Maternal undernutrition during periconceptional period affects whole-genome ovine muscle methylation in adult offspring

Abstract

Experimental and epidemiological studies suggest that maternal nutritional status during early pregnancy, including the period around the time of conception, may induce long-lasting epigenetic changes in the offspring. However, this remains largely unexplored in livestock. Therefore, the objective of this study was to evaluate if modification of the maternal diet of sheep (CTR: control; UND: 50% undernutrition) during the periconceptional period (42 d in total: −14/+28 from mating), would impact CpG methylation in muscle tissue (Longissimus dorsi) of adult offspring (11.5 mo old). Reduced representation bisulfite sequencing identified 262 (Edge-R, FDR < 0.05) and 686 (logistic regression, FDR < 0.001) differentially methylated regions (DMRs) between the UND and CTR groups. Gene ontology analysis identified genes related to development, functions of the muscular system, and steroid hormone receptor activity within the DMRs. The data reported here show that nutritional stress during early pregnancy leads to epigenetic modifications in the muscle of the resulting offspring, with possible implications for cardiac dysfunction, muscle physiology, and meat production.

Alteration of the maternal diet of sheep impacts the CpG methylation of genes related to muscular tissue development in muscle tissue of adult offspring.

Introduction

In utero fetal programming is well documented and supported by extensive epidemiological and experimental literature. Maternal undernutrition, specifically during early gestation, can result in permanent changes in the physiology and metabolism of the offspring (Godfrey and Barker, 2000; McMillen and Robinson, 2005), which in turn lead to an increased risk of cardiovascular and metabolic disorders in adulthood (Ravelli et al., 1998; Roseboom et al., 2000; Yan et al., 2013).

The epigenome constitutes the interface between the external environment and the genome, and suboptimal maternal nutrition may have long-term consequences for gene expression in the resulting offspring. The early developmental window, especially the periconceptional period, seems to be particularly sensitive to nutrient deficiencies (Van Soom and Fazeli, 2015; Sun et al., 2016). In this period, extensive epigenetic reorganization of the genome occurs (Reik et al., 2001). Shortly after fertilization, the mammalian genome undergoes extensive demethylation, then in preimplantation embryos, lineage-based methylation differences are established (Santos et al., 2002). As a result, different germ layers display distinct genome-wide DNA methylation patterns, which affect gene expression in the resulting tissues (Slieker et al., 2015). Therefore, the time around conception is crucial for the establishment of the adult epigenotype, and nutritional-induced epigenetic alterations can be maintained throughout life, influencing tissue-specific gene expression patterns (Lee, 2015).

Maternal undernutrition is frequently reported in livestock, due to seasonal and environmental variations in food quality and availability (Dunlap et al., 2015). Skeletal muscle has a low priority for nutrient allocation during fetal development (Redmer et al., 2004) and is vulnerable to reduced nutrient availability compared with other organs such as the brain or liver (Zhu et al., 2006; Yan et al., 2013). Skeletal muscle is therefore a good candidate tissue for assessing whether undernutrition induces long-lasting epigenetic changes.

There are several reports of altered methylation of the genome in muscle tissues of fetuses resulting from maternal undernutrition (Lan et al., 2013; Yan et al., 2013; Lie et al., 2014; Peñagaricano et al., 2014), while there are limited data regarding the epigenetic status after birth. Experimental evidence in sheep has shown that maternal undernutrition during the periconceptional period leads to poor metabolic outcomes in the resulting offspring due to altered regulation of the glucose–insulin axis (Yan et al., 2013; Dunlap et al., 2015; Oliver et al., 2020). Similarly in pigs, maternal undernutrition during pregnancy decreases GLUT4 expression in the muscle of adult offspring, which is most likely due to the increased methylation of the muscle GLUT4 promoter (Wang et al., 2016).

We explored the effect of the reduced diet (50% of the standard daily food ration) imposed on sheep during periconceptional period on the epigenetic programming of the skeletal muscle of adult offspring. Genome-wide methylation status of adult muscle tissues was explored using reduced representation bisulfite sequencing (RRBS) of adult offspring that had experienced periconceptional maternal undernourishment (UND) and normal feeding (CTR).

