Genetic factors for differentiated thyroid cancer in French Polynesia: new candidate loci

Abstract Background Populations of French Polynesia (FP), where France performed atmospheric tests between 1966 and 1974, experience a high incidence of differentiated thyroid cancer (DTC). However, up to now, no sufficiently large study of DTC genetic factors in this population has been performed to reach definitive conclusion. This research aimed to analyze the genetic factors of DTC risk among the native FP populations. Methods We analyzed more than 300 000 single nucleotide polymorphisms (SNPs) genotyped in 283 DTC cases and 418 matched controls born in FP, most being younger than 15 years old at the time of the first nuclear tests. We analyzed the genetic profile of our cohort to identify population subgroups. We then completed a genome-wide analysis study on the whole population. Results We identified a specific genetic structure in the FP population reflecting admixture from Asian and European populations. We identified three regions associated with increased DTC risk at 6q24.3, 10p12.2, and 17q21.32. The lead SNPs at these loci showed respective p-values of 1.66 × 10−7, 2.39 × 10−7, and 7.19 × 10−7 and corresponding odds ratios of 2.02, 1.89, and 2.37. Conclusion Our study results suggest a role of the loci 6q24.3, 10p12.2 and 17q21.32 in DTC risk. However, a whole genome sequencing approach would be better suited to characterize these factors than genotyping with microarray chip designed for the Caucasian population. Moreover, the functional impact of these three new loci needs to be further explored and validated.


Introduction
Fr enc h Pol ynesia ( FP ) is a Fr enc h Ov erseas Territory, onl y populated by the M āori people from Western Polynesia more than 1000 years ago , 1 up to the first contact with the Europeans, between the 16th and 18th centuries, depending on the arc hipela go. About 93% of the population born in FP still defined themselves as M āori or of mixed M āori origin in the 1988 census, the last census to record ethnic information. 2 From 1966 to 1974, France performed 44 atmospheric nuclear tests in the atolls of Mururoa and Fangataufa, in the southeastern part of FP, near the Gambier Islands. 3 Although only about 500 inhabitants permanently lived less than 500 km from nuclear sites, r adioactiv e fallouts of se v er al tests wer e observ ed in mor e distant islands and atolls, including Tahiti, wher e mor e than half of population of FP lived. 4 Since the beginning of the 2000s, we conducted a populationbased case-control study on differentiated thyroid cancer ( DTC ) risk factors in FP. This study was prompted by the high incidence of DTC recorded in the Polynesian islands, including FP, [5][6][7] and the known risks of r adiation-induced thyr oid health effects associated with r adioactiv e fallout fr om nuclear tests . 8 We ha ve shown that lifestyle-related risk factors such as obesity and a high number of pr egnancies ar e associated with DTC risk in European populations and could explain part of the high incidence of DTC in FP, due to their high pr e v alence in this territory. [9][10][11] A family history of thyroid cancer or drinking spring water was also shown to be associated with DTC risk, 12 , 13 whereas a traditional Pol ynesian diet lar gel y based on local vegetables and sea products was associated with a lower thyroid cancer risk. 10 Lastly, it was also suggested that there was a small, but significant, role of nuclear tests fallouts in the incidence of DTC in FP. 6 , 14 In order to investigate the potential role of genetic susceptibility in the high thyroid cancer incidence observed in FP, we previously performed a candidate gene-based analysis to confirm in the Fr enc h Pol ynesian population the r ole of 5 v ariants in 2q35, 9q22.33, 8p12 and 14q13.3 by Maillard et al . 11 that were identified in former GWAS in European populations. 13 , 15 We reported associations of similar magnitude to what was observed in the European population but sho w ed that the at-risk alleles of such variants were not more frequent in the French Polynesian population than in the European Population. 11 Maillard et al. performed an analysis on common variants at 9q22.33, 14q13.3, and ATM loci and DTC in the initial case-control study included in the present analysis ( about 50% of the population of the present study ) . 11 The findings of this pr e vious study highlighted an interaction between rs944289 in NKX-1 ( 14q13.3 ) and the radiation dose to the thyroid with a higher risk for the genotype "T/T" carriers. In addition, an association was found between increased DTC risk and rs965513 near FOXE1 . 11 More detailed results were reported through a multiethnic GWAS and a fine-mapping analysis that included the Melanesian population from New Caledonia and only a part of the Polynesian population included in the present article. 16 , 17 Indeed, we extended the pr e vious case-contr ol study conducted in FP by including thyroid cases diagnosed from 2004 to 2016. It resulted in a confirmation of our pr e vious r esults, although, slightly lo w ering the estimate of the risk associated with r adiation exposur e. 18 In this report, we performed the first GWAS on DTC for individuals exclusiv el y fr om FP to identify genetic DTC risk factors specific to this high-risk population.

