Maternal Obesity Induces Epigenetic Modifications to Facilitate Zfp423 Expression and Enhance Adipogenic Differentiation in Fetal Mice

Animal studies were conducted according to the protocol approved by the Animal Use and Care Committees of both the University of Wyoming and Washington State University. The diet-induced obesity model was used. Female C57BL/6J mice (4 weeks of age) were fed ad libitum either a control diet (D12450B; Research Diets, New Brunswick, NJ) with 10% energy from fat (Con) or a high-energy diet (D12451) with 45% energy (OB) from fat for 8 weeks at 12-h light/12-h dark cycles. Then, both Con and OB female mice were mated with C57BL/6J male mice (on the normal rodent diet), and mating was confirmed by the presence of a vaginal plug. During pregnancy, each female was individually housed. At E14.5, mice were killed and fetuses were collected and weighed. Then, under a dissecting microscope, fetal tissue was harvested after removing the head, heart, lung, and liver; surface gelatinous tissue; and spinal cord and primordial bones. Due to the small size, fetal tissue from the same litter was pooled, and the litter was considered as an experimental unit.

Real-time quantitative PCR was conducted as previously described. Primer sequences and the amplicon sizes are listed in Table 1. Relative expression of mRNA was determined after normalization to the 18S reference using the ΔΔ-Ct method (26).

Western blotting was conducted as previously described (26). Polyclonal antibody against Zfp423 was purchased from Santa Cruz, against EZH2 from Cell Signaling Technology (Boston, MA), and against β-actin from Earthox (San Francisco, CA).

Genomic DNA was extracted using the phenol-chloroform method. Genomic DNA was modified with sodium bisulfite using the EZ DNA Methylation Kit (Zymo Research, Irvine, CA). The MethPrimer software was used to search CpG islands and design primers (27). Primers specific to the bisulfite-modified sequences of the Zfp423 promoter region are listed in Table 1.

The Silica Bead DNA Gel Extraction Kit was purchased from Fermentas (Thermo Fisher Scientific, Waltham, MA). The pGEM-T Easy Vector System (Promega, Madison, WI) was used to clone bisulfite-modified sequences via TA cloning. Plasmids were extracted using a GeneJET Plasmid Miniprep Kit (Thermo Fisher Scientific). DNA methylation status was evaluated by comparing the cloned sequence against the genomic sequence, and the methylation map was generated using Quantification Tool for Methylation Analysis (QUMA, Kyoto, Japan) (28).

Mouse embryonic fibroblasts (MEFs) were separated from E14.5 fetal mice according to a common procedure, as previously described with slight modifications (29–32). In brief, fresh tissue separated under a stereomicroscope was minced and digested in medium containing 0.75 units/mL collagenase D (Roche, Pleasanton, CA) and 1.0 unit/mL Dipase type II (Roche) for 20 min at 37°C. The lysate was filtered sequentially through 70- and 40-µm cell strainers. The adipogenic differentiation was induced using a medium containing 0.5 mmol/L isobutyl-1-methylxanthine, 1 μmol/L dexamethasone, and 5 µg/mL insulin in Dulbecco’s modified Eagle’s medium with 15% FBS for 8 days and changed to insulin-only (5 µg/mL) Dulbecco’s modified Eagle’s medium with 10% FBS for 4 more days (26).

Progenitor cells were further sorted using surface markers, Sca-1⁺/CD45⁻, by Magnetic Cell Sorting System (MACS). The antibody MicroBeads kit and MACS separation column were bought from Miltenyi Biotec (Auburn, California). Sca-1⁺ marks a group of progenitor cells with high adipogenic potential (33,34), whereas CD45⁺ cells are of blood origin (33). Adipogenic differentiation was induced in sorted cells as described above.

Cells were fixed and stained with Oil Red O (Sigma-Aldrich, St. Louis, MO) as described previously (26). After microscopic observation, Oil Red O dye was solubilized with isopropanol, and the light absorbance was measured at 510 nm.

