figure a

Introduction

The ‘fetal programming’ hypothesis proposes that adulthood metabolic disease originates through adaptation of the fetus in early development [1]. These adaptations are tissue-specific, persist throughout life and may cause metabolic diseases in later life [2,3,4,5].

The classical events involved in fetal programming are of maternal origin [6,7,8,9,10,11,12] but paternal factors may also alter the epigenome and phenotype of offspring [13,14,15]. Feeding male rat founders a high-fat diet before mating induces impaired glucose tolerance in female offspring [16], possibly due to epigenetic adaptations in the pancreas and liver. Treatment approaches for paternal diet-induced adverse metabolic effects in offspring include physical activity, antioxidants and improvement in pre-mating diet [17, 18]. Previously, we fed a diet resembling an unhealthy ‘western’ diet (high-fat, high-sucrose and high-salt diet [HFSSD], often consumed by men) to male rats prior to mating and analysed the effect on glycaemic control in offspring. Since folate treatment in low-protein-diet-fed pregnant rats improves glycaemic control [19, 20], here we investigated the effects of paternal folate (folic acid) treatment before mating, as well as maternal folate treatment during pregnancy, on offspring phenotype. The unhealthy paternal diet was given before mating, during spermatogenesis, since it is known that a high-fat diet induces adverse effects on non-coding RNA in the sperm and has long-lasting adverse effects on the offspring [21,22,23].

Methods

Animals

The present study was performed in Sprague-Dawley rats of both sexes, including F0 generation (45 male rats, 32 female rats) and F1 generation animals. The F0 generation rats were purchased at the age of 4 weeks from Hunan SJA Laboratory Animal (Changsha, China). After acclimatisation to their new environment for 1 week, the rats were given a specific diet. The rats were housed in temperature-controlled chambers under control lighting with 12 h light–dark cycles. All rats were allowed free access to water and food. The experimental protocols were conducted in accordance with the ethical standards of the local ethics committee. See electronic supplementary material (ESM) Methods for further details.

Study design

F0 male rats were randomly divided into groups, each of which received one of the following diets: (1) control diet + tap water by oral gavage (n = 15); (2) HFSSD + tap water by oral gavage (n = 15) or (3) HFSSD + folate (HFSSD+F) at a daily dose of 3 mg/kg dissolved in tap water and given by oral gavage (n = 15). See ESM Table 1 for further details. The daily folate dosage of 3 mg/kg is in accordance with previous publications [24, 25].

The 14-week-old F0 founder male rats fed either the control diet or one of the two fat-rich diets were mated with F0 12-week-old, normal-weight, naturally cycling dams fed a control diet to produce F1 offspring. Depending on paternal diet before mating and on maternal diet after mating and throughout the gestational period, the F1 offspring of both sexes were allocated into one out of four study groups (Fig. 1):

  • PatCD/MatCD group—offspring of control-diet-fed founders and control-diet-fed dams (dams were fed a control diet from the beginning of gestation until delivery)

  • PatHFSSD/MatCD group—offspring of HFSSD-fed founders and control-diet-fed dams from 50% of obtained litters (dams were fed a control diet from the beginning of gestation until delivery)

  • PatHFSSD/MatCD+F group—offspring of HFSSD-fed founders and control-diet-fed dams from another 50% of obtained litters (dams were fed a control diet+folate [5 mg/kg daily in food] from the beginning of gestation until delivery, a folate intake comparable with that currently recommended for women in the UK before pregnancy and during the first trimester [20, 24])

  • PatHFSSD+F/MatCD group—offspring of HFSSD+F-fed founders and control-diet-fed dams (dams were fed a control diet from the beginning of gestation until delivery).

Fig. 1
figure 1

Study design for the F0 and F1 generations. F0 male rats were randomly divided into one of the following three study groups according to the diet type: (1) control diet + tap water by oral gavage (CD); (2) HFSSD + tap water by oral gavage; and (3) HFSSD + folate 3 mg/kg body weight daily, dissolved in tap water and given by oral gavage (HFSSD+F). F1 offspring of both sexes were allocated into one of four study groups: PatCD/MatCD—offspring of CD-fed founders and CD-fed dams (dams were fed a CD from the beginning of gestation until delivery); PatHFSSD/MatCD—offspring of HFSSD-fed founders and CD-fed dams from 50% of obtained litters (dams were fed a CD from the beginning of gestation until delivery); PatHFSSD/MatCD+F—offspring of HFSSD-fed founders and CD-fed dams from another 50% of obtained litters (dams were fed a CD+F [5 mg/kg body weight per day folate in food] from the beginning of gestation until delivery) and PatHFSSD+F/MatCD—offspring of HFSSD+F-fed founders and CD-fed dams (dams were fed a CD from the beginning of gestation until delivery)

