DNA methylation-independent growth restriction and altered developmental programming in a mouse model of preconception male alcohol exposure

ABSTRACT

The preconception environment is a significant modifier of dysgenesis and the development of environmentally-induced disease. To date, fetal alcohol spectrum disorders (FASDs) have been exclusively associated with maternal exposures, yet emerging evidence suggests male-inherited alterations in the developmental program of sperm may be relevant to the growth-restriction phenotypes of this condition. Using a mouse model of voluntary consumption, we find chronic preconception male ethanol exposure associates with fetal growth restriction, decreased placental efficiency, abnormalities in cholesterol trafficking, sex-specific alterations in the genetic pathways regulating hepatic fibrosis, and disruptions in the regulation of imprinted genes. Alterations in the DNA methylation profiles of imprinted loci have been identified in clinical studies of alcoholic sperm, suggesting the legacy of paternal drinking may transmit via heritable disruptions in the regulation of imprinted genes. However, the capacity of sperm-inherited changes in DNA methylation to broadly transmit environmentally-induced phenotypes remains unconfirmed. Using bisulphite mutagenesis and second-generation deep sequencing, we find no evidence to suggest that these phenotypes or any of the associated transcriptional changes are linked to alterations in the sperm-inherited DNA methylation profile. These observations are consistent with recent studies examining the male transmission of diet-induced phenotypes and emphasize the importance of epigenetic mechanisms of paternal inheritance beyond DNA methylation. This study challenges the singular importance of maternal alcohol exposures and suggests paternal alcohol abuse is a significant, yet overlooked epidemiological factor complicit in the genesis of alcohol-induced growth defects, and may provide mechanistic insight into the failure of FASD children to thrive postnatally.

Introduction

It has recently become clear that, through epigenetic means, environmental exposures before conception exert a significant impact on offspring health and development.¹ In particular, preconception male exposures to a range of environmental factors induce alterations to the developmental program of sperm, which can be correlated with increased rates of structural and metabolic defects in the next generation.^2-16 These studies challenge the singular importance of maternal in utero environmental exposures and implicate paternal exposure history as an additional and important mediator of both environmentally-induced disease and dysgenesis.^17,18 However, the capacity of preconception paternal exposures to broadly contribute to the development of environmentally-induced disease has not been rigorously explored, largely due to the common misconception that sperm do not transmit heritable information beyond the genetic code.

Epidemiologic studies demonstrate that alcohol is the most prevalent teratogen to which humans are exposed, with 6 to 17 children per 1000 live births diagnosed with some degree of fetal alcohol spectrum disorder (FASD).¹⁹ This condition is characterized by a spectrum of structural defects, central nervous system disorders, and growth deficits that persist well into postnatal life.^19,20 One of the major confounding elements in the study of this disorder is the enormous variation observed in FASD phenotypes and incidence.²¹ FASD-associated defects have a wide range of severity and, importantly, while many women who drink alcohol during pregnancy have affected children, a subset of the exposed offspring remain unaffected.²² The observed variance and inconsistency in FASD phenotypes suggest that multiple factors beyond the incidence of maternal ethanol exposure must play a significant role in the genesis of this disorder.

While alcohol exposure in utero is, undoubtedly, a significant element in the origins of FASD-associated growth defects, several independent studies have also emerged indicating a link between paternal alcohol consumption before conception and growth deficits consistent with those of FASDs.^18,23 The long-term persistence of FASD growth defects and emerging association with the development of chronic disease later in life, suggest alcohol has the capacity to heritably disrupt fundamental aspects of epigenetic programming.^24-28 These observations suggest a potential association between preconception male alcohol exposure, altered epigenetic programming in sperm, and the development of FASD-associated growth defects in the offspring. In support of this assertion, animal models of paternal alcohol exposure report alterations in the control of key enzymes regulating chromatin structure as well as changes in the DNA methylation profiles of alcohol-exposed sperm.^29-32

The majority of these studies have concentrated on alterations in DNA methylation occurring within the regulatory regions of imprinted genes, suggesting altered placentation may be at the root of the observed fetal growth restriction.^30-32 Indeed, disruptions in genomic imprinting have well-characterized impacts on the development of the placenta, and the faithful maintenance of these imprints through early development is crucial for both fetal and placental growth.^33,34 Importantly, similar alterations in the DNA methylation profiles of alcohol-exposed sperm can be identified in humans, suggesting the legacy of paternal drinking may transmit to the offspring via heritable disruptions in the regulation of imprinted genes.³⁵ However, while several instances of paternally-inherited alterations in developmental programming have been correlated with changes in the DNA methylation profiles of paternal sperm, an increasing number of cases suggest these heritable phenotypes transmit independent of this epigenetic modification.^3,13,14 Thus, the universal dependence of these phenotypes on DNA methylation-based mechanisms of epigenetic inheritance remains an unanswered question central to understanding the paternal transmission of environmentally induced traits.

Given the recent association between paternal environmental exposures and long-term metabolic dysfunction,¹⁷ we sought to examine the capacity of preconception male ethanol exposure to impact developmental programming in the male germline and determine if heritable alterations in the DNA methylation profiles of alcohol-exposed sperm contribute to the prenatal growth deficits associated with FASDs. To this end, we used an established mouse model of epigenetic programming, in which distinct single nucleotide polymorphisms between the maternal (C57BL/6J) and paternal [C57BL/6(CAST7)] strains allowed us to track allelic patterns of gene transcription within select imprinted loci.³⁶ Our studies associate preconception male ethanol exposure with prenatal growth restriction, decreased placental efficiency, disruptions in the genetic pathways controlling both lipid metabolism and hepatic fibrosis, and the abnormal expression of imprinted genes. However, similar to studies examining the inheritance of diet induced metabolic dysfunction,^3,13 we find evidence to suggest that the transmission of these phenotypes occurs independently of sperm-inherited alterations in DNA methylation.