Fetal tissue was homogenized in 1% formaldehyde solution and incubated for 10 min and centrifuged. The pellet was resuspended in PBS containing 125 mmol/L glycine. After centrifuge, the pellet was lysed in a cold lysis buffer (1% SDS, 10 mmol/L Tris-HCl, pH 8.0, 10 mmol/L NaCl, 3 mmol/L MgCl₂, and 0.5% NP-40) containing 10 µL protease inhibitor cocktail (Thermo Fisher Scientific), sonicated, and centrifuged. The supernatant was precleaned with chromatin immunoprecipitation (ChIP)–grade protein G (Cell Signaling Technology) and then incubated with an antibody against H3K27me3, H3K4me3, EZH2, H3K9ac, or H3K27ac (Cell Signaling), or normal rabbit IgG overnight at 4°C. Then, the antibody-chromatin complex was precipitated with protein G and further treated with RNaseA (fermentas) and then proteinase K for 2 h. DNA was purified with ChIP DNA Clean & Concentrator (Zymo Research) and used as templates for PCR using the primers listed in Table 1. Relative enrichment folds of detected proteins were determined after normalization to input DNA using the ΔΔ-Ct method.

Plasmid pMSCVFLAG-ZFP423 (catalog no. 24764) and pMKO.1-Zfp423 (catalog no. 35972) were obtained from Addgene (Cambridge, MA). Plasmid transfection was performed using PolyJet in vitro DNA transfection reagent (SignaGen Laboratory, Ijamsville, MD) according to the manufacturer’s instructions.

Each pregnancy was considered as an experimental unit. Data were analyzed as a complete randomized design using GLM (General Linear Model of Statistical Analysis System; SAS, 2000). The differences in the mean values were compared by Student t test, and mean ± SEs were reported. Statistical significance was considered as P < 0.05.

After 2 months feeding on a high-energy diet, OB mice were heavier than Con mice (Con vs. OB, 17.44 ± 0.21 vs. 23.18 ± 0.24 g, P < 0.01), which maintained until E14.5, when maternal mice were killed (Con vs. OB, 30.32 ± 1.08 vs. 35.40 ± 1.01 g, P < 0.05), consistent with our previous report (13). There was no difference in fetal weight between treatments (Con vs. OB, 0.50 ± 0.03 vs. 0.52 ± 0.03 g).

In this study, we chose E14.5 fetal tissue for analyzing adipogenesis and Zfp423 epigenetic modifications. Although white adipocytes are not detectable in fetal mice at this stage, molecular changes associated with the early stage of commitment to adipogenesis have initiated, and thus fetal tissue at this stage is ideal for testing preadipocyte commitment and the early events of adipogenic differentiation. Indeed, the expression of Zfp423, the initiator of adipogenic commitment, was very high in E14.5 fetal tissue but was absent when assayed on E9.5 and had only weak expression on postnatal day 0 (P0) (Fig. 1A). On the other hand, the expression of PPARγ2, a marker of differentiated adipocytes, was undetectable in fetal tissue at E14.5, showing that there were no differentiated adipocytes (Fig. 1A).

Compared with Con pregnancy, the Zfp423 mRNA expression was >3.6-fold greater in OB fetal tissue (Fig. 1B). Similarly, Zfp423 protein content was also higher (Fig. 1C and D), showing that the Zfp423 expression was enhanced in OB fetal tissue.

To analyze whether there was an overall difference in adipogenic commitment and adipogenesis, we prepared MEFs from fetal tissue; as demonstrated in previous studies, MEFs contain a large number of progenitor cells and have high adipogenic capacity (29–31), and primary MEFs do not express PPARγ (20). These cells were passaged once, and adipogenesis was induced with an adipogenic cocktail. The Zfp423 expression was higher in OB compared with Con MEFs during the whole 12 days of adipogenic differentiation. Correspondingly, the expression of PPARγ2 was clearly detectable after 6 days of differentiation, with OB higher than Con MEFs (Fig. 2A). After 12 days of adipogenic differentiation, many more adipocytes were detected in OB compared with Con MEFs (Fig. 2B and C). Because these cells have been cultured in vitro for numerous cell generations, the difference in their adipogenic differentiation strongly indicates that MO had long-term promoting effects on the adipogenic differentiation of MEFs in OB fetal tissue.

