It is widely accepted that thyroid cancer is strongly determined by the individual genetic background. In this regard, it is expected that sporadic thyroid cancer is the result of multiple low- to moderate-penetrance genes interacting with each other and with the environment, thus modulating individual susceptibility. In the last years, an important number of association studies on thyroid cancer have been published, trying to determine this genetic contribution. The aim of this review is to provide a comprehensive and critical evaluation of the associations reported so far in thyroid cancer susceptibility in case–control studies performed in both non-medullary (papillary and follicular) and medullary thyroid cancers, including their potential strengths and pitfalls. We summarize the genetic variants reported to date, and stress the importance of validating the results in independent series and assessing the functional role of the associated loci.
Thyroid cancer accounts for more than 1% of all malignancies, with an estimated annual incidence of 212 000 cases worldwide (Stewart et al. 2003, Ferlay et al. 2010). This number has been rapidly increasing in recent years (Liu et al. 2001), probably due to the improvement in diagnostic techniques (Davies & Welch 2006) and to the effect of environmental factors (Chen et al. 2009).
‘Thyroid cancer’ is a general term that includes two main groups of neoplasias, depending on the cell type affected by the malignant transformation. The most common thyroid cancers arise from follicular cells or thyrocytes, the thyroid hormone-producing cell lineage in the gland (DeLellis 2004). Among them, papillary thyroid carcinoma (PTC) and follicular thyroid carcinoma (FTC) represent the two most common subtypes (85 and 10% respectively; Kondo et al. 2006), generally displaying a good prognosis with relatively few deaths (Ferlay et al. 2010), in contrast with their poorly differentiated counterparts that, although less frequent (5%), show a poor outcome. PTC displays a sex ratio female:male of around 3:1, suggesting the influence of unknown hormonal factors. The only well-established environmental factor related to PTC is exposure to ionizing radiation (Williams 2009). The PTC incidence is remarkably high in developed countries, and when it spreads it usually metastasizes to local lymph nodes (DeLellis 2004). FTC lacks the morphological nuclear features of PTC, tends to be more aggressive, produces distant metastasis rather than lymph node invasion, and displays a more equilibrated sex ratio (around 2:1, female:male). It has been associated with a deficiency in iodine intake, thus being more frequent in developing countries (Woodruff et al. 2010).
The second group, referred to as medullary thyroid carcinoma (MTC), represents a minority of thyroid tumors (5%) and affects a different cell lineage, the parafollicular or C-cells, which are involved in calcitonin secretion (Randolph & Maniar 2000). This entity is histopathologically much more homogeneous than PTC and FTC, not only since the latter are subdivided into several morphological subtypes, but also with regard to the underlying genetic mechanisms driving the disease. MTC has a poor prognosis: up to 50% of patients develop local metastases to cervical and mediastinal nodal groups, and around 20% display distant metastases to the lung, liver, or bone at diagnosis (DeLellis 2004). Around 25% of MTC tumors are hereditary, with an autosomal dominant model of inheritance. Familial MTC (FMTC) appears as part of a rare inherited syndrome, multiple endocrine neoplasia 2 (MEN2; MIM#171400), which is subdivided into three clinically distinct forms called MEN type 2A, MEN2B, and FMTC. These familial forms have germline gain-of-function mutations in the RET proto-oncogene in common. Since there is a variable clinical expression among RET germline mutation carriers, the existence of additional genetic risk factors acting as modulators of phenotype has been suggested (Cranston & Ponder 2003). Actually, it is accepted that some risk factors could act both as phenotype modifiers in FMTC and as genetic determinants of the sporadic forms of MTC.
Aim and statistical approach
The aim of this review is to provide a comprehensive overview of the association studies published on sporadic thyroid cancers. An exhaustive search of case–control studies in follicular cell-derived tumors (PTC and FTC) and MTC was performed in PubMed, including papers published in English up to 1 October 2010. The present review is focused on reported positive associations, preferably replicated, of individual SNPs and disease, rather than on combinations of SNPs or haplotype analyses. For an exhaustive review of both positive and negative results published until May 2008, we strongly recommend reading the review by Adjadj et al. (2009).
Since an insufficient sample size is one of the main pitfalls of association studies (Houlston & Peto 2004, McCarthy et al. 2008, Kere 2010), we estimated the statistical power of each study reviewed in this paper, using ‘PS Power and Sample Size Calculations’, version 2.1.31 (Dupont & Plummer 1998). Given that the attributable risk of common susceptibility alleles is expected to be around 1.1- to 1.6-fold (Fletcher & Houlston 2010), we calculated the power for detecting an odds ratio (OR) of 1.5, at a significance level of α=0.05 in a χ2 test. The variables influencing the power of each study considered in this analysis were the number of cases included (sample size, n) and the frequency of the variant in the general population (minor allele frequency (MAF)). We subsequently divided the published articles into two categories: those with enough statistical power to detect an OR of 1.5 and those that lacked this power.
Follicular cell-derived thyroid cancer
Different genetic alterations involving the axis RET/PTC–RAS–BRAF of the MAP kinase signaling pathway have been described as causal somatic changes in PTC and FTC (Kondo et al. 2006). Although involving the same pathway, specific and mutually exclusive changes have been described in each subtype, involving rearrangements of RET and NTRK1 with several partners (Greco et al. 1992, Santoro et al. 1992, Nikiforov 2002) and point mutations in BRAF (Davies et al. 2002) in PTC, and PPARG rearrangements (Kroll et al. 2000), and mutations in RAS (Manenti et al. 1994, Nikiforova et al. 2003) associated with FTC.
In addition, it is well established that follicular cell-derived thyroid cancer has a strong genetic component, since it displays the highest risk in first-degree relatives of probands among neoplasias not displaying regular Mendelian inheritance (Goldgar et al. 1994, Pal et al. 2001, Hemminki et al. 2004). Familial risk ratio calculations for these individuals range from 8.6- to 10.3-fold increased susceptibility to develop a thyroid tumor, strongly suggesting an inherited genetic background predisposing to the development of the disease. This observation, along with the fact that around 5% of these so-called non-MTC (NMTC) show familial aggregation, first led to the idea that familial NMTC could be explained by mutations in high-penetrance genes. Several linkage studies, using non-syndromic families displaying PTC in several members, have explored this possibility, reporting putative susceptibility loci at 14q31 (Bignell et al. 1997), 19p13.2 (Canzian et al. 1998, Prazeres et al. 2008), 1q21 (Malchoff et al. 2000), 2q21 (McKay et al. 2001), 8p23.1-p22 (Cavaco et al. 2008), 8q24 (He et al. 2009), and 6q22 (Suh et al. 2009). These findings were not uniformly replicated, and it is interesting to note that no high-penetrance gene has been identified so far in these putative loci suggested by linkage studies, probably due to the heterogeneity of the disease. In fact, it seems that the commonly used terms ‘familial PTC’ or ‘familial NMTC’ include different entities. This is reinforced by the specific phenotypic features displayed in each family, including several histological subtypes of PTC (follicular variant of PTC (FVPTC), oxyphilic tumors) and different additional manifestations (e.g. multinodular goiters, melanoma). Overall, whether familial NMTC is a distinct entity (Capezzone et al. 2008) that could be explained by mutations in high-penetrance genes or alternatively is the result of the combined effect of low- to moderate-penetrance loci remains unclear. The genetic contribution to familial NMTC, its controversies and clinical implications have been extensively reviewed in the past couple of years (Nose 2008, Vriens et al. 2009, Bonora et al. 2010, Khan et al. 2010) and will not be further commented on here.
