Genetic susceptibility to hereditary non-medullary thyroid cancer

Non-medullary thyroid cancer (NMTC) is the most common type of thyroid cancer. With the increasing incidence of NMTC in recent years, the familial form of the disease has also become more common than previously reported, accounting for 5–15% of NMTC cases. Familial NMTC is further classified as non-syndromic and the less common syndromic FNMTC. Although syndromic NMTC has well-known genetic risk factors, the gene(s) responsible for the vast majority of non-syndromic FNMTC cases are yet to be identified. To date, several candidate genes have been identified as susceptibility genes in hereditary NMTC. This review summarizes genetic predisposition to non-medullary thyroid cancer and expands on the role of genetic variants in thyroid cancer tumorigenesis and the level of penetrance of NMTC-susceptibility genes.


Introduction
Thyroid cancer is the most common endocrine malignancy [1], with its global incidence rate increasing substantially in the past four decades [2]. Thyroid cancers can originate due to the accumulation of genetic mutations in para-follicular or follicular cells. Thyroid cancers originating from para-follicular calcitonin-producing C cells are known as medullary thyroid carcinoma (MTC) and account for 5% of all cases, whereas the more common type of thyroid cancer arises from follicular cells and is known as non-medullary thyroid cancer [1]. The majority of NMTC are differentiated thyroid cancers (DTC) which include papillary and follicular thyroid cancers. Papillary thyroid cancer (PTC) accounts for more than 85% of NMTC cases and follicular thyroid cancer (FTC) accounts for 10-15% of NMTC cases. The rare forms of NMTC are poorly differentiated thyroid carcinomas and anaplastic thyroid carcinomas [3,4]. Over 90% of all thyroid cancers are sporadic and arise from somatic genetic changes [5]. The remaining are familial forms of NMTC and MTC. Familial MTC (FMTC) has well-known genetic alterations and genotype-phenotype correlations. On the contrary, the genetic causes of familial NMTC (FNMTC), or familial follicular cell-derived carcinoma are poorly understood [6,7]. FNMTC is clinically defined as the presence of the disease in two or more first-degree relatives of the patient. FNMTC can further be classified as syndromic or non-syndromic, depending on whether the thyroid cancer is the primary cancer (non-syndromic) or as a part of one of many constellations of tumours in kindreds (syndromic FNMTC) [8]. Hereditary cancer syndromes associated with FNMTC account for 5% of all familial cases and include Familial adenomatous polyposis, Cowden syndrome, Carney complex, Werner syndrome, DICER1 syndrome, Ataxia-telangiectasia, Bannayan-Riley-Ruvalcaba syndrome, Li-Fraumeni syndrome, Peutz-Jeghers syndrome, and Pendred syndrome ( Fig. 1) [9].
According to the Swedish family cancer database, the proportion of cancer susceptibility accounted for by genetic factors was the highest for thyroid cancer among 15 cancer sites [10]. Additionally, family and twin studies from Utah and Sweden suggested thyroid cancer as one of the most heritable cancers displaying Mendelian inheritance, with reported risks of 8-to 12-fold higher for first-degree relatives of thyroid cancer patients compared to the general population [11][12][13] . A family cohort of the Norwegian cancer registry database estimated that the familial risk ratio of NMTC in affected first-degree relatives is 5.2 for men and 4.9 for women [14,15]. Similarly, Lin et al. identified the family structures of 38,686 NMTC patients in Taiwan. The prevalence of NMTC in the general population and in first-degree relatives of NMTC patients were 0.16% and 0.64%, respectively. This shows a 5.47-fold increased risk for NMTC for firstdegree family members [16]. FNMTC patients present with more aggressive disease at a younger age compared to sporadic cases, this includes larger tumours with more lymph node involvement [17][18][19]. However, studies have shown no significant increase in risk of recurrence or disease-related mortality in FNMTC cases compared to sporadic cases [20][21][22]. As a result, the evidence for a worst disease outcome in FNMTC compared to sporadic cases is conflicting [18,23]. Additionally, the second generation of FNMTC patients present at a younger age with more severe symptoms, indicating the presence of clinical anticipation [24]. In a prospective cohort study, at-risk relatives of twenty-five kindreds with two or more members affected by FNMTC were screened with neck ultrasound and fine-needle aspiration biopsy of thyroid nodules. The results indicated the presence of thyroid cancer in 4.6% of families with two affected members and 22.7% of families with three or more affected members. The tumours that were identified with screening were smaller in size, had less lymph node metastases, and required less extensive treatments. Therefore, the early detection of FNMTC can potentially improve the treatment outcome [25].
Despite the solid evidence for the heritability of thyroid cancer, only a handful of variants have been convincingly associated with a higher risk of this cancer. The high heritability of thyroid cancer is likely due to the contributions of rare but high-penetrance mutations in some cases or common but low-penetrance variants in others [26]. The present study aimed to summarize the literature regarding variants associated with higher risk of hereditary NMTC and provide a more extensive background on the penetrance, molecular function and functional consequences of these mutations, which can further clarify the etiology of thyroid cancer and aid in the identification of disease risk in family members of NMTC patients.

