Further Evidence That Defects in Main Thyroid Dysgenesis-Related Genes Are an Uncommon Etiology for Primary Congenital Hypothyroidism in Mexican Patients: Report of Rare Variants in FOXE1, NKX2-5 and TSHR

Mexico shows a high birth prevalence of congenital hypothyroidism (CH) due to thyroid dysgenesis (TD). PAX8 defects underlie only 1% of these cases and NKX2-1 does not seem to be involved. Here, we analyzed other TD-related genes in 128 non-related Mexican patients (females 77.3%; 6 months to 16.6 years) with non-syndromic CH-TD diagnosis established by clinical evaluation, thyroid hormone serum profiling, and scintigraphy (74%) or ultrasonography (26%). We performed Sanger sequencing of FOXE1, NKX2-5, and TSHR and evaluated copy number variations (CNVs) in TSHR, FOXE1, PAX8, and NKX2-1 by multiplex ligation-dependent probe amplification. Odds ratios for TD risk were explored for FOXE1 polyalanine stretches [polyAla-rs71369530] in cases and controls (N = 116). Five rare missense changes cataloged as benign (NKX2-5:p.(Ala119Ser)-rs137852684), of unknown significance (FOXE1:p.(Ala335Gly)-rs543372757; TSHR:p.(Asp118Asn)-rs1414102266), and likely pathogenic (FOXE1:p.(Gly124Arg)-rs774035532; TSHR:p.(Trp422Arg)-rs746029360) accounted for 1.5% (N = 2/128) of clinically relevant genotypes (supported in part by protein modeling) in CH-TD. No CNVs were identified, nor did polyAla > 14 alanines in FOXE1 significantly protect against TD. The present and previously published data collectively show that small clinically relevant germline variants in PAX8, FOXE1, and TSHR are found in only a very small proportion (2.5%) of isolated CH-TD Mexican patients.

The introduction of massively parallel or next-generation sequencing (NGS) has expanded the mutational spectrum of dyshormonogenesis and TD candidate genes for primary permanent CH in patients who were identified through newborn screening but lacked a formal thyroid imaging evaluation [6,7,13]. This approach has proven useful in documenting the wide allelic and locus heterogeneity underlying the thyroid hormone biosynthesis defects that predominate over TD in CH patients of some backgrounds, such as Chinese [6,13] and Russian [7] ones. To date, the five extensively studied TD candidate genes, FOXE1 (MIM*602617), NKX2-5 (MIM*600584), PAX8 (MIM*167415), TSHR (MIM+603372), and NKX2-1 (MIM*600635), appear to convincingly explain only <10% of the CH-TD cases, even when assessed using a NGS approach [3,7]. At least six additional novel loci involved in the TD etiology have recently been described, but some are responsible for syndromic TD forms, and together the known genes still explain only a minority of patients [3]. In addition, assessing of pathogenic copy number variations (CNVs), as well as further explorations in distinct populations to French ones regarding to association of TD phenotype with FOXE1 polyalanine stretch (polyAla) genotypes [14], has not been evaluated by the employed bioinformatics pipelines in these recent NGS studies [6,7,13]. Evaluation of CNVs by multiplex ligation-dependent probe amplification (MLPA) identified whole-exon heterozygous deletions in PAX8, TSHR, and FOXE1 as being relevant pathogenic changes in 11% of CH-TD Polish patients [15], but this has not been replicated in other populations [7,16].
Mexico shows one of the highest and increasing worldwide CH birth prevalences [2,8], with the disease mostly attributable to TD (~93%) [17]. However, the etiology of the disease in this population remain largely elusive [8,17]. Two Sanger sequencing-based reports found that heterozygous loss-of-function PAX8 genotypes account for the etiology of~1% of Mexican CH-TD patients [18], and that clearly deleterious small nucleotide changes in NKX2-1 did not appear to be a factor in this population [19]. To gain a more comprehensive picture of the main TD-related genes in a Mexican population, we herein studied 128 unrelated Mexican patients at pediatric age, all bearing a confirmed clinical, biochemical, and thyroid imaging diagnosis of non-syndromic CH-TD. In them, Sanger sequencing was applied to characterize small nucleotide changes in NKX2-5, TSHR, and FOXE1 (including genotyping of the polyAla stretch proposed as a TD susceptibility polymorphic variant [14]) and MLPA to assess CNVs in TSHR, FOXE1, PAX8, and NKX2-1. As has been described for most populations [3], our results indicate that germline defects in the studied TD-related genes are a very uncommon or even a debatable genetic cause for the majority of Mexican CH-TD patients, and suggest that a single-gene model does not explain the high birth prevalence of this congenital disorder in Mexico.
