Maternal smoking and high BMI disrupt thyroid gland development

The collection of fetal material was approved by the NHS Grampian Research Ethics Committee (REC 04/S0802/21). Women seeking elective, medical terminations of pregnancy were recruited with written informed consent. Only normally progressing pregnancies from women over 16 years of age and between 11 and 21 weeks of gestation were collected. Fetal plasma samples were collected by ex vivo cardiac puncture, and plasma was separated by centrifugation in heparin tubes and snap-frozen. Fetal thyroids were dissected, weighted, and snap-frozen (right lobe) or fixed in 4% formalin overnight and embedded in paraffin (left lobe). All frozen tissues were stored at − 80 °C. Fetal sex was determined by gonadal histology. Maternal smoke exposure was confirmed by cotinine measurements as previously described, and samples were categorized as control (C) or smoke-exposed (SE) [21]. Maternal BMI information was collected as part of the patient consent procedure, and women were categorized as normal weight (17.5 < BMI < 25) or overweight/obese (BMI ≥ 25). Different subsets from a total of 163 fetal plasma and fetal thyroid samples were used for (i) thyroid weights, (ii) circulating hormonal measurements, (iii) circulating thyroid hormone binding proteins, (iv) fetal thyroid morphological examination, (v) immunohistochemistry, and (vi) transcript measurements. Detailed maternal and fetal characteristics, as well as sample numbers in each category, are provided in Additional file 1: Figure S1. Only blood samples and thyroid tissues obtained within 30–60 min from the time of delivery were used in this study. The weights of 94 fetal thyroid glands were measured (smoke exposure study: 26 control females, 25 control males, 14 smoke-exposed females, and 29 smoke-exposed males; maternal BMI study: 24 normal weight females, 26 normal weight males, 16 overweight/obese females, and 27 overweight/obese males, Additional file 1: Figure S1A). Mean fetal age was not different among smoking or BMI groups in either sex, and mean BMI values did not differ by smoke exposure or sex among the fetal samples used for thyroid weight (Additional file 1: Figure S1A).

Total T3, total T4, and TSH were measured in a total of 60 fetal plasma samples from fetuses ranging between 12 and 20 weeks of gestation (Additional file 1: Figure S1B). Mean fetal age was not different among smoking or BMI groups in either sex, and mean BMI values did not differ by smoke exposure or sex among the fetal samples used for hormonal measurements (Additional file 1: Figure S1B). Total thyroid hormones were measured using a solid-phase extraction-liquid chromatography-tandem mass spectrometry method [22]. To 10-μL aliquots of fetal plasma, 2 mL of a 4% H₃PO₄ aqueous solution, 250 μL of a 60:40 methanol:water solution, and 50 μL of a 10 ppb solution of isotopically labeled [¹³C₆]-T4 and [¹³C₆]-T3 were added. Each sample was vortexed for 30 s and loaded on a SilicaPrepX SCX 60 mg/3 mL extraction cartridge preconditioned by adding 1 mL of methanol and 1 mL of the 4% H₃PO₄ aqueous solution. Cartridges were rinsed with 2 mL of water followed by 1 mL of methanol, and the compounds eluted with 1 mL of a methanol:ammonium hydroxide (95:5) solution. Solvents were evaporated to dryness in a Zymark Turbo Vap evaporator and the samples taken up in 500 μL of deionized (DI) water:methanol (50:50) containing 0.15% H₃PO₄. After vortex mixing, 250 μL of the resulting solution was transferred to a polypropylene tube for subsequent analysis by LC-MS/MS. A Xevo TQ-S electrospray ionization-triple quadrupole mass spectrometer coupled with an Acquity UPLC unit (UPLC-ESI-MS/MS, Waters, Manchester, U.K.) was used for thyroid hormone quantification. A 10-μL aliquot of the sample was injected into an ACE EXCEL C18-Ar column 50 × 2.1 mm, 2 μm. The flow rate was set to 0.4 mL/min and the column temperature to 30 °C. Initial conditions were 70% mobile phase A (0.1% acetic acid in DI-water) and 30% mobile phase B (methanol). The gradient went from 30% phase B to 80% within 4 min, followed by a wash for 30 s at 95% phase B and a return to initial conditions during the next 30 s. Thyroid hormones were monitored in positive ionization mode by MRM (multiple reaction monitoring). Selected transitions [M + H]+ were m/z 778 → m/z 732 (quantification) and m/z 778 → m/z 605 (qualifier) for T4 and m/z 652 → m/z 606 (quantification) and m/z 652 → m/z 479 (qualifier) for T3. Selected transitions [M + H]+ for isotopically labeled THs were m/z 784 → m/z 738 and m/z 784 → m/z 611 (qualifier) for [¹³C₆]-T4 and m/z 658 → m/z 612 and m/z 658 → m/z 485 (qualifier) for [¹³C₆]-T3. Standard curve ranged from 2.5 to 500 ng/mL for T4 and 0.25 to 50 ng/mL for T3. R² was greater than 0.995 for all standards, as determined by TargetLynx (Version 4.1, Waters). To assess the accuracy of T3 and T4 determination, two quality controls (QC) from Bio-Rad [Lyphocheck® Assayed Chemistry Control lot #14440: level 1—14,441 (T4 = 71 ng/mL) and level 2—14,442 (T4 = 208 ng/mL)] were included in the analytical batch (N = 5 for both QCs). The accuracy (%bias) was − 2.8% for QC 14441 and − 9.7% for QC 14442, whereas the precision (%CV) was 3.0% for QC 14441 and 2.0% for QC 14442. No T3 concentrations are provided for these standards, but mean T3 concentrations of 0.87 ng/mL and 3.51 ng/mL were measured in QC 14441 and QC 14442, respectively. The precision of T3 determination was 7.2% for QC 14441 and 1.9% for QC 14442.