Genetic variants associated with risk of nonmedullary thyroid cancer
Approximately 5-15% of NMTC cases occur due to germline mutations [17]. Genetic variants conferring risk of complex disorders such as cancer are either rare mutations with moderate to high penetrance or common variants with low penetrance. The genetic predisposition to NMTC seems to be relatively strong based on the previous case-control studies. To date, multiple susceptibility genes have been identified through genome-wide association studies (GWAS) ( Table 1). Methods such as family-based exome sequencing, next generation sequencing and linkage studies also identified several  susceptibility loci associated with NMTC. Genetic variants from these genes have been classified based on their level of penetrance using odds ratio (OR) compiled from previous case-control studies. In this review, variants with an OR lower than 1.5 were classified as low penetrance mutations, and those with an OR between 1.5 and 2.5 were classified as moderate penetrance and highly penetrant mutations were categorized as variants with an OR greater than 2.5 [34].

Moderate and high penetrant mutations FOXE1
The FOXE1 (forkhead factor E1) gene is located at chromosome 9q22.33 and encodes for the FOXE1 transcription factor (thyroid transcription factor 2, TTF-2), which regulates thyroglobulin and thyroperoxidase gene expression. In a genome-wide association study in a population of 192 and 37,196 thyroid cancer cases and controls, seven of the nine strongest association signals were in a similar linkage disequilibrium region as the FOXE1 gene. Further replication results from 241 patients in combination with results from a GWAS showed the strongest association signal for allele A of rs965513 with an OR of 1.75 in European populations from Iceland, Columbus, and Spain [35]. In a study by Landa et al. (2009), another single nucleotide polymorphism (SNP) (rs1867277) located in the 5'UTR region of FOXE1 gene was positively associated with thyroid cancer in Spanish and Italian cohorts. The authors proposed this variant as a causal SNP in susceptibility to thyroid cancer through DNA binding assays and transfection studies. The variant was observed to cause cancer susceptibility through the recruitment of USF1/USF2 transcription factors [36]. Furthermore, the first study between FOXE1 gene and susceptibility to FNMTC was reported in 2012. Nine exonic and promoter variants of FOXE1 gene were studied in a population of 60 Portuguese FNMTC probands and 80 sporadic cases with matched controls ( Table 2). As a result, rs965513 and rs1867277 were associated with increased risk of FNMTC. The authors also observed an association between FOXE1 polyalanine tract expansions and familial thyroid cancer risk (OR = 2.56) [37]. The same group identified a rare FOXE1 variant (p.A248G) which co-segregated with thyroid cancer in one family and was also present in a case of sporadic NMTC. Further In vitro studies showed that this variant promoted cell migration and proliferation [38]. A large familybased study with 672 subjects belonging to 133 pedigrees with FNMTC cases genotyped twenty-three variants on 11 loci. Only three variants of 9q22.33 near FOXE1 showed a positive association with FNMTC. FOXE1 gene variant rs1867277 had an OR of 3.17 under a recessive mode of inheritance. The other two variants also showed high penetrance under the recessive model (OR = 4.63 for rs10759944 and OR = 5.10 for rs965513) [39]. The rs965513 variant has been previously associated with increased tumour size and extrathyroidal   [37] expansions in PTC patients [40]. In another two-step association study involving 1820 DTC cases and 2410 controls in Europe, two moderate penetrant FOXE1 variants were identified, rs7028661 with an OR of 1.64 and rs7037324 with an OR of 1.54 [33]. He et al. showed that the rs965513 variant and 4 other variants in close proximity regulate FOXE1 and the PTC susceptibility candidate 2 (PTCSC2) gene transcriptional activity through regulatory enhancers [41]. PTCSC2 is novel long noncoding RNA (lncRNA) gene with its transcripts downregulated in PTC tumours. Later, myosin-9 (MYH9) was identified as a PTCSC2 binding protein with the ability to inhibit the promoter shared by FOXE1 and PTCSC2 in both directions. Thus, PTC risk is potentially conferred by the interaction between a lncRNA (PTCSC2), MYH9, and FOXE1 [42].