Written informed consent was obtained from parents of all included participants. This study was approved by the Research, Biosecurity and Ethics committees of the National Institute of Pediatrics, Mexico (Registries 083/2013 and 058/2019).
In order to explore a possible association between the polymorphic length of the polyAla tract of FOXE1 (rs71369530) and the CH-TD trait, we additionally genotyped the length of the polyAla stretches by Sanger sequencing in DNA samples of 116 (232 FOXE1 alleles) unrelated, healthy, and ethnically matched individuals. We used the 14-alanine allele and the 14/14 genotype as references, given that they showed the highest frequency among CH-TD patients and healthy controls. We then compared the allelic frequencies for <14 or >14 alanines or genotypes including ≤14/≤14 or ≥14/≥14 alanines between patients and healthy controls. The odds ratio (OR) with 95% confidence interval (CI) was calculated by Fischer's exact test, with a statistical difference assumed at a significance level of 0.05.
2.2.2. MLPA Analysis of TSHR, FOXE1, PAX8, and NKX2-1 To assess CNVs, each exon of these genes was evaluated by MLPA (SALSA ® MLPA ® probemix P319-B1 Thyroid; MRC Holland, Amsterdam, The Netherlands) according to the manufacturer's instructions. The kit do not included probes to analyze CNVs in the NKX2-5 gene. The obtained gene dosages were evaluated using the Coffalyser.Net Software (MRC Holland, Amsterdam, The Netherlands).

Structural Modeling of the Human Forkhead Box E1 (FOXE1) and Thyrotropin Receptor Isoform 1 Precursor (TSHR) Proteins
Through protein in silico modeling, we further assessed the potential deleterious effects of the identified missense variants cataloged as being of unknown significance (VUS) or likely pathogenic, and lacking evidence by functional assays. To construct protein models, we used the amino acid reference sequences NP_004464.2 and NP_000360.2 for human forkhead box E1 (FOXE1) and thyrotropin receptor isoform 1 precursor (TSH receptor), respectively. Only the sequence spanning amino acids 46 to 156 was considered when we created a model for FOXE1. The 3D model of this amino acid sequence was obtained using the I-TASSER server (http://zhanglab.ccmb.med.umich.edu/I-TASSER, accessed on 8 April 2021) [22], which is based in the Local MEta-Threading-Server (LOMETS) [23] and looks for structural templates similar to the query amino acid sequence deposited in the Protein Data Bank (PDB); these are then assembled as template fragments (threaded) into a complete protein model. To define the possible implications of the rare missense variants in FOXE1, we used the crystallographic structure of interleukin 1 enhancer binding factor in complex with DNA [24]. This structure showed high structural similarity with the FOXE1 model (RMSD for all the Cα atoms = 0.619 Å). Once the structures were aligned, the crystallographic coordinates of the interleukin 1 protein were deleted, and the FOXE1 and DNA structures were analyzed. We were unable to model the C-terminal domain of FOXE1, where p.(Ala335Gly) is located, due to the lack of crystallographic data.
For TSH receptor modeling, we used a fragment of crystallographic structure deposited in the PDB (code: 2xwt, aminoacyl residues 24 to 257) [25]. The rest of the amino acid sequence (residues 258 to 764) was then applied to construct a 3D model using the ITASSER server and following the above described protocol.
Once we obtained the best models for both proteins, we subjected them to energy minimization with the Molecular Modeling System UCSF Chimera software [26]. The new coordinates were manually inspected, evaluated for their geometric quality, and validated with MOLPROBITY (http://molprobity.biochem.duke.edu, accessed on 8 April 2021) [27]. The electrostatic potential of the TSH receptor model was determined with the PBEQ Solver server (http://www.charmm-gui.org/?doc=input/pbeqsolver&step=0, accessed on 8 April 2021) [28]. The aminoacyl sequence of the transmembrane region of the TSH receptor model was calculated with PROTTER (https://wlab.ethz.ch/protter/start, accessed on 8 April 2021) [29], and later embedded into the membrane structure generated by the VMD software [30].
Finally, in silico mutagenesis for missense variants of the FOXE1 and TSHR models was performed using the molecular graphics system, PyMOL (Molecular Graphics System, v.2.2.0 Schrödinger, LLC).

