Offspring body composition after maternal T3 treatment

Two-way ANOVA revealed a significant main effect of maternal T3 treatment during pregnancy on body weight (g), total fat mass, and gWAT mass relative to body weight (%) in adult offspring (P < 0.02, Fig. 1B-D). Subsequent pairwise comparisons stratified by sex revealed a significant effect on body weight specifically in female offspring (P = 0.017, Fig. 1B), with similar trends observed for both sexes and fat mass. Apart from this, serum levels of thyroid hormones (T3, T4, and TSH) were not altered in offspring of T3-treated dams compared with control offspring [23], which is in agreement with the findings of similar previous studies [40, 41].

Leptin expression and methylation levels in offspring after maternal T3 treatment

We observed a significant interaction effect between sex and maternal T3 treatment on Lep mRNA levels in offspring gWAT (F(1,16) = 11.18, P = 0.004, Fig. 1E), suggesting a sex-specific effect of maternal T3 treatment on Lep mRNA levels in the offspring. While significantly lower Lep mRNA levels were detected in the female offspring of T3-treated mothers than in the control offspring (P = 0.009, Fig. 1E), the Lep mRNA levels in the gWAT of the male offspring born to the T3-treated dams were, albeit higher, not significantly different (P = 0.167; Fig. 1E) from that of the control dams.

The gWAT DNA methylation at the previously reported enhancer region ~ 36 kb upstream of Lep revealed no significant interaction effect of sex and T3 treatment. However, DNA methylation was significantly higher in the female offspring of the T3-treated mothers (median methylation difference = 5.9%, P = 0.03; Fig. 1F) than in those of the control animals. No differential methylation was detected between male offspring born to T3-treated and control dams (median methylation difference = 0.3%, P = 0.714; Fig. 1F). For leptin serum levels in offspring, no significant interaction effect of sex and T3 treatment was found, but a simple main effect of T3 treatment was evident (P = 0.003; Fig. 1G). After stratification by sex, the difference was significant only for the male offspring (P = 0.034; Fig. 1G), but for the female offspring of the T3-treated mothers, a trend towards lower leptin serum levels was still observed (P = 0.058, Fig. 1G) than that of the controls.

Irrespective of maternal treatment, we observed a strong positive correlation of Lep mRNA expression in gWAT as well as leptin serum levels with gWAT mass (% of body weight), total fat mass (% of body weight), and body weight (Spearman rho > 0.6, FDR < 0.05; Fig. 1H) in female offspring. In male offspring, we observed this positive correlation only for leptin serum levels with gWAT mass (% of body weight), total fat mass (% of body weight), and body weight (Spearman rho > 0.8, FDR < 0.05; Fig. 1I), whereas Lep mRNA expression in gWAT did not correlate with any trait. In female offspring, Lep UE methylation was negatively correlated with body weight and leptin serum levels (Spearman rho < − 0.6, FDR < 0.05; Fig. 1H) but not significantly correlated with Lep mRNA expression in female offspring. In male offspring, Lep UE DNA methylation in gWAT did not correlate with any trait at all (all FDRs > 0.05; Fig. 1I).