HABP2
The Hyaluronan-Binding Protein 2 (HABP2) gene is located on chromosome 10q25.3 and encodes a member of the peptidase S1 family of serine proteases [43]. Mutations in this gene are associated with non-medullary thyroid cancer and susceptibility to venous thromboembolism. Gara et al. (2015) performed wholeexome sequencing of peripheral blood in 7 affected members of a FNMTC kindred and unaffected spouses as controls, a germline variant (G534E; rs7080536) was identified in the HABP2 gene. All affected family members were heterozygous for the variant in peripheral blood DNA. This mutation was found with an allele frequency of 2.2% in the ExAC database, and in 4.7% of 423 patients with sporadic thyroid cancer reported in Human Cancer Genome Atlas multiethnic database. Functional studies confirmed this loss of function variant's pathogenicity and showed normal HABP2 has tumour-suppressive functionality [43]. Zhou et al. [44], Sponziello et al. [45], and Tomsic et al. [46] later argued that the allele frequency of the G534E variant exceeds the filtering criterion used by Gara et al. (less than 1% in public databases) and the role of this variant in thyroid cancer requires more studies. A replication study by Zhang et al. with a cohort of 64 subjects from 29 kindred, identified G534E variant in 6 PTC patients from 4 independent kindreds. The prevalence rate of 13.8% was reported for this variant in the 29 kindreds, suggesting HABP2 as a susceptibility gene for hereditary thyroid cancer [47]. However, NMTC risk conferred by HABP2 G534E was not confirmed by an association study of over 2000 NMTC cases and over 5000 population controls from the British Isles. The frequency of HABP2 G534 variant was 4.2% in cases and 4.6% in controls (OR = 0.74; P = 0.017) [48]. The data from various ethnic populations with large sample sizes suggest that this variant is unlikely to be a moderate or high penetrance gene in NMTC patients. Multiple other groups were unable to verify an association between the G534E variant and hereditary thyroid cancer [49][50][51][52][53][54][55][56][57]. Additionally, targeted sequencing of 516 PTC cases failed to identify the G534E variant. However, three other HABP2 variants (rs138864377, rs2286742, and rs3740530) were identified that can potentially increase the risk of PTC. The rs2286742 and rs3740530 variants in HABP2 had odds ratio of 9.644 and 3.989 in a recessive model, respectively (Table 3) [56]. However, no replication studies have been performed to identify the pathogenicity of these three HABP2 novel variants in PTC.
Low penetrance mutations TITF1/ NKX2.1 and PTCSC3 TITF1/NKX2.1 consists of two exons that encode thyroid-specific transcription factor-1 (TTF-1). A GWAS in the Icelandic population by Gudmundsson et al. identified an association between intergenic variant (rs944289) on chromosome 14q13.3 and risk for DTC (OR = 1.37) ( Table 4) [35]. The closest gene to this variant is NKX2.1 [58]. This association was also confirmed by a case-control study in 1085 Korean DTC cases and 8884 controls that yielded an OR of 1.23 for rs944289 and an OR of 1.25 for the rs34081947 variant [29]. Another case-control study from the northern Chinese Han population also found an association between the rs944289 variant and PTC risk (OR = 1.23) [59]. On the   [56] contrary, a large family-based study of 672 subjects from 133 pedigree was not able to find any association between familial NMTC and rs944289 [39]. Studies on the rs944289 variant in different populations are summarized in Table 4 The mutation was not found among 349 healthy control subjects or among the 284 PTC patients who had no history of MNG [60]. In another study, none of the 63 familial PTC cases had the A339V mutation [61]. Jendrzejewski et al. identified a non-coding RNA gene named papillary thyroid carcinoma susceptibility candidate 3 (PTCSC3) in a transcriptome gene expression analysis from 46 PTC tumour and unaffected thyroid tissue samples. Interestingly, PTCSC3 is located 3.2 kb downstream of rs944289 at 14q.13.3 and has lower expression in PTC thyroid tumours, suggesting a tumoursuppressor role. PTCS3 expression is reduced by the T allele of the rs944289 SNP which affects promoter activation. As a result, the risk allele of rs944289 can potentially decrease PTCSC3 promoter activation and thereby acts as a predisposition to PTC [62] , [63].