Results
Sanger sequencing of the analyzed genes did not reveal any patently pathogenic variant in the 128 CH-TD patients analyzed. However, five rare and single-nucleotide missense changes cataloged as benign (NKX2-5, N = 1), VUS (FOXE1 and TSHR, N = 1 each), and likely pathogenic (FOXE1 and TSHR, N = 1 each) were identified in CH-TD patients (Table 1) and were absent from healthy controls. Genotypes and clinical and thyroid phenotypes documented in these four CH-TD patients, who did not have any other extra-thyroidal congenital abnormality, are summarized in Table 1    The asymptomatic carrier status for proven pathogenic FOXE1 alleles makes it difficult to propose any etiological role for heterozygous VUS in the development of CH-TD. Such heterozygosity was seen for the p.(Pro8Arg) and p.(Leu112Phe) variants recently identified in Chinese CH patients and their healthy parents [13], and the p.(Ala335Gly) [rs543372757] herein reported for HC-215 and her euthyroid mother ( Table 1). The p.(Ala335Gly) is an extremely infrequent allele worldwide (0.00076%); to date, it has been identified only in the heterozygous state (2/121660 alleles) in European (Non-Finnish)-descent populations (https://gnomad.broadinstitute.org/variant/9-100617200-C-G, accessed on 27 May 2021). The Ala335 position does not show a high degree of phylogenetic conservation, and it is located at the C-terminus of FOXE1, which is devoid of critical domains or known pathogenic variants. The p.(Ala335Gly) variant involves a conservative change, as alanine and glycine are small and hydrophobic residues; however, the in silico predictions of its effects were contradictory, ranging from tolerated to probable damaging. Unfortunately, we were unable to perform in silico modeling for the C-terminal domain of FOXE1, and thus could not use this information to consider its possible deleterious effects. Thus, future functional analysis and assessment of its cosegregation with the CH-TD trait in the homozygous state in other families are needed to determine the benignity or pathogenicity of this variant.
The knockout Nkx2-5 -/-mouse model develops thyroid bud hypoplasia [38], a feature that encouraged the study of its human homologue in CH-TD patients [3,38], especially those with associated congenital heart defects (CHD) [39]. However, although NKX2-5 has been extensively studied among different CH populations, including Chinese [13], Japanese [16], Brazilian [31,32], Czech [39], Dutch [40], and Iranian [41] patients, no fully penetrant pathogenic genotype has been found to condition CH-TD [40], with or without CHD [38][39][40], nor is TD a characteristic feature of patients with CHD attributable to NKX2-5 defects [40]. We did Fischer s exact test did not reveal any significant statistical difference in the distributions of the <14 versus >14 alanine alleles or the ≤14/≤14 or ≥14/≥14 genotypes between patients and healthy controls. The genotypic data and statistical associations regarding the polyAla FOXE1 polymorphic tract are presented in Table 2. acid for a positively charged residue and is predicted to alter the surface electrostatic potential (Figure 2), which largely determines the proper electrostatic interactions of TSH receptor with its ligand and/or blocking antibodies [25,46].

Discussion
Hispanics [1,2], including Mexicans [8], have the highest worldwide birth prevalence of CH. In contrast to the information available in the literature for Caucasian and Asian CH patients [3,6,7,13,16], relatively little is known about the participation of genetic factors in the etiology of CH-TD in Latin-American populations. To date, limited publications are available for Brazilian [31,32] and Mexican [18,19] populations. Given the predominance of TD (~93%) in the etiology of primary permanent CH in Mexicans [17], the aim of this work was to examine whether germline small-nucleotide (NKX2-5, FOXE1, and TSHR) and CNV-type (FOXE1, PAX8, NKX2-1, PAX8, and TSHR) changes in TD-related genes play an etiological role in some Mexican CH patients lacking clinical suspicion of a syndromic form and with a confirmed TD imaging-based diagnosis. Our results showed that~1.5% of studied CH-TD patients (N = 2/128) presented relevant FOXE1 (HC-266) and TSHR (HC-324) genotypes involving single-nucleotide, missense, and likely pathogenic classified variants (Table 1). We failed to identify any CNV-type change. Our results add to previous findings that 1% of CH-TD Mexican patients could be attributed to PAX8 missense defects [18], and there was no discernible participation of small-nucleotide NKX2-1 pathogenic changes [19]. Collectively, these data obtained in ≥100 Mexican patients suggest that germline small-nucleotide defects in the main TD-related genes, PAX8 [18], NKX2-1 [19], FOXE1, NKX2-5, and TSHR (this work), could explain~2.5% of the isolated CH-TD cases in Mexico, without any identified etiologic role for CNVs in PAX8, NKX2-1, FOXE1, or TSHR (NKX2-5 was not included in the employed MLPA kit). Although this explains only a small proportion of cases, it is higher than that documented in 90 Brazilian CH-TD patients, where PAX8, NKX2-5, TSHR, and HES1 (MIM * 139605) did not reveal any pathogenic or VUS change [31,32], and seems quite similar to the rate obtained in Japanese CH patients (2.0%, N = 2/102), at least 50% of whom carried a confirmed TD phenotype [16]. Future genetic studies for other novel TD candidate loci (e.g., GLIS3, JAG1, CDCA8, NTN1, TUBB1, and/or THRB) [3] could be warranted in our population.
