Investigation of common genetic risk factors between thyroid traits and breast cancer

Abstract

Breast cancer (BC) risk is suspected to be linked to thyroid disorders, however observational studies exploring the association between BC and thyroid disorders gave conflicting results. We proposed an alternative approach by investigating the shared genetic risk factors between BC and several thyroid traits. We report a positive genetic correlation between BC and thyroxine (FT4) levels (corr = 0.13, p-value = 2.0 × 10⁻⁴) and a negative genetic correlation between BC and thyroid-stimulating hormone (TSH) levels (corr = −0.09, p-value = 0.03). These associations are more striking when restricting the analysis to estrogen receptor-positive BC. Moreover, the polygenic risk scores (PRS) for FT4 and hyperthyroidism are positively associated to BC risk (OR = 1.07, 95%CI: 1.00–1.13, p-value = 2.8 × 10⁻² and OR = 1.04, 95%CI: 1.00–1.08, p-value = 3.8 × 10⁻², respectively), while the PRS for TSH is inversely associated to BC risk (OR = 0.93, 95%CI: 0.89–0.97, p-value = 2.0 × 10⁻³). Using the PLACO method, we detected 49 loci associated to both BC and thyroid traits (p-value < 5 × 10⁻⁸), in the vicinity of 130 genes. An additional colocalization and gene-set enrichment analyses showed a convincing causal role for a known pleiotropic locus at 2q35 and revealed an additional one at 8q22.1 associated to both BC and thyroid cancer. We also found two new pleiotropic loci at 14q32.33 and 17q21.31 that were associated to both TSH levels and BC risk. Enrichment analyses and evidence of regulatory signals also highlighted brain tissues and immune system as candidates for obtaining associations between BC and TSH levels. Overall, our study sheds light on the complex interplay between BC and thyroid traits and provides evidence of shared genetic risk between those conditions.

breast cancer, thyroid cancer, cross phenotype association, pleiotropy, thyroid traits

Introduction

Breast cancer is the most common cancer and the leading cancer-related cause of mortality in women worldwide [1]. Breast cancer risk has been shown to be influenced by age, hormonal and reproductive factors, lifestyle, and environmental factors as well as by genetic factors [2]. It has also been suggested that malignant or benign thyroid disorders may be linked to breast cancer risk [3]. Indeed, those diseases are predominant in women and the breast and thyroid tissues show some similarities as both glands are regulated by the hypothalamic-pituitary axis, and thyroid follicular cells and breast lactating cells store iodine through sodium iodine symporter-mediated iodine [4]. Thyroid peroxidase (TPO) which is an enzyme that plays an important role in thyroid hormone biosynthesis was also found to be expressed in breast tissue, and a higher prevalence of anti-TPO antibodies (TPOAb) among breast cancer patients compared to healthy controls suggests a link between autoimmune thyroid disorders and breast cancer [5]. It has also been shown that thyroid hormones such as triiodothyronine (T3) and thyroxine (T4) have a proliferating effect on follicular and papillary thyroid cells, but they also activate the estrogen receptors in mammary gland tissue and therefore may play a role in breast and thyroid cancers progression [4]. These intrinsic biological relationships could therefore be at the origin of common mechanisms between breast cancer and thyroid disorders [4].

The most common thyroid disorders in iodine-replete areas are autoimmune thyroid diseases (AITD), such as Graves’ disease and Hashimoto’s disease which are the most common cause of respectively hyperthyroidism and hypothyroidism [6]. Hyperthyroidism and hypothyroidism are characterized by respectively high and low levels of T3 and T4 (that could lead to respectively decreased or increased levels of thyroid-stimulating hormone (TSH)). However, observational studies reported conflicting results on the association between thyroid disorders and breast cancer risk. A meta-analysis on the association between thyroid dysfunctions and different cancers published before 2019 reported that hyperthyroidism was associated to a higher risk of thyroid and breast cancers, while hypothyroidism was not associated to breast cancer but to a higher risk of thyroid cancer [7]. Another recent meta-analysis [8] including 21 studies (with only a few overlapping studies with the previous one) confirmed the positive association between breast cancer risk and hyperthyroidism and also reported an increased risk of breast cancer associated to higher levels of TPOAb and history of thyroid cancer, as well as an inverse association with hypothyroidism. The most recent meta-analysis included 18 studies on the association between different thyroid disorders and breast cancer risk, and showed that goiter, autoimmune thyroiditis and Graves’ disease are associated to an increased risk of breast cancer while no association was found with hyperthyroidism or hypothyroidism [9]. For some of these associations, a significant heterogeneity was also reported which may be explained by the different definitions used for thyroid dysfunctions and the lack of information on potential important confounders in most studies, such as the use of thyroid medication or other treatment information [7].

On the other hand, several epidemiological studies reported that patients with prior breast cancer are more likely to develop thyroid cancer and vice versa [4, 10]. However, whether this phenomenon is an association or a cause-effect relation is not well understood yet. This association does not seem to be explained by surveillance bias or to be the consequence of the treatment but rather suggest common etiologies or biological mechanisms for breast and thyroid cancers.

Indeed, the potential link between thyroid disorders and breast cancer may also be due to common suspected etiological factors such as obesity, hormonal factors, iodine intake, radiation exposure, history of other diseases (such as diabetes and cardiovascular diseases) [4, 8, 10, 11] or genetic factors.

In this study, we proposed to investigate the shared genetic risk factors between breast cancer and thyroid traits (thyroid cancer, hyperthyroidism, hypothyroidism, levels of free T4 (FT4), and TSH levels) at different genomic scales (polymorphisms, gene and pathway levels) in order to help to elucidate the common mechanisms between those traits.

Material and methods

We used summary statistics from genome-wide association studies (GWAS) on breast cancer [12], thyroid cancer [13], thyroid dysfunction, and thyroid hormone levels [14, 15]. Detailed sample size and number of single nucleotide polymorphisms (SNPs) are indicated in the Supplementary Table S1. It should be mentioned that there was no sample overlap between breast cancer studies and any other thyroid traits GWAS. All GWAS datasets use the genome build hg19.

Breast cancer

Summary statistics of the breast cancer GWAS were based on 45 494 female controls and 61 282 female cases of European ancestry from 68 studies collaborating within the Breast Cancer Association Consortium (BCAC, version 2017) [12]. The data also included summary statistics according to ER status of breast cancer (38 197 ER+ cases, 9655 ER− cases). We used the summary statistics derived from the Infinium OncoArray-500 K BeadChip (Illumina), which is the same array used in EPITHYR. Information was available for 11.8 million SNPs after imputation.