SRGAP1
He et al. performed a genome-wide linkage analysis in 38 families with PTC and identified Slit-Robo GTPaseactivating protein 1 (SRGAP1) as a candidate gene on chromosome 12q14.2. The SNPs, rs781626187 (Q149H) and rs797044990 (A275T) were two loss-of-function mutations in the Fes/CIP4 homology domain that segregated with PTC in one family each. Additionally, a missense variant (rs114817817) in the RhoGAP domain (R617C) also occurred in only one family [64]. The protein encoded by this gene is a GTPase activator and mutations in this gene can severely impair the ability to inactivate CDC42. CDC42 can mediate multiple signaling pathways, and plays a role in PTC tumourigenesis [65,66]. To assess the frequency of the 4 missense variants in sporadic PTC cases and healthy controls, He et al. performed further targeted association studies on 2 large cohorts from Ohio and Poland which failed to confirm this association. In fact, Q149H and A275T were not found in 367 cases and 552 controls from Ohio or in the 432 cases and 424 controls from Poland. However, a SNP (rs2168411) located in intron 4 of SRGAP1 showed an association with PTC in both Ohio and Poland cohorts with a combined OR of 1.21 (95% CI 1.08-1.35, P = .0008). The rs114817817 variant was also identified in 4 of 742 sporadic cases of PTC in Ohio but in none of the 828 controls, which is suggestive of low penetrance. Future replication studies are required to confirm the candidacy of this variant [64].

NRG1
Previously, a SNP (rs2439302) on chromosome 8p12 was reported to be associated with PTC [27]. This association has been confirmed in multiple replication studies in Icelandic, Korean, Japanese, and Chinese populations (Table 5) [28,63,67]. The rs2439302 variant has been confirmed as a PTC risk variant with odds ratios ranging from 1.2 to 1.4. This variant has also been correlated with multifocality and lymph node metastasis in PTC patients [40]. Another variant of NRG1 locus (SNP rs2466076) was found to have an OR of 1.32 among 3001 NMTC cases and 287,550 controls [65]. Both SNPs  [29] are located in the intronic regions of the neuregulin 1 (NRG1) gene. The NRG1 gene encodes a membrane glycoprotein that mediates cell-cell signalling and plays a critical role in the growth and development of multiple organ systems. Additionally, NRG1 dysregulation is closely linked to PI3K-AKT and MAPK signalling pathways and has been demonstrated to be involved in tumourigenesis of both malignant and benign thyroid tumours [69,70]. He et al. evaluated candidate functional variants of NRG1. The [G] risk allele (rs2439302) was associated with higher expression of the three tested isoforms in normal thyroid tissue. The authors proposed these isoforms as contributing factors to higher PTC risk through allele-specific enhancer-mediated transcriptional regulation of NRG1 [71]. NRG1 expression was also shown to be essential for PTC cell proliferation through protection from reactive oxygen species (ROS) damage by nuclear factor E2-related factor 2 (NRF2). Therefore, NRG1 can also be useful as a potential therapeutic target for PTC patients [72,73]. Guibon et al. performed finemapping of the 8p12 (NRG1) locus in Europeans, Melanesians and Polynesians populations and identified rs2439304 associated with DTC (OR = 1.2). This variant had the highest posterior probability (PP) of causality in the three ethnic groups based on expression Quantitative Trait Locus (eQTL) data at this locus [74]. NRG1 variants show stronger association in the Korean population compared to the European populations, suggesting a potential Korean-specific marker for DTC [29,75]. Although the function of DIRC3 is still unknown, it is thought to have tumour suppressor activity [76]. In a GWAS with 561 Icelandic individuals with thyroid cancer cases and 40,013 controls, DIRC3 variants were associated both with thyroid cancer risk and thyroid stimulating hormone levels. One variant that was significantly correlated with PTC was rs966423 with an OR of 1.34 [27].  (Table 6) [29,30]. However, Mankickova et al. were unable to establish an association between rs966423 and thyroid cancer in a European population, suggesting inter-population heterogeneity in thyroid cancer susceptibility [33]. Patients homozygous for the T allele of rs966423 have a 6.4% higher mortality risk compared to CC/CT carriers (P = 0.017) [77]. Additionally, CT genotype carriers were associated with extrathyroidal extension and more advanced T stage [78]. On the contrary, a recent study in 1466 DTC patients reported no association between any genotype at the rs966423 SNP and overall mortality and response to therapy [79]. A recent GWAS analysis also identified five novel variants including rs11693806 as a non-coding variant located close to DIRC3 in a large sample of 2637 European ancestry cases and 134,811 European ancestry controls [28]. Another study by Guibon et al. identified rs16857609 as a novel variant located near DIRC3. This SNP was associated with DTC in the European population (OR = 1.4, p = 1.9 × 10 − 10 ) [74]. Future studies should replicate the findings of the known DIRC3 variants and confirm their association with PTC pathogenesis.