At least in the clinical setting, the European Society for Paediatric Endocrinology Consensus Guidelines on Screening, Diagnosis, and Management of CH recommends that molecular genetic study should be preceded by a careful clinical evaluation of the CH patients along with delineation of the thyroid phenotype [33]. However, this clinical aspect was not considered in recent genetic studies involving a larger number of Russian [7] and Chinese [6,13] samples. This precluded the authors from offering an updated and reliable proportion of molecular defects in TD-related genes and led to the predominant identification of dyshormonogenesis-related genotypes. In fact, we consider that the lower frequency of extra-thyroid abnormalities identified in the present study, compared to a previous report in Mexican CH patients (15.6 vs. 23%, respectively) [10], reflects that we performed clinical evaluation to exclude the known TD syndromic forms. The inclusion of such cases may have biased the accurate estimation of mutational spectrum underlying the most common isolated TD forms in both Caucasian [4,5] and Mexican [17] CH cases. Such a bias could be exemplified by the recurrent identification of brain-lung-thyroid syndrome (MIM#610978) in other studies [34,35].
CNV-type changes in TD-related genes have not been explored in Latin American CH populations [18,19,31,32]. In a small sample of Polish CH-TD patients directly subjected to MLPA, whole-exon heterozygous deletions in PAX8, TSHR, and FOXE1 genes surprisingly comprised 11% (N = 5/45) of the underlying TD pathogenic genotypes [15]. The two monoallelic PAX8 gene deletions identified in the Polish study agreed with the idea that a haploinsufficiency mechanism correlates with the reported thyroid ectopy and athyrosis phenotypes [3]. However, the authors found it difficult to explain the TD phenotype in the remaining three patients bearing monoallelic TSHR and FOXE1 deletions [15], especially as they did not identify a second pathogenic allele (i.e., by whole-gene sequencing) in these patients. The absence of these gene rearrangements in our larger sample of CH-TD Mexican patients agree with the data obtained from Japanese (PAX8, NKX2-1, and FOXE1) [16] and Russian (1/243 CH patients with PAX8 heterozygous deletion, no deletions in TPO, FOXE1, NKX2-1, or TSHR) [7] CH patients. Thus, with the exception of the Polish experience, CNVs seem to represent an uncommon pathogenic mechanism associated with TD in several CH populations.
A recent analysis of pooled literature supported the idea that heterozygous missense variants located at PAX8, TSHR, FOXE1, and NKX2-5 could increase the risk for developing  [36]. In our study, we identified a total of five rare missense single-nucleotide changes in FOXE1 (N = 2, homozygous and heterozygous genotypes), NKX2-5 (N = 1, heterozygous genotype), and TSHR (N = 2, compound heterozygous genotype); of them, only two were cataloged as "likely pathogenic" variants (Table 1).