Plasma TSH levels were measured (in 25 μl/fetus) using a single Milliplex® MAP Human Pituitary Magnetic Bead Panel 1 kit (ACTH, GH, TSH, CNTF, AGRP: Millipore Limited, Watford, UK) and analyzed using a BioPlex200 system (Bio-Rad Laboratories Ltd., Hemel Hempstead, UK). Intra- and inter-assay coefficients of variation were < 10% and < 15%, respectively, sensitivity was 0.02 μIU/mL, and cross-reactivity was negligible.

Fetal plasma proteins were measured in a total of 80 fetal plasma samples ranging from 12 to 20 weeks of gestation (Additional file 1: Figure S1C). Mean fetal age was not different among smoking or BMI groups in either sex, and mean BMI values did not differ by smoke exposure or sex among the fetal samples used for circulating protein measurements (Additional file 1: Figure S1C). Ten micrograms of fetal plasma proteins were diluted to a final volume of 100 μL in 50 mM NH4HCO3 (BioUltra grade, Sigma Aldrich). Proteins were digested in solution according to the PRIME-XS protocol (http://​www.​primexs.​eu/​protocols/​Public-Documents/​04%2D%2D-Protocols/​PRIME-XS-Protocol-NPC-In-Solution-Digestion.​pdf/​). Briefly, proteins were reduced in 2 mM dithiothreitol (Sigma Aldrich, > 99%) for 25 min at 60 °C and S-xalkylated in 4 mM iodoacetamide (Sigma Aldrich, > 99%) for 30 min at 25 °C in the dark, then digested by sequencing-grade modified trypsin (Promega, Southampton, UK, cat.no. V5111) at a 1:10 ratio of trypsin:protein overnight at 37 °C. The reaction was stopped by freezing at − 80 °C. Samples were then thawed, dried by vacuum centrifugation (SpeedVac Plus SC110A, Savant), and dissolved in 10 μL 2% acetonitrile/0.1% formic acid. The equivalent of 2 μg of peptides (assuming no losses) were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). The LC-MS system included a Thermo Scientific Dionex UltiMate 3000 RSLC nano-LC configured for pre-concentration onto a nano-column, coupled to a Q Exactive Plus hybrid quadrupole-Orbitrap mass spectrometer fitted with an EASY-Spray nano-ESI source (Thermo Scientific). Peptide samples were injected onto a C18 PepMap 100 pre-column (300 μm i.d. × 5 mm) in loading pump solvent (2% acetonitrile, 0.1% formic acid) at a flow rate of 10 μL/min for 5 min. The pre-column was then reverse-flushed to the analytical column (PepMap RSLC C18; 50 μm i.d. × 15 cm; Nano pump solvent A: 0.1% formic acid, Nano pump solvent B: 80% acetonitrile, 0.1% formic acid) at 0.3 μL/min using the nano-pump. Peptides were separated using a gradient of acetonitrile (LC gradient: 3–10% solvent B in 5 min, 10–40% solvent B in 30 min, 40–80% solvent B in 5 min, hold at 80% solvent B for 8 min, 80–3% solvent B in 1 min, hold at 3% solvent B for 15 min) while MS/MS data were acquired by the Q Exactive in data-dependent mode (Top10 method). Parameters for the full scan/data-dependent MS2 (Top10) method were as follows: full scan range 375–1750 m/z, resolution 70,000, AGC target 3e6, and maximum IT 50 ms; MS2 scan resolution 17,500, AGC target 5e4, maximum IT 100 ms, loop count 10, isolation window 1.6 m/z, NCE 26, underfill ratio 4%, charge states 2–5 included, peptide match preferred, exclude isotopes on, and dynamic exclusion 40 s.