Polygenic risk score
As reviewed in the previous section and summarized in  [81]. The PRS of 12 thyroid cancer-associated SNPs (rs11693806, rs2466076, rs1588635, rs368187, rs116909374, rs12129938, rs6793295, rs73227498, rs7902587, rs2289261, and rs56062135) was investigated in 2370 childhood cancer survivors with an European ancestry. Similar to previous findings, the hazard ratio for developing secondary thyroid cancer by one standard deviation increase in the PRS was 1.57 (95% CI = 1.25-1.83; P < 0.001) [82]. Likewise, in a phenome-wide association study of 472 thyroid cancer patients with European ancestry, a PRS of 9 SNPs exhibited a strong association with thyroid cancer (OR = 3.2) when the top PRS quartile was compared to the bottom quartile [83]. In a study by Wang et al., Individuals with African ancestry who were in the top PRS quintile of 5 SNPs had a 30% greater chance of thyroid cancer (OR = 1.3) than those in the lowest quintile [84]. Additionally, in a study with 495 thyroid cancer patients and 56,439 controls by Song et al., the PRS of 6 SNPs (rs6759952, rs13059137, rs7834206, rs72616195, rs1369535, rs11175834) increased thyroid cancer risk by a factor of 3.9 when comparing high PRS tertile with low PRS tertile [85]. Given the presented findings, PRS has the potential to identify individuals at a higher risk of thyroid cancer. However, studies with larger sample sizes and more inclusive PRS with wide varieties of SNPs are required for determining the optimal PRS model for thyroid cancer. Rare germline mutations in families with Nonsyndromic familial non-medullary thyroid cancer

Syndromic familial non-medullary thyroid cancer
Hereditary syndromes (syndromic FNMTC) with various clinical features may be associated with approximately 5% of familial non-medullary thyroid cancer cases (

Familial adenomatous polyposis (FAP) and Gardner's syndrome
Familial adenomatous polyposis (FAP) is an autosomal dominant disease caused by loss-of-function mutations of the APC tumour suppressor gene located on chromosome 5q21. The classic type of FAP is characterized by the development of multiple benign polyps lining the mucosa of the gastrointestinal tract, particularly the colon. Untreated polyps can become malignant with an early age of onset. Papillary thyroid carcinomas are seen in some families affected by FAP [110,111]. In fact, patients with FAP have a 160-fold greater risk of PTC compared to the general population. The prevalence of thyroid cancer among patients with FAP is 2.6%. These thyroid cancers have a unique cribriform pattern on histologic examination and occur more commonly at a young age (< 30 years) in women (95%) [112,113]. More than 60% of APC pathogenic mutations have been identified in the mutation cluster region between codons 1284 and 1580 [114,115]. Most female patients with FAP and PTC also have a RET somatic mutation in addition to APC germline mutations in their tumours [116]. The APC gene encodes a multidomain protein that plays a significant role in tumour suppression by negatively regulating the WNT signalling pathway. Loss of APC function results in inappropriate activation of this pathway which results in cancer progression [117].