Population
Cases were collected from the Cancer Incidence Registry of FP, private endocrinologists and clinics, as well as public hospitals. All patients diagnosed in 1981-2016 with DTC at the age of 55 years or less, born and resident in FP at the time of dia gnosis, wer e eligible. Contr ols wer e healthy individuals, selected from the native population using the registry of births that covers the whole territory. Cases and controls were matched on birth date ( around 3 months of maximum difference ) and sex. Full details about the first study are described elsewhere. 18 , 19 All participants answered a detailed questionnaire about education ( available as supplementary data ) , occupation, weight history, personal and familial history of thyroid diseases and cancers, gynecologic and r epr oductiv e history, and diet at the time of the interview and during childhood. Pathological and histological thyroid cancer details were also gathered from clinical records for each case with saliva DNA samples ( Oragene ® kit ) . The study has been a ppr ov ed by the "Commission Nationale de l'Informatique et des Libertés" ( CNIL a gr eements 996 007 and 900 124 ) and the Ethical Committee of Fr enc h Pol ynesia. Eac h participant signed an individual written consent.

Estimation of thyroid radiation dose
The methodology and results of the dose estimation process have already been described in former published studies . T hyroid radiation doses due to ( i ) intakes of 131I and of short-lived radioio-dine isotopes ( 132I, 133I, 135I ) and 132Te via inhalation and ingestion of foodstuffs and drinking water, ( ii ) external irradiation from gamma-emitting radionuclides deposited on the ground, and ( iii ) ingestion of long-lived 137Cs with foodstuffs were reconstructed for each study subject, mainly by using: r Results of radiation monitoring of the environment and foodstuffs performed by the SMSR ( Service Mixte de Sécurité Radiologique ) and the SMCB ( Service Mixte de Contrôle Biologique ) , the internal radiation protection services of Fr enc h Army and CEA after each test, which were declassified in 2013, instead of syntheses sent by France to UNSCEAR, as in the initial study. These reports made it possible to conduct a compr ehensiv e estimation of the gr ound deposition of 33 radionuclides at the time of arrival of fallout after each test, based on measurements of total ground deposition, measurements of total beta-concentration in air, or measurements of exposur e r ate at differ ent locations in FP. 20 r Integrating historical data on population lifestyle related to the period of the tests, which were collected in 2016-2017 using focus-group discussions and k e y informant interviews monitored in each of the 5 archipelagos of FP, in addition to declar ations r eported in self-questionnair es. 14 This dosimetric reconstruction led to an estimation of the mean lifetime thyroid dose among the study subjects of around 4.7 mGy. 21 Doses fr om131I intake r anged up to 27 mGy, while those from intake of short-lived iodine isotopes ( 132I, 133I, 135I ) and 132Te ranged up to 14 mGy. Thyroid doses from external exposur e r anged up to 6 mGy, and those from internal exposure due to 137Cs ingestion did not exceed 1 mGy. Intake of 131I was the main pathway of thyr oid exposur e accounting for 72% of the total dose. 22 Genotyping data DN A w as extracted from saliva samples at the CEPH-Biobank ( Centre d'Etude du Polymorphism Humain, Paris, France ) using an automated salt precipitation method on an Autopur e ( Qia gen ) . Genotyping was performed with a custom-made Infinium OncoArray-500K BeadChip EPITHYR ( Illumina ®) at the Centre National de Recherche en Génomique Humaine ( CNRGH/CEA, Evry, France ) , on an Illumina high-throughput genotyping platform automated according to manufacturer's instructions . T he OncoArra y-500K is described in detail in Amos et al . 2017 23 and contains 499 170 SNPs. An additional 13 759 custom markers wer e incor por ated, based on prior e vidence of association with r ele v ant biological pathways for thyroid cancer. 16 In total, 289 cases and 418 controls with a sufficient amount of DNA, from the first study and the extension, were genotyped. A threshold of 5% was applied for missing call rate per SNP and 10% per individual. Furthermor e, thr ee individuals wer e r emov ed because they were duplicates. Hardy-Weinberg equilibrium ( HWE ) was also assessed per SNP among controls, using a Fisher exact test with a p-v alue thr eshold of 10 −5 . Onl y SNPs with minor allele frequency ( MAF ) above 5% were considered in the analyses to ensure sufficient po w er. The quality control of the genotyping data was performed with Plink2 pr ogr am ( v ersion Alpha 2.3 ) . 24 As our genotyping array was originally designed based on the human assembly hg19, the SNPs positions wer e ma pped to the r efer ence genome hg38 using the Liftover tool from UCSC with default par ameters 25 ( av ailable her e: https:// genome.ucsc.edu/ cg i-bin/hgLiftOver) . The data was phased with Eagle v2.4.1 26 according to the 1000 Genomes Project phase 3 and the hg38 assembly as r efer ences, with a maxim um number of conditioning ha plotypes fixed at 10 000. Then imputation was performed with IM-PUTE2 27 , 28 for hits regions on c hr omosomes 6, 10, and 17 using the 1000 Genomes Project phase 3 ( accessed December 2019 ) 29 and the Human Genome Diversity Project ( accessed December 2019 ) 30 datasets as r efer ence panels. We imputed the region 500kb on both sides of the most significant SNP position for each of the three loci of interest. To avoid strand issues, we excluded palindromic SNPs. After imputation, SNPs with an info score < 0.85 were excluded.