Thyroid cancer

Summary statistics of the thyroid cancer GWAS were obtained from the Epidemiology of thyroid cancer consortium (EPITHYR) [13], which included 1957 controls (1508 women and 449 men), and 1551 cases of differentiated thyroid cancer (1276 women and 275 men) of European ancestry from seven case-control studies, which were genotyped using the Infinium OncoArray-500 K BeadChip. After imputation, information on 9.5 million SNPs was available. Summary statistics were available for the overall population and for women only (see Supplementary Table S1 for sample sizes).

TSH, FT4, hypothyroidism and hyperthyroidism

Summary statistics from GWAS on circulating levels of FT4, hypothyroidism and hyperthyroidism were obtained from the ThyroidOmics consortium [14]. This consortium included data from 22 cohorts for TSH levels, and 19 cohorts for FT4 levels [14]. GWAS results were available for TSH and FT4 levels for 54 288 and 49 269 individuals, respectively for European ancestry. These results were also available for respectively 27 380 and 30 097 women. Individuals with TSH levels above or below the cohort-specific reference range were classified as hypothyroidism cases (N = 3440) or hyperthyroidism cases (N = 1840) [14]. The control group consists of subjects with a TSH level within the cohort-specific reference range (N = 49 983). The exclusion criteria for all analyses were non-European ancestry, use of thyroid medication or previous thyroid surgery [14]. Information on 7.9 million SNPs were available.

For TSH levels, we used summary statistics for 22.4 million SNPs from a recent meta-analysis [15] that included up to 119 715 participants from the ThyroidOmics consortium, the Nord-Trondelag Health Study (HUNT study, N = 55 342, 54.4% women), and the Michigan Genomics Initiative (MGI, N = 10 085, 53.4% women).

Statistical analysis

We calculated the genome-wide correlation of the genetic effects to assess whether there was a global tendency for SNPs to have similar impacts between breast cancer and thyroid traits. Genetic correlation was estimated using the Linkage disequilibrium score regression (LDSC) [16] and a high-definition likelihood (HDL) method [17]. HDL was shown to give better estimations of the correlation, and smaller standard deviation than the LDSC [17]. We used both methods to estimate the genetic correlation between breast cancer and circulating levels of TSH and FT4, hypothyroidism and hyperthyroidism. As similar results were found using both methods, only estimations using HDL were shown. The LDSC method was used to estimate the genetic correlation between breast cancer and thyroid cancer because the limited sample size in EPITHYR did not allow to run HDL. GWAS polymorphisms were pruned using the default filters (imputation quality > 0.9, minor allele frequency (MAF) > 1%). We additionally removed palindromic sequence polymorphisms and non-bi-allelic SNPs.

We computed the PRS for breast cancer using a published list of 313 variants associated with breast cancer susceptibility [18], and tested its association with thyroid traits using the GWAS summary statistics and the function grs.summary() of the gtx R package [19]. After filtering the palindromic variants, 291 SNPs were kept to calculate the breast cancer PRS. The provided SNP lists and betas were separated into three categories: ER−, ER+, and all. The SNPs were the same between the three categories, but the weights associated with each SNP differed. Similarly, we calculated PRS for thyroid cancer (TC), TSH, FT4, hypo- and hyperthyroidism in the breast cancer GWAS. We used the list of SNPs from Liyanarachchi et al. [20] for the calculation of PRS for TC (10 SNPs). For TSH (59 SNPs), FT4 (31 SNPs), hypothyroidism (8 SNPs) and hyperthyroidism (8 SNPs), we used the list of SNPs included as instrumental variables in Yuan et al. [21]. A PRS was considered to be significantly associated to a trait if the p-value was lower than 0.05.

We investigated cross-phenotype association between breast cancer and each thyroid trait separately at the SNP level using PLACO [22]. More specifically, we tested the following pairs of traits: breast cancer with thyroid cancer, breast cancer with thyroid FT4 levels, breast cancer with TSH levels, breast cancer with hyperthyroidism and breast cancer with hypothyroidism. Each pair was studied for the three different subgroups for breast cancer: all, ER+ and ER−. The PLACO method tested the null hypothesis H0: at most one trait is associated with the genetic variant versus the alternative hypothesis H1: both traits are associated with the genetic variant. Only SNPs with a MAF > 1% and overlapping between all datasets were kept. For each SNP, the effect allele was harmonized between the datasets and Z-scores were calculated. SNPs with a Z² > 80 were removed as recommended for the use of PLACO, as extremely large effects can lead to show spurious signals. The total number of SNPs that was considered for PLACO analysis was 3 937 601 (for overall BC), 3 937 743 (for ER+ BC), and 3 937 789 (for ER− BC). After running PLACO, significant SNPs were defined as having a p-value P.placo lower than 5 × 10⁻⁸. We used the SNP2GENE function from FUMA [23] to clump all significant SNPs in a ± 500 kb region into a single SNP (lead SNP) using a linkage disequilibrium threshold of r² > 0.2.

We identified genes for which boundaries were ± 20 kb from this lead SNP to perform gene-set enrichment analyses (GSEA) using FUMA [23]. We considered two groups of genes identified by PLACO for analysis in GSEA: (1) genes associated to breast and thyroid cancers and (2) genes associated to breast cancer and thyroid hormone traits (hypothyroidism, hyperthyroidism, FT4 levels, TSH levels). For each group, we used the list of significant genes and performed GENE2FUNC function in FUMA tool using all genes as background genes, and with default values for the parameters (Ensembl version = v92, multiple test correction method = Benjamini-Hochberg FDR, Maximum adjusted P-value < 0.05, Minimum overlapping genes ≥2, Exclude the MHC region = Disabled).

To find evidence of a single causal variant among the loci highlighted by PLACO, we ran a Bayesian colocalization test of the breast cancer and thyroid traits summary statistics on all SNPs located at ±200 kb around the lead SNPs for each locus, using the R package coloc (v3.2-1) (function coloc.abf(), with defaults parameters) [24]. This test computes five posterior probabilities for each region: PP0 (posterior probabilities of no association with either disease), PP1 (posterior probabilities of association with breast cancer but not the thyroid trait), PP2 (posterior probabilities of association with the thyroid trait but not breast cancer), PP3 (posterior probabilities of association with both breast cancer and the thyroid trait, due to different causal SNPs), and PP4 (posterior probabilities of association with both breast cancer and the thyroid trait, due to the same causal SNP). As suggested in Ray et al. [22], we defined convincing evidence of a single causal SNP driving pleiotropic association at a locus if it shows PP3 + PP4 ≥ 0.9 and PP4/PP3 ≥ 3.

