 38 trans-active transcriptional repressor. Journal of Clinical Endocrinology 725–732. (https://doi.org/10.1007/s40618-015-0246-2) and Metabolism 92 1080–1087. (https://doi.org/10.1210/jc.2006- Kikuchi Y, Tsuji E, Yagi K, Matsusaka K, Tsuji S, Kurebayashi J, Ogawa T, 2106) Aburatani H & Kaneda A 2013 Aberrantly methylated genes in Li L, Choi JY, Lee KM, Sung H, Park SK, Oze I, Pan KF, You WC, human papillary thyroid cancer and their association with BRAF/RAS Chen YX, Fang JY, et al. 2012 DNA methylation in peripheral blood: mutation. Frontiers in Genetics 4 271. (https://doi.org/10.3389/ a potential biomarker for cancer molecular epidemiology. Journal of fgene.2013.00271) Epidemiology 22 384–394. (https://doi.org/10.2188/jea.JE20120003)

Li W, Zhang X, Lu X, You L, Song Y, Luo Z, Zhang J, Nie J, Zheng W, activity in human thyroid cancer cell lines using demethylating Xu D, et al. 2017 5-Hydroxymethylcytosine signatures in circulating agents and histone deacetylase inhibitors. Surgery 149 504–511. cell-free DNA as diagnostic biomarkers for human cancers. Cell (https://doi.org/10.1016/j.surg.2010.10.007) Research 27 1243–1257. (https://doi.org/10.1038/cr.2017.121) Mohammadi-asl J, Larijani B, Khorgami Z, Tavangar SM, Haghpanah V, Lim H, Devesa SS, Sosa JA, Check D & Kitahara CM 2017 Trends in Kheirollahi M & Mehdipour P 2011 Qualitative and quantitative thyroid cancer incidence and mortality in the United States, 1974– promoter hypermethylation patterns of the P16, TSHR, RASSF1A and 2013. JAMA 317 1338–1348. (https://doi.org/10.1001/ RARβ2 genes in papillary thyroid carcinoma. Medical Oncology 28 jama.2017.2719) 1123–1128. (https://doi.org/10.1007/s12032-010-9587-z) Lin CH, Hsieh SY, Sheen IS, Lee WC, Chen TC, Shyu WC & Liaw YF Mohandas T, Sparkes RS & Shapiro LJ 1981 Reactivation of an inactive 2001 Genome-wide hypomethylation in hepatocellular human X chromosome: evidence for X inactivation by DNA carcinogenesis. Cancer Research 61 4238–4243. methylation. Science 211 393–396. (https://doi.org/10.1126/ Lin J, Gilbert J, Rudek MA, Zwiebel JA, Gore S, Jiemjit A, Zhao M, science.6164095) Baker SD, Ambinder RF, Herman JG, et al. 2009 A phase I dose- Mon SY, Riedlinger G, Abbott CE, Seethala R, Ohori NP, Nikiforova MN, finding study of 5-azacytidine in combination with sodium Nikiforov YE & Hodak SP 2018 Cancer risk and clinicopathological phenylbutyrate in patients with refractory solid tumors. Clinical characteristics of thyroid nodules harboring thyroid-stimulating Cancer Research 15 6241–6249. (https://doi.org/10.1158/1078-0432. hormone receptor gene mutations. Diagnostic Cytopathology 46 CCR-09-0567) 369–377. (https://doi.org/10.1002/dc.23915) Lin Q, Hou S, Guan F & Lin C 2018 HORMAD2 methylation-mediated Moraes L, Galrão ALR, Rubió I & Cerutti JM 2016 Transcriptional epigenetic regulation of gene expression in thyroid cancer. Journal of regulation of the potential tumor suppressor ABI3 gene in thyroid Cellular and Molecular Medicine 22 4640–4652. (https://doi. carcinomas: interplay between methylation and NKX2-1 availability. org/10.1111/jcmm.13680) Oncotarget 7 25960–25970. (https://doi.org/10.18632/ Linnekamp JF, Butter R, Spijker R, Medema JP & van Laarhoven HWM