Searching for low-penetrance genes by association studies in follicular cell-derived thyroid cancer
The great majority of NMTC, around 95% of cases, behaves as a sporadic entity. In this regard, NMTC is widely considered as a complex disease, in which common genetic variants in low-penetrance genes (LPGs) may interact with each other and with the environment, modulating the individual susceptibility to this cancer (Adjadj et al. 2009, Sturgis & Li 2009). In this scenario, linkage analysis does not have the power to identify these LPGs. Thus, association studies, either genome-wide or carefully designed candidate gene approaches, may be more appropriate strategies to define genetic risk factors (Houlston & Peto 2004, Milne & Benitez 2008).
Moderate sample sizes lead to few replicated associations
The identification of LPGs related to follicular cell-derived carcinoma is an arduous task due to the heterogeneity and relative rarity of the disease. Indeed, many of the studies reviewed here were rather small in sample size, and only a few polymorphisms were replicated in independent cohorts of cases and controls. A summary of the positive associations published for individual polymorphisms and PTC/FTC susceptibility is shown in Tables 1 and 2 and Fig. 1A, and will be expanded throughout this review.
Variants associated with the development of NMTC in well powered and/or validated studies
|Chr location||Gene||Studied variant (synonyms)a||Cases||Controls||MAF cases; MAF controls (%)||OR (95% CI)b||P valueb||Best fitting model||Power to detect an OR=1.5||Power to detect an OR=2.0||Genotyping platform||Functional assessment||Matching or adjustment criteria||Ref.c||Replicated?|
|9q22.33||Near FOXE1 (57 kb)||rs965513 (g.29720641A>G)||962 (PTC+FTC)||38 923||48.0; 34.1||1.75 (1.59–1.94)||1.7×10−27||M||100.0||100.0||Illumina, centaurus, SNaPshot||Associated with lower serum TSH and T4 and higher T3||N/A||1||2, 3|
|9q22.33||Near FOXE1 (57 kb)||rs965513 (g.29720641A>G)||660 Rx PTC||1268||47.4; 35.7||1.65 (1.43–1.91)||4.8×10−12||M||98.5||100.0||Illumina +TaqMan||N/A||Age, gender, origin, radiation exposure||3||1|
|9q22||FOXE1||rs1867277 (c.−283 G>A)||984 PTC||1028||51.5; 42.8||1.49 (1.30–1.70)||5.9×10−9||M||99.5||100.0||Illumina +KASPar||Determines binding of USF1/2 factors||Age, gender, origin||2||1, 3|
|9q22||FOXE1||rs1867277 (c.−283 G>A)||660 Rx PTC||820||N/A||1.48 (1.27–1.71)||4.5×10−7||M||N/A||N/A||Illumina +TaqMan||N/A||Age, gender, origin, radiation exposure||3||2|
|14q13.3||Near NKX-1 (249 kb)||rs944289 (g.17649246C>T)||1006 (PTC+FTC)||38 345||35.7; 42.7||1.37 (1.24–1.52)||2.0×10−9||M||100.0||100.0||Illumina, centaurus, SNaPshot||Associated with lower serum TSH||N/A||1||No|
|8q24.21||Gene desert||rs6983267||485 PTC||1910||51.4; 47.6||1.37 (1.02–1.82)||0.04||R||97.8||100.0||RFLP (NmuCI)||N/A||N/A||4||Multisite cancer associated region|
|5q34||Pre-miR- 146a||rs2910164 (c.60G>C)||608 PTC||901||26.2; 23.9||1.62 (1.30–2.00)||7.0×10−6||O||93.3||100.0||Sequencing and/or SNaPshot||Influences miR-146a and its target genes||N/A||5||No|
|19q13.2||XRCC1||rs25487 (p.Arg399Gln)||255 PTC||596||32.7; 36.2||0.70 (0.59–0.93)||0.03||D||76.1||99.6||RFLP (MspI)||N/A||N/A||6||7|
|19q13.2||XRCC1||rs25487 (p.Arg399Gln)||251 (PTC+FTC)||503||27.3; 34.8||0.70 (0.50–1.00)||0.049||D||72.9||99.3||RFLP (NciI)||N/A||Age||7||6|
|19q13.2||XRCC1||rs1799782 (p.Arg194Trp)||283 (PTC+FTC)||469||34.1; 28.9||1.85 (1.11–3.07)||0.044||R||72.4||99.3||TaqMan||N/A||Origin||8||No|
|8q24||TG||rs1133076 (p.Arg2530Gln)||304 NMTC||396||44.7; 51.8||1.60 (1.10–2.30)||0.015||M||74.7||99.3||RFLP (TaqI)||N/A||Age, gender, ancestry||9||No|
|11p11.2||PTPRJ||rs4752904 (p.Asp872Glu)||299 PTC||339||41.1; 46.2||1.61 (1.15–2.25)||0.0053||R||70.7||98.8||Sequencing||N/A||N/A||10||No|
|14q32.3||XRCC3||rs861539 (p.Thr241Met)||109 (PTC+FTC)||217||44.0; 40.4||2.00 (1.10–3.60)||0.026||R||40.6||83.7||RFLP (NlaIII)||N/A||Age, gender, origin||11||12|
|14q32.3||XRCC3||rs861539 (p.Thr241Met)||134 NMTC (PTC +FTC+Hürthle)||161||40.7; 29.8||2.10 (1.30–3.40)||0.004||R||37.6||81.5||RFLP (NlaIII)||N/A||Age, gender, ancestry, radiation exposure, smoking and alcohol consumption||12||11|
|17p13.1||TP53||rs1042522 (p.Arg72Pro)||123 PTC||197||33.2; 23.0||1.70 (1.17–2.46)||0.006||M||35.2||78.8||TaqMan||N/A||Origin, age (±3 years)||6||13, 14|
|17p13.1||TP53||rs1042522 (p.Arg72Pro)||98 (PTC+FTC)||153||37.2; 34.3||7.02 (1.93–25.59)||0.0015||R||33.5||75.2||Allele-specific PCR||N/A||Age, gender||13||6, 14|
|17p13.1||TP53||rs1042522 (p.Arg72Pro)||68 Rx PTC||313||39.0; 33.1||2.00 (1.16–3.45)||<0.05||R||32.4||72.9||RT-PCR +sequencing||N/A||N/A||14||6, 13|
The table is sorted by power of the study, replication status, gene and variant; ars code for each SNP is always provided, as well as other aliases; bcalculated by us when not provided. MAF, minor allele frequency; OR, odds ratio; 95% CI, confidence interval 95%; PTC, papillary thyroid cancer; FTC, follicular thyroid cancer; Rx PTC, radiation-induced PTC; NMTC, non-medullary thyroid cancer. M, multiplicative; R, recessive; D, dominant; Ref., reference. cReferences: 1, Gudmundsson et al. (2009); 2, Landa et al. (2009); 3, Takahashi et al. (2010); 4, Wokołorczyk et al. (2008); 5, Jazdzewski et al. (2008); 6, Akulevich et al. (2009); 7, Ho et al. (2009); 8, Chiang et al. (2008); 9, Matakidou et al. (2004); 10, Iuliano et al. (2010); 11, Bastos et al. (2009); 12, Sturgis et al. (2005); 13, Granja et al. (2004); 14, Rogounovitch et al. (2006).