Raw mass spectrophotometric output files were processed by MaxQuant (v 1.5.3.30) [23]. MaxQuant runs were performed under the default parameters except that trypsin was set as the digestion enzyme. All peptide matching searches were performed against a FASTA file of the human proteome (20,437 canonical and isoform reviewed protein sequences, downloaded from Uniprot on the 6th of July 2018). Protein intensities across samples were normalized using the maxLFQ algorithm [24]. The normalized protein intensities for ALB, TTR, and TBG were extracted from the results file for the downstream statistical analysis.

Histological sections from 83 fetal thyroid glands ranging from 11 to 20 weeks of gestation (Additional file 1: Figure S1D) were stained with hematoxylin and eosin (H&E) prior to morphological assessment. Histological sections from 55 fetal thyroid glands, ranging from 11 to 20 weeks of gestation (Additional file 1: Figure S1E), were immunohistochemically stained for paired box 8 (PAX8), forkhead box A2 (FOXA2), calcitonin, Na⁺-I⁻ Symporter (NIS, gene name: SLC5A5), and thyroid transcription factor 1 (TTF1, gene name: NKX2-1). Mean fetal age was not different among smoking or BMI groups in either sex, and mean BMI values did not differ by smoke exposure or sex among the fetal samples used for H&E thyroid morphology scoring, but in the fetal samples used for immunohistochemical scoring, smoke-exposed females had higher mean maternal BMI (Additional file 1: Figure S1D-E). For each thyroid specimen, two sections were immunostained with the respective antibodies. Sections were dewaxed and rehydrated in Tris-based saline containing 0.05% Tween-20 (TBST), and endogenous peroxidase was inactivated with 3% H₂O₂ in methanol for 20 min at room temperature (RT). Heat-antigen retrieval was achieved with citrate buffer (pH 6.0) for 20 min, and non-specific binding sites were saturated with TBST containing 5% goat normal serum. Incubations were done overnight at 4 °C in blocking buffer (NIS and TTF1) and for 1 h at RT (PAX8). Horse-radish peroxidase-conjugated anti-mouse or anti-rabbit secondary antibodies (both Jackson Immunoresearch, West Grove, PA, USA) were used at 1:200 for 1 h at RT. Antibody information is provided in Additional file 2: Table S1. Specific binding was visualized using the 3,3′-diaminobenzidine tetrahydrochloride (DAB) substrate (Pierce Biotechnology, Rockford, IL). Nuclei were counterstained with hematoxylin. Slides were imaged using a Zeiss Axio Imager M.2 microscope with color camera and Zeiss Zen software (Zeiss, Jena, Germany). For PAX8, calcitonin, and NIS, immunostained sections were scored according to the number of positively stained cells ranging from −, no staining; (+), few cells; +, more cells; and ++, strong many cells. FOXA2-immunostained sections were scored according to staining localization ranging from −, no staining; (+), cytoplasmic staining; +, cytoplasmic and nuclear staining; and ++, nuclear staining. Thyroid morphology was examined on two 5-μm-thick sections which were sectioned 50 μm apart from each other. Thyroid sections were assessed at the glandular margin to avoid comparing central with peripheral parts of the thyroid gland and were independently examined by two observers blinded to the information on fetal age, sex, smoke exposure, and maternal BMI. The H&E morphology was observed throughout the spectrum of 11–21 gestational weeks.