Results

Preconception alcohol exposure does not impact male fertility or sire weight

Previous studies examining alterations in the DNA methylation profiles of alcohol-exposed sperm have used rodent models of alcohol exposure using oral gavage.^29,30,32 This technique is known to induce activation of the stress response,³⁷ which has previously been shown to modulate epigenetic programming in sperm.⁹ To avoid this confounding factor, we elected to use a voluntary model of consumption (Drinking in the Dark), where males were provided limited access to ethanol (EtOH) during a 4-hour window immediately after their sleep cycle.³⁸ Here, postnatal day 90, adult C57BL/6(Cast7) males were maintained on a 12-hour light/dark cycle and provided access to either a solution of 10% (w/v) EtOH plus 0.066% (w/v) Sweet'N Low (experimental) or 0.066% (w/v) Sweet'N Low alone (control) for 4 h a day. During the first week of the exposure paradigm, males acclimating to the EtOH treatment consumed 9% less fluid during the exposure window than the controls (P = 0.03, Fig. 1A). However, after the first week of treatment, no significant differences in fluid consumption were observed. Once consistent patterns of drinking were established (one week), males were maintained on this protocol for a period of 70 d, which corresponds to the length of approximately 2 complete spermatogenic cycles, thereby ensuring that both pre-meiotic and post-meiotic sperm were exposed to EtOH.^39,40 Using this model of exposure, experimental males consistently achieved blood EtOH concentrations of 210 mg/dL (Fig. 1B). As a reference, this concentration is approximately 2.5 times the legal limit, and reflects a range frequently obtained by binge drinkers and alcoholics.⁴¹ As previous studies have shown increased paternal weight can influence both epigenetic programming in sperm and fertility, we assayed these parameters at the conclusion of the exposure window.^7,42 Chronic EtOH consumption did not influence paternal body weight, fertility, or the sex distribution of the offspring (Fig. 1C–E).

Figure 1.

Chronic male alcohol exposure does not impact fertility or sire weight. (A) Average fluid consumption per gram body weight during the preconception treatment period (n = 8 control, 10 alcohol-exposed sires). (B) Average blood alcohol concentrations during the exposure period (n = 6 males). (C) Average body weight of sires at the end of the 70-day EtOH exposure period (n = 8 control, 10 alcohol). (D) Average litter size and (E) distribution of offspring sex between litters sired by preconception EtOH-exposed and control males (n = 15 litters control, 20 litters alcohol). Error bars represent SEM *P<0.05 and ***P<0.001 [comparisons between 10% (w/v) EtOH plus 0.066% (w/v) Sweet'N Low (alcohol) vs. 0.066% (w/v) Sweet'N Low alone (control) preconception treatments]. Data analyzed using either an unpaired t-test or a one-way ANOVA followed by Sidak post hoc analysis.

Male alcohol exposure associates with fetal growth restriction

Once the 70-day milestone was surpassed, males were mated to unexposed dams and fetuses collected at day 14.5 of gestation (GD14.5). During this dissection, the uterus was removed, and the maternal decidua completely separated from the fetal interface. The GD14.5-time point was selected based on our experience reliably excising the fetal component of the placenta away from maternal tissues, thus ensuring that the fetal component of the placenta is minimally contaminated with maternal cells. This approach and mouse model permit accurate assessment of placental patterns of imprinted gene expression.⁴³ At this developmental time point, chronic paternal EtOH exposure was determined to be significantly associated with a reduction in the weight of the gestational sac (11% in males, 8% in females, P < 0.01), a reduction in crown rump length (3.6% in males, 3.9% in females, P < 0.05), and, in female offspring only, a 7.4% reduction in fetal weight (Fig. 2A–C). In the female offspring of EtOH-exposed males, fetal weight was highly correlated with placental weight, and a 12% decrease in placental efficiency (P < 0.05) was observed (Fig. 2D–E). These data indicate that male preconception alcohol exposure is associated with fetal growth restriction, similar in magnitude to those observed in maternal in utero models of EtOH exposure.⁴⁴

Figure 2.

Preconception male alcohol exposure impacts fetal growth. (A) Average weight of the gestational sac, (B) average fetal crown-rump length, and (C) average fetal weights between male and female offspring sired by preconception EtOH-exposed or control males. All measures taken on gestational day 14.5 (control n = 46 males, 55 females, alcohol n = 56 males, 62 females). (D) Analysis of correlation between fetal weight and placental weight in female offspring sired by EtOH-exposed and control males; (E) placental efficiencies between male and female offspring sired by preconception EtOH-exposed and control sires. Efficiencies obtained by dividing fetal weight by placental weight (control n = 46 males, 55 females, alcohol n = 56 males, 62 females) Error bars represent SEM, ns = not significant, *P < 0.05, **P < 0.01 and *** P < 0.001 (comparisons between 10% (w/v) EtOH plus 0.066% (w/v) Sweet'N Low (alcohol) vs. 0.066% (w/v) Sweet'N Low alone (control) preconception treatments). Data analyzed using a one-way ANOVA followed by Sidak post hoc analysis.