Global Admixture analyses
To study genetic structure in our population we merged our data with published data from worldwide populations previously compiled and available from https://reichdata.hms.harvard.edu/ amh _ repo/( v29.1 ) . We filtered populations in the dataset keeping: Africans ( Yoruba ) , East-Asians ( Chinese , Vietnamese , and Ja panese ) , Centr al and South Asians ( Baloc hi, Br ahui, and Sindhi ) , North-Eastern Asians ( Yakut and Yukagir ) , other Oceanian populations ( Samoan and P a puan ) , Metr opolitan Fr enc h and ancient South Pacific individuals including three ancient French Polynesians who lived during the eighteenth-century AD. 31 Finally, data from 408 persons from the reference dataset and 56 316 shared SNPs were analyzed.
We first performed a principal component analysis ( PCA ) using the "smartpca" function from Eigensoft v7.2 32 on the study population alone and found a strong genetic structure ( Supplementary  Fig. 1 ) . The first 5 principal components explained 90% of the variance in our dataset. We then ran a PCA including worldwide diversity populations and projected the French Polynesians into the defined space. Based on the results from the PCA, we focused our analysis on three ancestries found in French Polynesia today: local Polynesian ancestry represented by ancient Fr enc h Pol ynesians, East Asian ancestry r epr esented by modern Chinese Han, and European ancestry represented by modern Fr enc h. We performed an analysis using the software ADMIXTURE 33 supervised by using onl y thr ee r efer ence populations, estimating the fr action of these different ancestries in our dataset.

Association analyses
Logistic mixed effects models implemented in the GENESIS R Bioconductor pac ka ge wer e used to test for association between genotype and DTC risk, using a penalized quasi-likelihood a ppr oximation to the generalized linear mixed model. To consider the population structure, GENESIS estimates a kinship matrix that accounts for cryptic and known relatedness between subjects. 34 This kinship matrix was deriv ed fr om a set of LD-pruned genotyped SNPs ( at threshold R 2 = 0.1 ) and was included as a random effect in the logistic mixed model. The three first principal components corresponding to our three founding populations were included as fixed-effect co variates . T he other covariables included in the model ar e a ge ( continuous ) , sex ( binary ) , thyroid cancer family history ( binary ) , goiter history ( binary ) , r adiother a py history ( binary ) , Mururoa site workers ( binary ) , BMI ( continuous ) , height ( continuous ) , educational le v el ( primary sc hool, secondary school ) , smoking status ( binary ) , among women the number of pregnancies longer than 7 months ( 0, 1-3, > 3 ) , dietary selenium and iodine intake during adulthood and childhood ( continuous ) and the internal radiation dose deliv er ed to the thyroid gland before the age of 15 ( continuous ) . The same model was also tested after the exclusion of the cases with non-inv asiv e micr ocarcinoma ( < 10 mm, n = 173 ) , and with an interaction term between the estimated dose deliv er ed to the thyr oid gland befor e the a ge of 15 and the genotype.
Besides, in an attempt to replicate former results from Maillard et al. , 11 we used an unconditional logistic model adjusted on the same variables as abo ve . Benjamin Hochberg FDR procedure was used for multiple testing correction 35 for the GWAS analysis, to better take into account the non-independence of the SNPs.