oncotarget.8416) 2017 Clinical and biological effects of demethylating agents on solid Moran S, Martínez-Cardús A, Sayols S, Musulén E, Balañá C, Estival- tumours – a systematic review. Cancer Treatment Reviews 54 10–23. Gonzalez A, Moutinho C, Heyn H, Diaz-Lagares A, de Moura MC, (https://doi.org/10.1016/j.ctrv.2017.01.004) et al. 2016 Epigenetic profiling to classify cancer of unknown Lister R, Pelizzola M, Dowen RH, Hawkins RD, Hon G, Tonti-Filippini J, primary: a multicentre, retrospective analysis. Lancet: Oncology 17 Nery JR, Lee L, Ye Z, Ngo QM, et al. 2009 Human DNA methylomes 1386–1395. (https://doi.org/10.1016/S1470-2045(16)30297-2) at base resolution show widespread epigenomic differences. Nature Murgo AJ 2005 Innovative approaches to the clinical development of 462 315–322. (https://doi.org/10.1038/nature08514) DNA methylation inhibitors as epigenetic remodeling drugs. Madakashira BP & Sadler KC 2017 DNA methylation, nuclear Seminars in Oncology 32 458–464. (https://doi.org/10.1053/j. organization, and cancer. Frontiers in Genetics 8 76. (https://doi. seminoncol.2005.07.004) org/10.3389/fgene.2017.00076) Nakamura N, Carney JA, Jin L, Kajita S, Pallares J, Zhang H, Qian X, Mancikova V, Buj R, Castelblanco E, Inglada-Pérez L, Diez A, De Sebo TJ, Erickson LA & Lloyd RV 2005 RASSF1A and NORE1A Cubas AA, Curras-Freixes M, Maravall FX, Mauricio D, Matias- methylation and BRAFV600E mutations in thyroid tumors. Guiu X, et al. 2014 DNA methylation profiling of well-differentiated Laboratory Investigation 85 1065–1075. (https://doi.org/10.1038/ thyroid cancer uncovers markers of recurrence free survival. labinvest.3700306) International Journal of Cancer 135 598–610. (https://doi.org/10.1002/ Neumann S, Schuchardt K, Reske A, Reske A, Emmrich P & Paschke R ijc.28703) 2004 Lack of correlation for sodium iodide symporter mRNA and Mani S & Herceg Z 2010 DNA demethylating agents and epigenetic protein expression and analysis of sodium iodide symporter therapy of cancer. Advances in Genetics 70 327–340. (https://doi. promoter methylation in benign cold thyroid nodules. Thyroid 14 org/10.1016/B978-0-12-380866-0.60012-5) 99–111. (https://doi.org/10.1089/105072504322880337) Marquez VE, Kelley JA, Agbaria R, Ben-Kasus T, Cheng JC, Yoo CB & Nguyen C, Liang G, Nguyen TT, Tsao-Wei D, Groshen S, Lübbert M, Jones PA 2005 Zebularine: a unique molecule for an epigenetically Zhou JH, Benedict WF & Jones PA 2001 Susceptibility of based strategy in cancer chemotherapy. Annals of the New York nonpromoter CpG islands to de novo methylation in normal and Academy of Sciences 1058 246–254. (https://doi.org/10.1196/ neoplastic cells. Journal of the National Cancer Institute 93 1465–1472. annals.1359.037) (https://doi.org/10.1093/jnci/93.19.1465) Massimino M, Tirrò E, Stella S, Frasca F, Vella V, Sciacca L, Pennisi MS, Niu H, Yang J, Yang K & Huang Y 2017 The relationship between Vitale SR, Puma A, Romano C, et al. 2018 Effect of combined RASSF1A promoter methylation and thyroid carcinoma: a meta- epigenetic treatments and ectopic NIS expression on undifferentiated analysis of 14 articles and a bioinformatics of 2 databases (PRISMA). thyroid cancer cells. Anticancer Research 38 