In order to characterize the SNPs with the highest causal PP4, we manually explored availability of any genes that are likely to be functionally implicated by the variant using Open Targets Genetics platform (https://genetics.opentargets.org/) that calculates a variant to gene (V2G) prediction score for every pair of variant-gene As a brief definition, to calculate a V2G score for each variant, the pipeline integrates evidences across 4 different data types: 1) molecular phenotype quantitative trait loci (QTL) experiments (expression (eQTLs), protein (pQTLs), and splicing (sQTLs)), 2) chromatin interaction experiments, promoter capture Hi-C (PCHi-C), 3) in silico functional predictions such as the variant effect predictor (VEP) from Ensembl, and 4) distance between the variant and each gene's canonical transcription start site (TSS) [25]. Then, the scores among these four data sources are harmonized and a weighting step is also provided. Finally, all scores are aggregated based on the mean weighted quantile approach and an overall V2G score is calculated for each variant-gene pair.

Results

Comparison of GWAS results

We first compared directly SNPs that reached genome-wide significance (p-value ≤ 5 × 10⁻⁸) in each GWAS independently using volcano plots and we observed that only breast cancer and thyroid cancer shared 29 genome-wide significant SNPs (Supplementary Figs S1 and S2). These SNPs were all located at the locus 2q35, within the non-coding RNA gene DIRC3.

Genetic correlation

Genome-wide correlations of the effects between breast cancer and thyroid traits are presented in Table 1 and the estimated genetic heritability are presented in the Supplementary Table S2. We observed a negative but not significant genetic correlation between breast and thyroid cancers. We also highlighted a significant positive correlation between breast cancer and FT4 levels (corr = 0.13, p-value = 2.0 × 10⁻⁴), which remained significant only for estrogen receptor-positive (ER+) breast cancer while stratifying by ER status (corr = 0.14, p-value = 2.8 × 10⁻⁵). Similar results were found while considering GWAS summary statistics on FT4 levels restricted to women. We also observed a negative correlation between breast cancer and TSH levels, which was significant only for ER+ breast cancer (corr = −0.09, p-value = 0.03) (Table 1). Since we estimated a heritability close to zero for hyperthyroidism and hypothyroidism (Supplementary Table S2), we did not compute their genetic correlation with breast cancer.

Table 1

Genome-wide genetic correlation between breast cancer (all breast cancer cases, estrogen receptor positive breast cancer, and estrogen receptor negative breast cancer) and thyroid traits.

	BC ALL				BC ER+				BC ER−
Trait	SNP	gencor	se	p-value	SNP	gencor	se	p-value	SNP	gencor	se	p-value
TC	1 038 703	−0.06	0.13	0.63	1 038 731	−0.06	0.12	0.64	1 038 746	−0.07	0.22	0.74
TC Women	1 038 529	−0.10	0.17	0.57	1 038 556	−0.08	0.16	0.62	1 038 570	−0.15	0.28	0.58
FT4	970 900	0.13	0.03	2.0E−04	970 900	0.14	0.03	2.8E−05	970 900	0.04	0.05	0.42
FT4 Women	970 675	0.10	0.05	0.04	970 675	0.11	0.05	0.03	970 675	0.02	0.07	0.80
TSH	968 203	−0.08	0.04	0.07	968 203	−0.09	0.04	0.02	968 203	−0.03	0.07	0.66

For thyroid cancer (TC), the LD score method was used while for the others the HDL method was used. BC ALL: all breast cancer cases, BC ER+: estrogen receptor positive breast cancer, BC ER−: estrogen receptor negative breast cancer, TC: thyroid cancer, FT4: levels of free thyroxine, TSH: thyroid-stimulating hormone levels, gencor: genome-wide genetic correlation, se: standard error.

Analysis of polygenic risk scores (PRS)

We evaluated the association between PRS for each thyroid trait with breast cancer risk. We observed that the PRS for FT4 was significantly positively associated to breast cancer risk (Odds Ratio (OR) = 1.07, 95% confidence interval (CI): 1.00–1.13, p-value = 0.03) while an inverse association was reported with the PRS for TSH (OR = 0.93, 95% CI: 0.89–0.97, p-value = 2.0 × 10⁻³) (Table 2a) in accordance with the results of genetic correlations (Table 1). Similar results were observed for ER+ and non-significant results were reported for ER− breast cancer. We found that the PRS for hyperthyroidism was significantly positively associated to breast cancer risk (OR = 1.04, 95% CI: 1.00–1.08, p-value = 0.04). We did not find any significant association between PRS for thyroid cancer or hypothyroidism and breast cancer risk (Table 2a). On the other hand, we evaluated the associations between the PRS for breast cancer with each thyroid trait, but none were significant (Table 2b).

Table 2

Polygenic risk score (PRS) association.

a. The associations between PRS of thyroid cancer, FT4 and TSH levels, hyperthyroidism, and hyperthyroidism, and breast cancer risk
		BC All			BC ER+			BC ER−
PRS for	nb SNPs	OR	95% CI	pval	OR	95% CI	pval	OR	95% CI	pval
TC	6	1.00	0.98–1.01	0.97	0.99	0.95–1.03	0.74	0.99	0.93–1.05	0.62
FT4	24	1.07	1.00–1.13	4.7E−02	1.08	1.00–1.17	2.8E−02	1.01	0.90–1.14	0.83
TSH	55	0.93	0.89–0.97	2.0E−03	0.93	0.88–0.99	3.0E−03	0.99	0.92–1.07	0.79
Hyperthyroidism	6	1.04	1.00–1.08	3.8E−02	1.05	1.01–1.09	9.5E−03	1.00	0.94–1.06	0.94
Hypothyroidism	8	0.97	0.93–1.00	0.14	0.98	0.94–1.02	0.45	0.96	0.91–1.02	0.21
b. The association between BC PRS and thyroid disorders
		PRS for BC All			PRS for BC ER+			PRS for BC ER−
Thyroid disorder	nb SNPs	OR	95% CI	pval	OR	95% CI	pval	OR	95% CI	pval
TC	272	0.93	0.83–1.05	0.23	0.93	0.83–1.05	0.21	0.99	0.86–1.14	0.87
FT4	215	1.02	1.00–1.04	0.10	1.02	1.00–1.04	0.09	1.01	1.00–1.03	0.30
TSH	233	1.00	0.98–1.02	0.68	1.00	0.98–1.02	0.59	1.00	0.98–1.02	0.62
Hypothyroidism	215	0.99	0.92–1.07	0.79	0.99	0.93–1.05	0.74	0.99	0.92–1.07	0.88
Hyperthyroidism	215	1.04	0.94–1.15	0.35	1.04	0.96–1.12	0.35	1.04	0.94–1.15	0.40

BC ALL: all breast cancer cases, BC ER+: estrogen receptor positive breast cancer, BC ER−: estrogen receptor negative breast cancer, nb SNPs: number of SNPs, TC: thyroid cancer, FT4: levels of free thyroxine, TSH: thyroid-stimulating hormone levels, pval: p-value, OR: odds ratio, CI: confidence interval.