Additional variants associated with the development of NMTC, which should be further validated
|Chr location||Gene||Studied variant (synonyms)a||Cases||Con.||MAF cases; MAF controls (%)||OR (95% CI)b||P valueb||Best fitting model||Power to detect an OR=1.5||Power to detect an OR=2.0||Genotyping platform||Functional assessment||Matching or adjustment criteria||Ref.d||Rep.?|
|10q11.2||RET||rs1800861 (p.Leu769Leu)||247 PTC||319||22.0; 24.0||2.22 (N/A)||<0.05||R||57.5||96.4||RFLP (TaqI)||N/A||Age, gender, origin||1||No|
|10q11.2||RET||rs1800860 (p.Ala432Ala)||101 NMTC||174||29.2; 26.4||1.64 (1.00–2.69)||0.049||D||32.7||74.6||RFLP (AseI)||N/A||Age, gender, ancestry||2||No|
|11q22-q23||ATM||rs1801516 (p.Asp1853Asn)||255 PTC||596||11.6; 15.7||0.69 (0.45–0.86)||0.03||M||57.4||96.4||RFLP (AflII)||N/A||N/A||3||No|
|11q22–q23||ATM||rs664677 (IVS22-77 T>C)||132 PTC||398||42.8; 38.9||1.84 (1.10–3.24)||0.03||R||52.0||93.1||RFLP (RsaI)||N/A||N/A||3||No|
|15q24.1||CYP1A1||rs1048943 (p.Ile462Leu)||202 PTC||511||30.9; 19.4||2.07 (1.49–2.88)||<0.0001||D||55.0||95.3||RFLP (BsaI)||N/A||Age, ancestry||4||No|
|15q24.1||CYP1A1||rs4646903 (CYP1A1m1, CYP1A1*2A)||136 PTC||277||13.2; 20.2||0.56 (0.36–0.89)||0.015||D||39.0||83.4||RFLP (MspI)||N/A||Age, gender, origin||5||No|
|6p12||VEGFA||rs699947 (−2578C>A)||297 (PTC +FTC)||249||28.1; 23.3||1.36 (1.02–1.81)||0.039||A||55.0||95.8||TaqMan||N/A||Age, gender||6||No|
|12p13.1-p12||CDKN1B||rs34330 (c.−79C>T)||129 FVPTC||385||31.0; 25.3||2.12 (1.09–4.15)||0.023||R||45.7||89.3||TaqMan||Reduces CDKN1B mRNA levels||Age, gender||7||No|
|22q13.1||CYP2D6||rs3892097 (CYP2D6*4)||187 PTC||256||13.4; 21.7||0.28 (0.09–0.93)||0.037||R||45.3||89.9||RFLP (BstNI)||N/A||N/A||8||No|
|8p22||NAT2||rs1801280 (p.Ile114Thr, NAT2*5)||152 (PTC +FTC)||148||44.1; 53.0||0.58 (0.35–0.98)||0.04||R||41.2||83.3||LightCycler real-time PCR||N/A||N/A||9||No|
|12p13-p12.2||RAD52||rs11226 (c.*744C>T)||210 PTC||228||31.2; 18.0||1.53 (1.03–2.28)||0.036||D||41.0||86.6||RFLP (HaeIII)||N/A||Age, ancestry||10||No|
|12p13-p12.2||RAD52||rs4987206 (p.Gln221Glu)||210 PTC||217||18.6; 1.4||17.58 (7.44–41.58)||<0.0001||D||8.5||16.7||RFLP (BfuA1)||N/A||Age, ancestry||10||No|
|22q11.23||GSTT1||GSTT1 null||47 FVPTC||247||25.3; 42.7||0.31 (0.10–0.97)||0.044||R||24.8||58.2||Multiplex PCR||N/A||N/A||8||4|
|22q11.23||GSTT1||GSTT1 null||223 PTC||513||N/A||3.48 (2.48–4.88)||<0.0001||R||N/A||N/A||Multiplex PCR||N/A||Age, ancestry||4||8|
|12q13.11||VDR||rs2228570 (rs10735810c, c.2T>A)||40 FTC||321||50.0; 34.6||3.75 (1.45–9.70)||0.04||R||23.2||54.7||RFLP (FokI)||N/A||N/A||11||No|
|12q13.11||VDR||rs7975232 (c.1025-49G>T)||40 FTC||321||57.5; 42.5||3.80 (1.45–9.98)||0.04||D||22.9||53.9||RFLP (ApaI)||N/A||N/A||11||No|
|22q12.1||CHEK2||rs17879961 (p.I157T)||173 NMTC||4000||4.3; 2.4||1.90 (1.08–3.24)||0.04||D||20.9||48.1||RFLP (PstI)||N/A||Origin||12||No|
|1p13.3||GSTM1||GSTM1 null||42 PTC (stage III/IV)||247||63.6; 76.1||0.50 (0.26–0.96)||0.036||R||13.4||28.9||Multiplex PCR||N/A||N/A||8||4|
|1p13.3||GSTM1||GSTM1 null||223 PTC||513||N/A||0.72 (0.52–0.99)||0.047||R||N/A||N/A||Multiplex PCR||N/A||Age, ancestry||4||8|
The table is sorted by power of the study, replication status, gene and variant; ars code for each SNP is always provided, as well as other aliases; bcalculated by us when not provided; cThis SNP was originally reported by the author, but was later merged into rs2228570. MAF, minor allele frequency; OR, odds ratio; 95% CI, confidence interval 95%; PTC, papillary thyroid cancer; FTC, follicular thyroid cancer; FVPTC, follicular variant of PTC; NMTC, non-medullary thyroid cancer; RFLP, restriction fragment length polymorphism. Con., controls; R, recessive; D, dominant; M, multiplicative; A, additive; Ref., reference; Rep., replicated. dReferences: 1, Lesueur et al. (2002); 2, Ho et al. (2005); 3, Akulevich et al. (2009); 4, Siraj et al. (2008a); 5, Bufalo et al. (2006); 6, Hsiao et al. (2007); 7, Landa et al. (2010); 8, Lemos et al. (2007); 9, Hernandez et al. (2008); 10, Siraj et al. (2008b); 11, Penna-Martinez et al. (2009); 12, Cybulski et al. (2004).
Models of inheritance and tumor types
Almost half of the studies reported a significant association fitting a recessive model, that is, when both alleles of the associated variant are present in homozygosis. This brings concern about the strength of some of these associations, since in several cases the number of polymorphic homozygous carriers was relatively small, conditioning the power of the studies and their reproducibility. Furthermore, and also influenced by sample size requirements, almost all studies reviewed here describe associations with PTC (accounting for 85% of cases) or with PTC+FTC considered together. Only one single study specifically reports an association between FTC and VDR variants (Penna-Martinez et al. 2009), although the low number of cases considered (n=40) requires a cautious interpretation.
Biases in population ancestry
In concordance with other diseases, most of the reported associations come from studies performed in populations of Caucasian ancestry. Some exceptions to this tendency are two papers published by a group studying patients from Taiwan (Hsiao et al. 2007, Chiang et al. 2008), who reported significant associations for XRCC1 (see below) and VEGFA, respectively, in PTC. Other examples are a study in a cohort from southern Tunisia (Rebai et al. 2009), and two reports based on Saudi Arabian populations (Siraj et al. 2008a,b), the latter reporting a significant twofold increased risk for a non-synonymous variant of the CYP1A1 gene in a group of 202 PTC patients.
Regarding genotyping methods, in contrast with other cancers and polygenic diseases, only a few studies relied on high throughput and/or array-based technologies (Hsiao et al. 2007, Chiang et al. 2008, Akulevich et al. 2009, Gudmundsson et al. 2009, Landa et al. 2009, Takahashi et al. 2010), and a restriction fragment length polymorphism approach was the predominantly used technique (more than 15 studies, see Tables 1 and 2). This fact supports the notion that much remains to be explored in the thyroid cancer susceptibility field.
Candidate gene versus genome-wide association studies approaches
Overall, the great majority of case–control studies are candidate gene or pathway-driven approaches, with the exception of two recently published genome-wide association studies (GWAS) in thyroid cancer (Gudmundsson et al. 2009, Takahashi et al. 2010), which will be commented on later on in this review. Among the pathways and metabolic processes studied, the most common ones are DNA repair due to the known relation between thyroid cancer and ionizing radiation (Ron et al. 1995, Williams 2002), cell cycle control along with the RET signaling pathway, and detoxification processes (related to the cytochrome P450 (CYP) and glutathione-S-transferase (GST) gene families).
