As a model of developmental overexposure to T3, we used mice genetically deficient in the type 3 deiodinase (DIO3). We have previously described that Dio3 -/- mice exhibit markedly elevated serum levels of T3 during fetal and early neonatal life (34). All experimental mice were on an outbred CD-1 genetic background to overcome the impaired fertility of Dio3-/- mice on inbred genetic backgrounds. The original mutant mouse strain was generated in a 129/SvJ genetic background (34, 39) and has been backcrossed on a CD-1 background for more than 12 generations. Due to the genomic imprinting and preferential expression of the Dio3 gene from the paternal allele (39–41), the colony has been maintained for more than eight generations by crossing wild type males with heterozygous females, so that the heterozygous mice generated for colony maintenance are phenotypically normal, as they carry the Dio3 mutation in the maternal allele, which is already largely suppressed due to genomic imprinting (39, 41). To avoid the influence of confounding factors and minimize biological variability, mothers of experimental animals were mated at two months of age with 2-3 month old males. Experimental animals were born before the mother was four months old. Experimental mice represent only first litters from three to seven different mothers, and litter size was limited to 8-12 pups at birth. Mothers of experimental mice were single-caged before giving birth to avoid paternal effects and a concurrent pregnancy while raising the pups. Animals for adult studies were weaned at the age of three weeks, caged in groups of three or four, and sacrificed at 18-20 weeks of age within the third and fifth hour of the light cycle (~Zeitgeber time 3-5). All mice were maintained on a 12 h light/dark cycle, and food (Envigo-Harlan Teklad T.2918.15 for non-breeders and T.2919.10 for breeders) and water were provided ad libitum. For fasting experiments, mice were fasted overnight for 16 h before euthanasia. Also, some neonatal determinations were performed in mice on a 50/50 mixed 129/SvJ/C57BL/6J genetic background, using the same experimental design. This subset of experimental animals was generated by mating 129/SvJ males of the appropriate genotype with wild type (WT) C57BL/6J females. All mice were euthanized by CO₂ asphyxiation at a rate of 30-50% volume of the euthanasia chamber per minute, according to the guidelines of the American Veterinary Medical Association. Tissues were harvested and immediately frozen on dry ice until further use. All experiments were approved by the MaineHealth Institute for Research Institutional Animal Care and Use Committee (IACUC), under current protocol number 2112.

Experimental mice were generated by crossing Dio3-/- males with Dio3+/+ females ( Figure 1 ). We have previously shown that Dio3-/- mice are overexposed to T3 during fetal and early neonatal life (34), and that the neonatal germ cells and adult sperm of Dio3-/- males show alterations in DNA methylation (36). The offspring of these males will be heterozygous for the Dio3 mutation (Dio3^m+/p-) and will have a mutated paternal Dio3 allele. To generate suitable control mice, we crossed wild type females with Dio3^m-/p+ males ( Figure 1 ). This cross will generate mice of exactly the same genotype (Dio3^m+/p-) as experimental mice, but in this case the father will be phenotypically normal (not overexposed to T3 during development (40)) as the mutated allele is the maternal one. Thus, both control (non-exposed fathers, NF mice) and experimental mice (exposed fathers, EF mice) had exactly the same genotype and were born to independent wild type females, the only difference between both groups being the T3 overexposure status of the fathers.

Fat and lean mass were measured in isofluorane-anesthetized mice using a Lunar PIXImus II DEXA Densitometer (Fitchburg, WI). During the 5 min of this scanning procedure, mice were maintained anesthetized with a constant airflow containing 3% isofluorane and 3% oxygen.