Placental alterations in the pathways controlling lipid transport and cellular proliferation

The placenta plays a critical role in controlling maternal-fetal resource allocation and mediating fetal growth.^33,34 A gross histological assessment of the placenta did not reveal any morphological abnormalities or alterations in the proportional sizes of the labyrinthine and spongiotrophoblast layers between preconception treatments (Fig. 3A–B). Deep-sequencing analysis of the placental transcriptome identified ∼1000 differentially expressed genes between the placentas of offspring sired by EtOH-exposed and control males. Analysis of these data sets identified disruptions in genetic pathways related to both lipid transport and cellular proliferation (Fig. 3C–H). As one of the predominant pathways to emerge from this analysis was lipid transport, we assayed the concentration of total cholesterol within the placenta and fetal liver. Total cholesterol was significantly increased in the fetal livers of offspring sired by the EtOH-exposed males (Fig. 3I–J). These results significantly correlate preconception male EtOH exposure with disruptions in the genetic pathways regulating lipid homeostasis, which is known to be associated with intrauterine growth restriction in clinical studies.⁴⁵

Figure 3.

Preconception male alcohol exposure alters genes controlling lipid transport and cellular proliferation in the placenta. (A) Periodic acid–Schiff stained histological sections of gestational day 14.5 placental tissues derived from offspring sired by preconception EtOH-exposed or control males; L = labyrinthine and S = spongiotrophoblast. (B) Ratio of labyrinthine to spongiotrophoblast surface area compared between preconception treatments (n = 5 males, 5 females). (C) Comparison of the placental transcriptome between offspring sired by EtOH-exposed and control males. Volcano plots between male and female offspring (n = 4). (D) Top 2 categories identified using Ingenuity Pathway Analysis. (E) qRT-PCR validation of candidate genes related to lipid transport in male offspring; (F) genes controlling cellular proliferation in male offspring; (G) genes regulating cell proliferation in female offspring, and (H) genes related to lipid transport in female offspring. Note the log scale. (I) Total cholesterol levels in the placenta and (J) fetal liver of offspring sired by sired by EtOH-exposed and control males (n = 5). For qRT-PCR analyses, measured Ct values were normalized to the geometric mean of Sdha, Hprt, and Mrpl1 and graphed relative to the control treatment. Graphs represent independent replicates (n = 6 male, 8 female), with 2 independent RT reactions and 3 qRT-PCR measurements for each RT. Error bars represent SEM * P < 0.05, ** P < 0.01 and *** P < 0.001 (comparisons between EtOH and control preconception treatments). Data analyzed using either an unpaired t-test or a one-way ANOVA followed by Sidak post hoc analysis.

Sex-specific impacts on the pathways regulating hepatic fibrosis

Given the observed changes in total cholesterol, we next assayed gene expression in the GD14.5 liver using deep sequencing. These analyses identified a small set of differentially expressed genes (∼350 in males, ∼500 in females), with a large proportion of the candidate genes participating in the genetic pathways regulating hepatic fibrosis and stellate cell activation (Fig. 4A–B).⁴⁶ Interestingly, of the 153 genes common between both sexes, 147 (96%) displayed alterations in transcription that were diametrically opposite between males and females. As examples, transcripts encoding Acta1 and Col18a1 were significantly increased in the livers of male offspring, while the same genes were downregulated in female offspring (Fig. 4C–D). Of the 34 identified genes associated with hepatic fibrosis, 20 (60%) increased transcription in the male offspring, while these same genes were downregulated in females (Fig. 4E)

Figure 4.

Preconception alcohol exposure activates the genetic pathways controlling hepatic fibrosis. (A) Comparison of the gestational day 14.5 liver transcriptome between offspring sired by EtOH-exposed and control males. Volcano plots between male and female offspring (n = 4). (B) Top catagory identified using Ingenuity Pathway Analysis. (C) qRT-PCR validation of candidate genes in male and (D) female offspring (n = 8). (E) Heatmap contrasting the abundance of transcripts encoding genes associated with hepatic fibrosis between male and female offspring. Expression patterns were normalized to the offspring of control animals. (F-I) qRT-PCR analysis of candidate genes between the smallest (bottom 25%), and largest (top 25%) weight offspring sired by EtOH-exposed and control males (n = 8). (J) Total levels of cellular hydroxyproline within the gestational day 14.5 livers of offspring sired by EtOH-exposed and control males (n = 4) (K) qRT-PCR analysis of transcripts encoding inflammatory cytokines between the offspring of EtOH-exposed and control males (n = 8). For qRT-PCR analyses, measured Ct values were normalized to the geometric mean of transcripts encoding Ywhaz, Hprt, and Mrpl1, and graphed relative to the control treatment. Graphs represent independent replicates, with 2 independent RT reactions and 3 qRT-PCR measurements for each RT. Data analyzed using either an unpaired t-test or a one-way ANOVA followed by Sidak post hoc analysis. Error bars represent SEM * P < 0.05, ** P < 0.01 and *** P < 0.001 (comparisons between alcohol and control preconception treatments).

To determine if alterations in these genes could be further correlated with the growth restriction phenotype, we separated samples into the smallest and heaviest offspring based on fetal weight, and assayed the expression of 2 pro-fibrotic candidate genes.⁴⁶ These experiments identified a strong correlation between altered patterns of transcription and preconception EtOH-dependent reductions in fetal growth (Fig. 4F–I). For example, in the smallest male fetuses, transcripts encoding Col18a1 and TgfB2 displayed preconception EtOH-dependent increases, while the abundance of these transcripts in the largest sized fetuses were not significantly different (Fig. 4F–G). In contrast, the smallest female offspring displayed preconception EtOH-dependent downregulation of both Col18a1 and TgfB2, with no detectable differences in the largest offspring (Fig. 4H–I). Interestingly, transcripts encoding TgfB2 exhibited differences in expression between the smallest and largest control offspring that were distinct between males and females (P > 0.05). These observations suggest a sex-specific interaction between fetal growth rate and the transcriptional control of this cytokine.