RNA was extracted from 78 whole fetal thyroid tissues ranging from 11 to 20 weeks of gestation (Additional file 1: Figure S1F) using Qiagen All-prep kits (#80004, Qiagen, Manchester, UK) and then reverse transcribed as described previously [25]. In order to identify the most appropriate housekeeping gene (HKG) for normalization of fetal thyroid mRNA transcript levels, the expression of five HKGs (SDHA, SFRS4, B2M, TBP, and PMM1) [25] was assessed using GeNorm and Normfinder algorithms [26, 27]. SDHA was found to be the most stable HKG using Normfinder while GeNorm identified a combination of SDHA and SFRS4. Transcript measurements were subsequently normalized against SDHA in all studies reported here. A list of all the primers used is provided in Additional file 3: Table S2. Mean fetal age for either sex was similar in the smoke exposure and maternal BMI groups. In the fetal samples used for thyroid transcript measurements, mean BMI values were similar between sexes and in the smoke exposure group (Additional file 1: Figure S1F).

All statistical analyses were performed in R statistical software (V3.4.0). Thyroid weight and hormonal, protein, and transcript measurements were analyzed using multivariate linear regression models; thyroid morphological scoring of H&E sections was compared among groups using multivariate generalized linear models for binomial data (mature vs immature); and immunohistochemical staining scores were analyzed using ordinal logistic regression. Covariates for all analyses were fetal age (continuous), fetal sex (categorical), maternal smoke exposure (categorical), and maternal BMI (categorical). Two major models were employed to examine the relationship between (i) fetal age, fetal sex, and maternal smoking and (ii) fetal age, fetal sex, and maternal BMI. Initially, the presence of three-way age-sex-smoke exposure (or maternal BMI) interaction was examined with the full model: Y~age*sex*smoking (or maternal BMI). For those cases where no statistical evidence for a three-way age-sex-smoking (or maternal BMI) interaction was found, the three-way interaction model was replaced with the model containing all two-way interaction: Y~age*sex +age*smoking (or maternal BMI) + sex*smoking (or maternal BMI). For those cases where no statistical evidence for two-way interactions was found, the simple one-way model was used: Y~age+sex+smoking (or maternal BMI). For those cases where significant interactions were detected (P < 0.05), the dataset was split by sex or smoke exposure (or maternal BMI) to examine the source of interactions and to simplify model interpretation. Correlations between TSH levels and fetal plasma cotinine levels (continuous), fetal T4 levels (continuous), and maternal BMI (continuous) were performed using TSH~cotinine, TSH~BMI, and TSH~T4 linear models, respectively (Pearson’s correlation). Associations between maternal BMI and TSH in the female fetuses were examined by stratifying females into control and smoke-exposed groups and then examining BMI associations using TSH~age+BMI(categorical) models for each smoke exposure group, ensuring absence of significant two-way interactions between age and BMI. For thyroid gland weight, hormonal measurements, protein, and transcript measurements, where multivariate linear regression was used, normal distribution of the residuals and homoscedasticity were visually assessed by quantile-quantile residual plots and scatter plot of residuals vs fitted model values, respectively. Data were log-transformed in those cases where model residuals departed from normality and/or showed heteroscedasticity. Statistical significance threshold was chosen at a P value of < 0.05. To ensure that maternal smoke exposure does not interact with maternal BMI, preliminary statistical models were used to ensure that there are no four-way (age*sex*smoking*maternal BMI), no three-way (age*smoking*maternal BMI and sex* smoking*maternal BMI), and no two-way interactions (smoking*maternal BMI) between smoke exposure and maternal BMI. Generated P values, effect sizes (where applicable), and interactions for the statistical tests performed are given in Additional file 3: Table S2, Additional file 4: Table S3 and Additional file 5: Table S4.

Total T3 and T4 levels in the fetal circulation increased with age, the T3/T4 ratio decreased with fetal age, and the thyroid hormone-binding proteins ALB and TBG increased with increasing fetal age in both males and females regardless of smoke exposure or maternal BMI models (Additional file 6: Figure S2A, Table 1). TSH levels did not change with fetal age in males and females not exposed to maternal smoking (Table 1). Male fetuses had higher T3 levels and a T3/T4 ratio compared to females (Table 1, Additional file 6: Figure S2A). There was no difference in thyroid gland weight between males and females, but male thyroids tended to show more immature thyroid gland morphology compared to female thyroids and this approached statistical significance (P = 0.057, Additional file 4: Table S3).