In this regard, several studies have reported associations between polymorphisms in DNA damage response genes and PTC. One of the validated variants in this pathway is rs861539, which leads to a p.Thr241Met amino acid change in the homologous recombination gene XRCC3; this gene is also associated with other malignancies, especially breast cancer (Economopoulos & Sergentanis 2010). The association with thyroid cancer was first reported in a hospital-based study that included 134 patients and 161 controls, reporting a 2.1-fold increased risk (95% confidence interval (CI)=1.3–3.4; P=0.004) for polymorphic homozygous carriers (Sturgis et al. 2005). These results were recently validated in another hospital-based study with 109 NMTC patients and 217 controls (OR=2.0; 95% CI=1.1–3.6; P=0.026; Bastos et al. 2009). Although the ORs for rs861539 are probably overestimated due to the relatively small sample size of both studies, the fact that two independent groups reported similar risks fitting the same model supports the involvement of XRCC3 in PTC susceptibility. Similarly, two groups reported an association between rs25487 (p.Arg399Gln amino acid change) in the XRCC1 gene and NMTC. Both groups reported a similar protective effect under a dominant inheritance model, in well-powered cohorts of 255 PTC versus 596 controls (OR=0.70; 95% CI=0.59–0.93; P=0.03; Akulevich et al. 2009) and 251 NMTC versus 503 controls (OR=0.70; 95% CI=0.50–1.00; P=0.049; Ho et al. 2009). Surprisingly, Akulevich et al. (2009) did not find any association with XRCC1, involved in the base excision repair pathway, when considering only radiation-exposed PTC cases and controls, but this could be an effect of the small sample size after stratification (n=123 vs 197). Another non-synonymous polymorphism (rs1799782, p.Arg194Trp) in XRCC1 was also associated with thyroid cancer in a cohort of 283 cases and 469 controls from Taiwan (OR=1.85; 95% CI=1.11–3.07; Chiang et al. 2008). The difference between this risk-conferring SNP and the protection associated with rs25487 could be due to a different effect of each amino acid change or to differences in population-specific allele frequencies (Consortium 2003). For additional comments about these associations between DNA repair genes and thyroid cancer, we recommend reading a recent editorial written by some of the authors who originally reported some of these results (Sturgis & Li 2009).
Another group of genes frequently studied in NMTC codes for proteins involved in cell cycle regulation. Common germline variations within the TP53 tumor suppressor gene were extensively studied (Granja et al. 2004, Bufalo et al. 2006, Rogounovitch et al. 2006, Akulevich et al. 2009) and focused on the polymorphism rs1042522 (p.Arg72Pro). First, arginine homozygosity was described as protective (OR=0.50; 95% CI=0.29–0.86) in a group of 68 radiation-induced PTC (Rogounovitch et al. 2006), that later was expanded (n=123) by the same group, reporting an increased risk (OR=1.70; 95% CI=1.17–2.46) for homozygous proline carriers (Akulevich et al. 2009). A risk associated to proline homozygosity had been already reported (Granja et al. 2004), although this was overestimated due to the low sample size and a very significant Hardy–Weinberg disequilibrium among controls. Overall, although confirmatory studies in TP53 with larger sample sizes may be desirable, there is evidence suggesting that variation at this locus could play a role in PTC pathogenesis. Furthermore, it is widely known that somatic mutations inactivating p53 are frequently found in thyroid tumors that become undifferentiated and have an aggressive clinical behavior (Kondo et al. 2006).
From gene desert regions to scarcity of functional studies
There are some other examples of positive associations for NMTC found in genes related to cell cycle control. In this regard, two apparently independent variants in the cell cycle checkpoint kinase ATM were associated with PTC (Akulevich et al. 2009). Another checkpoint regulator, CHEK2, was also associated with PTC in a multicancer study, mostly focused on breast and prostate neoplasias but also including 173 thyroid cancer cases, in which the change p.Ile157Thr seemed to confer a 1.90-fold increased risk (Cybulski et al. 2004). The same group also reported an association with PTC for rs6983267 (OR=1.37; 95% CI=1.02–1.82), located in the gene desert region 8q24, in a well-powered study that included an expanded series of 485 cases (Wokolorczyk et al. 2008).
Unfortunately, despite this tendency of studying genes and variants that a priori are supposed to have a straightforward biological effect, very few studies have assessed the functional role of the associated alleles (Jazdzewski et al. 2008, Landa et al. 2009, 2010).
In this regard, a study in a Spanish population found an association between the promoter variant rs34330 (c.−79C>T) of the cyclin-dependent kinase inhibitor CDKN1B (p27) and the risk of developing a specific subtype of thyroid cancer: the FVPTC (Landa et al. 2010). The relevance of this study mostly relies on the functional assays performed, which demonstrated an effect of the associated variant on the transcription rate of this gene, providing insight into the underlying mechanism that could drive the reported risk.
The role of polymorphic variants in non-coding RNA sequences, or microRNAs, has been explored in a single and elegant study, which associated in a large group of 608 cases the heterozygous status of SNP rs2910164, located in the pre-miR-146a, with an increased risk of developing PTC (Jazdzewski et al. 2008). This paper also provided a comprehensive view of the effect of this variant, which influences the stability and processing of pri-miR, the expression of mature miR-146a, and the levels of its relevant target genes, TRAF6, IRAK1, and CCDC6.
The importance of biological interpretation
As mentioned above, other pathways frequently assessed for NMTC are those involved in the detoxification of endogenous or xenobiotic compounds (related to CYP and GST gene families), and the RET signaling pathway.
Several detoxification genes have been associated with thyroid cancer, and some of the studies are worth mentioning. Regarding phase I enzymes, significant associations have been reported for rs4646903 of CYP1A1 in a group of 136 PTC cases (Bufalo et al. 2006) and for rs3892097 in CYP2D6 in 187 PTC cases (Lemos et al. 2007). About phase II enzymes, one study associated the GSTM1 null genotype to PTC of stages III/IV (n=42), and GSTT1 null status to FVPTC (n=47; Lemos et al. 2008). Both associations were also reported in a much larger study, with 223 Saudi Arabian PTC patients (Siraj et al. 2008b), although in an opposite direction for GSTT1. Combinations of GSTM1+T1 null genotypes have also been described in NMTC, although they should be interpreted with caution, since the consideration of multiple variants led to an important reduction in sample size (Morari et al. 2002, Granja et al. 2004, Ho et al. 2006). Finally, the genotype status of N-acetyltransferase (NAT2) also seemed to be associated with PTC (Hernandez et al. 2008, Guilhen et al. 2009). Overall, although there is some evidence for association of detoxification enzyme polymorphisms with thyroid cancer, the biological link remains unclear.
On the contrary, key thyroid follicular cell pathways such as the RET signaling pathway and iodine metabolism appear obvious processes to explore the genetic susceptibility to thyroid cancer. In fact, two studies have assessed the role of SNPs in RET, which is frequently rearranged in PTC tumors (Santoro et al. 1992, Nikiforov 2002), reporting associations for the synonymous SNPs rs1800861 and rs1800860 in 247 PTC (Lesueur et al. 2002) and 101 NMTC (Ho et al. 2005) patients respectively. Another example of how useful molecular knowledge can be for searching genetic variants is provided by a very recent study, which linked rs4752904 (p.Asp872Glu) in the PTPRJ gene with PTC susceptibility (OR=1.61; 95% CI=1.15–2.25; Iuliano et al. 2010). Interestingly, PTPRJ encodes a tyrosine phosphatase which is known to be able to dephosphorylate RET and which has an established role as a tumor suppressor gene in thyroid carcinogenesis (Trapasso et al. 2000, Iervolino et al. 2006). Similarly, the thyroglobulin gene (TG), due to its fundamental role in iodine metabolism and thyroid hormone production, is a priori a good candidate to study, and has been consequently associated with NMTC (Matakidou et al. 2004). In this work, performed in a group of 304 patients, the non-synonymous variant rs1133076 (p.Arg2530Gln) conferred a 1.6-fold increased risk of developing the disease.