Sex-specific induction of hepatic fibrosis and alterations in immune signaling

To determine if alteration of these gene cohorts could be associated with an increased fibrotic response, we next assayed the fetal liver for the abundance of hydroxyproline, which is a commonly used marker of hepatic fibrosis.⁴⁶ Male offspring displayed a 15% increase in hydroxyproline (P < 0.05), while female livers were identical to the controls (Fig. 4J). Given the link between intrauterine growth restriction, hepatic fibrosis, and inflammation in clinical studies,⁴⁷ we examined the expression of 4 key pro-inflammatory cytokines using qRT-PCR. These experiments identified significant alterations in transcripts encoding Infg in male and both Infg and Il6 in female offspring (Fig. 4K), while Il1B and Tnfα did not display any significant changes in either sex (data not shown).

Alterations in the regulation of imprinted genes

Given the link between imprinted genes, hepatic dysfunction, and fetal-placental growth,^33,34,48 we assayed the parent of origin-specific expression of 16 genes regulated through genomic imprinting within the placenta and fetal liver. These analyses identified decreased expression of Gnas and Grb10 within the placentas of male offspring and Sgce in the female offspring of EtOH-exposed males (Fig. 5A–B). In the fetal liver, we observed significant reductions in transcripts encoding Cdkn1c, Dcn, and Gnas in male offspring and Meg3 in female offspring (Fig. 5C–D). However, none of the examined imprinted genes displayed significant differences in the contribution from the normally silenced allele (Table 1). These observations associate preconception paternal EtOH exposure with decreased expression of select imprinted genes that is independent of compromised genomic imprinting.

Figure 5.

Altered patterns of imprinted gene expression within the offspring of alcohol-exposed males. (A) qRT-PCR analysis of imprinted genes in the placentas of male and (B) female offspring sired by EtOH-exposed and control sires. (C) qRT-PCR analysis of imprinted genes in the male and (D) female fetal liver. Graphs represent independent replicates (n = 8 male, 8 female), with 2 independent RT reactions and 3 qRT-PCR measurements for each RT. Data analyzed using an unpaired t-test. Error bars represent SEM *P < 0.05 and **P < 0.01 (comparisons between alcohol and control preconception treatments).

Table 1.

Analysis of single nucleotide polymorphisms within the mRNAs of genes regulated by genomic imprinting.

Gene	SNP Location	Increased Contribution of Silenced Allele?
Ascl2	chr7:142,968,971	No
Cdkn1c	chr7:143,460,109	No
Dcn	No SNP found	–
Dlk1	chr12:109,460,379	No
Gatm	chr2:122,594,926	No
Gnas	No SNP found	–
Grb10	chr11:11,954,907	No
Gtl2	chr12:109,545,837	No
H19	chr7:142,577,095	No
Igf2 (1)	chr7:142,652,037	No
Igf2 (2)	chr7:142,652,936	No
Igf2r	chr17:12,682,709	No
Mest	chr6:30,747,382	No
Ndn	chr7:62,348,457	No
Peg3 (1)	chr7:6,706,207	No
Peg3 (2)	chr7:6,706,217	No
Sgce	chr6:4,717,926	No
Slc38a4	chr15:96,995,974	No
Snrpn	chr7:59,983,415	No
Ube3a	chr7:59,228,878	No
Zac1	chr10:13,128,934	No
Zim1	chr7:6,675,637	No

Preconception EtOH exposure does not influence male fertility or alter levels of the one-carbon metabolite S-adenosylmethionine

To determine if chronic EtOH exposure could influence the male-inherited epigenetic program, breeder males were killed and both reproductive and epigenetic parameters examined. No significant differences in sperm concentration or the percentage of live sperm were observed between treatment groups (Fig. 6A–B). No treatment effects were observed on the weights of the paternal liver or epididymal fat (Fig. 6C–D). Previous studies have suggested that chronic exposure to alcohol disrupts one-carbon metabolism, and in doing so, has the capacity to alter epigenetic programming.⁴⁹ In our chronic model of exposure, no significant differences in the abundance of the one-carbon donor S-adenosylmethionine could be identified between the kidney, liver, or testis of EtOH-exposed and control males (Fig. 6E).

Figure 6.

Preconception male alcohol exposure does not impact base fertility or abundance of the one-carbon metabolite S-adenosylmethionine. (A) Concentration of sperm between EtOH-exposed and control males; (B) percentage of live sperm between EtOH-exposed and control males (n = 8). (C) Measures of the percentage of total body weight for the liver and (D) epididymal fat between EtOH-exposed and control males (n = 8). (E) Comparison of total S-adenosylmethionine in the kidney, liver and testis between EtOH-exposed and control males (n = 8). Physiological and reproductive parameters were analyzed using either an unpaired t-test or a one-way ANOVA followed by Sidak post hoc analysis.