Material and Methods

All experimental procedures involving animals were conducted in accordance with Animal Protection Regulations of Italy (DPR 27/1/1992) in conformity with European Community regulation 86/609. All animal experiments were performed with the approval of the local ethical committee (CEISA-comitato etico interistituzionale per la sperimentazione animale No PROT.UNCHD12-222/2014). All chemicals were obtained from Sigma-Aldrich unless otherwise stated.

Animal nutrition and breeding management

A commercial flock provided 12 multiparous Sarda breed sheep that were randomly assigned to 2 groups of 6 ewes and administered 1 of the 2 diets during the periconceptional period, from 14 d before mating (oocyte maturation) until 28 d after mating (early organogenesis), 42 d in total which corresponds to Carnegie stage 19 (Butler and Juurlink, 1987). CTR ewes were fed ad libitum with a balanced diet satisfying the nutritional requirements of pregnant ewes (300 g per animal per d). UND ewes were fed a restricted diet (50% of the requirements: 150 g per animal per d) of the same diet. The nutritional stress imposed on the animals was sufficient to induce a gross caloric restriction without affecting the health and well-being of the animals. All animals had a full clinical assessment each week (mucosal score, temperature, heart rate, and respiration). To optimize the pregnancy rate, estrus was synchronized by treating all ewes with intravaginal progesterone sponges (25 mg Chronogest; Intervet Ireland Ltd, Dublin) for 14 d. At the time of sponge removal, a ram of proven fertility was used to mate the two nutritional treatment groups, temporarily kept in a common paddock. From the time of mating onward, ewes were group-housed in pens under natural day length conditions with ad libitum access to water. On day 28 of gestation, all ewes were returned to an ad libitum diet, to meet daily energy and nutritional needs and ensure that all animals had regained sufficient fat reserves to maintain the pregnancy for the 149.7 d average gestation period for the Sarda sheep breed and for the onset and maintenance of lactation. Following parturition and birth-weight recording, all animals derived from a singleton pregnancy; six CTR (four males and two females) and six UND (three males and three females) were managed under standard farm conditions until 11.5 mo of age. A nonparametric one-way ANOVA (Kruskal–Wallis test) was used to analyze weight data using Graph Pad Prism software. Data are presented as mean ± SEM.

Muscle tissue collection

Muscle tissue (50 mg) from Longissimus dorsi were surgically sampled, between the 12th and 13th rib, under general anesthesia (acethyl-promazin IM followed by Thyopental Sodium) from all CTR and UND lambs at around 11.5 mo of age and stored at −80 °C for subsequent molecular analysis.

RRBS library preparation and sequencing

Genomic DNA from each muscle sample (n = 12; 6 CTR and 6 UND) was isolated using the NucleoSpin Tissue kit (Macherey-Nagel, Düren, Germany), following the manufacturer’s instructions. DNA concentration was measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). About 1 μg of DNA was digested with MspI (New England Biolabs, Ipswich, MA, USA) by overnight incubation at 37 °C, following the manufacturer’s instructions. Libraries were generated using a TruSeq DNA PCR-Free Library Preparation Kit (Illumina, San Diego, CA, USA). After adapter ligation, samples were converted using an EpiTect Bisulfite Kit (Qiagen, Venlo, The Netherlands) and finally PCR amplified with KAPA HiFi Uracil+ kit (Kapa Biosystems, Potters Bar, UK). RRBS libraries were sequenced on an Illumina Hiseq 3000 (San Diego, CA, USA) to generate 150-base paired-end reads.