The total mean (SEM) number of offspring per litter was 13.9 ± 0.3 and did not differ significantly between the groups. We sampled randomly 173 offspring (86 male offspring, 87 female offspring) out of a total of 540. Thus, the sample size was around one-third of the entire population of offspring. After delivery, all F1 offspring were fed a normal diet until adulthood. The composition of the normal diet was very similar to the control diet; the normal diet was used for reasons of economy (see ESM Methods for details). All the F0 founder male rats at the age of 18 weeks and all the F1 rats at the age of 15 weeks were killed after receiving deep anaesthesia. Blood samples were collected and the organs were harvested.

Real-time quantitative PCR

Expression levels of hepatic and pancreatic genes were assessed by real-time quantitative PCR using a Bio-Rad CFX96 cycler (Bio-Rad Laboratories, USA) and using standard protocol (see ESM Methods and ESM Table 2).

RNA sequencing in the liver

RNA sequencing was carried out in collaboration with Oebiotech (Shanghai, China). Total RNA was extracted from the livers of the PatCD/MatCD and PatHFSSD/MatCD F1 female offspring using the mirVana miRNA Isolation Kit (Ambion, Foster City, CA, USA) following the manufacturer’s protocol. RNA integrity was evaluated using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). See ESM Methods and ESM Table 3 for further details.

Global and gene-specific DNA methylation in the liver

DNA was extracted from the liver and the concentration and purity were assessed spectrophotometrically. Global DNA methylation was determined as per cent 5-methylcytosine (5-MC%) using MethylFlash Global DNA Methylation (5-mC) ELISA Easy Kit (Epigentek, Farmingdale, NY, USA). Gene-specific DNA methylation of CpG islands of the promoter region of Ppara, Lcn2 and Tmcc2 was analysed using MethylTarget, based on Illumina next-generation sequencing in combination with bisulfite treatment and DNA methylation mapping as previously described [26,27,28]. Illumina next-generation sequencing was carried out in collaboration with Genesky Biotechnologies (Shanghai, China). See ESM Methods and ESM Table 4 for further details.

Metabolic tests

When F1 offspring rats were aged 100 days, an OGTT was performed following 12 h of starvation. The offspring were administered 2 g/kg glucose (50% [wt/vol.] glucose solution) by gavage and serum glucose levels were measured (Sannuo glucometer; San Nuo, Changsha, China) at 0, 30, 60 and 120 min after glucose ingestion. Serum glucose, cholesterol, triacylglycerols, HDL-cholesterol, LDL-cholesterol, alanine aminotransferase, aspartate transaminase, blood urea nitrogen and creatinine were measured using Hitachi 7020 automatic biochemistry analyser (Hitachi High-Technologies, Tokyo, Japan). Serum insulin levels were determined by a rat insulin ELISA kit (Millipore, Billerica, MA, USA). See ESM Methods for further details.

Pancreas and liver morphology

Liver and pancreas samples were fixed, sectioned and stained with H&E for oil droplet visualisation and Sirius Red for fibrosis assessment and were immunostained for CD68 (ED-1, rabbit polyclonal anti-CD68 antibody, 1:50; Abcam, Cambridge, UK) to evaluate periportal inflammation. Pancreatic slices were H&E stained to assess islet density and number and were immunostained for insulin (guinea pig anti-insulin, 1:200; Abcam) to detect beta cell density per islet. The samples were examined with light microscopy using a BZ 9000 microscope (Keyence, Neu-Isenburg, Germany). See ESM Methods for further details.

Liver triacylglycerol measurement

Liver triacylglycerol levels were measured using an enzyme immunoassay kit (Kehua Bio-Engineering, Shanghai, China) and an automatic biochemistry analyser (Hitachi High-Technologies, Tokyo, Japan). See ESM Methods for details.