Cross-phenotype association at the SNP-level

We performed a cross-phenotype association analysis between breast cancer and each thyroid trait identifying both similar and contrasting effects for each 2 by 2 comparison using PLACO (Supplementary Fig. S3). We reported the significant loci in Table 3 and indicated if those loci were novel or previously reported (either by direct cross-phenotype analysis or by comparing results from individual trait analysis).

Table 3

Lead SNPs from each of the pleiotropic loci between breast cancer and thyroid traits identified by PLACO and results of the coloc colocalization posterior probability.

	Lead SNP from PLACO									coloc analysis of + −200 kb around lead SNP
										overall probabilities for the region					SNP with the highest causal probability
Thyroid traits	Locus number	SNP	position (hg19)	locus (hg19)	Novel Locusa	Nearest Gene	P.placo	Effect direction	SNP PP4	Number of SNPs	PP3	PP4	PP3 + PP4	PP4/PP3	SNP	position	P.placo	SNP PP4
Thyroid cancer	1	rs3821098	2:218292141	2q35	No	DIRC3	2.2E−16	concordant	0.31	725	0.01	0.99	1.00	115.0	Same as lead SNP
	2	rs17349706	3:27124261	3p24.1	Yes	NEK10	6.4E−12	discordant	0.51	438	0.03	0.27	0.30	7.9	Same as lead SNP
	3	rs77275268	6:151969198	6q25.1	Yes	ESR1	2.7E−08	concordant	0.47	735	0.03	0.14	0.17	5.1	rs9371545	6:151969740	4.1E−08	0.47
	4	rs189268208	8:76360637	8q21.11	Yes	HNF4G	9.6E−09	discordant	0.15	693	0.04	0.39	0.43	10.9	rs72658081	8:76328331	3.0E−08	0.31
	5	rs4752568	10:123330541	10q26.13	Yes	FGFR2	1.1E−08	concordant	0.99	614	0.04	0.17	0.21	4.2	Same as lead SNP
	6	rs1973765	11:1898664	11p15.5	Yes	LSP1	2.6E−11	discordant	0.53	762	0.03	0.63	0.66	20.1	Same as lead SNP
	7	rs1391721	12:115836132	12q24.21	Yes	LOC105370003	1.4E−08	concordant	0.54	764	0.03	0.12	0.15	3.9	Same as lead SNP
FT4 levels	1	rs7593049	2:135653355	2q21.3	Yes	ACMSD	2.3E−08	concordant	0.05	282	0.22	0.34	0.57	1.5	rs4954192	2:135632981	3.2E−08	0.07
	2	rs4571035	2:217861809	2q35	No	LOC101928278	1.9E−08	discordant	0	552	0.01	0.01	0.02	0.5	rs13034362	2:217929007	0.26	0.44
	3	rs73065147	3:46894939	3p21.31	Yes	MYL3	2.3E−09	discordant	0.01	427	0.05	0.34	0.40	6.4	rs6787229	3:46889187	3.8E−09	0.47
	4	rs3218020	9:21997872	9p21.3	Yes	CDKN2B-AS1	1.7E−09	discordant	0.05	536	0.06	0.27	0.33	5.0	rs3217992	9:22003223	1.8E−09	0.84
	5	rs525803	9:110923883	9q31.2	Yes	LOC105376214	6.7E−11	concordant	0.57	770	0.03	0.16	0.18	6.4	Same as lead SNP
	6	rs12380852	9:139118673	9q34.3	Yes	QSOX2	1.4E−10	concordant	0.01	403	0.004	0.04	0.05	10.2	rs2274114	9:139091460	7.1E−08	0.48
	7	rs12256551	10:21799726	10p12.31	Yes	SKIDA1	3.1E−08	concordant	0.08	268	0.09	0.58	0.67	6.5	rs7098100	10:21834536	4.4E−08	0.18
	8	rs10995201	10:64299890	10q21.2	Yes	ZNF365	4.0E−08	concordant	0.84	619	0.03	0.05	0.07	1.8	Same as lead SNP
	9	rs187679	16:52522717	16q12.1	Yes	TOX3	1.8E−10	concordant	0.01	436	0.01	0.08	0.09	9.6	rs9926539	16:52528727	1.5E−09	0.17
	10	rs13329835	16:80650805	16q23.2	Yes	CDYL2	5.1E−10	discordant	0.20	697	0.03	0.46	0.49	15.0	Same as lead SNP
	11	rs5750547	22:38546700	22q13.1	Yes	PLA2G6	3.4E−09	concordant	0.08	560	0.14	0.47	0.61	3.5	rs2284063	22:38544298	4.4E−09	0.09
TSH levels	1	rs11584323	1:51457904	1p32.3	No	CDKN2C	6.1E−10	discordant	0.03	315	0.18	0.17	0.34	0.9	rs11583886	1:51451499	8.6E−09	0.32
	2	rs870751	1:61470306	1p31.3	Yes	NFIA	7.1E−09	discordant	0.001	294	0.01	0.01	0.02	0.9	rs384893	1:61606167	0.06	0.53
	3	rs10211546	2:217892397	2q35	No	LOC101928278	1.4E−08	discordant	0.002	530	1.00	0	1.00	0.0	rs13034362	2:217929007	9.1E−05	0.49
	3	rs12989997	2:218266356	2q35	No	DIRC3	1.1E−12	discordant	0.36	734	1.00	0	1.00	0.0	Same as lead SNP
	3	rs9967835	2:217565967	2q35	No	IGFBP5	2.2E−09	discordant	0.01	473	0.09	0.01	0.10	0.1	rs12474719	2:217623723	2.5E−03	0.34
	3	rs888182	2:217571845	2q35	No	IGFBP-AS1	3.1E−09	discordant	0	466	0.09	0.01	0.10	0.1	rs12474719	2:217623723	2.5E−03	0.34
	4	rs1571583	9:4267209	9p24.2	Yes	GLIS3	4.0E−08	discordant	0	1011	0.01	0.01	0.02	0.7	rs7867224	9:4292152	4.3E−05	0.54
	5	rs532436	9:136149830	9q34.2	No	ABO	6.6E−11	concordant	0.03	793	0.03	0.06	0.09	2.5	rs529565	9:136149500	8.6E−10	0.12
	6	rs12825673	12:569945	12p13.33	Yes	B4GALNT3	6.4E−09	discordant	0.01	617	0.02	0.12	0.14	8.2	rs7955258	12:570947	1.0E−08	0.99
	7	rs3184504	12:111884608	12p11.22	Yes	SH2B3	8.1E−09	discordant	0.70	232	0.003	0.25	0.25	75.0	Same as lead SNP
	8	rs12885612	14:37127410	14q13.3	No	PAX9, LOC105370455	1.4E−08	concordant	0.06	639	0.56	0.08	0.64	0.1	rs7149262	14:37136545	1.7E−08	0.09
	9	rs8006015	14:81670558	14q31.1	Yes	GTF2A1, SNORA79	2.2E−10	discordant	0	492	0.06	0.01	0.08	0.2	rs2288497	14:81659731	9.0E−06	0.27
	10	rs3809453	14:105219467	14q32.33	No	SIVA1, LOC107987209	2.7E−15	discordant	0.13	532	0.02	0.96	0.98	49.0	rs12590163	14:105223525	3.6E−15	0.23
	11	rs11639111	15:49749735	15q21.2	No	FGF7, FAM227B	4.7E−12	discordant	0.71	467	0.01	0.27	0.28	22.8	Same as lead SNP
	12	rs76119208	15:91535329	15q26.1	No	PRC1	2.8E−08	discordant	0.13	752	0.01	0.13	0.14	13.3	rs12594752	15:91531995	3.2E−08	0.13
	13	rs1045476	16:4015313	16p13.3	No	ADCY9	4.5E−10	discordant	0.39	512	0.09	0.51	0.59	5.9	Same as lead SNP
	14	rs7499149	16:80648327	16q23.2	No	CDYL2	8.9E−10	discordant	0.15	697	0.02	0.30	0.32	16.3	rs13329835	16:80650805	1.3E−09	0.15
	15	rs199502	17:44862347	17q21.31	No	WNT3, LRRC37A2	3.3E−11	discordant	0.25	289	0.04	0.60	0.64	14.5	rs116956554	17:44699851	4.6E−10	0.40
	15	rs35354512	17:43515927	17q21.31	No	PLEKHM1	4.5E−09	discordant	0.02	230	0.10	0.85	0.95	8.3	rs34363898	17:43515846	5.3E−09	0.02
	16	rs3761446	22:38595483	22q13.1	Yes	MAFF	2.7E−08	discordant	0.05	573	0.07	0.11	0.17	1.6	rs5995543	22:38579026	1.3E−07	0.08
	16	rs55844816	22:40803112	22q13.1	Yes	SGSM3	3.9E−11	discordant	0	406	0.21	0.12	0.32	0.5	rs5995870	22:40921816	4.3E−10	0.05
hypothyroidism	1	rs3184504	12:111884608	12q24.12	No	SH2B3	5.2E−09	discordant	0.12	232	0.004	0.17	0.17	44.3	rs597808	12:111973358	1.3E−08	0.48
	2	rs12936639	17:53235982	17q22	Yes	STXBP4	1.2E−09	discordant	0.01	545	0.03	0.11	0.14	3.5	rs244294	17:53228543	2.0E−07	0.14
hyperthyroidism	1	rs16856873	2:217927278	2q35	Yes	LOC101928278, LOC105373873	1.6E−10	discordant	0.01	535	0.10	0.17	0.27	1.6	rs13034362	2:217929007	2.7E−09	0.46
	1	rs2372966	2:217967719	2q35	Yes	LOC101928278	1.2E−09	concordant	0	542	0.10	0.17	0.26	1.7	rs13034362	2:217929007	2.7E−09	0.46
	2	rs606677	8:128334485	8q24.21	No	CASC8, CASC21	6.2E−10	discordant	0.003	668	0.10	0.14	0.24	1.5	rs587948	8:128341680	1.2E−07	0.20
	3	rs3088440	9:21968159	9p21.3	Yes	CDKN2A, CDKN2A-DT	1.9E−08	concordant	0.04	549	0.05	0.02	0.07	0.3	rs3217992	9:22003223	0.24	0.73
	4	rs10419404	19:18578654	19p13.11	Yes	ELL	9.8E−09	concordant	0.02	666	0.03	0.18	0.21	5.2	rs8103660	19:18566395	2.2E−08	0.06