Most alleles conditioning the Bamforth-Lazarus syndrome are missense changes that impair the DNA-binding activity of the FOXE1 forkhead domain (amino acid positions 53 to 141) [37]. The Gly124 position is located just inside this domain and shows high phylogenetic conservation from human to zebrafish. The non-conservative amino acid substitution identified in HC-266 changes a small hydrophobic amino acid (glycine) to a positively charged and larger residue (arginine). Our protein modeling indicated that Gly124 is in close contact with DNA, and its drastic substitution with an arginine causes clash contacts (Figure 1), leading to a possible destabilization of the FOXE1 DNA-binding site. This modeling correlates with the in silico predictions made by the PolyPhen, PROVEAN, and Pmut programs, which unanimously predicted the p.(Gly124Arg) FOXE1 variant as deleterious with high confidence scores. To date, the p.(Gly124Arg) FOXE1 allele (rs774035532) exhibits a very low worldwide allelic frequency (0.0016%) and has been identified exclusively in four heterozygous Latin American individuals according to gnomAD database (https://gnomad.broadinstitute.org/variant/9-100616566-G-C, accessed on 27 May 2021). Although there have been a few reported cases of Bamforth-Lazarus syndrome, a wide variation in clinical expressivity has been noted. For example, not all described patients have choanal atresia or epiglottic bifidity, and a minority shows a thyroid hypoplasia instead of athyrosis [37], as seen in our patient HC-266. Attempts to correlate this variable clinical expressivity with the residual functional activity of mutant FOXE1 protein have not been conclusive [37], and functional studies based on gel mobility-shift and transfection assays are needed to corroborate if p.(Gly124Arg) along with a 16 alanine polyAla tract represents a hypomorphic FOXE1 allele that retains considerable residual function to prevent the full development of Bamforth-Lazarus syndrome and only leave CH-TD as the phenotypic manifestation. To the best of our knowledge, our patient HC-266 could represent the first case of a non-syndromic CH-TD diagnosed by newborn screening and bearing a likely pathogenic biallelic FOXE1 genotype, which expands the phenotypic spectrum for FOXE1related disorders. On the other hand, it is generally assumed that the heterozygous parents of Bamforth-Lazarus syndrome patients are healthy and euthyroid [37], as was documented in the parents of HC-266. Unfortunately, we lacked information on the thyroid morphology in these obligate carriers and thyroid image evaluations could not be performed to discard any asymptomatic congenital thyroid defect. In addition to the 16/16 polyAla genotype in HC-266, we noted homozygosity at the rs201181824, rs139551528, rs371516340, and rs3021526 sites (data not shown). This could suggest an identity-by-descent for the 9q22.33 region, potentially reflecting endogamy or consanguinity of the parents (although such antecedents were denied), as has been described for most Bamforth-Lazarus syndrome families [37].
The asymptomatic carrier status for proven pathogenic FOXE1 alleles makes it difficult to propose any etiological role for heterozygous VUS in the development of CH-TD. Such heterozygosity was seen for the p.(Pro8Arg) and p.(Leu112Phe) variants recently identified in Chinese CH patients and their healthy parents [13], and the p.(Ala335Gly) [rs543372757] herein reported for HC-215 and her euthyroid mother ( Table 1). The p.(Ala335Gly) is an extremely infrequent allele worldwide (0.00076%); to date, it has been identified only in the heterozygous state (2/121660 alleles) in European (Non-Finnish)-descent populations (https://gnomad.broadinstitute.org/variant/9-100617200-C-G, accessed on 27 May 2021). The Ala335 position does not show a high degree of phylogenetic conservation, and it is located at the C-terminus of FOXE1, which is devoid of critical domains or known pathogenic variants. The p.(Ala335Gly) variant involves a conservative change, as alanine and glycine are small and hydrophobic residues; however, the in silico predictions of its effects were contradictory, ranging from tolerated to probable damaging. Unfortunately, we were unable to perform in silico modeling for the C-terminal domain of FOXE1, and thus could not use this information to consider its possible deleterious effects. Thus, future functional analysis and assessment of its cosegregation with the CH-TD trait in the homozygous state in other families are needed to determine the benignity or pathogenicity of this variant.
TSHR-related disorders display a broad clinical spectrum due to wide allelic heterogeneity. Monoallelic or biallelic loss-of-function variants of TSHR condition partial or complete TSH resistance, respectively. The first is generally manifested as mild postnatal hyperthyrotropinemia (hyperTSH) associated with an apparently euthyroid state accomplished by a normal-sized and orthotopic gland; the second is generally characterized as severe non-autoimmune CH due to thyroid hypoplasia [43]. This complete TSH resistance leads to CH identified by TSH-based newborn screening [43], as occurred with HC-324: This patient was diagnosed at 1 month of age through a marked elevation of serum TSH (>150 µUI/, reference value: 0.4-4-0 µUI/mL) along with severe hypothyroidism (total T3:  the large N-terminal extracellular domain of the TSH receptor, where several pathogenic missense changes have been described as altering ligand-binding capacity and/or reducing the receptor's cell surface expression [43]. It is well-known that G protein-coupled receptors, like TSH receptor, can exist as dimers and higher-order complexes [44], and the substitution of tyrosine 116 (located at 3.9 Å from Asp118) to serine was reported to totally abrogate the formation of multimers [45]. Since Asp118 is found on the same convex surface of LRR region 1, it seems plausible that this residue could participate in the formation of TSH receptor multimers. Additionally, the p.(Asp118Asn) variant involves the substitution of a negatively charged surface amino acid for a positively charged residue and is predicted to alter the surface electrostatic potential (Figure 2), which largely determines the proper electrostatic interactions of TSH receptor with its ligand and/or blocking antibodies [25,46].