Because MEFs contain cells at various stages of early commitment and differentiation, we further sorted MEFs using Sca-1⁺/CD45⁻ as surface markers. Sca-1⁺ marks a group of progenitor cells with high adipogenic potential (33) whereas CD45⁺ cells are of blood origin (34). After 12 days of adipogenic differentiation, more adipocytes were detected in OB compared with Con sorted progenitor cells (Fig. 2D and E), consistent with data obtained from MEFs. Of note, the adipogenic capacity of sorted cells was lower than that of MEFs (Fig. 2B and D), showing that there were additional populations of cells with high adipogenic capacity in fetal tissue. To be inclusive for all cells, for further studies, we used MEFs or fetal tissue, which represents the overall adipogenic commitment in E14.5 fetuses due to MO. This approach fits our objective to explore the impact of MO on Zfp423 expression, associated epigenetic modifications, and the adipogenic commitment during fetal development.

We have demonstrated that MO enhanced Zfp423 expression in OB fetal tissue. The remaining question is whether high Zfp423 expression is critical for enhanced adipogenic differentiation during fetal development. To test this, we prepared MEFs from both Con and OB fetal tissue, which were then transfected with either a vector overexpressing Zfp423 or an shZfp423 plasmid to knockdown Zfp423 mRNA. After 48 h, Zfp423 expression was dramatically increased in Zfp423-transfected cultures, accompanied by enhanced PPARγ expression. In contrast, knockdown of Zfp423 decreased PPARγ expression compared with MEFs transfected with enhanced green fluorescent protein (P < 0.05) (Fig. 3A). The same pattern of changes was observed after 6 days (Fig. 3B). After 12 days of adipogenic differentiation, many more adipocytes were visually detected in MEFs overexpressing Zfp423 while reduced by Zfp423 knockdown (Fig. 3C). Furthermore, comparing Con and OB MEFs, overexpression of Zfp423 was sufficient to enhance adipogenic differentiation of Con MEFs while knockdown of Zfp423 reduced the high adipogenic capacity of OB MEFs (Fig. 3D). These data showed that Zfp423 regulates adipogenesis of MSCs and suggest that Zfp423 mediates the effects of MO on the adipogenic differentiation of fetal progenitor cells during development.

The next question is why Zfp423 expression was enhanced in OB compared with Con tissue. Through the analysis of ChIP sequencing data, we found that the Zfp423 promoter and its 5′ upstream region had an abundant presence of H3K27me3 and H3K4me3, showing that the Zfp423 promoter is in a bivalent state (Fig. 4A). The promoters of genes with bivalent histone modifications may lose active histone modifications and convert to stable DNA methylation when lacking the stimuli to activate gene expression (25). To analyze whether there was a change in DNA methylation, we further analyzed the abundance of CpG sites in the Zfp423 promoter. As expected, exceptionally rich CpG sites were detected in the Zfp423 promoter (Fig. 4B). The location of CpG islands overlays with the presence of H3K27me and H3K4me3. To elucidate whether there is a difference in DNA methylation, we compared DNA methylation status using the sodium bisulfite sequencing. Interestingly, DNA methylation was largely absent in CpG sites proximate at the transcription start site (TSS) in both Con and OB tissue, and a higher density of CpG methylation was detected upstream of TSS. Overall, DNA methylation was more abundant in the Zfp423 promoter of the Con compared with the OB tissue (Fig. 4C), consistent with the lower Zfp423 expression in Con fetal tissue.