Statistics

All data are presented as means ± SEM. Two-way analysis of variance with Bonferroni post hoc test was used to analyse the body weight gain and OGTT data. For all other data, one-way analysis of variance followed by least significant difference test was applied. To account for potential bias, additional ANCOVA models considering litter size as a covariate were calculated. A p value of <0.05 was considered statistically significant. The data were analysed using SPSS version 20.0 (SPSS, Chicago, IL, USA) and GraphPad Prism version 5 (GraphPad Software, San Diego, CA, USA).

Results

F0 founders

Detailed data on body weight gain and selected metabolite serum levels in the F0 founders are presented in ESM Table 5. The number of pups was similar in all groups; no effect of HFSSD or folate treatment on the number of pups was detectable.

F1 offspring

Body weight and length

In the F1 generation, a paternal HFSSD diet resulted in body weight increase, which appeared to be more pronounced in female offspring (ESM Fig. 1). The male and female PatHFSSD/MatCD offspring were 8% and 6% heavier than their control littermates, respectively (p ˂ 0.05 in male offspring, p = 0.066 in female offspring). Folate intervention given to pregnant dams and folate food supplementation given to founders before mating both failed to reduce F1 progeny body weight.

At the end of the study, male and female PatHFSSD/MatCD offspring were 3.7% and 3.2% longer than their control littermates, respectively (p ˂ 0.01 for both sexes; Tables 1 and 2). In male and female offspring, folate given to pregnant dams decreased body length (by 3.0% and 2.7%, p < 0.05 and p < 0.01 vs PatHFSSD/MatCD, respectively). Folate given to founders before mating failed to reduce F1 progeny body length (Tables 1 and 2). The liver weight was increased in female PatHFSSD/MatCD offspring (p < 0.01 vs PatCD/MatCD) but not in the male offspring.

Table 1 Body length, organ weights, pancreas and liver morphology and serum metabolites in male F1 offspring
Table 2 Body length, organ weights, pancreas and liver morphology and serum metabolites in F1 female offspring

Glucose tolerance

Feeding the F0 founders with HFSSD impaired the glucose homeostasis of their F1 offspring in a sex-specific manner. Female F1 progeny were more susceptible to glucose intolerance, as assessed by OGTT. When F1 female offspring of the PatHFSSD/MatCD group were 100 days old, serum glucose levels were elevated 60 min after glucose intake in the OGTT compared with levels in the PatCD/MatCD offspring (p < 0.001; Fig. 2b) and the AUC for glucose following the OGTT was also increased (p < 0.01 for PatHFSSD/MatCD vs PatCD/MatCD; Fig. 2d). In contrast to the F1 females, there were no significant alterations in glucose homeostasis in male F1 offspring except for serum glucose levels, which were elevated 30 min after glucose intake in the PatHFSSD/MatCD group compared with levels in the PatCD/MatCD offspring (p < 0.05; Fig. 2a, c).

Fig. 2
figure 2

Blood glucose levels (a, b) and AUC (c, d) following an OGTT in the F1 male (a, c) and female (b, d) offspring at 100 days of age. Brown circles, PatCD/MatCD (n = 21 and n = 34 in males and females, respectively); pink squares, PatHFSSD/MatCD (n = 35 and n = 23 in males and females, respectively); green triangles, PatHFSSD/MatCD+F (n = 15 and n = 18 in males and females, respectively); blue inverted triangles, PatHFSSD+F/MatCD (n = 18 and n = 15 in males and females, respectively). Values are shown as means ± SEM. *p < 0.05, **p < 0.01 and ***p < 0.001 vs PatCD/MatCD; p < 0.05 and ††p < 0.01 vs PatHFSSD/MatCD

In female F1 offspring, folate intervention given to pregnant dams (group) restored glucose tolerance (p < 0.05 for PatHFSSD/MatCD+F vs PatHFSSD/MatCD, 60 min after glucose intake, Fig. 2b) and significantly decreased glucose AUC following the OGTT (p < 0.01 for PatHFSSD/MatCD+F vs PatHFSSD/MatCD, Fig. 2d). Folate supplementation given to F0 founder male rats (PatHFSSD+F/MatCD) failed to restore glucose tolerance in their F1 offspring (Fig. 2a, b).

With the exception of lowered fasting glucose in PatHFSSD+F/MatCD vs PatHFSSD/MatCD male F1 offspring (p < 0.01), there were no significant differences in fasting glucose and insulin levels between the groups of male and female F1 offspring (Tables 1 and 2).