Local ancestry Inference using RFMix
Local ancestry inference, that infers the regional ancestral origin of a c hr omosomal segment in admixed populations, was performed using RFMix version 2.03, 36 with default parameters . T hen, the RFMix local ancestry calls were converted to Plink TPED files, where the encoded dosage is defined as having 0, 1, or 2 copies for each of the ancestry populations at a particular local ancestry segment.

Admixture mapping
The main aim of this analysis is to test for association between the local ancestry at a genomic region and DTC. These tests were performed using the linear mixed-effect models implemented in the EMMAX softwar e pac ka ge . T hese models included local ancestry count as the main effect and a Balding-Nichols kinship matrix as a random effect. This step was run for each of the three identified ancestries: European, South Asian, and East Asian. We adapted the correction for multiple testing from Gao et al ., 37 to be used in the estimation of the number of effective tests in a Bonferroni corr ection for m ultiple testing. 38 This number of effective tests is the sum of PC explaining more than 99.5% of genetic variance per c hr omosome . T he significance threshold for our admixtur e ma pping analysis is 1.13 × 10 −4 based on 442 effective tests . T his appr oac h has already been adopted in a similar study. 39

Single-SNP heritability and posthoc power calculation
Based on the summary r esults fr om the association analysis we estimated the heritability and the post-hoc po w er for each tested SNP. The heritability was calculated as the proportion of phenotypic variance explained by a given SNP. The following formula was used to assess single-SNP heritability h 2 from genome-wide association study r esults, wher e i is a given polymorphism: 40 where MAF i is the minor allele frequency for the SNP i , and allelic causal effect size is β i , β i .Se is the standard error of β i and N is the population size . T hen, based on h 2 values , we estimated the observed po w er ( post-hoc po w er ) for each tested SNP i : Observedpo w e r i = 1 − G ( t, λ, 1 ) where G is the cumulative distribution function of non-central Chi-square distribution, and the non-centrality parameter ( λ) depends on sample size ( N ) and the proportion of phenotypic variance that is explained by the SNP i and calculated above ( h 2 i ) : Two different values of significance threshold α were considered: α 1 = 0.05/total number of tested SNPs = 1.66 × 10 −7 and α 2 = 0.05/number of independent SNPs tested ( D' < 0.7 in windows of 1000 SNPs ) = 1.11 × 10 −5 .

Demographic and epidemiologic characteristics
At the end of quality control of the genotyping data, the present study included 283 cases and 418 controls and 300 908 SNPs were available for the analysis ( study flowchart in Supplementary Fig.  1 ) . Main c har acteristics of all cases and contr ols ar e described in Table 1 : more than 80% of cases were women. The mean age of the thyr oid cancer dia gnosis w as 44 y ears, and 56% of the cases were smokers. Anthropometric factors ( BMI and body height ) and dietary intake of selenium and iodine during c hildhood wer e found to significantly differ between cases and controls . T he distributions of the thyroid doses for the cases and for the controls are presented in the Supplementary Fig. 2.

Population structure
PCA of the Fr enc h Pol ynesian population r e v ealed a significant genetic structure in the population ( Supplementary Fig. 3 ) . When the Fr enc h Pol ynesian population was pr ojected on a PCA defined b y w orldwide populations w e identified tw o ancestry clines showing a group of Fr enc h Pol ynesians that wer e dr awn to w ar ds East Asians compared to the position of ancient Polynesians while a second group was drawn to w ar ds Europeans ( Fig. 1 A & B ) . Our admixture test confirmed East Asian-related and Europeanr elated admixtur e in Fr enc h Pol ynesia ( Fig. 2 , Supplementary Fig.  4 ) . PCA results including African and other Asian ethnicities are not shown.