Bioinformatic analysis

The preliminary quality control of raw sequence reads was carried out with FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Illumina raw sequences were then filtered with Trimmomatic software to remove adapters and low-quality bases at the ends of reads, using a sliding window approach (Bolger et al., 2014). Data are available in the Sequence Reads Archive (SRA), BioProject accession number, PRJNA757927. Bismark software v.0.17.0 was used to align each read to a bisulfite-converted sheep genome (Oar_v4.0) with option-N 1. Methylation calls were extracted using the Bismark methylation_extractor function. Seqmonk software (version 1.47.1) (http://www.bioinformatics.babraham.ac.uk/projects/seqmonk/) was used for the visualization and analysis of the Bismark output. Only cytosines present in all 12 animals (6 CTR and 6 UND) with coverage of 10X were imported and analyzed by splitting the reference genome into 200 bp sliding windows with a 100 bp offset. The methylation percentage was calculated for each window if at least three cytosines were detected. Differentially methylated regions (DMRs) were identified by comparing CTR with UND groups. DMRs were calculated using both Edge-R (FDR < 0.05) and a logistic regression (LR) filter in R (FDR < 0.001) with differences in methylation level between CTR and UND groups above 10%. LR uses a binomial distribution and is a simple yet efficient method for the estimation of differential methylation (Huh et al., 2019). Edge-R uses empirical Bayesian methods to model gene-specific biological variation between biological replicates through the negative binomial dispersion parameter. Edge-R linear models fit the total read count (methylated + unmethylated) along loci. At each locus, the ratio of methylated reads is modeled indirectly as an over-dispersed binomial-like distribution. This has been demonstrated to be an efficient solution for methylation studies even when only a few biological replicates are available (Chen et al., 2017). To compute shared DMRs and DMR-associated genes across comparisons, commonality analysis was run and visualized through Venn diagrams, using the Venn Diagrams software (http://bioinformatics.psb.ugent.be/webtools/Venn/). DMR-associated genes were classified according to classical Gene Ontology (GO) categories, using the Cytoscape plug-in ClueGO which integrates GO and enhances the biological interpretation of large lists of genes (Bindea et al., 2009). In ClueGO, the P-value was calculated with the Fisher exact test corrected using the Bonferroni step-down method.

Reverse transcription-quantitative real-time PCR

Total RNA was extracted from muscle tissues using TRIzol (Invitrogen, Carlsbad, CA, USA) and cleaned up with a NucleoSpin miRNA kit (Macherey-Nagel, Dueren, Germany), following the protocol which gives small and large RNA in one fraction (total RNA). RNA was reverse transcribed into cDNA in a total reaction volume of 18 μL. About 9 μL of RNA was added to 0.5 μL of random hexamers, 0.5μL oligodT, 0.5μL dNTPs, and incubated at 65 °C for 5 min, then placed at 4 °C. About 4 μL RT buffer (5X), 1 μL of DTT (0.1 M), 0.5 μL RNase inhibitor, and 1 μL of SuperScript II Reverse Transcriptase (Thermo Fisher, Waltham, MA, USA) were added. Reverse transcription was carried out at 25 °C for 5 min, 42 °C for 1 h, and 70 °C for 15 min. The expression level of PPARGC1α PPARδ, PTP1N, and IRS2 was assessed by reverse transcription-quantitative real-time PCR (RT-qPCR) using the expression of GADPH, ACTB genes as reference. Primers were designed from specific exon–exon junctions to avoid amplifying genomic DNA (Supplementary Table S1). RT-qPCR was performed with three technical replicates, using 5 μL of the Power SYBR Green Master Mix (Applied Biosystems, Carlsbad, CA, USA), 0.5 μL of forward and reverse primers (final concentration 400 nM), and 4 μL of diluted cDNA (1:20 Vol), with QuantStudio 6 Flex Real-Time PCR Systems (Applied Biosystems, Carlsbad, CA, USA). Relative expression of genes between CTR and UND groups and P-value (t-test) were calculated using the PCR R package (Ahmed and Deok, 2018).

Results

Offspring weight

Lamb birth weights were similar between maternal dietary groups (CTR: 3.5 ± 0.32 kg; UND: 3.6 ± 0.28 kg) and were appropriate, given the size of the ewe and plane of nutrition during pregnancy. At around 1 yr of age, at the time of muscle tissue collection, there was no difference in body weight between the two groups (11.5 mo; CTR: 44.20 ± 2.93 kg; UND: 40.63 ± 2.10 kg).

Muscle methylation

The average number of sequence reads per muscle sample was 20.4 M (ranging from 12.9 to 31.3 M) with a mapping efficiency of about 60%. RRBS had an average level of cytosine methylation in the CpG context of 59.2% across all muscle samples in both CTR and UND animals (Supplementary Table S2). A total of 112,852 methylated regions were detected across all samples using SeqMonk software. All these regions were evaluated using two different statistical methods to identify methylation variation between the two groups greater than 10% (Figure 1). Pairwise comparison of muscle methylation between experimental groups identified 262 and 686 DMRs in CTR vs. UND calculated using Edge-R (FDR < 0.05) and LR (FDR < 0.001), respectively; annotation of which identified 99 and 253 associated genes for Edge-R and LR, respectively.