p.placo: p-value of the PLACO test, testing the association between the SNPs and both traits, SNP PP4: posterior probability of association of the SNP with both breast cancer and the thyroid trait of interest, PP3: posterior probability of association of the region with both breast cancer and the thyroid trait of interest, due to different causal SNPs, PP4: posterior probability of association of the region with both breast cancer and the thyroid trait of interest, due to the same causal SNP.

^aNovel Locus: locus that has been previously reported in the literature satisfying one of these conditions: i) it was significantly associated with both traits in a single study, ii) it was significantly associated with each trait through two different studies.

For breast and thyroid cancers, we identified 7 loci significantly associated to both traits (4 loci in the same direction and 3 in opposite direction), including 2q35 which was previously reported (Table 3) While analyzing ER+ and ER− breast cancer separately, no additional locus was reported to be associated to both ER+ breast cancer and thyroid cancer (Supplementary Table S3), while 2 additional significant loci (one previously reported locus 5p15.33 and a novel locus 8q22.1) were reported for ER− breast cancer (Supplementary Table S4).

Analyzing common susceptibility loci between breast cancer and TSH levels led to 16 independent significant loci (2 in the same direction and 14 in opposite directions) of which 7 were novel (Table 3). The analyses restricted to ER+ breast cancer led to 3 additional reported loci (3p25.2, 5q13.3, and 11q12.1), while analysis restricted to ER− led to 5 additional loci (1p36.13, 6q14.3, 15q15.1, 19p13.11, and 20p11.21).

For breast and FT4 levels, we found 10 novel and one known loci (5 in the same direction and 6 in opposite directions) (Table 3). Analyses by ER status led to 2 additional reported loci (11q12.1 and 11q13.3) for ER+ breast cancer (Supplementary Table S3) and 2 other additional loci for ER− breast cancer (1q32.1 and 6q16.2) (Supplementary Table S4).

For breast cancer and hypothyroidism we identified 2 loci (both in opposite directions) in which one (17q22) was not previously reported (Table 3). In the case of breast cancer and hyperthyroidism, 3 novel and one known loci were found (2 in the same direction, 1 in opposite directions and one mixed) (Table 3). Analysis by ER status highlighted an additional locus (15q21.1) for ER+ breast cancer and hyperthyroidism (Supplementary Table S3).

Interestingly, the locus 2q35 was found as a common locus for all thyroid traits except hypothyroidism. Besides, three loci (9q34, 16q23.2, and 22q13.1) were in common for both TSH and FT4 levels and one common loci (9p21.3) was reported for both FT4 levels and hyperthyroidism (Table 3).

Gene set enrichment analysis

We identified a total of 130 genes within the 49 loci detected as pleiotropic between breast cancer and thyroid traits (Supplementary Tables S5 and S6).

The 15 genes mapped to the 7 pleiotropic loci reported for breast and thyroid cancers (Supplementary Table S5) were significantly enriched in GWAS catalog reported genes for breast cancer, thyroid cancer and cancers in general as expected, as well as for cardiovascular factors (Supplementary Table S7). No pathway from Reactome, Biocarta, KEGG or Gene Ontology databases were significantly enriched for these genes.