Association and admixture mapping results
The GWAS allo w ed us to identify 3 loci as highlighted in the Manhattan plot presented in Fig. 3 A. The genomic inflation factor λ was = 1.03, confirming a lack of systematic genetic bias between our cases and controls ( Fig. 3 B )  A third susceptibility loci was observed at 17q21.32, which contain a large linkage disequilibrium region, mostly in the SKAP1 gene . T he minor allele of the lead SNP rs7211607 ( tested allele frequency = 0.16, genotypic count detailed in Supplementary Table 1 ) in the SKAP1 region had an OR allelic = 2.37, ( p-value of 7.19 × 10 −7 and FDR of 0.026 ) . Same associated r egions wer e found to have the lo w est p-values when excluding cases with noninv asiv e micr ocarcinoma fr om the anal yses ( Results not shown ) . The lead best SNP in this analysis was rs28390243, but not significant ( OR allelic = 2.04, FDR = 0.49 ) .
Results from local imputation in each of the three regions on c hr omosomes 6, 10, and 17 found that the original lead SNPs in the three regions remained the most significant.
The estimates of the radiation doses before the age of 15 years to the thyroid gland were low, with a maximal dose of 22 mGy among cases, and 13 mGy among controls.
In the interaction analysis, none of the SNPs reached the significance threshold. The most significant was observed in many SNPs ( in LD: R 2 between 0.85 and 1, D' = 1 ) located in 21q21.1, all in the MIR99AHG region. These results did not change when we excluded individuals with a r adiother a py history and when noninv asiv e micr ocar cinoma cases w ere excluded.
We compared our findings with results from our previous study that were based on a subpopulation the current study ( 137 cases and 218 controls ) ( Maillard et al. 11 ) . We failed to find the same significant interaction in rs944289. Ho w ever, the main effect of the genotype "T/T" had the same direction as in Maillard et al . findings 11 ( OR T/T genotype = 1.45, 95% Confidence Interval ( CI ) = 0.26-7.98 ) ( Allelic ORs are listed in Supplementary Table  2 ) .
According to our significance threshold for the admixture mapping anal ysis, onl y two loci in c hr omosomes 14 and 15 wer e found to be significantly associated with DTC risk in a population of East-Asian ancestry. The locus 15q14 ( Table 3.