Preconception EtOH exposure does not influence the DNA methylation profile of sperm

We quantified the DNA methylation profiles of alcohol-exposed and control sperm using bisulphite mutagenesis and second-generation deep sequencing. In these analyses, we could not detect any differences in the global levels of DNA methylation between any of the preconception treatments, in any of the CpG, CHG or CHH contexts (Fig. 7A). Informatics analysis performed using the RStudio Bioconductor package methylKit identified extremely strong correlation between treatment groups and clustering analysis failed to separate sperm samples by preconception treatment (Fig. 7B–C). Setting a cutoff of 30x coverage, a minimal change in DNA methylation of 25%, a q-value of 0.01 and P value of less than 0.01,⁵⁰ we were only able to identify 2 high-confidence differentially methylated regions (Filip1l and Rn45s) between the EtOH-exposed and control males. When we relaxed these criteria to include regions with only 5x coverage (remaining parameters kept the same), we could expand this list to 18 candidate CpGs and 124 tiled loci. Unexpectedly, 68% of these differentially methylated regions mapped to gene bodies (37% exonic and 31% intronic), while only 3% could be mapped to gene promoters (Fig. 7D). When we randomly selected 4 of these lower confidence candidate loci to contrast measures between Methyl-seq and PCR-based bisulphite sequencing approaches, only 2 displayed significant alterations in DNA methylation using both methods (Fig. 7E). Using PCR-based bisulphite sequencing, the 2 significant regions identified both displayed less than a 25% change. Importantly, no differences in the DNA methylation profiles of any imprinted genes were observed and no correlative changes in the expression of the nearest mappable genes in either the fetal liver or placenta for any candidate loci could be identified.

Figure 7.

Preconception male alcohol exposure does not impact the DNA methylation profiles of paternal sperm. (A) Global DNA methylation profiles of paternal sperm (n = 3) across CpG, CHG and CHH contexts. (B) Pearson correlation analysis between genomic DNA methylation profiles of paternal sperm (n = 3). (C) Clustering analysis between EtOH-exposed and control DNA methylation profiles of paternal sperm. (D) Association of differentially methylated loci with genomic features. (E) Comparison of low confidence differentially methylated loci between Methyl-Seq data sets and PCR-based Bisulphite sequencing. Bisulphite sequencing data were compared using an unpaired t-test. Error bars represent SEM * P < 0.05, ** P < 0.01 and **** P < 0.0001.

Discussion

The objective of this study was to determine the impact preconception male drinking exerts on the incidence of FASD growth deficits, and the relevance the reported alterations in the DNA methylation profiles of EtOH-exposed sperm have on the development of these phenotypes. We find that preconception paternal EtOH exposure associates with fetal growth restriction, as measured on day 14.5 of gestation, as well as proportional increases in the placental weight of female offspring (reduced placental efficiencies). These phenotypes are accompanied by alterations in imprinted gene expression, altered cholesterol trafficking, and sex-specific indicators of hepatic fibrosis. Interestingly, while hepatic fibrosis has long been one of the clinical manifestations of alcoholic liver disease in adults,⁵¹ 3 isolated case reports have also identified a similar condition in FASD children.^52-54 The identification of increased lipids and markers of hepatic fibrosis in these published case reports, as well as our work, suggest some aspects of this pathology may involve alterations in developmental programming. However, we do not know if the indicators of fibrosis we observed in this model were a primary consequence of altered developmental programming (programming of the sperm to impart a heightened fibrotic state) or a secondary consequence of fetal growth restriction. Indeed, intrauterine growth restriction alters the hepatic control of glucose transport, has been linked to abnormal lipid profiles in the offspring, and has also been associated with the development of adolescent non-alcoholic fatty liver disease (NAFLD).⁵⁵ NAFLD shares many common elements with alcoholic liver disease, including both fibrosis and the accumulation of lipids.^45,56 Each of these conditions are also accompanied by alterations in immune signaling, which may account for the sex-specific phenotypes we observe here.⁵⁷ We suspect that, similar to other models of developmental programming, a “second hit” will be required to unmask the full extent of the programmed dysfunction and reveal the breadth of pathologies in our model.⁵⁵

Alterations to the developmental program have been proposed to arise through environmentally-induced changes to the epigenome during crucial periods of plasticity.⁵⁸ Numerous studies support the notion that gametogenesis is a labile period and that both placental and hepatic function are profoundly impacted by alterations in the developmental program.^1,56 Similar to studies of preconception male alcohol exposure, rodent studies examining maternal periconceptional alcohol exposure also report EtOH-induced changes in fetal growth as well as sex-specific impacts on metabolic programming.^59,60 Importantly, both our model of preconception male alcohol exposure and the periconceptional model of maternal exposure reported by the Moritz group yield fetal growth deficits similar in magnitude to those reported in rodent models of in utero exposure.⁴⁴ Collectively, these studies reveal preconception alcohol exposure to be a significant, yet under-recognized contributor to FASD biology and an important factor in the broad variation of FASD growth and neurocognitive defects.

One common element to emerge from epigenetic focused studies of both preconception and gestational models of EtOH exposure has been disruptions in the transcriptional control of imprinted genes.^24,30-32 Similar to these works, we also observe altered expression of imprinted genes in our preconception male model of exposure, including Gnas, Grb10 (males) and Sgce (females) in the placenta, as well as Cdkn1c, Dcn, Gnas (males), and Meg3 in the fetal liver. Cdkn1c, Dcn, Gnas, and Grb10 are all expressed predominantly from the maternally-inherited allele and act to regulate both fetal and placental growth.^33,34 Meg3 is a maternally expressed non-coding RNA and in the liver, its overexpression has been associated with the inhibition of stellate cell activation and liver fibrosis.⁴⁸ However, in our model, all of these candidate genes exhibited decreased expression with no evidence of inappropriate contributions from the normally silent paternal allele. Moreover, using methyl-seq with an average of 90 million reads per sample, we could find no evidence to suggest that these, or any other transcriptional changes, were due to alterations in the sperm-inherited DNA methylation profiles. However, we cannot rule out the possibility that a higher dose of alcohol could influence the profile of sperm-inherited DNA methylation.