FOXE1: thyroid related, extensively validated, and functionally assessed
Finally, in the last 2 years, three multistage, well-powered case–control studies, including two GWAS (Gudmundsson et al. 2009, Takahashi et al. 2010) and a candidate gene approach (Landa et al. 2009), have strongly pointed out the involvement of the forkhead box E1 factor, encoded by the FOXE1 gene, in PTC susceptibility. These findings probably represent the best validated genetic factor identified in PTC so far, and represent a good example of an a priori biologically relevant locus that has been extensively identified as a LPG. FOXE1 (formerly known as ‘thyroid transcription factor 2’) is a pioneer transcription factor, whose action is essential for the development, differentiation, and hormone responsiveness of the thyroid gland (Cuesta et al. 2007). FOXE1 is also necessary for the maintenance of the differentiated state of the thyroid, as it is involved in regulating the transcription of thyroid-specific genes such as the TG and TPO (Zannini et al. 1997).
The first GWAS ever performed on thyroid cancer reported a top association for tagSNP rs965513, located in an intergenic region at 9q22.33, 57 kb upstream of FOXE1 (Gudmundsson et al. 2009). The authors described an OR=1.75 (95% CI=1.59–1.94; P=1.7×10−27) based on 962 PTC+FTC patients, fitting a multiplicative model, that is, a model in which the risk conferred by an allele is increased r-fold for heterozygotes and r2-fold for homozygotes (Lewis 2002). They also correlated the risk allele with low concentrations of TSH and thyroxine (T4), and high levels of tri-iodothyronine (T3). In parallel, another group, based on a cohort of 984 PTC, associated the rs1867277 (c.−283G>A) variant, located in the FOXE1 5′ UTR, with an increased risk of developing the disease (OR=1.49; 95% CI=1.30–1.70; P=5.9×10−9; Landa et al. 2009). These latter authors also proposed a functional mechanism, in which the associated risk allele of rs1867277 affects FOXE1 transcription through a differential recruitment of USF1/USF2 transcription factors. Finally, a third study, involving 660 radiation-induced PTC from the Chernobyl accident (Takahashi et al. 2010), extended and validated the association of both rs965513 (OR=1.65; 95% CI=1.43–1.91; P=4.8×10−12) and rs1867277 (OR=1.48; 95% CI=1.27–1.71; P=4.5×10−7). The large sample sets studied, the similar OR and models of inheritance reported in these three studies, as well as the functional assessment of FOXE1 variants, highlight the importance of this locus as a LPG in PTC. Nevertheless, since those variants are quite common in the general population, it remains a challenge to identify other genetic factors that could complete the picture and elucidate the still unexplained heritability in PTC.
A schematic representation of the multiple factors affecting NMTC is shown in Fig. 2.
In addition, some variants were shown not to be associated with NMTC forms. Among them, SNPs in ERCC2 (Silva et al. 2005), MTF1 (Akulevich et al. 2009), FAS (Ho et al. 2008), APEX1 (Chiang et al. 2008), PARP1 (Chiang et al. 2008), GSTO1 (Granja et al. 2005, Bufalo et al. 2006), GSTP1 (Hernandez et al. 2003, Gaspar et al. 2004), TSHR (Matakidou et al. 2004, Lonn et al. 2007), RAD51 (Sturgis et al. 2005, Bastos et al. 2009), RAD52 (Sturgis et al. 2005), BRCA1 (Sturgis et al. 2005), BRCA2 (Sturgis et al. 2005), and XRCC7 (Sturgis et al. 2005), were included in relatively well-powered studies and seem not to play a role in the disease. On the other hand, studies on CDKN2A (Debniak et al. 2006), EGFR (Rebai et al. 2009), and MYCL1 (Yaylim-Eraltan et al. 2008), lacked the necessary power and led to inconclusive results that should be further expanded.
Medullary thyroid cancer
As described above, about 75% of MTC, the thyroid cancer subtype arising from parafollicular cells, appear as a sporadic disease (Randolph & Maniar 2000). Those forms, for which there are no environmental causes described, could also behave as a complex disease, driven by multiple polymorphic variants that might determine individual susceptibility.
Association studies in MTC
MTC, although very homogeneous, which avoids the need for stratification, is much less frequent than thyroid cancers derived from follicular cells. The resulting difficulty in recruiting enough patients to reach the necessary power has probably determined the relative rarity of association studies on sporadic MTC (sMTC) published so far. Case–control studies on MTC have unanimously adopted candidate gene or pathway-driven approaches, with several studies assessing the role of the same variants in different populations. So far, variations of the RET proto-oncogene and its effectors have monopolized the efforts in looking for low-penetrance loci. Positive associations are summarized in Table 3 and Fig. 1B, and commented on below.
Variants associated with the development of MTC
|Chr location||Gene||Studied variant (synonyms)a||Cases||Con.||MAF cases; MAF controls (%)||OR (95% CI)b||P valueb||Best fitting model||Power to detect an OR=1.5||Power to detect an OR=2.0||Genotyping platform||Functional assessment||Matching or adjustment criteria||Ref.c||Rep.?|
|10q11.2||RET||rs1800862 (p.Ser836Ser)||32 MTC||250||9.3; 3.6||2.97 (1.08–8.16)||0.04||D||11.6||21.6||RFLP (AluI)||N/A||Race, origin||1||2, 3|
|10q11.2||RET||rs1800862 (p.Ser836Ser)||49 MTC||70||9.2; 3.7||N/A||0.03||N/A||8.2||14.7||RFLP (AluI)||N/A||Race, origin||2||1, 3|
|10q11.2||RET||rs1800862 (p.Ser836Ser)||81 MTC||80||10.5; 3.1||3.98 (1.39–11.40)||0.01||D||7.7||14.4||RFLP (AluI)||N/A||Gender, age||3||1, 2|
|10q11.2||RET||rs1799939 (p.Gly691Ser)||120 MTC||529||27.1; 17.7||1.72 (1.24–2.39)||0.001||C||40.1||83.9||TaqMan||Does not affect splicing of RET||Age, gender, ethnicity||4||5|
|10q11.2||RET||rs1799939 (p.Gly691Ser)||106 MTC||106||27.8; 18.8||1.60 (1.03–2.49)||0.039||M||23.0||57.8||RFLP (BanI)+ sequencing||Does not influence RET mRNA expression||Age, gender, ethnicity, origin||5||4|
|10q11.2||RET||rs1800863 (p.Ser904Ser)||125 MTC||528||26.8; 17.7||1.70 (1.23–2.35)||0.001||C||41.0||84.9||TaqMan||N/A||Age, gender, ethnicity||4||No|
|10q11.2||RET||rs3026782 (c.*388G>A)||123 MTC||522||26.4; 17.5||1.68 (1.21–2.32)||0.002||C||40.3||84.2||TaqMan||N/A||Age, gender, ethnicity||4||No|
|10q11.2||RET||rs2565206 (IVS1–126 G>T)||104 MTC||100||50.0; 35.0||N/A||0.003||N/A||29.1||68.2||Fluorescent SSCP+RFLP (MspI and NlaIII)||N/A||Age, gender, race||6||7, 8|
|10q11.2||RET||rs2565206 (IVS1–126 G>T)||58 MTC||100||50.0; 35.0||N/A||0.012||N/A||22.7||54.5||Fluorescent SSCP+RFLP (MspI and NlaIII)||N/A||Age, gender, race||8||6, 7|
|10q11.2||RET||rs2565206 (IVS1–126 G>T)||51 MTC||100||48.0; 35.0||N/A||0.039||N/A||21.4||51.3||Fluorescent SSCP+RFLP (MspI and NlaIII)||N/A||Age, gender, race||7||6, 8|
|10q11.2||RET||rs2472737 (IVS14-24)||45 MTC||79||27.7; 6.3||6.04 (2.49–14.63)||<0.001||D||10.2||20.8||Sequencing||N/A||Age, gender||9||No|
|10q26.11||GFRα1||rs1061413 (c.*946C>G)||129 MTC||523||33.7; 26.1||1.54 (1.04–2.27)||0.029||D||49.1||91.6||TaqMan||N/A||Age, gender, origin||4||No|