Blood from adult and neonates was collected from the inferior vena cava, allowed to clot at 4 °C for at least four hours. After 10 minutes centrifugation at 3000 g, serum was taken from the supernatant and stored at -70 °C until further used. Serum levels of T3 and T4 were measured in 10 and 1 μl of serum, respectively, as previously described (42–44) by highly sensitive specific radioimmunoassays using in-house generated antibodies. Serum determinations of leptin, adiponectin and ghrelin were performed using commercially available ELISA kits according to the manufacturer’s instructions. Leptin and Adiponectin kits were purchased from R&D Systems (Bio-Techne, Minneapolis, MN, USA); ghrelin kit was purchased from EMD Millipore Corporation (Billerica, MA, USA).

Mouse tissues were harvested and subsequently frozen on dry ice. Total RNA was extracted using the RNeasy kit from Qiagen (Valencia, CA). Total RNA (1 µg) was reverse transcribed with M-MLV reverse transcriptase in the presence of random decamers (both from Thermo Fisher Scientific, Waltham, MA) at 65°C for 5 min, then 37°C for 50 min. The 20 μl reverse transcription reactions were DNAse treated and diluted by adding 230 μl of RNase free water. An aliquot of each sample was mixed together for an internal standard and diluted fourfold. Real-time PCR reactions were set up in duplicate with gene-specific primers and SYBR Select Master Mix (Thermo Fisher Scientific, Waltham, MA), and run on the CFX Connect from Bio-Rad (Hercules, CA), where they underwent an initial 10 min denaturing step, followed by 36 cycles of a denaturing step (94°C for 30 s) and an annealing/extension step (60°C for 1 min). For each individual sample, expression was corrected by the expression of one or more (geometric mean) housekeeping genes (Gapdh, Actb, Ppia or Rn18s), if they did not exhibit any significant difference in expression between experimental groups. Expression data are shown in arbitrary units and represented as fold-change over the mean value in the control group, or as fold-change over baseline values for fasting experiments. The sequences of the primers used for each gene determination are shown in Supplementary Table 1 .

Statistical analyses were performed using the statistical tools of GraphPad Prism 6 (GraphPad Software, Inc.). Given the sexual dimorphisms in body weight and endocrine parameters, adult females and males were analyzed separately. A Student’s t-test was used to determine statistical significance at baseline between the two experimental groups. We used one-way ANOVA and Tukey’s post hoc test to determine statistical significance in comparisons between experimental groups or treatments. Unless otherwise indicated, responses to fasting (gene expression and hormone concentrations) were calculated as a ratio between post-fasting and corresponding baseline values, or as an increment (body composition) between post-fasting and corresponding baseline. Statistical significance was defined as P<0.05. Data are represented as the individual values of independent biological samples and the mean ± SEM. The number of mice (independent samples) used per experimental group is indicated in each figure and typically varied between 17-22 for baseline body composition, 5-12 for the rest of the adult studies, and 14-23 for neonatal studies.

We examined weight and body composition in mice (EF mice) fathered by males overexposed to T3 during development, and compared them with control mice (NE mice) of the same genotype (see Methods) fathered by phenotypically normal males ( Figure 1 ). We observed that EF females showed a marked increase in body weight that was the result of increased adiposity and not of changes in lean mass ( Figure 2A ). Compared to control males, EF males did not exhibit significant differences in body weight or lean mass, but we noted a mild increase in fat mass and a significant increase in fat percentage ( Figure 2B ).

We subjected a subset of animals to a 16 h fasting and determined individual changes in body composition. Both EF and NE female lost total weight after fasting, and this was largely due to a decrease in lean mass, not fat mass ( Figure 3A ). However, this post-fasting loss in body weight and lean mass was significantly more pronounced in EF females than in NE females ( Figure 3A ). Regarding males, both EF and NF males showed similar reductions in lean mass after fasting ( Figure 3B ), but total body weight loss was larger in EF males ( Figure 3B ). This difference, in contrast to females, was accounted for by a lack in fat mass increase, which was noted in NF males ( Figure 3B ).