Figure 1.

Venn diagram of shared differentially methylated regions identified with Edge-R and logistic regression methods.

Both statistical approaches identified genes that were hyper- and hypomethylated between CTR and UND groups. We investigated their function by GO analysis (Table 1). The GO classes showed enrichment for genes that had functions predominantly related to muscle development, growth, and maintenance, including actin cytoskeleton organization, myoblast proliferation, muscle cell formation and contraction, regulation of carbohydrate metabolic process, steroid hormone activity, and negative regulation of the ERK1 and ERK2 cascade (Table 1).

Table 1.

Gene ontology analysis of annotated genes found to be differentially methylated between the UND and CTR group using two Edge-R and LR approaches and considering the regions that were hyper-methylated in UND and CTR group

GO-ID	Associated genes found	GO-term	P-value*
Total DMRs (Edge-R)
2000291	[MEIS2, PPARD, ZNF609]	Regulation of myoblast proliferation	5.97E-04
2027	[CACNA1C, CACNA1G, DMD, DSP, IRX5]	Regulation of heart rate	1.30E-03
51450	[MEIS2, PPARD, ZNF609]	Myoblast proliferation	1.52E-03
31532	[EZR, NTF3, PARVB, PHACTR1, PTPN1]	Actin cytoskeleton reorganization	1.94E-03
14909	[BCL2, ITGB3, PPARD, PPARGC1A]	Smooth muscle cell migration	3.91E-03
70373	[DMD, EZR, PTPN1]	Negative regulation of ERK1 and ERK2 cascade	3.94E-03
86003	[CACNA1C, CACNA1G, DSP]	Cardiac muscle cell contraction	3.94E-03
86001	[CACNA1C, CACNA1G, DMD, DSP]	Cardiac muscle cell action potential	4.00E-03
10518	[ADCYAP1R1, ESR1, NTF3]	Positive regulation of phospholipase activity	6.74E-03
14812	[BCL2, ITGB3, PPARD, PPARGC1A]	Muscle cell migration	6.81E-03
86091	[CACNA1C, CACNA1G, DSP]	Regulation of heart rate by cardiac conduction	7.33E-03
51899	[ADORA2A, BCL2, CACNA1C, CACNA1G]	Membrane depolarization	7.85E-03
14910	[BCL2, PPARD, PPARGC1A]	Regulation of smooth muscle cell migration	8.93E-03
86065	[CACNA1C, CACNA1G, DSP]	Cell communication involved in cardiac conduction	1.09E-02
1900274	[ADCYAP1R1, ESR1, NTF3]	Regulation of phospholipase C activity	1.09E-02
10863	[ADCYAP1R1, ESR1, NTF3]	Positive regulation of phospholipase C activity	1.09E-02
3707	[ESR1, NR2C2, PPARD]	Steroid hormone receptor activity	1.12E-02
10676	[ADCYAP1R1, KAT2B, PPARGC1A]	Positive regulation of cellular carbohydrate metabolic process	1.14E-02
86002	[CACNA1C, CACNA1G, DSP]	Cardiac muscle cell action potential involved in contraction	1.25E-02
Hyper-methylated UND (Edge-R)
14909	[BCL2, ITGB3, PPARD, PPARGC1A]	Smooth muscle cell migration	7.31E-04
86001	[CACNA1C, CACNA1G, DMD, DSP]	Cardiac muscle cell action potential	7.70E-04
14812	[BCL2, ITGB3, PPARD, PPARGC1A]	Muscle cell migration	1.50E-03
70373	[DMD, EZR, PTPN1]	Negative regulation of ERK1 and ERK2 cascade	1.51E-03
86003	[CACNA1C, CACNA1G, DSP]	Cardiac muscle cell contraction	1.51E-03
86091	[CACNA1C, CACNA1G, DSP]	Regulation of heart rate by cardiac conduction	1.73E-03
14910	[BCL2, PPARD, PPARGC1A]	Regulation of smooth muscle cell migration	2.26E-03