|10q26.11||GFRα1||rs45568534 (c.−193C>G)||31 MTC||31||14.5; 0.0||N/A||<0.05||N/A||N/A||N/A||Sequencing||Tendency towards higher GFRA1 mRNA and protein expression||Race||10, 11||No|
|1p33-p32||ARNT||rs3762422 (c.−797A>T)||132 MTC||520||32.2; 22.5||1.93 (1.31–2.84)||0.0008||D||47.0||90.2||TaqMan||N/A||Age, gender, ethnicity||4||No|
|9p21||CDKN2B||rs643319 (c.612-11596C>A)||403 MTC||880||39.7; 44.9||0.75 (0.58–0.96)||0.02||D||92.0||100.0||Illumina +TaqMan||N/A||Age, gender, origin||12||No|
|9p21||CDKN2B||rs2106119 (c.612-11882A>G)||170 MTC||241||55.6; 45.6||1.47 (1.12–1.94)||0.006||M||52.3||93.0||Illumina +TaqMan||Influences CDKN2B transcription through HNF1α binding||Age, gender, origin||12||No|
|5p15.33||TERT||rs2075786 (c.2654+269T>C)||402 MTC||898||31.8; 36.9||0.69 (0.54–0.87)||0.002||D||91.5||100.0||Illumina +TaqMan||N/A||Age, gender, origin||12||No|
|20q13.2||AURKA||rs911160 (c.566+478C>G)||400 MTC||888||30.0; 23.0||1.47 (1.21–1.78)||1.07× 10−4||M||84.9||99.9||Illumina +TaqMan||N/A||Age, gender, origin||12||12|
|22q11.21||COMT||rs165849 (g.34407G>A)||386 MTC||553||28.5; 34.6||0.76 (0.62–0.92)||0.006||M||84.7||99.9||Illumina +TaqMan||N/A||Age, gender, origin||12||No|
|7q21-q22||CDK6||rs8 (c.234-4184G>A)||399 MTC||878||15.0; 21.4||0.61 (0.47–0.79)||1.69× 10−4||D||83.2||99.9||Illumina +TaqMan||N/A||Age, gender, origin||12||12|
|2q32.2||STAT1||rs10173099 (c.373-1443G>A)||299 MTC||678||37.0; 29.9||1.55 (1.18–2.05)||0.002||D||79.5||99.8||Illumina +TaqMan||N/A||Age, gender, origin||12||No|
|18q21.3||BCL2||rs1801018 (c.21A>G)||260 MTC||378||48.0; 39.6||1.44 (1.14–1.82)||0.002||M||70.4||99.0||Illumina +TaqMan||N/A||Age, gender, origin||12||No|
|11p15.5||HRAS||rs8176332 (g.4774C>A)||226 MTC||362||14.4; 10.1||1.47 (1.04–2.09)||0.031||M||35.8||80.6||Illumina +TaqMan||N/A||Age, gender, origin||12||No|
The table is sorted by gene and variant; ars code for each SNP is always provided, as well as other aliases; bcalculated by us when not provided. MAF, minor allele frequency; OR, odds ratio; 95% CI, confidence interval 95%; MTC, medullary thyroid cancer; RFLP, restriction fragment length polymorphism; SSCP, single-strand conformational polymorphism. Con., controls; D, dominant; C, codominant; M, multiplicative; Ref., reference; Rep., replicated. cReferences: 1, Ruiz et al. (2001); 2, Gimm et al. (1999); 3, Siqueira et al. (2010); 4, Cebrian et al. (2005); 5, Elisei et al. (2004); 6, Fernandez et al. (2004); 7, Borrego et al. (2003); 8, Fernandez et al. (2006); 9, Baumgartner-Parzer et al. (2005); 10, Gimm et al. (2001a); 11, Gimm et al. (2001b); 12, Ruiz-Llorente et al. (2007).
RET polymorphisms in sMTC
Given the essential role of both germinal (95%) and somatic (30–60%) RET mutations, researchers have focused their efforts on the putative role of polymorphic variants within this locus. Moreover, because polymorphisms are relatively common in the population, they may present a much higher attributable risk in the general population than rare mutations in high-penetrance cancer susceptibility genes such as RET (de Groot et al. 2006). Several RET SNPs have been postulated as modifiers of phenotype in the familial forms of the disease (Gil et al. 2002, Robledo et al. 2003, Lesueur et al. 2006, Tamanaha et al. 2009, Shifrin et al. 2010). This modifier effect on familial forms has already been reviewed (Weber & Eng 2005) and will not be further commented on here. In addition, some of those variants were studied as LPGs in sMTC.
A major limitation in these studies is the sample size, with only some series reaching a 100 patients, and none of them including an intra-study validation set of the results. Another major pitfall is the few and controversial functional results for these variants. This, along with the LD pattern of the RET locus, makes it difficult to interpret if the associated SNPs are in fact markers of other unknown functional variants, or are themselves influencing RET function.
In the first work studying RET SNPs, the synonymous variant rs1800862 (p.Ser836Ser) was significantly associated with MTC, and correlated with the presence of the RET somatic mutation M918T (Gimm et al. 1999). This association was reproduced in two additional, rather small, populations (Ruiz et al. 2001, Siqueira et al. 2010). Nevertheless, several other groups failed to replicate this signal (Wiench et al. 2004, Baumgartner-Parzer et al. 2005, Cebrian et al. 2005), thus leading to the conclusion that the RET rs1800862 variant may only be a risk factor in certain populations.
On the other hand, the non-synonymous variant rs1799939 (p.Gly691Ser) has been associated with about a 1.7-fold risk per allele for developing sMTC in two larger studies (Elisei et al. 2004, Cebrian et al. 2005). These two studies postulated, through a functional assessment of RET transcription and splicing, that rs1799939 could be the functional variant, but the results were inconclusive. Although one study did not replicate this association (Fugazzola et al. 2008), others reported significant results for rs1800863 (p.Ser904Ser; Cebrian et al. 2005), which is in LD with p.Gly691Ser, thus reinforcing this association. Furthermore, this variant was identified as an independent predictor of a higher basal calcitonin synthesis rate in sMTC patients (Cardot-Bauters et al. 2008).
The picture gets even more complicated since studies with other RET variants, such as the synonymous rs1800861 (p.Leu769Leu) and 3′UTR rs3026782, have led to contradictory results (Wiench et al. 2004, Cebrian et al. 2005, Weinhaeusel et al. 2008, Sromek et al. 2010). In addition, some intronic polymorphisms have been associated with sMTC, such as rs2565206 (IVS1–126 G>T; Borrego et al. 2003, Fernandez et al. 2004, 2006) and in LD with p.Ser836Ser, and rs2472737 (IVS14–24; Baumgartner-Parzer et al. 2005).
Overall, it seems that there are two independent association signals in the RET gene, which are the carriers of either G691S/S904S or S836S/IVS1–126G>T haplotypes, and which are more prone to develop sMTC through a still unknown mechanism that could involve these and/or other functional variants. Nonetheless, a comprehensive study of unselected RET variants and their LD pattern in several unbiased populations will be desirable to better define the associated risks in this locus.
Other genes related to MTC
Besides the RET proto-oncogene, some other loci have been studied in sMTC. One of them is the GDNF family receptor α 1, which acts as a co-receptor of RET, and is encoded by the GFRA1 gene. The promoter variant rs45568534 (c.−193C>G) was first associated with sMTC development in a small group of patients, who also showed a tendency toward a higher expression of tumoral GFRA1 at both the mRNA and the protein levels when carrying the risk alleles (Gimm et al. 2001b). This association was not reproduced in larger case–control studies (Borrego et al. 2002, Cebrian et al. 2005), although another variant in the 3′UTR (rs1061413, c.*946C>G) was significantly associated (Cebrian et al. 2005). In this study, based on a UK population, an association was found for rs3762422 (c.−797A>T) in the ARTN gene, which encodes a ligand of RET.