We then investigated hormone levels and gene expression of relevance to the thyroid axis. We observed no differences between EF and NF females in serum levels of T3 and T4 ( Figure 4A ). In males, no difference in serum T4 was noted, but EF males showed significantly increased serum T3 ( Figure 4B ). After fasting, females and males of both experimental groups manifested a reduction of serum T4 from baseline levels, but this reduction was comparable between EF and NF mice of either sex ( Figure 4C ).

We measured gene expression in the thyroid gland of markers thyroid-stimulating hormone receptor (Thsr), thyroid peroxidase (Tpo) and iodotyrosine deiodinase (Iyd). We observed that the expression of all three markers at baseline was significantly elevated in EF females ( Figure 5A ), but significantly decreased in EF males ( Figure 5B ). Although the expression of these markers was higher in EF female thyroid glands, their expression after fasting was more pronouncedly reduced from baseline levels than in NF females. ( Figure 5C ). In contrast, fasting elicited comparable reductions in the expression of these markers in EF and NF males, and we only noted that Tpo fasting response was not as large in EF males as in NF males ( Figure 5D ).

Pituitary and hypothalamic gene expression did not show major differences between EF and NF females. We observed increased pituitary Trhr1expression in EF females, and a slight, not significant, increased in Tshb expression but no changes in type 2 deiodinase (Dio2), Dio3, or T3 target kruppel-like factor 9 (Klf9) ( Figure 6A ). EF females also exhibited decreased hypothalamic expression of thyrptropin-releasing hormone (Trh) at baseline, but no differences in the expression of Dio2, Dio3 or T3-target hairless (Hr) ( Figure 6B ). We observed no differences between EF and NF females in the fasting response of pituitary expression of Klf9, Dio2 or Dio3. However, pituitary thyrotropin-releasing hormone receptor 1 (Trhr1) expression robustly increased in NF females after fasting, but this response was not observed in EF females ( Figure 6C ). In addition, a post-fasting reduction in pituitary thyroid-stimulating hormone beta chain (Tshb) expression was noted in both female groups, but this response was significantly more pronounced in EF females ( Figure 6C ). Regarding hypothalamic expression, we observed a differential response to fasting in the expression of Dio2 and Dio3, while the increased expression of T3-target Hr was similar in both groups ( Figure 6D ).

In males, thyroid axis-related pituitary and hypothalamic expression did not show remarkable differences between EF and NE groups ( Supplementary Figure 1A–D ). At baseline, pituitary expression of Dio3 was reduced in EF males, a trend also observed in Trhr1 expression ( Supplementary Figure 1A ), while experimental groups showed no differences in the hypothalamic expression of the genes examined ( Supplementary Figure 1B ). After fasting, male pituitary expression of Dio2, Dio3, Trhr1 and Tshb was decreased to different extents compared to baseline levels, but these reductions were similar between EF and NF males ( Supplementary Figure 1C ). In the hypothalami of males, fasting elicited minor or no changes in gene expression, and no differential responses were observed between EF and NF males ( Supplementary Figure 1D ).

We also investigated hormone and molecular endpoints of the leptin-melanocortin system that may be associated with the adiposity phenotypes. Consistent with their body composition, serum leptin was markedly elevated in EF females, while we observed no differences in serum adiponectin ( Figure 7A ). After fasting, leptin levels were slightly reduced in both experimental groups of females, but the response was comparable in both experimental groups ( Figure 7B ). Serum adiponectin did not change in either group after fasting ( Figure 7B ). Interestingly, we observed that serum ghrelin, an orexigenic hormone, was significantly reduced at baseline in EF females ( Figure 7A ). After fasting, as expected, serum ghrelin was markedly increased in both experimental groups, but the increase was significantly larger in EF females ( Figure 7B ).

We observed no differences between EF and NF males in serum levels of leptin, adiponectin or ghrelin ( Supplementary Figure 2A ). However, the post-fasting decrease in serum leptin was significantly more pronounced in EF males than in NF males ( Supplementary Figure 2B ). Fasting adiponectin in EF males also trended lower than in NF males, although it did not reach statistical significance ( Supplementary Figure 2A ). The changes in serum ghrelin after fasting were large in both groups of males and of similar magnitude.