10676	[ADCYAP1R1, KAT2B, PPARGC1A]	Positive regulation of cellular carbohydrate metabolic process	3.25E-03
86065	[CACNA1C, CACNA1G, DSP]	Cell communication involved in cardiac conduction	3.28E-03
3707	[ESR1, NR2C2, PPARD]	Steroid hormone receptor activity	3.53E-03
86002	[CACNA1C, CACNA1G, DSP]	Cardiac muscle cell action potential involved in contraction	4.02E-03
Total DMRs (LR)
2000291	[GDNF, MEIS2, PPARD, ZNF609]	Regulation of myoblast proliferation	1.19E-03
51450	[GDNF, MEIS2, PPARD, ZNF609]	Myoblast proliferation	4.90E-03
2000288	[GDNF, MEIS2, PPARD]	Positive regulation of myoblast proliferation	9.70E-03
31532	[BAIAP2, EZR, FARP2, NTF3, PARVB, PHACTR1, PTPN1]	Actin cytoskeleton reorganization	1.75E-02
7422	[ALDH3A2, DMD, GDNF, NTF3, PLXNA4, RUNX1]	Peripheral nervous system development	1.84E-02
48813	[BAIAP2, CUX1, KIF1A, MAP2, NSMF, PHACTR1, SEMA4D, SHANK1]	Dendrite morphogenesis	2.54E-02
86001	[CACNA1C, CACNA1G, DMD]	Cardiac muscle cell action potential	4.89E-02
Hyper-methylated UND (LR)
86001	[CACNA1C, CACNA1G, DMD]	Cardiac muscle cell action potential	1.55E-02
36037	[BCL2, RUNX1, SH3RF1]	CD8-positive, alpha-beta T cell activation	2.14E-02
50855	[FOXP1, LYN, RUNX1]	Regulation of B cell receptor signaling pathway	2.28E-02
30521	[FOXP1, NCOR2, PMEPA1, PPARGC1A]	Androgen receptor signaling pathway	2.29E-02
10522	[ADCYAP1R1, BCL2, CACNA1C, DMD, LYN]	Regulation of calcium ion transport into cytosol	2.33E-02
50854	[EZR, FOXP1, LYN, RUNX1]	Regulation of antigen receptor-mediated signaling pathway	2.46E-02
19395	[ACOXL, ALDH3A2, PPARD, PPARGC1A, SOAT2]	Fatty acid oxidation	2.55E-02
31532	[BAIAP2, EZR, FARP2, PHACTR1, PTPN1]	Actin cytoskeleton reorganization	2.59E-02
34440	[ACOXL, ALDH3A2, PPARD, PPARGC1A, SOAT2]	Lipid oxidation	2.76E-02
2548	[CCL16, CCL17, LYN]	Monocyte chemotaxis	2.78E-02
40014	[BCL2, EZR, FTO, RAI1]	Regulation of multicellular organism growth	2.88E-02
14909	[BCL2, ITGB3, PPARD, PPARGC1A]	Smooth muscle cell migration	2.94E-02
70373	[DMD, EZR, LYN, PTPN1]	Negative regulation of ERK1 and ERK2 cascade	2.99E-02
7422	[ALDH3A2, DMD, PLXNA4, RUNX1]	Peripheral nervous system development	3.53E-02
60612	[FTO, PPARD, PPARGC1A]	Adipose tissue development	4.01E-02
32623	[EZR, HDAC7, RUNX1]	Interleukin-2 production	4.01E-02
14812	[BCL2, ITGB3, PPARD, PPARGC1A]	Muscle cell migration	4.28E-02
71526	[FARP2, PLXNA4, SEMA4D]	Semaphorin-plexin signaling pathway	4.43E-02
1901505	[SLC25A48, SLC29A3, SLC35A2]	Carbohydrate derivative transmembrane transporter activity	4.60E-02
50919	[ITGB3, PLXNA4, SEMA4D]	Negative chemotaxis	4.61E-02
5089	[ARHGEF10L, ARHGEF28, FARP2]	Rho guanyl-nucleotide exchange factor activity	4.75E-02
14910	[BCL2, PPARD, PPARGC1A]	Regulation of smooth muscle cell migration	4.75E-02
Hyper-methylated CTR (LR)
2000291	[GDNF, MEIS2, ZNF609]	Regulation of myoblast proliferation	1.38E-04
51450	[GDNF, MEIS2, ZNF609]	Myoblast proliferation	2.97E-04
50885	[CAMTA1, RBFOX1, SHANK1]	Neuromuscular process controlling balance	2.59E-03
48512	[ADORA2A, BTBD9, ZFHX3]	Circadian behavior	2.82E-03

Corrected with Bonferroni step down.