6653–6662. (https://doi. Medicine 96 e8630. (https://doi.org/10.1097/MD.0000000000008630) org/10.21873/anticanres.13032) Park SY, Seo AN, Jung HY, Gwak JM, Jung N, Cho NY & Kang GH 2014 Melck A, Masoudi H, Griffith OL, Rajput A, Wilkins G, Bugis S, Alu and LINE-1 hypomethylation is associated with HER2 enriched Jones SJM & Wiseman SM 2007 Cell cycle regulators show diagnostic subtype of breast cancer. PLoS ONE 9 e100429. (https://doi. and prognostic utility for differentiated thyroid cancer. Annals of org/10.1371/journal.pone.0100429) Surgical Oncology 14 3403–3411. (https://doi.org/10.1245/s10434-007- Portela A & Esteller M 2010 Epigenetic modifications and human 9572-8) disease. Nature Biotechnology 28 1057–1068. (https://doi.org/10.1038/ Melki JR, Vincent PC & Clark SJ 1999 Concurrent DNA nbt.1685) hypermethylation of multiple genes in acute myeloid leukemia. Provenzano MJ, Fitzgerald MP, Krager K & Domann FE 2007 Increased Cancer Research 59 3730–3740. iodine uptake in thyroid carcinoma after treatment with sodium Miasaki FY, Vivaldi A, Ciampi R, Agatel L, Collecchi P, Capodanno A, butyrate and decitabine (5-aza-dC). Otolaryngology: Head and Neck Pinchera A & Elisei R 2008 Retinoic acid receptor β2 re-expression Surgery 137 722–728. (https://doi.org/10.1016/j. and growth inhibition in thyroid carcinoma cell lines after 5-aza-2′- otohns.2007.07.030) deoxycytidine treatment. Journal of Endocrinological Investigation 31 Qi M & Xiong X 2018 Promoter hypermethylation of RARβ2, DAPK, 724–730. (https://doi.org/10.1007/BF03346422) hMLH1, p14, and p15 is associated with progression of breast Mitmaker EJ, Griff NJ, Grogan RH, Sarkar R, Kebebew E, Duh QY, cancer: a PRISMA-compliant meta-analysis. Medicine 97 e13666. Clark OH & Shen WT 2011 Modulation of matrix metalloproteinase (https://doi.org/10.1097/MD.0000000000013666)

Qiu YY, Mirkin BL & Dwivedi RS 2002 Differential expression of DNA- Epigenetics and Chromatin 10 16. (https://doi.org/10.1186/s13072- methyltransferases in drug resistant murine neuroblastoma cells. 017-0123-7) Cancer Detection and Prevention 26 444–453. (https://doi.org/10.1016/ Smiraglia DJ, Rush LJ, Frühwald MC, Dai Z, Held WA, Costello JF, S0361-090X(02)00116-2) Lang JC, Eng C, Li B, Wright FA, et al. 2001 Excessive CpG island Qiu YY, Mirkin BL & Dwivedi RS 2005 Inhibition of DNA hypermethylation in cancer cell lines versus primary human methyltransferase reverses cisplatin induced drug resistance in malignancies. Human Molecular Genetics 10 1413–1419. (https://doi. murine neuroblastoma cells. Cancer Detection and Prevention 29 org/10.1093/hmg/10.13.1413) 456–463. (https://doi.org/10.1016/j.cdp.2005.05.004) Smith JA, Fan CY, Zou C, Bodenner D & Kokoska MS 2007 Methylation Reik W, Collick A, Norris ML, Barton SC & Surani MA 1987 Genomic status of genes in papillary thyroid carcinoma. Archives of imprinting determines methylation of parental alleles in transgenic Otolaryngology: Head and Neck Surgery 133 1006–1011. (https://doi. mice. Nature 328 248–251. (https://doi.org/10.1038/328248a0) org/10.1001/archotol.133.10.1006) Riesco-Eizaguirre G & Santisteban P 2016 ENDOCRINE TUMOURS: Sorm F & Vesely J 1968 Effect of 5-aza-2′-deoxycytidine against leukemic Advances in