Liver morphology

Male and female F1 offspring of HFSSD-fed F0 founder rats revealed mild hepatic steatosis (p < 0.01 and p < 0.05 vs PatCD/MatCD, respectively; Fig. 3a, b) manifesting as microvesicular steatosis (Fig. 3c, d). Minor hepatocyte degeneration was observed in all paternally programmed groups (data not shown). In PatHFSSD/MatCD offspring, the hepatic steatosis score was increased by 57% (p < 0.01) and 52% (p < 0.05) vs control male and female counterparts, respectively (Fig. 3a, b). Neither of the folate intervention groups showed any significant beneficial effects regarding steatosis.

Fig. 3
figure 3

Hepatic steatosis score in the male (a) and female (b) F1 offspring and typical photomicrographs of H&E staining for hepatic steatosis in male (c) and female (d) F1 offspring. Brown circles, PatCD/MatCD (n = 8 and n = 10 in males and females, respectively); pink squares, PatHFSSD/MatCD (n = 12 and n = 10 in males and females, respectively); green triangles, PatHFSSD/MatCD+F (n = 12 and n = 11 in males and females, respectively); blue inverted triangles, PatHFSSD+F/MatCD (n = 8 and n = 5 in males and females, respectively). Values are shown as means ± SEM. Scale bars, 100 μm. *p < 0.05 and **p < 0.01 vs PatCD/MatCD

In female F1 offspring born to male rats exposed to the HFSSD diet prior to mating liver fibrosis was increased (65% increase in periportal fibrosis score; p < 0.01 PatHFSSD/MatCD vs PatCD/MatCD; Fig. 4b). The increase in liver fibrosis in the female offspring was abolished by folate treatment of the dams during pregnancy (p < 0.01 PatHFSSD/MatCD+F vs PatHFSSD/MatCD, Fig. 4b). None of the study groups in the F1 male progeny showed any fibrotic changes around portal vessels. There were no statistical differences in parenchymal fibrosis levels and inflammation between any of the study groups in male offspring (Fig. 4a and Table 1). Liver triacylglycerol content did not significantly differ between the groups of F1 offspring, regardless of the sex (Tables 1 and 2).

Fig. 4
figure 4

Hepatic periportal fibrosis score in male (a) and female (b) F1 offspring and typical photomicrographs of Syrius Red staining for hepatic periportal fibrosis in male (c) and female (d) F1 offspring. Brown circles, PatCD/MatCD (n = 6 and n = 9 in males and females, respectively); pink squares, PatHFSSD/MatCD (n = 12 and n = 9 in males and females, respectively); green triangles, PatHFSSD/MatCD+F (n = 12 per study group); blue inverted triangles, PatHFSSD+F/MatCD (n = 8 and n = 7 in males and females, respectively). Values are shown as means ± SEM. Scale bars, 100 μm. **p < 0.01 vs PatCD/MatCD; ††p < 0.01 vs PatHFSSD/MatCD

Hepatic gene expression

We selected genes from studies showing that glucose tolerance in offspring was clearly differently regulated as a result of paternal high-fat diet during roughly two full rounds of spermatogenesis before mating [16, 29, 30]. Hepatic expression levels of Lcn2 mRNA were affected in both male and female F1 offspring of fathers fed HFSSD; folate treatment of parent rats normalised Lcn2 expression levels (Fig. 5b). Hepatic expression of Ppara mRNA was elevated in female F1 offspring only (1.51-fold, p ˂ 0.05) and was restored by folate treatment given to both dams and founders (p ˂ 0.05), whereas Ppara mRNA revealed no significant regulation in male F1 progeny (Fig. 5a). The analysis of hepatic mRNA expression of other genes (selected either based on a hypothesis-driven approach or based on micro-array data) revealed only minor differences between study groups in both male and female F1 progeny (Table 3).