The 116 genes mapped to the 43 independent pleiotropic loci related to breast cancer and thyroid hormones traits (Supplementary Table S6) were significantly enriched in GWAS catalog reported genes for 119 traits that included breast cancer, thyroid hormones levels as expected but also traits related to other cancers, aging or cognitive functions, body size, alcohol consumption, sex hormones, blood pressure and cardio-vascular diseases, glycemia, lung functions, auto-immune disorders, etc. (Supplementary Table S8). Using curated gene sets defined by Reactome, Biocarta, Wikipathways or Gene Ontology, we found that these genes were enriched in several pathways involved in cell cycle regulation, DNA repair, oxidative stress induced senescence, oncogene induced senescence and thyroid hormones production (Supplementary Table S9). While using the GTEx V8 data on 54 tissue-specific differential gene expression, these genes were found to be significantly enriched in brain tissues (amygdala and anterior cingulate cortex), whole-blood and skeletal muscle (Supplementary Fig. S4).

Bayesian colocalization analysis

To identify the loci with the highest posterior probability to be associated to both breast cancer and thyroid traits among the 49 loci identified by PLACO, we performed a Bayesian colocalization analysis of the ±200 KB regions around the lead SNPs. As suggested in Ray et al. 2020 [22], we considered the loci identified by PLACO validated by the colocalization analysis if PP3 + PP4 ≥ 0.9 and PP4/PP3 ≥ 3 (see Methods section).

Only two out of seven loci showed high probability of containing SNPs causally associated to both breast and thyroid cancers. At 2q35, rs3821098 and rs16857611 had the highest causal probabilities and showed the same direction pleiotropy signal between thyroid cancer and respectively breast cancer and ER+ breast cancer. They are located within the non-coding gene DIRC3. At 8q22.1, rs16917117 demonstrated the highest probability of being associated to both ER− breast cancer and thyroid cancer, but in opposite directions. This SNP is located in gene NDUFAF6 (Table 4).

Table 4

Significant loci in PLACO with convincing evidence of a single causal SNP driving pleiotropic association after the colocalization analysis.

Cross-Phenotype   analysis	Genomic   locus	SNP	position (hg19)	a1	a2	Beta   BC	Pval   BC	Beta   GWAS2	pval   GWAS2	P.placo	Located  in gene
BC All/TC	2q35	rs3821098	2:218292141	T	C	−0.062	1.2E−09	−0.350	1.93E−10	2.2E−16	DIRC3
BC ER+/TC	2q35	rs16857611	2:218296732	C	T	0.065	1.1E−08	0.353	1.24E−10	8.0E−16	DIRC3
BC ER−/TC	8q22.1	rs16917117	8:95909498	T	C	−0.093	1.8E−05	0.262	6.90E−05	1.7E−08	NDUFAF6
BC All/TSH	14q32.33	rs3809453	14:105219467	A	G	0.040	1.0E−05	−0.038	6.72E−16	2.7E−15	SIVA1 (+3 bp)
BC All/TSH	17q21.31	rs35354512	17:43515927	C	T	−0.049	3.5E−05	0.031	1.33E−06	4.5E−09	PLEKHM1

BC ALL: all breast cancer cases, BC ER+: estrogen receptor positive breast cancer, BC ER−: estrogen receptor negative breast cancer, TC: thyroid cancer, TSH: thyroid-stimulating hormone levels, a1: reference allele, a2: alternate allele, beta: beta value from the GWAS summary statistics data, pval: p-value from the GWAS summary statistics data, P.placo: p-value for the PLACO output.

We also reported two loci convincingly associated to breast cancer and TSH levels: at 14q32.33 and 17q21.31, and the highest causal probabilities in these regions were reported for rs3809453 (close to the gene SIVA1) and rs35354512 (intronic region of gene PLEKHM1), respectively. Both SNPs showed associations in opposite directions between breast cancer and TSH levels (Table 4).

Regional plots of PLACO results for these four regions are shown in Supplementary Fig. S5.

Investigation of regulatory signals for five SNPs with potentially pleiotropic effect

At 2q35, no V2G score was observed for rs3821098 while rs16857611 shows a V2G score only for the gene IGBP5 based on a PCHi-C signal in fetal thymus cells (Supplementary Tables S10–S12).

For rs16917117 located in an intron of NDUFAF6 at 8q22.1, the highest V2G score was observed for NDUFAF6 based on eQTL in blood, monocyte, and muscle tissues, sQTL and PCHi-C data (Supplementary Tables S10–S12). This variant also shows elevated V2G scores for INTS8, TP53INP1, and CCNE2 based on eQTL in multiple tissues such as adipose, blood, lymphoblastoid cell lines (LCL), heart, rectum, and monocyte and PCHI-C data (Supplementary Tables S10–S12).

For rs3809453, located in the 5′-UTR (untranslated region) of SIVA1 at 14q32.33, the first five highest V2G scores were observed for ADSS1, INF2, SIVA1, AKT1, and ZBTB42 (Supplementary Table S10). The SNP is a sQTL for the first three genes and eQTL for all five genes in multiple tissues. In particular, rs3809453 was reported to affect the expression of AKT1 and ZBTB42 in thyroid tissue (Supplementary Table S11).

At 17q21.31, rs35354512 is an intronic variant in PLEKHM1. The first five highest V2G scores for this variant were observed for ARHGAP27, PLEKHM1, LRRC37A, KANSL1, and CRHR1 (Supplementary Table S10). The SNP was associated with the expression of PLEKHM1, ARHGAP27, LRRC37A, KANSL1, and CRHR1 in multiple tissues including blood, adipose tissue, cells of the immune system and brain (Supplementary Table S11). There was also some evidence of PCHi-C signal for the gene ARHGAP27 and it showed regulatory signals in 17 human primary hematopoietic cell types (Supplementary Table S12).

Discussion

In this study, we aimed to analyze the shared genetic risk factors between breast cancer risk and several thyroid traits including thyroid cancer, TSH and FT4 levels, hypothyroidism and hyperthyroidism. We explored genetic correlations between those traits and association with the genetically predicted risk of the different traits using PRS. We also investigated the nature of the association using several approaches, including a SNP-based approach, functional enrichment analysis, and evidence of regulatory signals to emphasize possible common biological mechanisms between breast cancer and thyroid cancer on one side and between breast cancer and thyroid hormones traits on the other.

Common genetic risk factors between breast and thyroid cancers

We reported a non-significant negative genetic correlation between breast cancer and thyroid cancer, and we did not find any association between the PRS for thyroid cancer and breast cancer risk or between PRS for breast cancer and thyroid cancer risk. However, using a SNP-based approach, we reported several SNPs associated to both diseases that were mapped to seven loci, which were enriched in genes reported by GWAS on cancers and on arterial pressure. Using Bayesian colocalization analysis, we shortlisted two loci (2q35 and 8q22.1) with high probability to be associated to both cancers.