Analysis of immunohistochemistry showed that TTF1 was invariably present in the nuclei of follicular cells of all sections examined while PAX8 staining ranged from absent to progressively present in more follicular cell nuclei (Fig. 1a, b). Calcitonin staining ranged from absent to increasingly present in parafollicular cells, and NIS staining ranged from absent to mostly cytoplasmic in a few follicular cells to more plasma membrane-localized in larger number of cells (Fig. 1c, d). FOXA2 staining ranged from absent to predominantly cytoplasmic, to cytoplasmic and nuclear, and predominantly nuclear (Fig. 1e). Irrespective of the statistical model used (age-sex-smoke exposure or age-sex-maternal BMI), PAX8 immunohistochemical scores remained constant in males but increased by gestation age in females (Additional file 6: Figure S2B, Additional file 4: Table S3). NIS immunostaining, in contrast, increased by gestation age in males and remained constant in females (Additional file 6: Figure S2B, Additional file 4: Table S3). Calcitonin immunostaining scores on fetal thyroid tissues increased with fetal age and the FOXA2 immunostaining score remained constant in males and females (Additional file 4: Table S3).

Levels of transcripts encoding AHR, BAX, BCL, ESR1, FOXA2, and TPO increased with fetal age, irrespective of fetal sex, smoke exposure, or maternal BMI (Additional file 5: Table S4 and Additional file 7: Table S5). In contrast, levels of GATA4 transcripts decreased with fetal age. Male fetal thyroids had lower levels of AHR, ARNT, and AR transcripts than females while the BAX/BCL2 ratio increased with fetal age in males (Additional file 6: Figure S2C, Additional file 5: Table S4 and Additional file 7: Table S5), indicative of increased apoptosis.

Maternal smoke exposure accelerated the developmental increase of T4 regardless of fetal sex, and this was associated with a developmental increase in TSH levels in smoke-exposed female fetuses (Fig. 2a, Table 1). TSH levels did not, however, correlate with fetal plasma cotinine levels in either sex (Additional file 8: Figure S3Ai) and did not correlate with T4 in males (Additional file 8: Figure S3Aii). In contrast, fetal TSH levels did show a negative correlation with fetal T4 levels in control females and a positive correlation in smoke-exposed females (Additional file 8: Figure S3Aii). This developmentally dependent increase in TSH in smoke-exposed females was evident in both the BMI < 25 and BMI > 25 subgroups (Additional file 8: Figure S3Cii). Maternal smoke exposure did not affect thyroid weight and thyroid morphology in either sex (Additional file 4: Table S3), but reduced PAX8 immunostaining scores across gestation in smoke-exposed females (Fig. 2b, Additional file 4: Table S3). Smoke exposure also abolished the normal developmental increase in AHR transcripts, caused an age-dependent reduction in ESR2 transcripts, and reduced NKX2-1, GATA6, and TSHR transcript levels in thyroid tissue of both sexes (Fig. 2c, Additional file 5: Table S4).

High maternal BMI was associated with increased thyroid weight and with sex-specific alterations in thyroid morphology, with female thyroids from mothers with BMI ≥ 25 exhibiting a higher percentage of immature morphology (Fig. 3b, Additional file 4: Table S3). The fetal thyroid tissues that were scored as immature showed alterations resembling delayed development of angio-follicular units, with areas of dense thyroid tissue composed of predominantly immature thyroid follicles, scarcity of organized thyroid follicles, and marked blood vessel dilations (Fig. 3a). Maternal BMI status did not affect T3 and T4 levels but increased TSH levels in female fetuses (Fig. 4a, Table 1) with maternal BMI positively correlated with fetal TSH levels in females (P = 0.0004, R² = 0.37, Additional file 8: Figure S3B). To further understand the relationship between TSH, BMI, and smoke exposure in the females, the TSH levels were examined in relation to BMI in the control and smoke-exposed populations separately (Additional file 8: Figure S3C). The results showed that, irrespective of smoke exposure, increased maternal BMI was associated with increased fetal TSH (controls: P = 0.049, R² = 0.28; smoke-exposed: P = 0.003, R² = 0.48) and that, irrespective of maternal weight, smoke exposure was associated with increased developmental trajectory of TSH (Additional file 8: Figure S3Ciii-iv).