the molecular pathogenesis of thyroid cancer: lessons and hemopoietic tissues in AKR mice. Neoplasma 15 339–343. from the cancer genome. European Journal of Endocrinology 175 Sorm F, Piskala A, Cihak A & Vesely J 1964 5-Azacytidine, a new, highly R203–R217. (https://doi.org/10.1530/EJE-16-0202) effective cancerostatic. Experientia 20 202–203. (https://doi. Riggs AD 1975 X inactivation, differentiation, and DNA methylation. org/10.1007/BF02135399) Cytogenetic and Genome Research 14 9–25. (https://doi. Stephen JK, Chitale D, Narra V, Chen KM, Sawhney R & Worsham MJ org/10.1159/000130315) 2011 DNA methylation in thyroid tumorigenesis. Cancers 3 Rodríguez-Rodero S, Fernandez AF, Fernandez-Morera JL, Castro-Santos P, 1732–1743. (https://doi.org/10.3390/cancers3021732) Bayon GF, Ferrero C, Urdinguio RG, Gonzalez-Marquez R, Suarez C, Stephen JK, Chen KM, Merritt J, Chitale D, Divine G & Worsham MJ Fernandez-Vega I, et al. 2013 DNA methylation signatures identify 2015 Methylation markers for early detection and differentiation of biologically distinct thyroid cancer subtypes. Journal of Clinical follicular thyroid cancer subtypes. Cancer and Clinical Oncology 4 Endocrinology and Metabolism 98 2811–2821. (https://doi. 1–12. (https://doi.org/10.5539/cco.v4n2p1) org/10.1210/jc.2012-3566) Stephen JK, Chen KM, Merritt J, Chitale D, Divine G & Worsham MJ Russo D, Manole D, Arturi F, Suarez HG, Schlumberger M, Filetti S & 2018 Methylation markers differentiate thyroid cancer from benign Derwahl M 2001 Absence of sodium/iodide symporter gene nodules. Journal of Endocrinological Investigation 41 163–170. (https:// mutations in differentiated human thyroid carcinomas. Thyroid 11 doi.org/10.1007/s40618-017-0702-2) 37–39. (https://doi.org/10.1089/10507250150500649) Suelves M, Carrió E, Núñez-Álvarez Y & Peinado MA 2016 DNA Ryder M, Ghossein RA, Ricarte-Filho JCM, Knauf JA & Fagin JA 2008 methylation dynamics in cellular commitment and differentiation. Increased density of tumor-associated macrophages is associated Briefings in Functional Genomics 15 443–453. (https://doi.org/10.1093/ with decreased survival in advanced thyroid cancer. Endocrine-Related bfgp/elw017) Cancer 15 1069–1074. (https://doi.org/10.1677/ERC-08-0036) Swain JL, Stewart TA & Leder P 1987 Parental legacy determines Saghafinia S, Mina M, Riggi N, Hanahan D & Ciriello G 2018 Pan-cancer methylation and expression of an autosomal transgene: a molecular landscape of aberrant DNA methylation across human tumors. Cell mechanism for parental imprinting. Cell 50 719–727. (https://doi. Reports 25 1066.e8–1080.e8. (https://doi.org/10.1016/j. org/10.1016/0092-8674(87)90330-8) celrep.2018.09.082) Timp W, Bravo HC, McDonald OG, Goggins M, Umbricht C, Zeiger M, Saiselet M, Floor S, Tarabichi M, Dom G, Hebrant A, van Staveren WC & Feinberg AP & Irizarry RA 2014 Large hypomethylated blocks as a Maenhaut C 2012 Thyroid cancer cell lines: an overview. Frontiers in universal defining epigenetic alteration in human solid tumors. Endocrinology 3 133. (https://doi.org/10.3389/fendo.2012.00133) Genome Medicine 6 61. (https://doi.org/10.1186/s13073-014-0061-y) Schagdarsurengin U, Gimm O, Hoang-Vu C, Dralle H, Pfeifer GP & Tomasetti C, Li L & Vogelstein B 2017 Stem cell divisions, somatic Dammann R 2002 Frequent epigenetic silencing of the CpG island mutations, cancer etiology, and cancer prevention. Science 355 promoter of RASSF1A in thyroid carcinoma. Cancer Research 62 1330–1334. (https://doi.org/10.1126/science.aaf9011) 3698–3701. Toraño EG, Petrus S, Fernandez AF & Fraga MF 2012 Global DNA Schagdarsurengin U, Gimm O, Dralle H, Hoang-Vu C & Dammann R hypomethylation in cancer: review of validated methods and clinical 2006 CpG island methylation of tumor-related promoters occurs significance. Clinical Chemistry and Laboratory Medicine 50 preferentially in undifferentiated carcinoma. Thyroid 16 633–642. 1733–1742. (https://doi.org/10.1515/cclm-2011-0902) (https://doi.org/10.1089/thy.2006.16.633) Toyota M, Ahuja N, Ohe-Toyota M, Herman JG, Baylin SB & Issa JP Segura-Pacheco B, Perez-Cardenas E, Taja-Chayeb L, Chavez-Blanco A, 1999 CpG island methylator phenotype in colorectal cancer. PNAS Revilla-Vazquez A, Benitez-Bribiesca L & Duenas-González A 2006 96 8681–8686. (https://doi.org/10.1073/pnas.96.15.8681) Global DNA hypermethylation-associated cancer chemotherapy Tuncel M, Aydin D, Yaman E, Tazebay UH, Güç D, Doğan AL, resistance and its reversion with the demethylating agent Taşbasan B & Uğur O 2007 The comparative effects of gene hydralazine. Journal of Translational Medicine 4 32. (https://doi. modulators on thyroid-specific genes and radioiodine uptake. Cancer org/10.1186/1479-5876-4-32) Biotherapy and Radiopharmaceuticals 22 281–288. (https://doi. Sharma S, Kelly TK & Jones PA 2010 Epigenetics in cancer. Carcinogenesis org/10.1089/cbr.2006.319) 31 27–36. (https://doi.org/10.1093/carcin/bgp220) van Staveren WC, Solis DW, Delys L, Duprez L, Andry G, Franc B, Shi D-Q, Ali I, Tang J & Yang W-C 2017 New insights into 5hmC DNA Thomas G, Libert F, Dumont JE, Detours V, et al. 2007 Human modification: generation, distribution and function. Frontiers in thyroid tumor cell lines derived from different tumor types present a Genetics 8 100. (https://doi.org/10.3389/fgene.2017.00100) common dedifferentiated phenotype. Cancer Research 67 8113–8120. Siraj AK, Hussain AR, Al-Rasheed M, Ahmed M, Bavi P, Alsobhi SADS, (https://doi.org/10.1158/0008-5472.CAN-06-4026) Al-Nuaim A, Uddin S & Al-Kuraya K 2011 Demethylation of TMS1 Venkataraman GM, Yatin M, Marcinek R & Ain KB 1999 Restoration of gene sensitizes thyroid cancer cells to TRAIL-induced apoptosis. iodide uptake in dedifferentiated thyroid carcinoma: relationship to Journal of Clinical Endocrinology and Metabolism 96 E215–E224. human Na+/I- symporter gene methylation status. Journal of Clinical (https://doi.org/10.1210/jc.2010-0790) Endocrinology and Metabolism 84 2449–2457. (https://doi. Skvortsova K, Zotenko E, Luu PL, Gould CM, Nair SS, Clark SJ & org/10.1210/jcem.84.7.5815) Stirzaker C 2017 Comprehensive evaluation of genome-wide Villar-Garea A, Fraga MF, Espada J & Esteller M 2003 Procaine is a DNA- 5-hydroxymethylcytosine profiling approaches in human DNA. demethylating agent with growth-inhibitory effects in human cancer