Werner's syndrome
Werner's syndrome is an autosomal recessive disease characterized by premature aging, scleroderma-like skin changes, cataracts, subcutaneous calcifications, muscular atrophy, diabetes, and a high incidence of neoplasms, including thyroid neoplasms. Werner's syndrome has been linked to mutations of the WRN gene on chromosome 8p11-21. This gene encodes a member of the RecQ subfamily of DNA helicase proteins that is important in maintaining genome stability by regulating DNA repair, replication, transcription, and telomere maintenance [118]. Thyroid cancer was observed in 16% of 189 patients with Werner syndrome in a Japanese case series. Follicular thyroid cancer was more common, followed by papillary and anaplastic thyroid cancers among these patients [119].

Carney complex
Carney complex (CNC) is an autosomal dominant disease caused by mutations in the PRKAR1 tumour suppressor gene mapped to chromosome 17q22-24 [120]. A loss of function mutation in PRKAR1A can lead to increased PKA signalling [121]. Additionally, this gene can Unknown 4q32 Linkage analysis and targeted deep sequencing identified an ultra-rare mutation (SNP) in chromosome 4q32 in a large pedigree affected by FNMTC.   fuse to the RET proto-oncogene by gene rearrangement and forming a thyroid tumour-specific chimeric oncogene known as PTC2. A loss of function mutation in PARKARI causes increased PKA signalling, leading to AMP-activated kinase (AMPK) activation through LKB1 kinase and increasing mTOR signalling [122]. As a result, patients may present with acromegaly, spotty skin pigmentation, an increased risk of cardiac and mucocutaneous myxomas, and a variety of tumours involving endocrine organs. Additionally, about 60% of patients affected by CNC will develop thyroid tumours that range from follicular hyperplasia to multiple types of thyroid cancer, with follicular adenoma as the most common finding [7]. In a study by Stratakis et al. the prevalence of thyroid nodules and cancers in a series of 338 Carney's complex patients was 5%, including follicular adenomas, PTC, follicular variant PTC (FvPTC), and FTC [123]. Patients affected by CNC should undergo surveillance using frequent ultrasound and biopsies to increase the likelihood of treatment success.

DICER1 syndrome
DICER1 syndrome, also known as pleuropulmonary blastoma syndrome and dysplasia syndrome, is an autosomal dominant genetic disorder that predisposes individuals to various conditions, including benign and malignant tumours of different origins. Germline mutations of the DICER1 gene located on 14q32.13 are detected in endocrine tumours (thyroid, parathyroid, pituitary, pineal gland, endocrine pancreas, paragangliomas, medullary, adrenocortical, ovarian, and testicular tumours).
The DICER1 gene is a member of the ribonuclease III (RNaseIII) family involved in the generation of micro-RNAs (miRNAs) and modulates gene expression by interfering with mRNA function. DICER1 germline lossof-function mutations disrupt the correct timing and expression of miRNA production necessary for normal thyroid differentiation and function [124,125]. Khan et al. investigated the risk of thyroid cancer in 145 individuals with DICER1 germline mutations and 135 family controls from 48 families. This group reported a 16-fold increased risk of thyroid cancer, with all the cases harbouring germline and somatic pathogenic DICER1 mutations [126].
Thyroid abnormalities are common in DICER1 syndrome with multinodular goiter seen frequently in many families with a germline DICER1 mutation. Thus, familial MNG is highly suggestive of DICER1 syndrome. In contrast, differentiated thyroid carcinoma (DTC) was infrequently seen in pedigrees with germline DICER1 mutation. However, multiple differentiated thyroid carcinomas have been found in three children with a history of prior chemotherapy and radiation exposure for the treatment of pleuropulmonary blastoma (PPB). As a result, there has been considerable speculation on a possible link between chemotherapeutic agents and an increased risk of differentiated thyroid cancer due to somatic DICER1 mutations [127]. More recently, a family study reported differentiated thyroid cancer and MNG in six individuals from a family with DICER1 pathogenic mutations and no history of chemotherapy [128].  [129,130]. Approximately 6 to 38% of PHTS patients develop thyroid cancer with a median age of diagnosis of 31-37 years, indicating a risk 51 to 72 times higher than those without PHTS [130]. Therefore, the presence of a PTEN mutation justifies surveillance with annual neck palpation and ultrasound imaging starting at age 10 [131,132].