Genome-wide expression changes in offspring gWAT after maternal T3 treatment

To further explore the molecular determinants of altered body composition in offspring of the T3-treated vs. control dams, we performed microarray gene expression analysis in gWAT of adult offspring. While the effect sizes of maternal T3 treatment on the adult offspring might not be sufficient to detect genome-wide significant differences in gene expression after correction for multiple testing, we identified several DEGs using relaxed cut-offs with |log2 FC|> 0.5 and an unadjusted P < 0.01 (females: 291 DEGs, Suppl. Table 2; males: 862 DEGs, Suppl. Table 3). A summary of the most significant DEGs can be found in volcano plots (Suppl. Figure 1). Notably, Lep was among the most significantly altered genes in female offspring (log2 FC = -0.84, P = 0.0026), which was consistent with our quantitative real-time PCR results. To gain further insight, we performed a Wikipathway overrepresentation analysis on upregulated and downregulated genes in adult offspring following maternal T3 treatment. The analysis of the female offspring revealed that upregulated genes (Creb1, Foxo1, Gata2, Nr2f2, Rora, Klf5) are involved in ‘white fat cell differentiation’ (WP2872, FDR = 0.048; Suppl. Figure 2A, Suppl. Table 4). Additionally, six (Mvk, Dhcr7, Lss, Idi1, Hmgcr, and Sqle) of the fifteen genes of the ‘cholesterol biosynthesis’ pathway (WP103, FDR = 0.001) were among the downregulated genes in the female offspring of T3-treated dams compared with those of the control animals (Suppl. Figure 2A, Suppl. Table 4). In gWAT of male offspring, this analysis revealed that gene expression of 20 out of 268 ‘non-odorant G-Protein coupled receptors’ (GPCRs) was upregulated (WP1396, FDR = 0.035), whereas the expression of 16 genes from 81 encoding ‘cytoplasmic ribosomal proteins’ (WP 163, FDR = 2 × 10–6) was downregulated in offspring of the T3-treated dams compared with those of the control dams (Suppl. Figure 2B, Suppl. Table 5).

In vitro hypomethylation of Lep upstream enhancer using a dCas9-Suntag-TET1 system

To investigate the functional role of DNA methylation at the Lep UE in vitro, the target region was hypomethylated in immortalised murine epididymal preadipocytes using a system consisting of a dCas9–SunTag and scFv–TET1 catalytic domain (Fig. 2A). The workflow of the experiment is illustrated in Fig. 2B. After transfection, we achieved an approximately 20% reduction of Lep UE DNA methylation in successfully transfected cells (mean DNA methylation ± SD = 71.0 ± 4.6%) compared with that in untreated cells (mean DNA methylation ± SD = 90.4 ± 6.4%) on the day of induction (= day 0). We further observed no significant increase in DNA methylation during adipocyte differentiation in the transfected cells (Suppl. Figure 3), indicating a stable reduction in DNA methylation during differentiation. After combining DNA methylation at all time points, we observed significantly different DNA methylation rates between the treatments (P = 5.1 × 10–4, Fig. 2C), and subsequent pairwise comparisons confirmed significantly lower DNA methylation in the cells transfected with the vector targeting the Lep UE (median DNA methylation [IQR] = 73.6 [71.4–74.3] %) than in the untreated cells (median DNA methylation [IQR] = 91.4 [83.8–96.2] %, FDR < 0.001) and the cells transfected with the plasmid control without gRNA (median DNA methylation [IQR] = 88.5 [75.9–94.1] %, FDR < 0.001). We observed no significant difference in DNA methylation between the cells transfected with the plasmid control and the untreated cells (FDR = 0.87), indicating that there was no significant global effect due to increased amounts of the TET1 catalytic domain per se.

Fig. 2

In vitro Lep UE hypomethylation and effect on adipocyte differentiation. A Principal scheme of targeted in vitro hypomethylation using a dCas9-Suntag-TET1 system showing in the top the ‘all in one’ vector which contains dCas9 peptide array (linker length: 22 amino acids), antibody-sfGFP-TET1CD, and gRNA expression system. Scheme was generated using biorender.com. B Workflow of the experiment, which is described in detail in the methods section. For the hypomethylation, epididymal preadipocytes were transfected by electroporation with all-in-one vectors including three different gRNAs targeting the Lep UE (= Lep UE gRNA, red). For control, we used untreated cells (= UT, dark grey) and cells transfected with the all-in-one vector without gRNA (= plasmid control, light grey). Workflow was generated using biorender.com. C Boxplot shows DNA methylation level [%] of Lep UE across all three analysed CpG sites by bisulphite pyrosequencing. Significance of differences between treatments was calculated using Kruskal–Wallis test, followed by pairwise comparisons by Wilcoxon rank-sum test corrected for multiple testing by FDR (***FDR < 0.001, ns FDR > 0.05). Dots represent mean methylation levels of days 0, 2, 4, 6, and 8 of differentiation for each experiment (n = 3 experiments × 5 time points). D Barplot shows the mean of log transformed Lep mRNA levels normalised to Rplp0 mRNA levels for days 4, 6, and 8 of adipocyte differentiation from n = 3 experiments. Results of mixed two-way ANOVA used to assess the effect of treatment and time on Lep expression are shown on top of the graph. Pairwise comparisons of treatment effects using paired Student's t-test (paired by experiment number) are depicted in the graph. No significant differences were found after correction for multiple testing by FDR. Shown significance level are uncorrected for multiple testing: #P < 0.05. E Barplot shows the lipid amounts in cells during differentiation on days 0, 4, and 8 between treatments. Shown is the average ratio of Adipored and Hoechst fluorescence intensity from n = 12 wells in n = 3 experiments. Results of mixed two-way ANOVA used to assess the effect of treatment and time on lipid accumulation are shown on top of the graph. Significance levels from pairwise comparison of time points using paired Student's t-test (paired by experiment number) and corrected for multiple testing by FDR are depicted in the graph (***FDR < 0.001, **FDR < 0.01, and *FDR < 0.05). No significant differences are found when comparing treatments