Fig. 5
figure 5

(a, b) Expression levels of hepatic Ppara (a) and Lcn2 mRNA (b) in male and female F1 generation offspring. mRNA expression data are given as fold change relative to PatCD/MatCD; mRNA expression levels are normalised to β-actin. Brown circles, PatCD/MatCD (n = 10 per study group); pink squares, PatHFSSD/MatCD (for Ppara, n = 9 and n = 10 in males and females, respectively; for Lcn2, n = 8 and n = 10 in males and females, respectively); green triangles, PatHFSSD/MatCD+F (for Ppara, n = 10 and n = 9 in males and females, respectively; for Lcn2, n = 10 per study group); blue inverted triangles, PatHFSSD+F/MatCD (for Ppara, n = 9 and n = 10 in males and females, respectively; for Lcn2, n = 10 per study group). Values are shown as means ± SEM. *p < 0.05 vs PatCD/MatCD; p < 0.05 and ††p < 0.01 vs PatHFSSD/MatCD. (c, d) Heat maps of group-specific inter-CpG site correlation coefficients of DNA methylation of Ppara (c) and Lcn2 (d) in F1 offspring (both male and female)

Table 3 Relative hepatic gene expression profile in male and female F1 offspring

To identify yet unknown genes that are differentially expressed in offspring of male rats given an unhealthy diet prior to mating, we performed a liver whole-genome array RNA sequencing approach. We chose the most promising candidate genes based on p values and fold change in the arrays and conducted real-time quantitative PCR (Table 3, Fig. 5a, b). The most prominent alterations were seen in the expression of Lcn2 and Tmcc2 genes which showed elevation in female offspring born to male rats exposed to an unhealthy diet (p ˂ 0.05, Fig. 5b and Table 3). Similar alterations in Lcn2 gene expression were seen in male offspring.

Pancreas morphology

Pancreatic islets in the F1 offspring of both sexes born to male rats exposed to the HFSSD diet prior to mating had a decreased beta cell density (−20% in male and −15% in female F1 offspring, (p < 0.001 PatHFSSD/MatCD vs PatCD/MatCD; Fig. 6a, b). In female F1 progeny, folate given to either pregnant dams or F0 founders resulted in elevation of beta cell density when compared with their counterparts born to parents with no folate intervention (4.3% and 3.3% after folate supplementation given to dams and founders, respectively, p < 0.05 vs PatHFSSD/MatCD, Fig. 6b). However, neither of the parental folate interventions restored pancreatic beta cell density in the F1 male progeny (Fig. 6a). There were no significant differences in total number of islets, islet size distribution or islet area per mm2 pancreas section between the study groups of F1 offspring (Tables 1 and 2).

Fig. 6
figure 6

Beta cell density per pancreatic islet (% of insulin-positive area) in the male (a) and female (b) F1 offspring and typical photomicrographs of immunostaining of the pancreatic tissue for insulin in male (c) and female (d) F1 offspring. Brown circles, PatCD/MatCD (n = 16 and n = 20 in males and females, respectively); pink squares, PatHFSSD/MatCD (n = 30 and n = 21 in males and females, respectively); green triangles, PatHFSSD/MatCD+F (n = 21 and n = 19 in males and females, respectively); blue inverted triangles, PatHFSSD+F/MatCD (n = 16 and n = 17 in males and females, respectively). Values are shown as means ± SEM. Scale bars, 100 μm. ***p < 0.001 vs PatCD/MatCD; p < 0.05 vs PatHFSSD/MatCD

Pancreatic gene expression

Pancreatic candidate genes were selected as described for the liver, see above. The analysis of mRNA expression profiles of selected genes (Pparg, Ikbke, Ppara, Foxo1 and Fos) in the pancreas revealed only minor differences between study groups in both male and female progeny of the F1 generation (Table 4).

Table 4 Relative pancreatic gene expression profile in male and female F1 offspring

Global DNA methylation in the liver

Regardless of the F1 offspring sex, the rate of global DNA methylation in the liver was 1.52-fold elevated in PatHFSSD/MatCD offspring compared with their PatCD/MatCD control counterparts (p ˂ 0.05, Fig. 7a, b). In female F1 offspring, the treatment of the F0 dams with folate restored global methylation to a normal level (p ˂ 0.05 PatHFSSD+F/MatCD vs PatHFSSD/MatCD, Fig. 7b). However, effects of folate treatment of either the F0 founders or dams on the reduction in global DNA methylation rate was not statistically significant in the male offspring (Fig. 7a).