PcG protein EZH2 catalyzes inhibitory histone modification, H3K27me3, and further acts as a recruitment platform for DNA methyltransferases (35). To explore mechanisms leading to the reduction of Zfp423 DNA methylation in OB fetal tissue, we analyzed H3K27me3 and H3K4me3 in the Zfp423 promoter. OB decreased H3K27me3 modification in the GC-rich region of the Zfp423 promoter (2.6-fold, P = 0.01), showing that PRC2 was likely involved in the epigenetic modifications (Fig. 5A and B). On the other hand, active H3K4me3 histone marker was significantly higher in the OB fetal tissue (Fig. 5A, C, and D), consistent with reduced Zfp423 methylation and its enhanced expression, as well as reduction of H3K27me3 in the OB tissue (Fig. 4). These data further confirm Zfp423 as a member of genes with bivalent histone modifications, which are known to be critical for early cell commitment and differentiation (24).

In addition, the binding of Suz12, EZH2, and RNG1B, components of PRC2, was highly enriched in the Zfp423 promoter (Fig. 6A). EZH2 catalyzes H3K27me3, and the mRNA expression and protein content of EZH2 were higher in OB compared with Con tissue (Fig. 6B and C). This was unexpected but consistent with a previous report, where EZH2 expression was shown to be elevated during adipogenic differentiation (36). Despite the increase in overall EZH2 content, MO reduced the binding of EZH2 to the Zfp423 promoter of OB compared with Con fetal tissue, corroborating H3K27me3 and DNA methylation data (Fig. 6D).

Active gene expression is associated with enhanced histone acetylations, which were further analyzed (H3K9ac and H3K27ac). There was no difference in H3K27ac, but H3K9ac was higher in the Zfp423 promoter of OB compared with Con tissue (Fig. 7A and B), accentuating the role of these active epigenetic marks in the heightened Zfp423 expression in OB compared with Con fetal tissue.

Based on these data, we propose that MO reduced the binding of Polycomb group proteins to the Zfp423 promoter, which reduced inhibitory H3K27me3 histone modification and DNA methylation in the promoter, activating Zfp423 expression in fetal progenitor cells to predispose these and their derived cells to adipogenic differentiation (Fig. 7C).

Obesity is associated with excessive adipose tissue accumulation in three major depots in the body, including visceral, subcutaneous, and intermuscular/intramuscular depots. However, fat accumulation is not evenly distributed to these depots, and the accumulation of fat in the visceral depot is correlated with insulin resistance and the risk for T2D (37,38). In addition, fatty accumulation inside skeletal muscle leads to muscle insulin resistance, which has not been appreciated until recently (39).

Adipose tissue, including both white and brown adipose tissue, is mainly developed during the early developmental stages. Enhancing adipogenic differentiation during fetal development forms more preadipocytes and white adipocytes, which lead to adiposity in specific depots in offspring. In this study, we observed that adipogenic differentiation was enhanced in OB compared with Con fetal tissue, which is consistent with previous studies. MO enhanced adipogenic differentiation in fetal tissue (40,41), and increased intramuscular adipose tissue in MO offspring (42), as well as central adiposity (12,13). Of note, we did not observe an increase in fetal body weight at E14.5, which could be due to the smallness and immaturity of fetuses at this stage. Using the same dietary-induced obesity model, the body weight of neonatal mice of OB was higher than that of Con mice (13), and MO leads long-term obesity in offspring (12,13).