The p.(Trp422Arg) variant is a non-conservative substitution that changes a polar, hydrophobic (tryptophan) residue to a positively charged (arginine) one; it affects a phylogenetically invariable position and could disturb the hydrophobic milieu of the first transmembrane domain (TM helix 1, amino acids 416 to 442, Figure 2), where substitutions close to Trp422 have been reported to interfere with phospholipase C-activating G-protein coupled receptor signaling [43,47]. Regardless of the classification of these variants according to ACMG/AMP criteria (Table 1) and protein modeling predictions (Figure 2), further functional studies performed in vitro will be needed to determine if they impair some of the described cellular mechanisms governed by TSH receptor.
Profound thyroid hypoplasia without thyroid Tc99 m uptake is characteristic of complete TSH resistance [43], which contrasts with the athyrosis phenotype documented in HC-324 by ultrasonography at 18 months of age. Unfortunately, the lack of serum thyroglobulin in HC-324 does not allow to disclose the absence of thyroid tissue, and several factors (i.e., younger age) could interfere with the accurate delineation of a thyroid phenotype by ultrasonography, where it has been estimated that up to 18% of the cases could be misdiagnosed mainly as athyrosis [48].
To further support the pathogenicity of the TSHR variants identified in HC-324, TSHR genotyping and complete thyroid profiling should be performed in both parents to assure their obligate heterozygous status, and assessment for possible partial and non-autoimmune TSH resistance [43] should be performed. Even if we assume that the identified p.
Finally, we explored whether the polyAla FOXE1 tract was associated with the TD trait in our population. In a previous case-control study and transmission disequilibrium test in French Caucasians, the 16 alanine allele (either 14/16 heterozygous or 16/16 homozygous status; allele dose effect) was identified as a protective TD factor (0.39, 95% CI = 0.22-0.68, p = 0.0005) [14]. This contrasted with a TD risk of OR 3.02 for the 14/16 genotype observed in Sicilian primary permanent CH patients [52]. Despite these contradictory data, which suggest that the polyAla FOXE1 tract should be studied in other populations, recent NGS-based studies in Chinese [6,13] and European [7,34] CH cohorts did not explore the association of polyAla FOXE1 alleles and/or genotypes with the risk of TD. This may reflect: the inherent difficulties of using NGS bioinformatics approaches to achieve accurate variant annotation in repetitive sequences, such as polyAla stretches; the high GC content of the FOXE1 gene (~73%), which could lead to a low sequencing depth and coverage (i.e., 51.7%) [13]; and/or the inclusion of CH patients regardless of their thyroid phenotype [6,7,13]. Unlike the analysis carried out by Carré A et al. [14], which considered only the 14 and 16 alanine alleles and genotypes, our association approach encompassed all identified polyAla alleles and genotypes, including a novel 10 alanine allele and a rare 17 alanine allele. While a tendency for a protective effect was noted for alleles larger than 14 alanines and their genotypes, similar to that described for French Caucasian patients [14], this effect did not achieve statistical significance (Table 2). Thus, alleles larger or shorter that 14 alanines or their genotypes did not appear to influence the risk for developing CH-TD in our population. Future association analyses performed with larger samples of patients and controls among different ethnicities may identify strong and specific associations between polyAla FOXE1 stretches and specific thyroid phenotypes [14,52,53], sex [14], extra-thyroidal congenital defects [52], and/or familial CH-TD antecedents [53].

•
Although the Mexican population has one of the highest worldwide CH birth prevalences, our results along with the previously published findings confirm that small nucleotide and clinically relevant germline variants in the main TD-related genes of PAX8 Although we did not evaluate CNVs in NKX2-5, our data support the idea that it has negligible etiological participation in TD along with a benign character for its p.(Ala119Ser) allele [40].

•
The protective effect on TD risk previously described for FOXE1 polyAla alleles larger than 16 alanines or their genotypes in French Caucasians [14] was not significantly supported in our Mexican CH-TD population.