Fetal thyroid tissues from mothers of BMI ≥ 25 had reduced PAX8 scores while female thyroid tissues from mothers of BMI ≥ 25 were more likely to have lower FOXA2 immunohistochemical scores (Fig. 4b, Additional file 4: Table S3). Maternal BMI ≥ 25 was associated with altered development of AHR and ESR1 transcript levels in both sexes (Fig. 4c, Additional file 7: Table S5) and reversed the developmental trajectory of FOXE1 transcripts in female fetuses (Fig. 4c, Additional file 7: Table S5).

The lack of developmental changes in SLC5A5, TSHR, PAX8, NKX2-1, FOXE transcript levels; the developmental increase of TPO transcripts; the immunohistochemical localization of TTF1; and the developmental shift of NIS staining from perinuclear to basolateral observed here (Additional file 5: Table S4 and Additional file 7: Table S5; Fig. 1) are consistent with previous observations on the human fetal thyroid [28]. In addition, data described here show that the human fetal thyroid gland is sexually dimorphic and that the sexes respond differently to maternal smoking or increased BMI. The developmentally regulated transcription factors FOXA1, FOXA2, GATA4, GATA6, PAX8, NKX2-1, and SOX17 did not exhibit any statistically significant sex-specific differences in transcription (Additional file 4: Table S4 and Additional file 7: Table S5). Nevertheless, the male thyroids tended to show a more immature morphology (did not reach statistical significance), irrespective of smoke exposure or maternal BMI (Fig. 3b, Additional file 4: Table S3). This trend to an underlying phenotypic difference between the sexes requires further investigation. However, we observed two potential contributing factors: (i) immunoreactivity levels of PAX8, which is essential for the formation and organization of follicular cells in the thyroid gland [29, 30], were constant between 11 and 21 weeks in males but showed an age-dependent increase in female thyroid glands (Additional file 6: Figure S2B); (ii) the transcriptional expression of both AHR and ARNT, two key mediators of xenobiotic and detoxification responses, were reduced in male thyroid tissues (Additional file 5: Table S4 and Additional file 7: Table S5), suggesting that the male thyroids are more susceptible to potentially harmful environmental contaminants. Male fetuses also had higher circulating T3 levels and a higher T3/T4 ratio which may be the consequence of both increased NIS levels during gestation and increased rates of T4-to-T3 deiodination in fetal tissues (Additional file 6: Figure S2A and S2B). Consistent with the latter, male fetuses are known to have a higher metabolic rate than females [31]. Overall, the data suggest more rapid organizational development in the female with greater sensitivity to disruption and changes in thyroid hormones in the males.

Of the key thyroid transcripts levels quantified, maternal smoking downregulated the transcription factors GATA6 and NKX2-1 in both male and female thyroids (Fig. 2c, Additional file 5: Table S4), suggesting a disruption of fetal thyroid development, which may affect T4 production and release. TTF1 (gene name: NKX2-1) is under the control of GATA6 [32], and a GATA6-TTF1 protein complex directly affects the expression of various genes [33]. The TTF1 transcription factor also controls the expression of TSHR [34], and here, the reduced thyroid TSHR transcript (Additional file 5: Table S4) levels is a possible consequence of reduced GATA6 and NKX2-1 transcripts. Even though the downregulation of GATA6 and/or NKX2-1 transcripts by maternal smoking during the second trimester may contribute to neonatal phenotypes such as thyroid enlargement and/or reduction of TSH levels [18, 19], it is uncertain whether the small changes for both GATA6 and NKX2-1 (Additional file 5: Table S4) can, alone, account for profound alterations on the thyroid.