Fig. 7
figure 7

Global DNA methylation in the liver of male (a) and female (b) F1 offspring. The data are given as fold change relative to PatCD/MatCD. Brown circles, PatCD/MatCD; pink squares, PatHFSSD/MatCD; green triangles, PatHFSSD/MatCD+F; blue inverted triangles, PatHFSSD+F/MatCD. Values are shown as means ± SEM, n = 5 per study group. *p < 0.05 vs PatCD/MatCD; p < 0.05 vs PatHFSSD/MatCD

DNA methylation of specific target genes in the liver

The methylation rate of CpG islands in the promoter region of Ppara, Lcn2 and Tmcc2 genes was analysed. The differences among the groups were not significant, most likely due to the limited number of samples analysed (n = 5 per group) (ESM Table 6). We also investigated the potential correlation between the mRNA expression and the methylation rate of CpG islands in the promoter region of Ppara, Lcn2 and Tmcc2. Interestingly, the methylation rate of some CpG islands in the promoter region of Lcn2 showed significant negative correlation with Lcn2 mRNA expression in female offspring (ESM Table 7). Moreover, to investigate whether the methylation rate of a given CpG site is correlated with other CpG sites within the promoter region, Pearson correlation matrices were calculated and plotted as heat maps for each group (Fig. 5c, d and ESM Table 8). The resulting group-specific correlation patterns were clearly different. Regarding the correlation matrices of the methylation rate of CpG islands within the Ppara promoter, positive correlations, indicated by dark blue, were more predominant in the offspring born to fathers on a normal diet when compared with the offspring born to fathers on an unhealthy diet, even following folate treatment of both parents. On the other hand, regarding the correlation matrices of the methylation rate of CpG islands within the Lcn2 promoter, positive correlations were more predominant in the offspring born to fathers on an unhealthy diet when compared with the offspring born to fathers on a normal diet or those born to fathers fed an unhealthy diet where fathers or mothers had been treated with folate.

ANCOVA models considering litter size

As stated above, litter size was not significantly different between the groups. However, to investigate the influence and the potential bias of litter size, ANCOVA models, considering litter size as a covariate, were calculated for variables that were significantly different in the ANOVA analyses. Indeed, litter size was associated with several readouts, yet results overall were not affected by litter size (ESM Table 9).

Discussion

Several studies suggest that exposure of a male parent to environmental adverse factors during spermatogenesis can influence the development of traits in their offspring [14]. A model of paternal high-fat diet prior to mating is usually used. We exposed male rats to a high-fat, high-carbohydrate and high-salt diet mimicking an unhealthy fast-food diet often eaten by young men. We tested the hypothesis that the adverse effects of an unhealthy paternal diet before mating on the offspring could be ameliorated by folate treatment of either the dams or the founders before mating. Our study demonstrated that folate treatment of dams ameliorated the adverse effects on female offspring’s glucose metabolism. This might be partially due to folate-induced beta cell preservation in the female offspring combined with a normalisation of hepatic connective tissue density. Furthermore, folate treatment reversed dysregulated Ppara, Lcn2 and Tmcc2 gene expression and normalised liver total DNA methylation.

Sex-dependent effects of adverse paternal diets before mating on the offspring

The effects of an adverse paternal diet on the offspring’s phenotype were sex-specific. Only female offspring developed an impaired glucose tolerance. This is in agreement with the findings of published studies [16, 31, 32]. Gene expression shows sex-specific differences, which are detectable in the pre-implanted embryo, long before gonadal development and sex hormone production [33]. When comparing the consequences of parental nutritional insults with respect to the offspring, paternal pre-conception stimuli were shown to display a stronger effect on female offspring [16, 31, 32].

Folate treatment and fetal programming

Folate treatment during pregnancy prevents adverse developmental programming [19, 20, 25]. A recent study showed that the effects on the offspring of folate during pregnancy are sex-specific [34]. We demonstrated that folate treatment of pregnant dams prevents adverse metabolic effects of paternal programming. In other words, a disadvantageous diet in the male parent prior to mating can be corrected by folate treatment of the dam after mating.

Folate treatment of male rats exposed to an unhealthy diet improved pancreatic beta cell density in female offspring (Fig. 6). Since we only analysed the effect of one folate dose, the very next task should be to establish the dose-dependency of this effect. The same is true for the supplementation of the dams.

Whereas the paternal unhealthy diet-induced fat accumulation in the liver of male and female offspring, liver fibrosis and systemic inflammation was only seen in female offspring (Tables 3 and 4). The fat accumulation in the liver of male offspring is not associated with inflammation and is thus benign, in contrast to the findings in female offspring. Beta cell density in offspring was reduced by paternal HFSSD feeding and was improved slightly by folate treatment in female offspring only. This effect in female offspring might contribute to the folate-treatment-related improvement in the OGTT measurements.