RT-qPCR evaluation of PPARGC1α, PPARδ, PTP1N, and IRS2 expression in muscle

The expression level of PPARGC1α, PPARδ, and PTP1N was lower in the UND vs. CTL group while IRS2 higher in UND animals. PPARGC1α showed the greatest variation in gene expression (Supplementary Figure S1).

Interestingly, the expression of these genes was inversely correlated with the methylation level of DMRs in which they were found (Supplementary Table S3). PPARGC1α and PPARδ DMRs had higher methylation and IRS2 lower methylation in UND animals compared with CTL. Conversely, PTPN1 had both a higher methylation level and a slightly higher expression in UND animals (Supplementary Table S3).

Discussion

To date, diet-induced effects on the development of muscle have been assessed using morphological parameters, such as size, weight, fiber number, and composition (Fahey et al., 2005; Quigley et al., 2005; Sharples et al., 2016), however, the molecular mechanisms controlling differences remain largely unexplored. This study found that the 50% diet restriction imposed on sheep around the time of conception lead to epigenetic alterations in the muscle of the adult offspring. This is consistent with other reports that have shown that reduced nutrition during early fetal life is associated with abnormal methylation status in fetal muscle tissues (Lan et al., 2013; Namous et al., 2018).

In the present study, the methylation frequency of CpG sites of Longissimus dorsi muscle tissues from 11.5 mo old lambs ranged between 56% and 62% with no difference in overall methylation levels between UND and CTR groups. Similar results have been reported for sheep Longissimus dorsi muscles of 8-mo old lambs with about 50% to 55% of the total CpG sites methylated (Couldrey et al., 2014). A small increase of 5-mC content in muscle DNA in pigs receiving a methyl supplemented diet has been suggested, although the difference was not significant (Braunschweig, 2009).

In the present study, analysis of differential methylation between UND and CTR groups found several DMRs which were consistently identified using two different statistical methods. The two methods identified alterations in the methylation status of specific genes, the functions of which are mainly related to muscular development and activity. GO analysis of annotated DMRs identified the “negative regulation of the ERK1 and ERK2 cascade” pathway, which has recently been associated with the switch between fast-to-slow muscular fiber types (Boyer et al., 2019). We found several genes containing DMRs that have also been identified previously in diet-restriction studies. These include the peroxisome proliferator-activated receptor-γ coactivator-1α (PPARGC1α), which is involved in glucose and fatty acid metabolism in the liver and muscle (Liang and Ward, 2006). A rat model of intrauterine growth restriction showed an increase in the DNA methylation of specific CpG sites in PPARGC1α, and a relative decrease in the transcriptional level in muscle tissues with a positive correlation with fasting insulin concentration (Xie et al., 2015). We also found increased methylation in PPARδ and PPARGC1α, and a lower, although not significant, expression of these two genes in UND ewes. Interestingly, both PPARGC1α and PPARδ have been previously reported to regulate muscle fiber composition. PPARGC1a is a principal factor regulating muscle fiber type determination by activating genes implicated in slow fiber gene expression (Lin et al., 2002). Endurance exercise training promotes PPARδ expression by inducing an increase in the number of type I muscle fibers (Wang et al., 2004). Fiber composition within a muscle is important for the growth potential of skeletal muscle, endurance fitness, and adaptability to environmental stress. Maternal nutrition during pregnancy may influence the type and amount of adult muscle fibers which are determined in utero (Zhu et al., 2004; Fahey et al. 2005; Quigley et al., 2005). Interestingly, despite the lack of muscle structural analysis, in this work, PPARGC1α and PPARδ both showed changes in methylation level and expression in lambs born to ewes that had undergone diet restriction.

PPARGC1and PPARδ also impact the aerobic physical fitness and insulin sensitivity in humans (Stefan et al., 2007). The long-term effects of undernutrition of sheep during the periconceptional period have been reported to impair glucose–insulin axis function (Todd et al., 2009; Dunlap et al., 2015), affect the insulin-signaling pathway (Nicholas et al., 2013) as well as glucose uptake (Dunlap et al., 2015) in skeletal muscle of adult offspring. This suggests diet affects the metabolic health of the next generation.