Effects of upstream enhancer hypomethylation on Lep mRNA levels during adipocyte differentiation in epididymal preadipocytes

The effects of UE hypomethylation on the Lep mRNA level and lipid accumulation during differentiation were evaluated. Since Lep mRNA levels were not detectable on the day of induction or 2 days after induction (Ct values > 33), only the mRNA levels on day 4 and later were considered. No significant interaction effect of treatment and time on the Lep mRNA level was detected (ANOVA, F(4,12) = 0.19, P = 0.94; Fig. 2D). A simple main effect of time on the Lep mRNA level was found, indicating that the Lep mRNA level increased over the course of adipocyte differentiation (ANOVA, F(2,12) = 110.119, P = 1.9 × 10–8). We did not observe a significant effect of treatment on the Lep mRNA level (ANOVA, F(2,12) = 0.309, P = 0.745). After pairwise comparisons, we observed nominal lower Lep mRNA levels in the transfected cells than in the untreated cells at days 4 and 8 (P < 0.05, Fig. 2D). However, this trend of lower Lep expression was also observed in the cells transfected with the plasmid control compared to untreated cells (P < 0.05). On day 6, we observed slightly lower Lep expression in treated cells (Lep UE gRNAs) than in plasmid control cells (P < 0.05). The observed differences were moderate, and none of these comparisons were statistically significant different after correction for multiple testing. Overall, Lep expression remained low even after differentiation (day 8 Ct values > 28). Consistently, leptin secretion into the cell culture medium was minimal, making ELISA measurements unreliable and not robustly evaluable.

Effects of upstream enhancer hypomethylation on lipid accumulation during adipocyte differentiation in epididymal preadipocytes

The effect of Lep UE hypomethylation on lipid accumulation during adipocyte differentiation was assessed via AdipoRed staining of lipids. No significant interaction effect on the lipid accumulation was detected between hypomethylation and time (ANOVA, F(2.08, 6.24) = 0.24, P = 0.8; Fig. 2E). A significant increase of lipids during the course of adipocyte differentiation (ANOVA, F(1.04,6.24) = 23.998, P = 0.002) was confirmed by pairwise comparisons (FDR < 0.05, Fig. 2E), which supported the accumulation of lipids during differentiation. However, no significant difference was detected in cells after hypomethylation of the Lep UE compared with the control cells (Fig. 2E).

Effects of in vitro hypomethylation of the Lep UE in other adipocyte cell lines

Additionally, we performed experiments in 3T3L1 cells as well as in immortalised female cells from the inguinal (subcutaneous) fat depot. In 3T3L1 cells, we achieved on average a 35% hypomethylation of the Lep UE in transfected vs. untreated cells (Suppl. Figure 4A). In these cells, on day 4, a significant difference in the Lep mRNA levels was detected between hypomethylated and untreated cells (FDR < 0.01; Suppl. Figure 4B), but not compared with that in the plasmid control cells, and this difference was not detected in the following days of differentiation. Additionally, in 3T3L1 cells, we observed no effect of Lep UE hypomethylation on lipid accumulation (Suppl. Figure 4C). In the inguinal cells from female mice, on average 21% hypomethylation of Lep UE (transfected vs. untreated inguinal cells, Suppl. Figure 5A) was observed at all time points. Although we were not able to detect Lep mRNA expression despite successful adipocyte differentiation, we observed a trend towards reduced lipid accumulation after hypomethylation of the Lep UE in these cells (Suppl. Figure 5B/C).