PTEN hamartoma tumour syndromes
PTEN is a phosphatase that counteracts the phosphatidylinositol 3-kinase (PI3K)/AKT signalling pathways. The tumour-suppressor activity of PTEN is thought to be associated with lipid dephosphorylation at the plasma membrane. An inactivated PTEN gene may increase PIP3 levels leading to AKT activation and mTOR signalling which in turn upregulates cell proliferation and survival while decreasing apoptosis [133,134].
Patients with Cowden syndrome may suffer from breast, endometrium, colon, thyroid, and kidney tumours in addition to NMTC due to PTEN mutations. At least two-thirds of patients with this syndrome are affected by thyroid disease, often before the age of 20. In addition, approximately 10% of patients with Cowden syndrome will develop thyroid cancer (FTC or PTC) in their lifetime [7], [135]. In a study of 664 patients with Cowden syndrome (CS) or Cowden-like syndrome (CLS), 55.1% of the thyroid cancer cases were of classical papillary subtype. In this cohort, 5.4% of the CS and CLS patients had PTEN germline mutations. About 4% of the patients that did not harbor PTEN mutations tested positive for SDHB-D mutations and 2.3% tested positive for KLLN promoter methylation [136]. The SDHB-D genes located on chromosome 1p36.13 encodes succinate dehydrogenase (SDH). Its germline variants can result in the upregulation of the AKT and MAPK pathways, similar to PTEN mutations that can drive tumour formation [137]. KLLN is a tumour suppressor gene located upstream of PTEN. KLLN promoter methylation downregulates its transcription and disrupts p53-mediated activation of KLLN [138]. In another study, targeted sequencing of 91 CS and CLS probands without PTEN, SDHB-D and KLLN mutations, revealed PIK3CA germline mutations in 8.8% and AKT1 germline mutations in 2.2% of cases [139]. PIK3CA is a gene located on chromosome 3q26.32 and encodes p110a, the catalytic subunit of PI3K, which adds a phosphate to phosphatidylinositol-4,5-biphosphate (PIP2) to form phosphatidylinositol-3,4,5-triphosphate (PIP3) at the cellular membrane. PIP3 recruits AKT1 to the cell membrane. Subsequently, activated AKT phosphorylates downstream protein effectors, including the mammalian target of rapamycin (mTOR), which has an established role in human cancers [140].
Whole exome sequencing of a CS proband with the other family members affected with thyroid cancer across 4 generations was performed in 2015. Although all individuals tested negative for PTEN, SDHB-D, KLLN hypermethylation, PIK3CA and AKT1, several novel candidate genes were identified. All family members with CS shared 3 genes with heterozygous missense variants, C16orf72 (c.253 T > C,p.Ser85Pro), PTPN2 (c.1204G > A,p.Ala402Thr) and SEC23B (c.1781 T > G, p.Val594Gly). All 3 genes were sequenced in 96 unrelated CS probands with thyroid cancer, and germline heterozygous SEC23B variant was detected in 3 probands (3.1%). SEC23B encodes Sec23 Homolog B, a component of coat protein complex II (COPII) responsible for transporting proteins from the endoplasmic reticulum (ER) to the Golgi apparatus [141].