Although previous reports examining alcohol-exposed sperm have identified alterations in DNA methylation within imprinted loci, these changes tend to be very modest.^30,32,35 For example, Knezovich et al. reported EtOH-dependent decreases of 1.75% for CpG # 7 within the H19 regulatory region and a 2.2% decrease for CpG # 10 within the Snrpn locus. Similarly, Liang and colleagues identified a 3–6% EtOH-dependent decrease within the H19 regulatory region, while the Peg3 regulatory region displayed a 7% increase. These reported changes tend to be very minor and often only involve single cytosines, while the levels of DNA methylation for the overall loci remain unchanged.^30,35 Each sperm carries a single haploid genome and a change of 10% or less suggests that an incredibly small proportion of the total population was impacted. For example, a loci that transitions from ∼90% methylation down to 80% methylation implies that instead of 1 out of 10 sperm being affected by alcohol, the populations shifts to 2 in 10 being impacted. For changes of less than 5%, the induced change is even less frequent. Yet, each of these cited studies demonstrate measurable changes in offspring growth or behavior across the total population.^29,30,32 Given that the profiles of EtOH-exposed sperm reported in both these studies, as well as our genome-wide analysis, exhibited less than a 20% overall change, it is very challenging to reconcile how the observed differences in sperm DNA methylation, especially at single CpG resolution, could influence the resulting phenotypes. Further, while changes in the somatic DNA methylation profile of the resulting offspring have been reported, they again tend to be very modest (less than 10%) and cannot be correlated with alterations in the transcriptional control of affected genes.³² Previous studies examining the transmission of metabolic dysfunction through sperm have provided evidence to both support and refute the involvement of sperm-inherited changes in DNA methylation in the transmission of paternally-inherited phenotypes.^3,5,13 Collectively, our results, along with previously published data,^29,30,32 strongly suggest a DNA methylation independent mechanism in the transmission of epigenetic errors induced by preconception paternal alcohol exposure.

Recently, it has been revealed that mature sperm carry populations of non-coding RNAs (ncRNAs) that include microRNAs (miRNAs), tRNA fragments and endogenous short interfering RNAs.⁶¹ Interestingly, these ncRNAs appear to have functional roles during embryonic development and the profile of sperm-inherited RNAs can be modulated by paternal diet.^62,63 Many ncRNAs, including miRNAs, are influenced by alcohol exposure and have roles in the development of FASD phenotypes.⁶⁴ As previous studies have identified a link between paternally inherited miRNAs and stress-induced phenotypes in the offspring,⁹ an important next step will be to determine if alcohol exposure can impact the ncRNA population of sperm, and if these RNAs represent the transmissible epigenetic factor driving development of the observed growth restriction phenotypes.

Despite the implementation of numerous community outreach and educational programs, alcohol-related birth defects remain a significant public health concern, and a tremendous economic burden.¹⁹ The identification of one of the defining FASD-associated defects in an established mouse model of preconception male exposure strongly suggest the lifestyle choices of the father are extremely relevant to the genesis of this debilitating disorder. Given that 70% of men drink and 40% drink heavily,⁶⁵ further investigation is clearly warranted.

Materials and methods

Animal work

All experiments were conducted under AUP 2014–0087 and approved by Texas A&M University IACUC. The C57BL/6(CAST7) strain of mice were generated in the Bartolomei laboratory and were selected to possess portions of a Mus musculus castaneus (CAST) chromosome 7 and chromosome 12 (where at least 5 imprinting domains and more than 30 imprinted genes reside) bred onto a C57BL/6J background.³⁶ When using F1 hybrid crosses between the B6(CAST7) strain and a C57BL/6J strain, we can distinguish the maternal and paternal alleles of select genes using C57BL/6(CAST7) and C57BL/6J polymorphisms that we have identified by either primary sequence or database analysis.⁴³ C57BL/6(CAST7) males were used in a chronic, moderate-dose, voluntary alcohol exposure paradigm referred to as ‘Drinking in the Dark’.³⁸ Here, individually caged, postnatal day 90, adult males were provided limited access to EtOH during a 4-hour window immediately after their sleep cycle.³⁸ Males were maintained on a 12-hour light/dark cycle and provided access to either a solution of 10% (w/v) EtOH (Sigma, catalog# E7023) plus 0.066% (w/v) Sweet'N Low (Cumberland Packing Corp, Brooklyn, NY) vs. 0.066% (w/v) Sweet'N Low alone for 4 h a day. Prolonged exposure to a 10% Sweet'N Low solution has been shown to drive the development of glucose intolerance through functional alterations to the intestinal microbiota.⁶⁶ Although the experimental paradigm reported here used a 0.066% Sweet'N Low solution, which is 150-fold lower than those cited in these previous experiments,⁶⁶ we were careful to ensure that mice in both preconception treatment groups received equivalent exposures. Once the 70-day preconception treatment was achieved, 2 naturally cycling females were placed into a new cage along with each exposed male. During these matings, males were not provided access to the alcohol/control preconception treatments. The next morning, matings were confirmed by the presence of a vaginal plug and both the male and female mice returned to their original cages. Males were allowed a 24-hour rest period, during which the preconception exposure was resumed and then used in a subsequent mating. This procedure was repeated until a minimum of 3 confirmed matings had been achieved, at which point sires were killed and their reproductive tracts isolated.