cells. Cancer Research 63 4984–4989. (https://doi.org/10.3389/ Xing M, Cohen Y, Mambo E, Tallini G, Udelsman R, Ladenson PW & fpsyg.2014.00533) Sidransky D 2004 Early occurrence of RASSF1A hypermethylation Vitale G, Dicitore A, Pepe D, Gentilini D, Grassi ES, Borghi MO, and its mutual exclusion with BRAF mutation in thyroid Gelmini G, Cantone MC, Gaudenzi G, Misso G, et al. 2017 tumorigenesis. Cancer Research 64 1664–1668. (https://doi. Synergistic activity of everolimus and 5-aza-2′-deoxycytidine in org/10.1158/0008-5472.CAN-03-3242) medullary thyroid carcinoma cell lines. Molecular Oncology 11 Yane K, Konishi N, Kitahori Y, Naito H, Okaichi K, Ohnishi T, 1007–1022. (https://doi.org/10.1002/1878-0261.12070) Miyahara H, Matsunaga T & Hiasa Y 1996 Lack of p16/CDKN2 Vivaldi A, Miasaki FY, Ciampi R, Agate L, Collecchi P, Capodanno A, alterations in thyroid carcinomas. Cancer Letters 101 85–92. (https:// Pinchera A & Elisei R 2009 Re-differentiation of thyroid carcinoma doi.org/10.1016/0304-3835(96)04117-1) cell lines treated with 5-aza-2′-deoxycytidine and retinoic acid. Yang X, Gao L & Zhang S 2016a Comparative pan-cancer DNA Molecular and Cellular Endocrinology 307 142–148. (https://doi. methylation analysis reveals cancer common and specific patterns. org/10.1016/j.mce.2009.03.020) Briefings in Bioinformatics 18 bbw063. (https://doi.org/10.1093/bib/ Wang C, Mirkin BL & Dwivedi RS 2001 DNA (cytosine) bbw063) methyltransferase overexpression is associated with acquired drug Yang Z, Wong A, Kuh D, Paul DS, Rakyan VK, Leslie RD, Zheng SC, resistance of murine neuroblastoma cells. International Journal of Widschwendter M, Beck S & Teschendorff AE 2016b Correlation of Oncology 18 323–329. (https://doi.org/10.3892/ijo.18.2.323) an epigenetic mitotic clock with cancer risk. Genome Biology 17 205. Wang P, Pei R, Lu Z, Rao X & Liu B 2013 Methylation of p16 CpG (https://doi.org/10.1186/s13059-016-1064-3) islands correlated with metastasis and aggressiveness in papillary Yang XY, Zhang J, Yu XL, Zheng GF, Zhao F & Jia XJ 2018 Death- thyroid carcinoma. Journal of the Chinese Medical Association 76 associated protein kinase promoter methylation correlates with 135–139. (https://doi.org/10.1016/j.jcma.2012.11.007) clinicopathological and prognostic features in nonsmall cell lung Weisenberger DJ, Siegmund KD, Campan M, Young J, Long TI, cancer patients: a cohort study. Journal of Cancer Research and Faasse MA, Kang GH, Widschwendter M, Weener D, Buchanan D, Therapeutics 14 S65–S71. (https://doi.org/10.4103/0973- et al. 2006 CpG island methylator phenotype underlies sporadic 1482.158197) microsatellite instability and is tightly associated with BRAF Yegnasubramanian S, Haffner MC, Zhang Y, Gurel B, Cornish TC, Wu Z, mutation in colorectal cancer. Nature Genetics 38 787–793. (https:// Irizarry RA, Morgan J, Hicks J, DeWeese TL, et al. 2008 DNA doi.org/10.1038/ng1834) hypomethylation arises later in prostate cancer progression than White MG, Nagar S, Aschebrook-Kilfoy B, Jasmine F, Kibriya MG, CpG island hypermethylation and contributes to metastatic tumor Ahsan H, Angelos P, Kaplan EL & Grogan RH 2016 Epigenetic heterogeneity. Cancer Research 68 8954–8967. (https://doi. alterations and canonical pathway disruption in papillary thyroid org/10.1158/0008-5472.CAN-07-6088) cancer: a genome-wide methylation analysis. Annals