We then determined the hypothalamic expression of genes related to the leptin-meanocortin system and energy balance. We observed that EF females show significant decreases in the hypothalamic expression of leptin-melanocortin system components, including pro-opiomelanocortin (Pomc), agouti-related peptide (Agrp), neuropeptide Y (Npy) and melanocortin receptor 4 (Mc4r) ( Figure 8A ), while no differences were noted in the expression of other determinants of energy balance like uncoupling protein 2 (Ucp2) and bone morphogenetic protein receptor 1a (Bmpr1a). After overnight fasting, responses relative to their corresponding baselines were significantly more pronounced in EF females, with larger changes in all the markers measured ( Figure 8B ).

These results were sexually dimorphic, as they were not observed in EF males. At baseline, hypothalamic expression of all those markers was not different between EF and NF males ( Supplementary Figure 3A ). In this regard, fasting responses in EF and NF males were similar for Pomc, Agrp, Npy and Ucp2, while we observed a modest difference for Mc4r and Bmpr1a. For these genes, the post-fasting expression levels were slightly different between experimental groups, with their expression values in EF males being higher than in NF males due to an absence of post-fasting decrease or the presence of a post-fasting increase, respectively ( Supplementary Figure 3B ).

Given the increased adiposity in EF females, we explored relevant gene expression in white adipose tissue (WAT). Compared to control females, the sub-cutaneous WAT of EF females showed increased expression of adiposity markers mesoderm-specific transcript (Mest) and leptin (Lep) (the latter did not reach statistical significance), but no differences in the expression of adiponectin (Adipoq), adipocyte progenitor marker delta-like non canonical notch ligand 1 (Dlk1) or T3-generating, beiging marker Dio2 ( Figure 9A ). However, gonadal WAT manifested remarkable changes in gene expression. In this fat depot, adiposity markers Mest and Lep were strongly up-regulated, while Dlk1 expression trended higher ( Figure 9B ). Importantly, we also observed that Dio2 expression was markedly suppressed in EF gonadal fat, but this decrease was not accompanied by reduced expression of T3 target genes Klf9, Hr and glycerol phosphate dehydrogenase 2 (Gpd2) ( Figure 9B ).

We then examine serum levels of leptin and T4 during late neonatal development. At postnatal day 14 (P14), serum leptin was markedly reduced in EF mice compared to NF mice ( Figure 10A ). The lower leptin in EF mice was comparable to that in their T3-overexposed fathers (Dio3-/- males) at the same age, while the higher leptin levels in NF mice were similar to those in genetically intact Dio3+/+ mice ( Figure 10A ). At P20, serum levels were significantly higher in EF mice than in NF mice ( Figure 10A ), suggesting a pronounced disruption in serum leptin ontogeny in EF animals. At P14, we observed no differences in serum adiponectin between EF and NF mice, and values were similar to those in Dio3+/+ animals ( Figure 10B ). Similarly, EF, NF and Dio3+/+ mice showed comparable levels of serum T4 at P14, while values at this age for Dio3+/+ mice were markedly suppressed ( Figure 10C ).

Interestingly, using the same experimental design with animals on a 50/50 mixed genetic background (129/SvJ with C57BL/6J, see Methods), we observed that both EF and NF mice have reduced serum T4 compared to Dio3+/+ mice at P4 ( Figure 10D ). By P15, we still observed lower serum T4 in EF and NF mice compared to Dio3+/+ values, but there was a pronounced divergence in serum T4 between EF and NF mice, with the latter trending higher and closer to Dio3+/+ values, while T4 in EF mice remained strongly suppressed, with values comparable to those in Dio3-/- mice ( Figure 10D ).