Sex determination

Genomic DNA was isolated using the DNeasy Blood and Tissue Kit (Qiagen, catalog # 69504) and PCR amplification of the Zfy and Xist genes conducted.

Measurement of physiological parameters

Blood alcohol concentrations were measured using an Ethanol Assay Kit (BioAssay Systems; catalog # ECET100) according to manufacturers protocol. Total cholesterol levels were determined using the Total Cholesterol Assay Kit (Cell Biolab, Inc., catalog # STA- 384), according to the recommended protocol. The levels of hydroxyproline were determined using the Hydroxyproline Assay Kit (Sigma-Aldrich; catalog # MAK008). The concentration of S-adenosylmethionine within the paternal liver, testis and kidney were measured using the Bridge -It® S-Adenosyl Methionine (SAM) Fluorescence Assay Kit according to the recommended protocol (Mediomics, catalog # 1–1–1003B).

RNA analyses

Total RNA was isolated from E14.5 fetal liver and placenta using the RNeasy Plus Mini Kit (Qiagen. catalog # 74134), according to manufacturer's instructions. Samples were randomized before RNASeq library preparation. Libraries were generated from 10ng of RNA using the TruSeq RNA Sample Preparation kit (Illumina; catalog # RS-122–2001) and pooled for sequencing on an Illumina HiSeq 2500 at Whitehead Genomic Services (Cambridge, MA). Sequencing data were demultiplexed, aligned using STAR with default parameters⁶⁷ and referenced against the Mus musculus genome (UCSC version mm10).

RNA deep sequencing data analysis, selection of candidate mRNAs, and functional enrichment

Following deep sequencing analysis of 50-bp length paired-end reads, Bowtie and Tophat were used to align the reads into transcripts based on the Mouse Reference Genome. To measure the relative abundance of each transcript, the resulting aligned reads were analyzed using the Cufflinks Suite. Expression was quantified as the number of reads mapping to a gene divided by the gene length in kilobases and the total number of mapped reads in millions, and designated as fragments per kilobase of exon per million fragments mapped (FPKM). To select differentially expressed transcriptomes, the volcano plot measuring statistical significance and magnitude of fold-change was generated based on the log2 fold-change (X-axis) and −log10 P value from Cuffdiff analysis within the Cufflinks suite (Y-axis). Differentially expressed mRNAs were selected on the basis of linear P value cut off of at 0.05, which was considered significant and highlighted by colored dots in the volcano plot. Subsequently, functional clusters were identified by applying Ingenuity Pathway Analysis (IPA, Ingenuity System Inc.).⁶⁸

Real-time qRT-PCR analysis of gene expression

Total RNA was isolated from isolated E14.5 fetal liver and placenta using the RNeasy Plus Mini Kit (Qiagen, catalog # 74134), according to manufacturer's instructions. One microgram of purified total RNA was treated with amplification grade DNase I (Sigma, catalog # AMPD1), according to the manufacturer's recommendations, and 250 ng RNA seeded into a reverse transcription reaction using the SuperScriptII system (Invitrogen, catalog # 18064–071) by combining 1 μL random hexamer oligonucleotides (Invitrogen, catalog # 48190011), 1 μL 10 mM dNTP (Invitrogen, catalog # 18427- 013), and 11 μL RNA plus water. This mixture was brought to 70°C for 5 min and then cooled to room temperature. SuperScriptII reaction buffer, DTT, and SuperScriptII were then added according to manufacturer's protocol, and the mixture brought to 25°C for 5 min, 42°C for 50 min, 45°C for 20 min, 50°C for 15 min, and then 70°C for 5 min. Relative levels of candidate gene transcripts were analyzed using the Dynamo Flash mastermix (Thermo Scientific, catalog # F-415XL), according to the recommended protocol. Reactions were performed on a Bio-Rad CFX38. Primers are listed in Table S1 – Primer Sequences.

Sperm collection and analysis

After a minimum of 3 matings, breeder males were killed and sperm collected using a modified swim-up procedure.⁶⁹ Briefly, human tubal fluid (HTF) medium (Life Global Group, catalog # GMHT-100) supplemented with 4 mg/ml of BSA fraction V (Sigma Aldrich, catalog # A5611) and 1 μl/ml of Gentamicin (Invitrogen, catalog # 15750–060) was covered with mineral oil, the dish set in an incubator at 37°C in a 5% CO2, 5%O2, 90% N2 atmosphere for 3 h before males were to be killed. The male reproductive tract was surgically excised and placed into a petri dish containing warmed fertilization medium. Forceps were used to force the sperm through cuts in the epididymis. Concentrated sperm were placed in 5 mL of HTF medium in a polystyrene round bottom tube (BD Falcon, catalog # 353058) at a 45° angle in the incubator for 1 h. The top 1.5 ml were removed and sperm counted using a hemacytometer. For all experiments, sperm samples were judged to be >99% pure, as assessed by microscopy.

Freshly washed (in PBS) sperm were incubated 1:1 with a lysis buffer containing 20 mM TrisCl (pH 8), 20mM EDTA, 200 mM NaCl and 4% SDS, supplemented before use with 100 mM DTT and 250 ug/ml Proteinase K. Incubation was performed for 4 h at 55°C with frequent vortexing. Samples were combined with 200 uL of 100% ethanol and 200 uL of DNeasy lysis buffer (DNeasy Blood and Tissue Kit, Qiagen catalog # 69504), then this mixture added to the DNA isolation columns. The remaining purification steps were performed according to the Qiagen DNeasy Blood & Tissue Kit instructions.