The beneficial effects of folate administration during pregnancy were associated with decreased global DNA methylation in the liver of female offspring. This is counterintuitive to previous findings suggesting that folate is a major source of methyl groups required for DNA methylation and, hence, increases DNA methylation. However, recent studies support our findings: in one study, folate supplementation was associated with genome-wide loss of methylation [35] and in another study [36] maternal plasma folate during pregnancy was associated with a decreased methylation of 416 CpGs (94%) in newborns and increased methylation of 27 CpGs (6%). Alterations of microRNAs (miRNAs) in sperm are known to affect paternal programming [21, 22]. Folate can alter miRNA status via regulation of gene expression, possibly altering synthesis/effects of DNA methyltransferases or enzymes involved in the folate-dependent one-carbon metabolism pathway, leading to decreased DNA methylation [37].

To identify underlying molecular mechanisms, we performed a candidate gene approach [29, 30, 38] as well as a whole-genome array approach. Both approaches are necessary, since open non-hypothesis-driven technologies are not yet capable of discovering all underlying alterations in gene expression [39,40,41]. The key finding of the whole-genome array was the identification of the dysregulated hepatic genes Lcn2 and Tmcc2 in female offspring (Fig. 5c, d). A high-fat, high-fructose diet was found to upregulate hepatic Lcn2 expression in mice [42]. Importantly, lipocalin 2 levels correlate with obesity, impaired insulin sensitivity and diabetes and have been suggested as a potential prognostic biomarker of non-alcoholic fatty liver disease [43]. Notably, a recent study showed that Lcn2 expression is altered by maternal nutrition during the development of the fetal liver [35]. In our study, the increased hepatic expression seen in offspring born to F0 founders fed an unhealthy diet was normalised in female offspring of folate-treated F0 founders on an unhealthy diet. In contrast, treatment of pregnant dams with folate had no significant effect on hepatic Lcn2 expression in female offspring. Tmcc2 plays a role in Alzheimer’s disease; however, its role in the pathogenesis of paternal unhealthy diet-induced liver damage is unknown so far.

The candidate gene approach revealed normalisation of liver Ppara expression seems to play a key role in the effects of maternal folate treatment. Notably, a recent study in mice found that a fast-food-induced increase in hepatic fibrosis was associated with an increase in hepatic Ppara expression [44]. Successful treatment of the liver fibrosis normalised hepatic Ppara expression, like in our study. Maternal high-fat diet also results in dysregulated fetal hepatic Sirt1 expression, sirtuin 1 protein level and activity and a concomitant dysregulation of Sirt1-associated genes, including Ppara [45, 46]. However, in the current study the paternal unhealthy diet was not associated with a significant effect on Sirt1 expression, indicating other underlying mechanisms of Ppara regulation.

The methylation level of certain CpG islands in the promoter region of Lcn2 showed significant negative correlation with Lcn2 mRNA expression exclusively in the female offspring, suggesting that liver Lcn2 expression could be controlled through DNA methylation in a sex-dependent manner. Because it has been suggested that there may be correlations between the methylation states of neighbouring and/or functionally related CpG sites [47, 48], we analysed treatment-group-specific inter-CpG site correlation of DNA methylation and plotted the resulting correlation coefficients as heat maps. Interestingly, different patterns were observed when comparing treatment groups. Different correlation patterns between the degree of DNA methylation of one CpG site to another could result in a different net effect on gene expression [47]. However, as methylation is just one of several epigenetic modifications, other mechanisms such as histone modifications might be of importance [45]. Recent studies highlight a key role for paternal sperm-cell-derived small non-coding RNAs [21, 49]. One study, investigating effects of a paternal high-fat diet, demonstrated sex-specific metabolic disturbances in female offspring. The authors showed that a high-fat diet alters the expression of miRNA let-7c in the sperm of F0 rats and their F1 offspring. This finding is further substantiated by studies showing that the microinjection of sperm-derived RNAs into oocytes can transmit environmentally induced paternal phenotypic changes to the resulting offspring [21, 49, 50].

In conclusion, folate treatment of pregnant dams, but not the male founders, reverses detrimental effects on female offspring’s glucose metabolism induced by pre-conceptional male founder high-fat, high-carbohydrate and high-salt diet. This effect might be at least partially due to a folate-induced beta cell preservation in the female offspring combined with a partial improvement of the liver abnormalities.