To define the mechanisms leading to the enhanced adipogenic differentiation in OB fetal tissue, we first analyzed whether Zfp423 regulates adipogenic commitment in fetal tissue. To our knowledge, our data, for the first time, suggest that Zfp423 regulates adipogenic commitment and adipogenesis in the developing fetal tissue. The remaining question becomes how MO regulates Zfp423 expression during fetal development. The Zfp423 promoter presents exceptionally rich CpG sites and islands, meeting the characteristics of a “key developmental gene” with high CpG density promoters (24), and previous chromatin immunoprecipitation and parallel sequencing (ChIP-Seq) results showed that during embryonic development, these rich CpG sites recruit PRC2, which catalyzes H3K27me3 in MEFs (43). These data strongly support the notion that MO regulates Zfp423 expression and adipogenesis through altering DNA methylation and other epigenetic modifications. To test this, we designed primers corresponding to the CpG islands overlaying the H3K27me3 peak in the Zfp423 promoter. Indeed, we observed that MO reduced DNA methylation in the Zfp423 promoter. However, DNA methylation in these CpG-rich regions was low and largely unmethylated, especially for regions located on TSS, which was unexpected but in agreement with previous reports that the majority of CpG islands of high CpG genes remain unmethylated or at a low methylation status during differentiation (24), and DNA methylation mainly occurs in the shore area of CpG islands (22). Indeed, we found that the methylation level was much higher in those regions located upstream of TSS. As expected, the DNA methylation was lower in OB compared with Con fetal tissue and negatively associated with the higher expression of Zfp423 in OB fetal tissue.

Stem cells maintain their pluripotency through reversible inhibition of lineage-specific genes while allowing genes needed for self-renewal to express. Conversely, during differentiation, lineage-specific genes are expressed while pluripotency genes are inhibited (25). PRCs are mainly responsible for reversible inhibition of genes. There are two well-characterized PRCs, PRC1 and PRC2. PRC2 tends to bind promoters with rich GC sites, which attracts PRC1 binding (22,25). Because the Zfp423 promoter has exceptionally rich GC sites, PRC2 is positioned as a key mediator of Zfp423 expression and adipogenic commitment, which induces H3K27me3 (21). De novo DNA methylation serves to convert plastic gene inhibition by PRCs to permanent silencing (25). Our data show that the H3K27me3 and EZH2 levels in the Zfp423 promoter were lower in OB compared with Con fetal tissue, consistent with the lower DNA methylation and the high expression of Zfp423. However, it is surprising that total EZH2 expression in OB was higher than in Con tissue; nevertheless, this result is consistent with a previous report (36). These data suggest that the specific binding of EZH2 to the Zfp423 promoter, not the EZH2 content, is critical for the epigenetic modifications in the Zfp423 promoter, which warrants further studies.

TrxG catalyzes H3K4 trimethylation (H3K4me3), which activates gene transcription. It appears that H3K4me3 is transient, and only induced when gene expression is needed to counter the inhibitory effect of the Polycomb group (23,44). The level of H3K4me3 in the Zfp423 promoter was slightly higher in OB fetal tissue, which indicates that trxG was also involved in the control of Zfp423 expression due to MO but appeared minor compared with PRC2.

In summary, for the first time, we observed that MO reduced DNA methylation in the Zfp423 promoter, which was correlated with higher Zfp423 expression and enhanced adipogenesis of progenitor cells. We further show that histone modification, H3K27me3, a reaction catalyzed by PRC2, was lower in OB fetal tissue, which provides a mechanism for the reduced DNA methylation in the Zfp423 promoter, and enhanced Zfp423 expression and adipogenesis in OB fetal tissue. Because DNA methylation is stable, the reduced DNA methylation in the Zfp423 promoter and enhanced Zfp423 expression in fetal progenitor cells likely enhance the adipogenic capacity of their derived cells in offspring adipose tissue, predisposing offspring to adiposity and the accompanied metabolic dysfunction, which might partially explain the increasing rates of obesity and even T2D in teenagers and even children.

This work was funded by National Institutes of Health grant R01-HD-067449.

No potential conflicts of interest relevant to this article were reported.

Q.-Y.Y. researched data and wrote the manuscript. J.-F.L. researched data. C.J.R. and J.-X.Z. contributed to discussion. M.-J.Z. contributed to discussion and reviewed and edited the manuscript. M.D. wrote the manuscript and designed the experiments. M.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.