Reduced representation bisulfite sequencing (RRBS)

For each sample, 300 ng of DNA was digested for 2 h with Msp1 enzyme 20U/sample at 37°C followed by 2 h with TaqαI 20U/sample at 65°C. The digested DNA was size selected for DNA fragments larger than 300 bp, which represent CGI enriched fragments and subjected to bisulfite treatment using the Methylamp DNA Bisulfite Conversion Kit (EpiGentek, catalog # P-1001). DNA was verified to be >99% converted before moving forward. Library Preparation was performed by DNA end polishing and adaptor ligation followed by library amplification using indexed primers and library purification. Purified library DNA was then eluted in 12 μl of water. Library was verified on a Bioanalyzer and by KAPA Library Quantification. Sample libraries (10 nM) were subjected to next generation sequencing on an Illumina HiSeq 2500 (EpiGenTek, Farmingdale, New York).

Fastq files from multiple replicates were merged to a single fastq file, analyzed for quality by Fastqc and reads trimmed by TrimGalore (-q 20, adaptor AGATCGGAAGAGC, –length 20, –rrbs). Reads were processed through Babraham Bioinformatic's program, Bismark (www.bioinformatics.babraham.ac.uk/projects/bismark/Bismark_User_Guide.pdf). The reference bisulfite genome was produced via “bismark_genome_preparation,” and “bismark” was run against the converted genome (–non_directional, –bowtie1, -n 1, -q). Resulting BAM files were sorted and indexed by “samtools sort” and “samtools index” programs. Sorted/indexed BAM files were converted to SAM files via “samtools view” program. The rest of the analysis was performed through methylkit in RStudio, with a minimum 30X coverage, minimum difference of 25%, and both P-value and q-value cutoffs of 0.01.

Bisulfite PCR

DNA was isolated using a DNeasy Blood & Tissue Kit (Qiagen, catalog # 69504) and quantified using a NanoDrop 2000c. DNA input (300ng) was bisulfite converted using the EZ DNA Methylation Kit (Zymo Research, catalog # D5001). Primers were designed through the online program “Bisulfite Primer Seeker” available through the Zymo Research website. Gel extraction of PCR product was performed using the Qiagen QIAEX II Gel Extraction Kit (Qiagen, catalog # 20021). PCR products were cloned into the pGEM-T Easy Vector System (Promega, catalog # A1360) and sequenced at the Texas A&M's Plant Genome Technologies core. Analysis was performed using BiQ Analyzer (http://biq-analyzer.bioinf.mpi-inf.mpg.de/). Primers are listed in Table S1 – Primer Sequences.

Statistical analysis

For all experiments, statistical significance was set at α = 0.05. In this study, the effect of 2 independent variables (sex vs. preconception treatment) were assessed using an analysis of variance test (ANOVA), and differences among the means evaluated using Sidak's post-hoc test of contrast. No interactions (P > 0.05) were observed between fetal weight and sire weight, fetal weight and dam weight, nor litter size and fetal weight.

For analysis of gene expression, the replicate cycle threshold (Ct) values for each transcript were compiled and normalized to the geometric mean of 3 validated reference genes. For placental tissues, transcripts encoding succinate dehydrogenase complex, subunit A (Sdha – NM_023281), mitochondrial ribosomal protein L1 (Mrpl1 – NM_053158) and hypoxanthine-phosphoribosyl transferase (Hprt – NM_013556) were measured. For liver, we measured transcripts encoding tyrosine 3-monooxygenase / tryptophan 5-monooxygenase activation protein zeta (Ywhaz – NM_011740), mitochondrial ribosomal protein L1 (Mrpl1 – NM_053158) and hypoxanthine-phosphoribosyl transferase (Hprt – NM_013556). Each of these reference genes were validated for stability across treatment groups as described previously.⁷⁰ Normalized expression levels were calculated using the ΔΔCt method described previously.⁷¹ Relative fold change values from each biologic replicate were transferred into the statistical analysis program GraphPad (GraphPad Software, Inc., La Jolla, CA) where data sets were first verified for normality using the Brown-Forsythe test. For comparisons including sex and preconception treatments, an analysis of variance (ANOVA) was used and Sidak's analysis applied to comparisons with P-values<0.05. For single comparisons, an unpaired student's t-test was applied. In all instances, we have marked statistically significant differences with an asterisk. For genes demonstrating significant changes, multiple primer sets targeting the same transcript were used. Before averaging the results from 3 primer sets, a mixed 2-way ANOVA was applied to ensure no interactions between primer sets. A Sidak post-test was used to identify differences between the preconception EtOH and control treatment groups.

In our mouse model, distinct single nucleotide polymorphisms between the maternal (C57BL/6J) and paternal [C57BL/6(CAST7) – Mus musculus castaneus bred onto a C57BL/6J (B6) background] strains³⁶ allowed us to track allelic patterns of gene transcription for multiple imprinted genes. For RNA sequence-based comparisons of allelic patterns of imprinted gene expression, the proportion of identified single nucleotide polymorphisms were analyzed using either Chi-Squared analysis or, if read counts were less than 5, a Fisher's Exact test.

Funding Statement

This work was supported by the National Institutes of Health, Grant 1R21AA022484.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

This work was supported by the National Institutes of Health, Grant 1R21AA022484. W.M.S. and Y.S.B. were supported through the Texas A&M University College of Veterinary Medicine and Biomedical Sciences Graduate Student Research Trainee Award. We thank Emily Zuber for assistance with the histological analysis of the placenta.