Rare syndromes associated with NMTC
Studies involving patients with syndromic FNMTC have the power to add to the list of possible thyroid cancer susceptibility loci and help the identification of key players in thyroid tumorigenesis. Case reports and familial studies have identified multiple rare syndromes associated with a risk for NMTC. Li-Fraumeni syndrome caused by mutations in TP53 gene and presents with a high risk of cancers with bone, breast, adrenal gland, and nervous system origins, with a lifetime cancer risk of > 70% for men and > 90% for women [143] [144].
Pendred syndrome is an autosomal recessive disorder characterized by bilateral sensorineural deafness and goitre caused by mutations in the SLC26A4 (PDS) gene (7q12-34) [145]. The protein product of the SLC26A4 gene is pendrin, a surface anion channel found on the apical membrane of thyroid follicular cells. A loss of function mutation in SLC26A4 may disrupt iodine transport and result in goitre and hypothyroidism [146]. Additionally, follicular thyroid cancer, Hürthle cell adenoma, MNG, and fvPTC have been observed in families affected by Pendred syndrome [147][148][149][150]. Untreated congenital hypothyroidism, chronic stimulation by thyroidstimulating hormone, and additional genetic alterations may also be involved in the formation of thyroid cancer in pendred patients [145,149,151].
Ataxia-telangiectasia syndrome is an autosomal recessive disorder caused by mutations in the (Ataxia Telangiectasia, Mutated) ATM gene on 11q22-23. The ATM gene encodes a member of the phosphatidylinositol 3-kinase family and plays a role in cellular responses to DNA breaks and oxidative stress. Patients with Ataxia-telangiectasia may present with cerebral ataxia, immunodeficiency, telangiectasia, radiation sensitivity, thymic atrophy, and various malignancies, particularly those with lymphoid origin [152]. Furthermore, mutations in the ATM gene have been implicated in PTC and fvPTC [153][154][155][156][157]. A Danish population-based study of 10,324 individuals identified an association between heterozygosity at ATM Ser707Pro and thyroid/endocrine cancer (HR = 10) [158]. Additionally, using whole-genome sequencing and genome-wide linkage analysis, Wang et al. identified ATM variants in 2 of 17 families affected by FNMTC [92].
Peutz-Jeghers syndrome (PJS) is an autosomal dominant disorder characterized by hamartomatous polyps, mucocutaneous hyperpigmentation, and a 4-fold increase in cancer risk compared to the general population [159]. Mutations in the STK11 (serine/threonine-protein kinase 11 alias LKB1) gene (19p13.3) have been implicated as a causative agent for PJS. The protein product of the STK11 gene is a serine-threonine kinase involved in second messenger signal transduction and AMPK inhibition [160]. Additionally, PJS has been associated with multiple cases of thyroid cancer of PTC, FTC, tall cell variant PTC, and fvPTC subtypes [159,161,162]. Papillary Renal Neoplasia (PRN) [101] and McCune-Albright syndrome [163] are two other rare disorders associated with risk of thyroid cancer. Nevertheless, only a few families have been affected by both thyroid cancer and the mentioned syndromes.

Conclusion
Non-medullary thyroid cancer originates from follicular cells of the thyroid gland and accounts for the majority of thyroid cancers. The genetic component of NMTC tumourigenesis is strong but poorly understood, especially for familial NMTC. This review aimed to summarize the current understanding of genetic predisposition to NMTC by looking at genetic variants implicated in familial and sporadic NMTC. Increasing evidence suggests that mutations in the FOXE1 gene have moderate to high penetrance. On the other hand, there is a lack of strong evidence for the role of HABP2 mutations. Thus, further research is needed to clarify its role as a susceptibility gene in NMTC. Most of the mutations in TITF1/ NKX2.1, PTCSC3, SRGAP1, NRG1, DIRC3 genes are low penetrant mutations. Although each low penetrant mutation does not seem to have clinical significance alone, a combination of these mutations could have clinical importance regarding hereditary NMTC.
Different germlines variants are only observed in small groups of FNMTC patients and may not be present in all affected family members within a kindred. Likewise, due to the lack of interventional screening programs, there are no genetic tests available to identify individuals at risk of FNMTC. As a result, the National Comprehensive Cancer Network (NCCN), the American Thyroid Association (ATA), and the European Society for Medical Oncology (ESMO) provide no recommendations on using genetic testing for screening at-risk family members of FNMTC patients [164][165][166]. We also do not think that we have enough evidences supporting the application of genetic screening for certain genes among patients with FNMTC, unless patients medical history and family history suggest a syndromic NMTC that should be tested for the related gene(s).
Further multi-center studies with larger cohorts and stricter inclusion criteria using targeted sequencing or whole exome/genome sequencing are needed to better understand the clustering pattern seen in the families with NMTC. Identification of NMTC susceptibility genes could potentially result in determining targeted treatment options for NMTC patients. Likewise, alternative hereditary mechanisms such as epigenetic modifications may also be involved in the pathogenesis of FNMT C and requires further research. Additionally, identifying new NMTC-associated genetic loci and research on the known implicated variants can improve our understanding of NMTC tumorigenesis in general, which could eventually result in earlier diagnosis and more effective treatment options for sporadic NMTC.