Maternal undernutrition is known to have a transgenerational effect, causing fetal hyperinsulinemia, increased diabetic risk, and obesity in both first- and second-generation offspring of rats (Aerts and Van Assche, 2006). Transgenerational transmission requires alterations in the DNA of the germ cell line. Although general demethylation occurs during early embryo cleavage, a proportion of genomic domains are resistant to early embryo methylation reprogramming (Li et al., 2018). In mice, the intergenerational transmission of glucose intolerance is the result of altered expression of ATP-binding cassette subfamily C member 8 (ABCC8) coding for the sulfonylurea receptor 1 (SUR1) protein in β-cell islets (Jimenez-Chillaron et al., 2009). Sperm of in utero undernourished offspring has been shown to have specific DMRs in ABCC8 that are retained in the F2 population (Radford et al., 2014). In our study, specific alterations in ABCC8 methylation were also observed in lamb muscle tissue. In addition, methylation patterns of other key genes related to glucose metabolism including insulin receptor substrate 2 (IRS2), fat mass and obesity-associated protein or (FTO), ATPase phospholipid transporting (ATP11A and ATP13A5), and protein tyrosine phosphatase, non-receptor type 1 (PTP1N) differed between CTR and UND progeny.

Whole-genome bisulfite sequencing of human sperm revealed methylation differences in PPARGC1α and FTO and other family members including insulin receptor substrate 1 (IRS1) and ATPase phospholipid transporting 10B (ATP10) in patients with type 2 diabetes mellitus (Chen et al., 2019). PTPN1 also showed variations in methylation in peripheral blood mononuclear cells in patients with type 2 diabetes mellitus (Huang et al., 2017). The effect of maternal undernutrition is known to cause insulin resistance in skeletal muscle by reducing glucose uptake stimulated by insulin and promoting lipid accumulation in muscle cells, which may be ascribed to methylation changes in lamb muscle (Phielix and Mensink, 2008).

In conclusion, the present study shows that maternal periconceptional undernourishment of sheep impacts the muscle methylome in adult offspring. Due to the limited number of samples and an unbalanced sex ratio, we are not able to define gender-related differences as reported in a previous work analyzing the periconceptional undernutrition effect on the phenotype of adult sheep (Todd et al., 2009; Jaquiery et al., 2012). However, specific genes involved in muscle fiber structure and muscular energy metabolism are affected, suggesting that early life events have long-term consequences for adult life. Interestingly, in the most DMRs, many genes responsible for muscle development were identified, suggesting that undernutrition may increase the risk of metabolic disease. The investigation of the expression of genes associated with metabolism in DMRs showed a decreased expression in UND. A follow-up of muscle methylation status vs. metabolic profiles and structural differences of muscle in the progeny of ewes with different nutritional diets are needed, to shed light on the biological mechanisms linking undernutrition, epigenomics, and muscle physiology.

Glossary

Abbreviations

5-mC

5-methylcytosine

ABCC8

ATP-binding cassette subfamily C member 8

ATP10B

ATPase phospholipid transporting 10B

ATP11A

ATPase phospholipid transporting 11A

ATP13A5

ATPase phospholipid transporting 13A5

CTR

control

DMRs

differentially methylated regions

ERK1

extracellular signal-regulated kinase 1

ERK2

extracellular signal-regulated kinase 2

FDR

false discovery rate

FTO

fat mass and obesity-associated protein

GLUT4

glucose transporter type 4

GO

gene ontology

IRS1

insulin receptor substrate 1

IRS2

insulin receptor substrate 2

LR

logistic regression

PBMCs

peripheral blood mononuclear cells

PPARGC1α

peroxisome proliferator-activated receptor-γ coactivator-1α

PPARδ

peroxisome proliferator-activated receptor-δ

PTP1N

protein tyrosine phosphatase non-receptor type 1

RRBS

reduced representation bisulfite sequencing

SLC2A4

solute carrier family 2 member 4

SUR1

sulfonylurea receptor 1

UND

undernutrition

Conflict of interest statement

The authors declare no conflict of interest.