Sex-specific differences in leptin expression and methylation levels in human visceral adipose tissue

Bisulphite sequencing of a corresponding LEP UE region in human OVAT (N = 52) revealed significantly higher methylation levels in female (N = 31, median methylation [IQR] = 65.3 [60.8–68.5] %) than in male individuals (N = 21, median methylation [IQR] = 59.1 [53.2–65.5] %, P = 0.031, Table 1). In the entire cohort, in female subjects, we observed a significantly lower OVAT LEP mRNA expression (N = 518, P = 0.045, Table 1), whereas leptin serum levels were significantly higher (N = 485, P < 0.0001, Table 1) than in males.

Sex-specific causal relationship of human LEP upstream enhancer methylation with body fat percentage

Correlation analysis with DNA methylation at the LEP upstream enhancer revealed a significant positive correlation with body fat (% of body weight) in female individuals with obesity (Spearman rho = 0.61, FDR < 0.05; Fig. 3A). We also observed a positive correlation between LEP methylation and leptin serum levels in female individuals (Spearman rho = 0.38, FDR < 0.1; Fig. 3A). We found no significant association between LEP UE methylation and LEP mRNA level (FDR > 0.1). In male individuals with obesity, no correlation of the LEP UE methylation with any trait was identified (all FDRs > 0.1; Fig. 3B). Additionally, LEP mRNA levels in OVAT and leptin serum levels correlated with no other trait in males. Since LEP UE methylation did not correlate with LEP mRNA expression in OVAT, we addressed another causal relationship of DNA methylation with body fat percentage. First, we tested in a mediation analysis whether LEP UE DNA methylation functions as a mediator for the sex-specific difference in body fat percentage (Fig. 3C, Table 2). In our model, the sex-specific effect on body fat percentage was fully mediated via DNA methylation at three of the six analysed CpG sites (CpG 4–6: ADE P > 0.1, ACME P < 0.05; Table 2) and partially mediated at CpG site 1 (ADE P < 0.1, ACME P < 0.05). The percentage of mediation ranged from 30.8% (CpG site 1) to 48.1% (CpG site 4). Since LEP UE methylation correlated with the body fat percentage only in females, we further used a second mediation analysis to test whether this association was mediated via the leptin serum level in females (Fig. 3D, Table 3). The second mediation analysis revealed that for DNA methylation at CpG sites 1, 4, and 6 as well as for the mean DNA methylation across all analysed CpG sites, the effect on body fat percentage was partially mediated via the leptin serum level (ADE P < 0.01, ACME P < 0.05; Table 3). The percentage of the mediation effect ranged from 27.6% (CpG site 6) to 32.5% (CpG site 4).

Fig. 3

Relationship between sex-specific differences in leptin expression and enhancer DNA methylation levels in human visceral adipose tissue with body fat percentage. Heatmaps show Spearman correlation coefficients of enhancer DNA methylation, mRNA level, and serum levels of leptin with anthropometric phenotypes for female (A) and male (B) individuals with obesity. Positive correlations are shown in petrol and negative correlations in brown. Non-significant correlations (FDR < 0.1) are crossed out. C Scheme for causal mediation analysis exploring the mediation of sex-specific differences in body fat percentage via DNA methylation at the LEP upstream enhancer. Results of the mediation analysis are shown in Table 2. D Scheme for causal mediation analysis in female individuals exploring the mediation of the effect of DNA methylation at the LEP upstream enhancer on body fat percentage via leptin serum level. Results of the analysis are shown in Table 3

Table 2 Causal mediation analysis of sex > LEP UE methylation > body fat percentageTable 3 Causal mediation analysis of LEP UE methylation > LEP serum level > body fat percentage (in female individuals)