of Surgical Yim JH, Choi AH, Li AX, Qin H, Chang S, Tong S-WT, Chu P, Kim BW, Oncology 23 2302–2309. (https://doi.org/10.1245/s10434-016-5185-4) Schmolze D, Lew R, et al. 2019 Identification of tissue-specific DNA Wilson AS, Power BE & Molloy PL 2007 DNA hypomethylation and methylation signatures for thyroid nodule diagnostics. Clinical human diseases. Biochimica et Biophysica Acta 1775 138–162. Cancer Research 25 544–551. (https://doi.org/10.1158/1078-0432. (https://doi.org/10.1016/j.bbcan.2006.08.007) CCR-18-0841) Wu Ct. & Morris JR 2001 Genes, genetics, and epigenetics: a Yoo SK, Lee S, Kim SJ, Jee HG, Kim BA, Cho H, Song YS, Cho SW, correspondence. Science 293 1103–1105. (https://doi.org/10.1126/ Won JK, Shin JY, et al. 2016 Comprehensive analysis of the science.293.5532.1103) transcriptional and mutational landscape of follicular and papillary Wu SC & Zhang Y 2010 Active DNA demethylation: many roads lead to thyroid cancers. PLoS Genetics 12 e1006239. (https://doi.org/10.1371/ Rome. Nature Reviews: Molecular Cell Biology 11 607–620. (https://doi. journal.pgen.1006239) org/10.1038/nrm2950) You JS & Jones PA 2012 Cancer genetics and epigenetics: two sides of Wu X & Zhang Y 2017 TET-mediated active DNA demethylation: the same coin? Cancer Cell 22 9–20. (https://doi.org/10.1016/j. mechanism, function and beyond. Nature Reviews: Genetics 18 ccr.2012.06.008) 517–534. (https://doi.org/10.1038/nrg.2017.33) Zafon C, Obiols G, Castellví J, Ramon y Cajal S, Baena JA & Mesa J 2008 Wu W, Yang SF, Liu FF & Zhang JH 2015 Association between p16 Expression of p21Cip1, p27Kip1, and p16INk4a cyclin-dependent promoter methylation and thyroid cancer risk: a meta-analysis. kinase inhibitors in papillary thyroid carcinoma: correlation with Asian Pacific Journal of Cancer Prevention 16 7111–7115. (https://doi. clinicopathological factors. Endocrine Pathology 19 184–189. (https:// org/10.7314/APJCP.2015.16.16.7111) doi.org/10.1007/s12022-008-9037-z) Wu W, Zhang L, Lin J, Huang H, Shi B, Lin X, Huang Z, Wang C, Qiu J Zhang S, Wang Y, Chen M, Sun L, Han J, Elena VK & Qiao H 2017 & Wei X 2016 Hypermethylation of the HIC1 promoter and aberrant CXCL12 methylation-mediated epigenetic regulation of gene expression of HIC1/SIRT1 contribute to the development of thyroid expression in papillary thyroid carcinoma. Scientific Reports 7 44033. papillary carcinoma. Oncotarget 7 84416–84427. (https://doi. (https://doi.org/10.1038/srep44033) org/10.18632/oncotarget.12936) Zuo H, Gandhi M, Edreira MM, Hochbaum D, Nimgaonkar VL, Zhang P, Xing M, Usadel H, Cohen Y, Tokumaru Y, Guo Z, Westra WB, Tong BC, DiPaola J, Evdokimova V, Altschuler DL & Nikiforov YE 2010 Tallini G, Udelsman R, Califano JA, et al. 2003 Methylation of the Downregulation of Rap1GAP through epigenetic silencing and loss thyroid-stimulating hormone receptor gene in epithelial thyroid of heterozygosity promotes invasion and progression of thyroid tumors: a marker of malignancy and a cause of gene silencing. tumors. Cancer Research 70 1389–1397. (https://doi. Cancer Research 63 2316–2321. org/10.1158/0008-5472.CAN-09-2812)

Received in final form 23 April 2019 Accepted 29 April 2019 Accepted Preprint published online 29 April 2019