At 2q35, we reported two SNPs: rs3821098 which does not seem to affect the function of any gene, and rs16857611 that may affect the function of IGBP5. This finding is in accordance with previous fine-mapping analyses of breast cancer [26] and thyroid cancer GWAS [27] that highlighted rs16857609 (r² = 0.99 and D′ = 0.99 with rs16857611 in EPITHYR) as a possible causal SNP at 2q35 through the deregulation of IGFBP5 expression that is suspected to affect cells proliferation in breast and thyroid tissues.

At 8q22.1 locus, rs16917117 had the highest posterior probability to be associated to breast cancer ER− and thyroid cancer. Among the genes that are likely to be functionally influenced by this variant, we reported NDUFAF6, INTS8, TP53INP, and CCNE2. NDUFAF6 has no known role in cancer etiology. INTS8 plays a role in the regulation and the activation of protein-coding genes, and its overexpression in multiple tumors including breast cancer was previously reported in an experimental study [28]. TP53INP1 is a tumor suppressor gene and is a major regulator of p53 in response to oxidative stress and its expression was associated to progression of several tumors [29]. More specifically, low expression of TP53INP1 in mammary tissue was positively associated with breast cancer progression [30], while a higher expression of TP53INP1 was reported in the blood of patients with papillary carcinoma compared to healthy controls [31]. Interestingly, we found that rs16917117 was associated to breast and thyroid cancers but in opposite directions. CCNE2 was suggested to act as a proto-oncogene in cancers, especially in breast cancer [32] and was shown to be over-expressed in thyroid tumors compared to non-cancerous thyroid tissues [33].

So, our result suggests that the two most convincingly pleiotropic loci we highlighted may affect both cancers through the expression of multiple genes involved in cell proliferation such as IGBP5, INTS8, TP53INP1 and CCNE2. These results may explain part of the bidirectional association reported by a recent review of epidemiological papers [10]. A more recent meta-analysis of 5 epidemiological studies reported a 1.28-fold increased risk of breast cancer among patients with thyroid cancer [8]. This latter study also detected a publication bias suggesting that more studies are required to conclude. A recent MR analysis [34] on the bidirectional association between these diseases suggests that breast cancer increases the risk of thyroid cancer while thyroid cancer slightly decreases the risk of breast cancer. However, this MR analysis was based on a small sample size GWAS for thyroid cancer (649 cases and 431 controls) and these results need to be confirmed.

Common genetic risk factors between breast cancer and thyroid hormones traits

We reported a positive genetic correlation between breast cancer and FT4 levels and a negative correlation with TSH levels. In accordance with these results, we found a positive association between breast cancer risk and PRS for FT4 levels, while an inverse association was observed with PRS for TSH levels. We also reported a positive association between breast cancer risk and PRS for hyperthyroidism, and no association with PRS for hypothyroidism. All these results were similar while restricting the analysis to ER+ breast cancer, whereas non-significant results were reported for ER− breast cancer.

We conducted a gene enrichment analysis of 116 genes (mapped to 43 loci) related to SNPs associated to both breast cancer and FT4 or TSH levels. Hyperthyroidism or hypothyroidism showed enrichment in genes previously reported in GWAS related to several cancer types, blood pressure traits, aging, cognitive function, and autoimmune disorders. Expression of these genes was found to be significantly enriched in brain tissues (amygdala and anterior cingulate cortex), whole-blood, and skeletal muscle using the GTEx data. We also found significant enrichment of these genes in biological pathways relative to oncogene induced senescence, cell cycle pathway, DNA damage, and oxidative stress induced senescence.

From the results of PLACO analysis, we looked at the lead SNPs of the candidate pleiotropic loci between breast cancer and thyroid hormone traits. We found 12 loci (1p32.3, 8q24.21, 9q34.2, 12q24.12, 14q13.3, 14q32.33, 15q21.2, 15q26.1, 16p13.3, 16q23.2, 17q21.31, and 17q21.31) for which association with breast cancer [12, 36–47], and thyroid hormone traits [14, 41, 48, 49–55] have been previously reported. After colocalization analysis for shortlisting the number of loci, we found two loci with high probability to be associated to both TSH levels and breast cancer (14q32.33 and 17q21.31).

At 14q32.33, we reported rs3809453 that potentially affects the function of the following genes: ADSS1, INF2, SIVA1, AKT1, and ZBT42. Of these genes, no specific function associated with breast cancer or thyroid-related traits was reported previously for ZBT42. A role of ADSS1, INF2, or SIVA1 was previously reported in carcinogenesis but not in thyroid hormones. Indeed, association of ADSS1 with breast cancer has been previously reported with rs10623258 [12] which is in high linkage disequilibrium with rs3809456 (D′ = 0.98, r² = 0.94 in European population from 1000 Genomes). INF2 was reported to be associated with thyroid cancer through a tumor suppressor function [56] and to act as an oncogene in breast cancer migration and invasion [57]. For SIVA1, experimental studies have shown that it regulates cell proliferation [58] it has been associated to early-onset breast cancer incidence in a previous GWAS study [59]. Conversely, AKT1 (protein kinase B) is the key mediator of the PI3K/Akt signaling pathway which is involved in several cellular processes including cell proliferation and could be activated by thyroid hormones [60]. It also has a role as a mediator of thyroid autoregulation that is induced by iodine [61]. Inhibition of AKT1 has been shown to promote breast cancer cell invasion and metastasis in experimental studies [62].

At 17q21.31, we highlighted rs35354512 which was reported to potentially affects the function of PLEKHM1, ARHGAP27, CRHR1, LRRC37A and KANSL1. To our knowledge, LRRC37A PLEKHM1, ARHGAP27 were not previously reported to be associated to breast cancer nor thyroid hormones. However, ARHGAP27 and PLEKHM1 were reported as susceptibility genes for epithelial ovarian cancer [63]. The gene PLEKHM1 has also a Protein-Protein interaction with RAB7A which is involved in the innate immune system and autophagy pathway. Zhang et al. recently showed that loss of RAB7/PLEKHM1 impaired the fusion of autophagosomes and lysosomes, resulting in autophagosome accumulation [64]. Lysosomes and autophagosomes provide an environment that supports key events in the immune response and interestingly rs35354512 is an eQTL for PLEKHM1 in the immune system tissue. CRH1 was found to induce apoptosis in breast cancer cells in experimental studies [65] but no link with thyroid hormones levels was reported. KANSL1 was associated with early-onset breast cancer risk [59]. It was also highlighted by a GWAS on thyroid function and dysfunction that identified a variant, rs199461 (r² = 0.60 and D′ = 0.92 with rs35354512) that was an eQTL for KANSL1 in the thyroid tissue [14]. In our study, rs35354512 was an eQTL of KANSL1 in the immune cells such as monocyte and CD4+ T cells, but not in the thyroid tissue.