Discussion
The main purpose of this study was to explore the role of genetic factors in DTC occurrence in the FP population. Compared to the GLOBOC AN statistics , the sex ratio in thyroid cancer cases worldwide is 3.87, which is similar to our study. 41 Smoking habits are mor e fr equent in our study population compar ed to the gener al population in FP ( 36.5% of men and 41.4% of women smokers ) . 42 Our r esults fr om the GWAS suggest potential DTC susceptibility loci at 6q24.3, 10p12.2, and 17q21.32. While all these regions hav e been pr e viousl y linked to cancer, none of them was pr e viously associated with DTC risk.
At locus 6q24.3, all v ariants wer e in non-coding regions, the closest genes were GRM1 , RAB32, and ADGB . These genes have been associated with the risk and prognosis of ovarian cancer, 43 gastric cancer 44 and breast cancer. 45 In addition, this region is also associated with autoimmune thyroiditis 46 and in the regulation of pr olifer ation and invasion of thyroid cancer cells 47 and with familial lung cancer risk. 48 The lead SNP ( rs1358902 ) in c hr omosome 10 is located near a promoter for the PIP4K2A gene in the 10p12.2 locus . T his locus has been associated with childhood B-cell acute lymphoblastic leukemia occurrence by a previous GWAS, 49 in the European ancestry population, and in multiple other populations including Hispanic , African, and Chinese . 50 , 51 Variants in this gene were also found to be associated with acute lymphoblastic leukemia prognosis. 52 , 53 The results from a cross-ancestry meta-analysis identified an association between a variant in the PIP4K2A gene region and breast cancer susceptibility in European and Asian populations. 54 Although the role of PIP4K2A is not well-defined and mainly depends on the biological context, this gene has r ecentl y been suggested to be involved in o xidati v e str ess, cellular senescence, and tumor growth. 55 , 56 It has also been suggested to play an oncogenic role and was found to be frequently overexpressed in multiple types of leukemia and some types of solid cancers in cancer cohorts including TCGA. 55 , 56 Another signal was identified on c hr omosome 17 in the region of SKAP1 ( SRC kinase-associated phosphoprotein 1 gene ( Fig. 3 A ) ) which had been previously reported to be associated with ovarian cancer in the European and Chinese Han populations. 57 , 58 The expr ession le v els of these genes wer e found to be associated with ovarian, endometrial, and colorectal cancer risk. [59][60][61] The 17q21.32 locus is additionally associated with endometrial, prostatic, and renal cancers . [62][63][64][65] T he biological function of SKAP1 related to these cancer types is not yet well elucidated.
The observed po w er under the α 1 assumption varied from 39% to 51% for the three leading SNPs of the region. This po w er is between 78% and 81% if we consider α 2 . In view of the small size of the general PF population ( almost 300 000 ) , the present study included an important proportion of the total DTC cases occured in FP.
In r adiation-r elated DTC studies, onl y two v ariants in FOXE1 and ATM were associated with the disease in two different populations. 66 In FOXE1 , only the association of rs965513 with radiationrelated DTC could be replicated in a population exposed to higher le v els of r adiation caused b y Chornob yl disaster. 67 Notably, the same variant rs1801516 in ATM could have an opposite effect between the FP population 11 and the Belarusian population. 68 This could highlight the importance of taking into account ancestry when considering the genetic susceptibility to DTC, and the importance of further exploration of the concerned genetic regions, as the causal variant could be a different one in the same region.
We did not observe a significant interaction between thyroid radiation dose due to atmospheric nuclear tests and genetic factors, but this result was expected given the very low level of exposure resulting in a limited statistical po w er for testing this interaction.
The interaction found by a former study with different doses estimation and genotyping technology 11 was not replicated in the current study. This difference could be caused by the different thyroid IR doses between both studies, the partially shared study population, and the difference between genotyping technologies, as about 10% of the genotyping results are different between the first and the second studies.
A mor e r ecent study focused on the genomic profile of r adiation-r elated pa pillary thyr oid cancer after the Chornobyl accident, which included normal and tumor tissues, with wholegenome, mRN A, and microRN A sequencing; DN A methylation pr ofiling; and genotyping arr a ys . T he germline DNA analyses yielded three possible associations with the radiation dose of variants in SMAD3, TERT, and 9q22.33. 69 Ho w e v er, none of these genes/loci were found in our study. T hus , our r esults fr om genetic susceptibility analysis identified novel genes related to DTC with many of them linked to hormonal-dependent cancers. Results fr om admixtur e ma pping highlighted two r egions: 15q14 ( 34 830 119-34 903 911 ) and 14q21.2 ( 43 863 839-43 935 211 ) to be associated with DTC risk in a population with East Asian ancestry. These results should be taken with precaution because of the limited sample size of the study and the high admixture of this population. Altogether, results from this analysis show that the outline of the genetic regions associated with DTC risk is relativ el y differ ent depending on the ethnic origins.
The significant associations identified in our study need to be replicated and functionally validated to understand the biological mec hanisms involv ed. An incr ease in the sample size in the future will permit the exploration of further orders of interaction such as the interaction between gene sets.
In addition to genetic susceptibility results, our study identified significant admixture in FP related to Asian and European populations. Ho w e v er, the dominating ancestry in FP is related to ancient local Fr enc h Pol ynesians. Our study suggests that our results were not biased by population structure or admixture, especially because of the use of a mixed model. Futur e explor ation of ancestry in FP could enable the study of the potential correlation between geogr a phic r egions ( arc hipela go and atolls ) , ancestry, and DTC. A larger study population could help to better understand these potential interconnections, especially comparing to classic GWAS study population sizes, though increasing the sample size may be difficult due to the small size of the Fr enc h Pol ynesian population and ther efor e the r elativ el y small number of DTC cases in this population. Besides, the use of non-ethnically ada pted micr oarr ays str ongl y r educed the number of anal yzable SNPs and genomic regions, as these micr oarr ays ar e designed for European population and doesn't cover the same variants as in FP population. Exploring genetic factors of DTC in non-European populations will allow to propose better polygenic risk scores by identifying other variants.

Conclusion
The Fr enc h Pol ynesian population has genetic, geogr a phic, and cultur al specificities. Her e, we attempted to study genetic DTC risk factors while accounting for genetic structure in an admixed population. Our results are suggestive of a role of three loci: 6q24.3, 10p12.2, and 17q21.32. The genes present at these loci were not previously associated with DTC, but rather with other hormone-dependent cancers or psyc hiatric disorders. Giv en that the micr oarr ay used for genotyping in the present study was designed for the population of European ancestry, it would be more adapted to perform whole genome sequencing in the Polynesian population.