Since the limited volumes of fetal plasma available for the analyses performed in this study precluded measurement of free T3 (fT3) and free T4 (fT4) in the fetal circulation, we also measured fetal circulating binding proteins: ALB, TTR, and TBG (Table 1). Levels of ALB, TTR, and TBG in the fetus were unaffected by maternal smoking (Table 1), and so it is likely that changes in total T3 and total T4 levels measured here reflect changes in fT3 and fT4. As T3 and the T3/T4 ratio were unaffected by smoke exposure, the increase in the developmental trajectory of T4 is unlikely, however, to reflect altered deiodination to T3 (Fig. 2a, Table 1). Instead, it may suggest an increased sensitivity of the fetal thyrocytes to TSH, although THSR transcript levels are reduced in smoke-exposed fetuses (Additional file 5: Table S4). It is also possible that the apparent age-dependent increase in TSH levels in smoke-exposed female fetuses (Fig. 2a) may be because TSH is initially suppressed by maternal smoking (Fig. 2a, Table 1). The age-dependent increase in TSH levels in smoke-exposed female fetuses (Fig. 2a) may account for the accelerated rate of T4 production, although no such effect was observed in males (Table 1), suggesting a complex regulatory network. It is likely, therefore, that the hypothalamic-pituitary-thyroid axis (HPT) in the fetus is not only sexually dimorphic but also affected by smoke exposure in a sex-specific manner. The sexual dimorphism of the HPT axis and in particular TSH levels are influenced by many factors, including sex steroids and adrenal function, and are subject to diurnal variation, as has previously been demonstrated in rats and humans [35‐39]. Our studies show that AR is expressed in the human fetal thyroid and that male fetal thyroids have reduced AR transcript levels (Additional file 6: Figure S2C, Additional file 5: Table S4 and Additional file 7: Table S5). In addition, smoke exposure altered the developmental trajectory of thyroid ESR2 transcripts (Fig. 2c, Additional file 5: Table S4), suggesting that sex-specific effects of maternal smoking on the fetal thyroid may be mediated by fetal sex hormones. In support of this, non-smoke-exposed female fetuses show an inverse relationship between TSH and T4 in agreement with current population data [40], which is absent in males and is reversed in smoke-exposed females (Additional file 8: Figure S3Aii). Finally, our studies show that maternal smoking reduced immunoreactive PAX8 in female fetuses only, which may affect the differentiation as well as the TSH responsiveness of hormone-producing fetal thyrocytes.

In this study, a maternal BMI above 25 was associated with a small but significant increase in fetal thyroid weights of both sexes although total circulating thyroid hormone levels remained unchanged (Fig. 2b, Table 1, Additional file 4: Table S3). High maternal BMI coincided with an immature histological phenotype, preferentially affecting the thyroid gland of female fetuses (Fig. 2b). Dysregulation of morphogenic transcription factors such as FOXA1, FOXA2, and FOXE1 can cause morphological deviations, and in this study, the immunoreactivity of FOXA2 and the expression of FOXE1 were affected in female thyroids from mothers with BMI ≥ 25 (Fig. 4b, c, Additional file 4: Table S3 and Additional file 7: Table S5). It is possible that the developmental changes in FOXA2 and FOXE1 expression control thyroid folliculogenesis and that deviations seen in the female fetuses from mothers of BMI ≥ 25 contributed to the perturbed thyroid folliculogenesis. The FOXA2 transcription factor is expressed in the C cell precursors of the mouse fetal thyroid, and it may modulate the expression of TTF1 while also playing a role in thyroid regeneration [41‐43]. FOXA2 distribution in the human thyroid suggests similar developmental functions, with FOXA2 staining sometimes resembling the distribution of calcitonin-positive C-cells (Fig. 1e, middle panel compared to Fig. 1c, right panel).

As in the case of smoke exposure, changes in the expression of fetal thyroid ESR1 transcripts in overweight mothers (Fig. 4c, Additional file 7: Table S5) may have affected responsiveness to sex steroids and may, therefore, contribute to sex-specific responses to maternal BMI. The increased circulating TSH levels, with unaffected thyroid hormone levels, in female fetuses from mothers with BMI ≥ 25 are consistent with data that have been reported for overweight and obese adults [44, 45]. Taken together, this suggests altered hypothalamic-pituitary function and/or reduced fetal thyroid sensitivity to TSH in female fetuses.

The fetal thyroid gland plays an essential role in the control of fetal metabolic rate, cardiac output, and brain development while normal thyroid gland function is essential for post-natal health. Maternal lifestyle factors, such as smoking and increased BMI, are associated with adverse outcomes for the child, and here, we examined the effects of those factors on the human fetal thyroid system. Unfortunately, this study lacks knowledge of the thyroid status of the consenting women in this study which could have explained, at least in part, fetal thyroid alterations observed here in response to smoking and BMI. Nevertheless, this is the first study to demonstrate adverse effects on the developing thyroid system associated with maternal lifestyle during the second trimester. Critically, this is the period when the fetal thyroid begins to secrete hormones, and it is likely that changes in thyroid function at this time will have a significant impact on fetal development with implications for the fetal programming of adult diseases.