Finally, a two-step, pathway-driven study including a Spanish (n=266) and a British (n=155) group of patients, reported several genes associated with sMTC (Ruiz-Llorente et al. 2007). Although not uniformly replicated, significant associations were described for variants within the genes STAT1, AURKA, BCL2, CDKN2B, CDK6, COMT, TERT, and HRAS. In addition, the potential role of CDKN2B was confirmed by a functional assay showing a role of the rs7044859 variant in the promoter region in altering the binding of the transcription factor HNF1.
Association studies in sMTC reported negative results, apart from for several RET SNPs (Fugazzola et al. 2008, Fernandez et al. 2009), for variants in GFRA2 (Gimm et al. 2001b, Borrego et al. 2002, Cebrian et al. 2005), GFRA3 (Gimm et al. 2001b, Borrego et al. 2002, Cebrian et al. 2005, Ruiz-Llorente et al. 2007), GFRA4 (Cebrian et al. 2005), NTRK2 (Gimm et al. 2001a), and NTRK3 (Gimm et al. 2001a), among many others, assessed in the study by Ruiz-Llorente et al. (2007). Some of these genes should not be completely discarded, however, due to the limited power of some of the studies.
Overall, the potential role of the different polymorphisms in the development of sMTC needs to be further characterized, and the molecular background of these polymorphisms needs to be elucidated. In addition, studying the genetic susceptibility to MTC could benefit from an as yet not performed GWAS, which could potentially unveil new genes, probably not directly linked to the RET pathway, but playing a role in the disease.
Conclusions and future perspectives
In the last 10 years, an important number of case–control studies have been published on thyroid cancer, although few loci have been consistently associated. Some good examples of the achievements in this field are, for example, the involvement of the FOXE1 gene in PTC, or RET variants in MTC.
A vast proportion of risks attributable to genetic factors, however, remain unexplained. An evident solution, especially relevant for thyroid cancer, which is relatively rare, relies on the establishment of international consortia, able to recruit enough patients to reach the required statistical power to detect low- to moderate-penetrance loci, especially those involving rare (MAF<10%) variants. It would be interesting to evaluate the joint effect of the top reported variants in a large population, and thus evaluate the proportion of risk that they explain when considered together.
The establishment of international consortia leads us to another key feature: the need for tumor stratification. In fact, the heterogeneity of NMTC has not been properly addressed so far. It seems logical that if ‘PTC, FTC’, and each of the PTC subtypes (‘classic PTC, FVPTC’, etc.) are clinically different, and, in some cases, display distinct somatic alterations, they should also involve different germinal low-penetrance loci. Nonetheless, associations in follicular cell-derived thyroid cancer usually refer to ‘NMTC’ or to the predominant ‘PTC’.
Another major pitfall in this field is the lack of functional data that could explain the underlying mechanisms of the reported polymorphisms. This also applies to other complex or multifactorial diseases, where it remains a challenge to elucidate the functional link between the great majority of associated variants and phenotypic traits (Frazer et al. 2009, Katsanis 2009). The path between an association signal and the functional variant is hardly ever straightforward, but it is essential to understand the cancer biology, and potentially improve its prognosis and treatment. In this regard, it will be interesting to take advantage of the new massive sequencing technologies, and resequence those regions or genes associated with thyroid cancer, but still unexplained from a functional point of view.
Finally, and keeping in mind the multifactorial nature of these cancers, it is unlikely that they can be explained by single, individual genes. Because of this, a more comprehensive view should be adopted, considering gene–gene and gene–environment interactions. So far, studies on the combination of several genotypes, such as those involving GST genes (Morari et al. 2002, Hernandez et al. 2003, Ho et al. 2006), were mainly performed due to the lack of strong individual associations, with the exception of the study by Gudmundsson et al. (2009) who showed an OR=5.7 for individuals carrying the homozygous genotypes for the individually associated variants near FOXE1 (TITF2) and NKX2-1 (TITF1). Still, the epistatic contribution to thyroid cancer, that is, the conditional phenotypic effect that several variants provide, not resulting from the additive individual effects (Dimas & Dermitzakis 2009), remains unexplored. In this regard, a large study, specifically designed to assess epistasis, which takes into account the reported associated genes, but also considers randomly additional loci as well as environmental factors, could provide the necessary integrative view of susceptibility to thyroid cancer.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the review reported.
This work was supported in part by the Fondo de Investigaciones Sanitarias (FIS), project PI080883, and Fundación Mutua Madrileña AP2775/2008. I L is a predoctoral fellow of FIS (FI07/00326).
Baumgartner-ParzerSMLangRWagnerLHeinzeGNiederleBKasererKWaldhauslWVierhapperH2005Polymorphisms in exon 13 and intron 14 of the RET protooncogene: genetic modifiers of medullary thyroid carcinoma?Journal of Clinical Endocrinology and Metabolism906232–6236.doi:10.1210/jc.2005-1278.
BignellGRCanzianFShayeghiMStarkMShugartYYBiggsPMangionJHamoudiRRosenblattJBuuP1997Familial nontoxic multinodular thyroid goiter locus maps to chromosome 14q but does not account for familial nonmedullary thyroid cancer. American Journal of Human Genetics611123–1130.doi:10.1086/301610.
BorregoSWrightFAFernandezRMWilliamsNLopez-AlonsoMDavuluriRAntinoloGEngC2003A founding locus within the RET proto-oncogene may account for a large proportion of apparently sporadic Hirschsprung disease and a subset of cases of sporadic medullary thyroid carcinoma. American Journal of Human Genetics7288–100.doi:10.1086/345466.
CapezzoneMMarchisottaSCantaraSBusoneroGBrilliLPazaitou-PanayiotouKCarliAFCarusoGTotiPCapitaniS2008Familial non-medullary thyroid carcinoma displays the features of clinical anticipation suggestive of a distinct biological entity. Endocrine-Related Cancer151075–1081.doi:10.1677/ERC-08-0080.
CavacoBMBatistaPFSobrinhoLGLeiteV2008Mapping a new familial thyroid epithelial neoplasia susceptibility locus to chromosome 8p23.1-p22 by high-density single-nucleotide polymorphism genome-wide linkage analysis. Journal of Clinical Endocrinology and Metabolism934426–4430.doi:10.1210/jc.2008-0449.
CebrianALesueurFMartinSLeylandJAhmedSLuccariniCSmithPLLubenRWhittakerJPharoahPD2005Polymorphisms in the initiators of RET (rearranged during transfection) signaling pathway and susceptibility to sporadic medullary thyroid carcinoma. Journal of Clinical Endocrinology and Metabolism906268–6274.doi:10.1210/jc.2004-2449.
World Health Organization Classification of Tumours. Lyon: IARC Press.
EliseiRCosciBRomeiCBotticiVSculliMLariRBaraleRPaciniFPincheraA2004RET exon 11 (G691S) polymorphism is significantly more frequent in sporadic medullary thyroid carcinoma than in the general population. Journal of Clinical Endocrinology and Metabolism893579–3584.doi:10.1210/jc.2003-031898.
GilLAzanedoMPollanMCristobalEArribasBGarcia-AlbertLGarcia-SaizAMaestroMLTorresAMenarguezJ2002Genetic analysis of RET, GFR alpha 1 and GDNF genes in Spanish families with multiple endocrine neoplasia type 2A. International Journal of Cancer99299–304.doi:10.1002/ijc.10298.