Our results on the genetic correlation between FT4 and TSH levels and breast cancer are in agreement with a recent MR study on thyroid hormones traits and risk of breast cancer which confirmed a positive causal association between breast cancer risk and hyperthyroidism or circulating FT4 levels, and a negative causal association with TSH levels, mainly for ER+ tumors [21]. In this study, the negative association with TSH levels was also reported with risk of thyroid cancer and multiple cancers. Our findings based on genetic correlation and PRS on a positive association between breast cancer risk and hyperthyroidism but no association with hypothyroidism, are also consistent with two recent meta-analyses of epidemiological studies which reported ORs between 1.12 and 1.20 for hyperthyroidism, and non-significant association for hypothyroidism [7, 8]. Interestingly, we reported that the potential 43 pleiotropic loci were enriched in the genes highlighted by GWAS on several cancer types and on autoimmune diseases. Moreover, the expression of 116 genes located in these 43 loci were enriched in multiple tissues including brain and immune system. This might be explained by the fact that the production of thyroid hormones is regulated by the hypothalamus-pituitary-thyroid (HPT) axis. Besides, thyroid hormones are involved in regulating metabolism and energy balance, as well as growth and development and hence have a role in multiple tissues. [66]. We finally shortlisted two most convincingly pleiotropic loci at 14q32.33 and 17q21.31 and highlighted variants that can affect the function of several genes, including AKT1 and KANSL1. AKT1 has been found to play a role as a mediator of thyroid autoregulation induced by iodine [61]. Abnormal overexpression or activation of AKT1 has been observed in various cancers, including ovarian, lung, and pancreatic cancers, and is associated with enhanced proliferation and survival of cancer cells [67]. Consequently, targeting AKT1 holds great promise as an important approach for cancer prevention and therapy. In ovarian cancer, overexpressed KANSL1 has been found to be associated with the lymphocyte profile, which serves as a biomarker for response to histone deacetylase (HDAC) inhibition. Furthermore, it is postulated that alterations in the KANSL1 gene may lead to reduced mRNA expression of genes which results in decreased levels of tumor-killing CD8+ T-cells. Fejzo et al. have also highlighted KANSL1 in a transcriptome-wide association studies of ovarian and breast cancers [68]. The identification the importance of AKT1 and KANSL1 in other types of cancer and their associations with immune response genes, as well as their appearance as candidate genes in our study might open up new researches for the development of novel prevention and treatment strategies. For instance, by exploring the precise genetic mechanisms underlying AKT1 and KANSL1 roles in breast cancer and thyroid hormone-related traits, we may uncover valuable insights that can contribute to improved clinical outcomes for patients affected by these diseases.

Our results also suggest a possible role of the immune system, as some of the highlighted genes are expressed in the immune cells.

Study strengths and limitations

We proposed a comprehensive approach to identify common genetic risk factors for breast cancer and several thyroid traits using different complementary approaches at different scales. Analyses based on genetic correlation and PRS permitted to have a global association but if there are multiple pleiotropic SNPs with different directions of associations then genetic correlation or association with PRS could be diluted. The SNP-based approach was also limited by the difference in sample sizes of the initial GWAS datasets which lead to a difference of power between the comparisons. In particular, GWAS on hyperthyroidism, hypothyroidism and thyroid cancer had limited sample sizes. However, our study is an important complement to the epidemiological studies on the shared risks between breast cancer and thyroid traits suggesting some biological mechanisms underlying these associations.

Conclusion

Our study shows common genetic risk factors between breast cancer and thyroid cancer or thyroid hormones levels. Our approach allowed us to confirm a known pleiotropic locus at 2q35 and to propose an additional one at 8q22.1 that could affect breast and thyroid cancers through genes involved in cell proliferation. We also proposed two new pleiotropic loci at 14q32.33 and 17q21.31 that were associated to both TSH levels and breast cancer risk. Identification of common biological mechanisms between thyroid conditions and breast cancer risk may have important implications for the development of new prevention and treatment strategies.

Author contributions

T.T. conceptualized and coordinated the project. F.L., A.B., E.O., F.d.V., M.Z., P.G., J.F.D., M.C.B.R. and G.S. were involved in subject’s recruitment and biological material collection for EPITHYR data. E.A.L. and M.K. curated the data. E.A.L., M.K. and Y.A. performed the statistical analyses. E.A.L., Y.A., P.E.S. and T.T. supervised the statistical analyses. E.A.L., Y.A., P.E.S., B.L. and T.T. interpreted the results. E.A.L., Y.A. and T.T. drafted the manuscript. All authors reviewed the manuscript and approved the final version of the paper.

Conflict of interest statement

The authors declare no competing interests.

Funding

This study was supported by INSERM Itmo Cancer. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. We thank the Breast Cancer Association Consortium, the Thyroidomics consortium and the EPITHYR consortium for providing summary statistics from GWAS. The breast cancer genome-wide association analyses were supported by the Government of Canada through Genome Canada and the Canadian Institutes of Health Research, the ‘Ministère de l’Économie, de la Science et de l’Innovation du Québec’ through Genome Québec and grant PSR-SIIRI-701, The National Institutes of Health (U19 CA148065, X01HG007492), Cancer Research UK (C1287/A10118, C1287/A16563, C1287/A10710) and The European Union (HEALTH-F2-2009-223175 and H2020 633784 and 634935). All studies and funders are listed in Michailidou et al. [12]. TSH, FT4, hypothyroidism, and hyperthyroidism data were conducted within the ThyroidOmics Consortium (http://www.thyroidomics.com). Extended acknowledgments and study-specific acknowledgments are provided in Supplementary Note 4 of Teumer et al. paper [14]. The EPITHYR genome-wide association study was founded by INCA (#9533) and ARC (#PGA120150202302). This work is also part of the Inserm Cross-Cutting Project GOLD.

Data availability

GWAS summary statistics data for breast cancer is accessible through the webpage of the BCAC consortium (https://bcac.ccge.medschl.cam.ac.uk/). GWAS summary statistics data for thyroid cancer is available on request from the corresponding author. FT4, hypothyroidism and hyperthyroidism GWAS summary statistics data are accessible via the ThyroidOmics consortium. GWAS summary statistics data for TSH levels is available at http://csg.sph.umich.edu/willer/public/TSH2020/.

Ethics statement

Participants from studies included in EPITHYR provided written informed consent and each study was approved by their governing ethics committee.

Disclaimer

Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy or views of the International Agency for Research on Cancer/World Health Organization.

Author notes

Elise A. Lucotte, Yazdan Asgari equally contributed.