GimmODziemaHBrownJde la PuenteAHoang-VuCDralleHPlassCEngC2001aMutation analysis of NTRK2 and NTRK3, encoding 2 tyrosine kinase receptors, in sporadic human medullary thyroid carcinoma reveals novel sequence variants. International Journal of Cancer9270–74.doi:10.1002/1097-0215(200102)9999:9999<::AID-IJC1153>3.0.CO;2-N.
GimmODziemaHBrownJHoang-VuCHinzeRDralleHMulliganLMEngC2001bOver-representation of a germline variant in the gene encoding RET co-receptor GFRalpha-1 but not GFRalpha-2 or GFRalpha-3 in cases with sporadic medullary thyroid carcinoma. Oncogene202161–2170.doi:10.1038/sj.onc.1204289.
IervolinoAIulianoRTrapassoFVigliettoGMelilloRMCarlomagnoFSantoroMFuscoA2006The receptor-type protein tyrosine phosphatase J antagonizes the biochemical and biological effects of RET-derived oncoproteins. Cancer Research666280–6287.doi:10.1158/0008-5472.CAN-06-0228.
LandaIRuiz-LlorenteSMontero-CondeCInglada-PerezLSchiaviFLeskelaSPitaGMilneRMaravallJRamosI2009The variant rs1867277 in FOXE1 gene confers thyroid cancer susceptibility through the recruitment of USF1/USF2 transcription factors. PLoS Genetics5e1000637doi:10.1371/journal.pgen.1000637.
LandaIMontero-CondeCMalangaDDe GisiSPitaGLeandro-GarciaLJInglada-PerezLLetonRDe MarcoCRodriguez-AntonaC2010Allelic variant at −79 (C>T) in CDKN1B (p27Kip1) confers an increased risk of thyroid cancer and alters mRNA levels. Endocrine-Related Cancer17317–328.doi:10.1677/ERC-09-0016.
LesueurFCorbexMMcKayJDLimaJSoaresPGriseriPBurgessJCeccheriniILandolfiSPapottiM2002Specific haplotypes of the RET proto-oncogene are over-represented in patients with sporadic papillary thyroid carcinoma. Journal of Medical Genetics39260–265.doi:10.1136/jmg.39.4.260.
LesueurFCebrianARobledoMNiccoli-SirePSvenssonKAPinsonSLeylandJWhittakerJPharoahPDPonderBA2006Polymorphisms in RET and its coreceptors and ligands as genetic modifiers of multiple endocrine neoplasia type 2A. Cancer Research661177–1180.doi:10.1158/0008-5472.CAN-05-2995.
LonnSBhattiPAlexanderBHPinedaMADoodyMMStruewingJPSigurdsonAJ2007Papillary thyroid cancer and polymorphic variants in TSHR- and RET-related genes: a nested case–control study within a cohort of U.S. radiologic technologists. 16174–177.doi:10.1158/1055-9965.EPI-06-0665.
MalchoffCDSarfaraziMTendlerBForouharFWhalenGJoshiVArnoldAMalchoffDM2000Papillary thyroid carcinoma associated with papillary renal neoplasia: genetic linkage analysis of a distinct heritable tumor syndrome. Journal of Clinical Endocrinology and Metabolism851758–1764.doi:10.1210/jc.85.5.1758.
NikiforovaMNLynchRABiddingerPWAlexanderEKDornGWIITalliniGKrollTGNikiforovYE2003RAS point mutations and PAX8–PPAR gamma rearrangement in thyroid tumors: evidence for distinct molecular pathways in thyroid follicular carcinoma. Journal of Clinical Endocrinology and Metabolism882318–2326.doi:10.1210/jc.2002-021907.
PalTVoglFDChappuisPOTsangRBrierleyJRenardHSandersKKantemiroffTBaghaSGoldgarDE2001Increased risk for nonmedullary thyroid cancer in the first degree relatives of prevalent cases of nonmedullary thyroid cancer: a hospital-based study. Journal of Clinical Endocrinology and Metabolism865307–5312.doi:10.1210/jc.86.11.5307.
RuizAAntinoloGFernandezRMEngCMarcosIBorregoS2001Germline sequence variant S836S in the RET proto-oncogene is associated with low level predisposition to sporadic medullary thyroid carcinoma in the Spanish population. Clinical Endocrinology55399–402.doi:10.1046/j.1365-2265.2001.01328.x.
Ruiz-LlorenteSMontero-CondeCMilneRLMoyaCMCebrianALetonRCasconAMercadilloFLandaIBorregoS2007Association study of 69 genes in the ret pathway identifies low-penetrance loci in sporadic medullary thyroid carcinoma. Cancer Research679561–9567.doi:10.1158/0008-5472.CAN-07-1638.
ShifrinALOgilvieJBStangMTFayAMKuoYHMatulewiczTXenachisCZVernickJJ2010Single nucleotide polymorphisms act as modifiers and correlate with the development of medullary and simultaneous medullary/papillary thyroid carcinomas in 2 large, non-related families with the RET V804M proto-oncogene mutation. Surgery1481274–1280;.discussion 1280–1271doi:10.1016/j.surg.2010.09.004.
SiqueiraDRRomittiMda RochaAPCeolinLMeottiCEstivaletAPunalesMKMaiaAL2010The RET polymorphic allele S836S is associated with early metastatic disease in patients with hereditary or sporadic medullary thyroid carcinoma. Endocrine-Related Cancer17953–963.doi:10.1677/ERC-09-0312.
SirajAKIbrahimMAl-RasheedMAbubakerJBuRSiddiquiSUAl-DayelFAl-SaneaOAl-NuaimAUddinS2008bPolymorphisms of selected xenobiotic genes contribute to the development of papillary thyroid cancer susceptibility in Middle Eastern population. BMC Medical Genetics961doi:10.1186/1471-2350-9-61.
SromekMCzetwertynskaMSkaskoEZielinskaJCzapczakDSteffenJ2010The frequency of selected polymorphic variants of the RET gene in patients with medullary thyroid carcinoma and in the general population of central Poland. Endocrine Pathology21178–185.doi:10.1007/s12022-010-9125-8.
World Cancer Report. Lyon: IARC Press.
SuhIFilettiSVriensMRGuerreroMATuminoSWongMShenWTKebebewEDuhQYClarkOH2009Distinct loci on chromosome 1q21 and 6q22 predispose to familial nonmedullary thyroid cancer: a SNP array-based linkage analysis of 38 families. Surgery1461073–1080.doi:10.1016/j.surg.2009.09.012.
TakahashiMSaenkoVARogounovitchTIKawaguchiTDrozdVMTakigawa-ImamuraHAkulevichNMRatanajarayaCMitsutakeNTakamuraN2010The FOXE1 locus is a major genetic determinant for radiation-related thyroid carcinoma in Chernobyl. Human Molecular Genetics192516–2523.doi:10.1093/hmg/ddq123.
TamanahaRCamachoCPPereiraACda SilvaAMMacielRMCeruttiJM2009Evaluation of RET polymorphisms in a six-generation family with G533C RET mutation: specific RET variants may modulate age at onset and clinical presentation. Clinical Endocrinology7156–64.doi:10.1111/j.1365-2265.2008.03491.x.
TrapassoFIulianoRBocciaAStellaAViscontiRBruniPBaldassarreGSantoroMVigliettoGFuscoA2000Rat protein tyrosine phosphatase eta suppresses the neoplastic phenotype of retrovirally transformed thyroid cells through the stabilization of p27(Kip1). Molecular and Cellular Biology209236–9246.doi:10.1128/MCB.20.24.9236-9246.2000.
ZanniniMAvantaggiatoVBiffaliEArnoneMISatoKPischetolaMTaylorBAPhillipsSJSimeoneADi LauroR1997TTF-2, a new forkhead protein, shows a temporal expression in the developing thyroid which is consistent with a role in controlling the onset of differentiation. EMBO Journal163185–3197.doi:10.1093/emboj/16.11.3185.