Molecular and Organismal Changes in Offspring of Male Mice Treated with Chemical Stressors

Abstract

Both gene methylation changes and genetic instability have been noted in offspring of male rodents exposed to radiation or chemicals, but few specific gene targets have been established. Previously, we identified the gene for ribosomal RNA, rDNA, as showing methylation change in sperm of mice treated with the preconceptional carcinogen, chromium(III) chloride. rDNA is a critical cell growth regulator. Here, we investigated the effects of paternal treatments on rDNA in offspring tissue. A total of 93 litters and 758 offspring were obtained, permitting rigorous mixed-effects models statistical analysis of the results. We show that the offspring of male mice treated with Cr(III) presented increased methylation in a promoter sequence of the rDNA gene, specifically in lung. Furthermore polymorphic variants of the multi-copy rDNA genes displayed altered frequencies indicative of structural changes, as a function of both tissue type and paternal treatments. Organismal effects also occurred: some groups of offspring of male mice treated with either Cr(III) or its vehicle, acidic saline, compared with those of untreated mice, had altered average body and liver weights and levels of serum glucose and leptin. Males treated directly with Cr(III) or acidic saline presented serum hormone changes consistent with a stress response. These results establish for the first time epigenetic and genetic instability effects in a gene of central physiological importance, in offspring of male mice exposed preconceptionally to chemicals, possibly related to a stress response in these males.

INTRODUCTION

Exposures and experiences of fathers before mating can have a variety of effects on their offspring, ranging from epigenetic and metabolic changes to altered risk for cancer and other abnormalities (Campbell and Perkins, 1988; Anderson et al., 2000; Nomura, 2003). Such male-mediated preconceptional effects have been documented for multiple animal species. Similar phenomena are suspected in humans, based on epidemiology (Kaati et al., 2002, 2007; Pembrey et al., 2006; Franklin and Mansuy, 2009; Matthews and Phillips, 2010). Occupational metals are among the agents implicated in some studies of human fathers and cancer in offspring (Robison et al., 1995). Pursuant to this observation, we tested the metals present in welding fumes in mice (Anderson et al., 1994). Of six metals tested, Cr(III) chloride, administered to male mice two weeks before mating, resulted in a significant increase in neoplasms in their offspring. This finding was confirmed in a second, larger study (Yu et al., 1999).

Since Cr(III) enters cells poorly and is generally non-toxic (Eastmond et al., 2008), we localized this metal in the testes of exposed males. Cr(III) was not detected in sperm or in the germinal epithelium (Bench et al., 2000; Ortega et al., 2001). Absence of Cr(III) within the putative target cells led to the proposal that the metal was having indirect effects, possibly hormonally mediated, and that changes in sperm and offspring were likely to be epigenetic rather than mutational. In confirmation of this, we found, using methylation-sensitive representational difference analysis, that the gene for ribosomal rRNA, rDNA, displayed an alteration in methylation in the sperm of Cr(III)-treated male mice (Cheng et al., 2004; Shiao et al., 2005). Changes in rRNA are of considerable interest, in view of its central role in growth control, metabolic economy, and cancer (Moss, 2004; Montanaro et al., 2008; Drygin et al., 2010).

The next question was whether similar epigenetic changes occurred in the offspring of Cr(III)-treated males. A large study of rDNA in offspring tissues was conducted, involving several batches of treated males, and many litters and individual offspring. In addition to untreated controls, intraperitoneal treatment with acidic (pH 4) saline (AS) was included as an intended second control, since Cr(III) chloride was administered in AS.

The results of this study revealed, first of all, that the multicopy rDNA genes normally undergo both epigenetic and genomic tissue-specific changes during ontogeny (Shiao et al., 2011). Methylation of the gene was higher in sperm than in lung, and higher in lung than in liver for both males and females. Furthermore, there were several alternate sequence forms of the rDNA gene repeats, involving both the main gene promoter and the upstream spacer-promoter, and the relative frequencies of these variants defined by single nucleotide polymorphisms (SNPs) differed in a tissue-specific manner. Analysis of the within-litter variance of these alternate genotypes, including embryos as well as the three adult tissues, strengthened the suggestion that ontogenetic programmatic genomic structure change resulted in altered rDNA SNP and haplotype ratios. Several of the alternate genotypes had a significant negative association with percent gene methylation.

A second observation in that publication (Shiao et al., 2011) was that paternal exposures to either Cr(III) or AS impacted on some of the ontogeny-related outcomes. The endpoints influenced by paternal treatments included the within-litter variances for percent rDNA gene methylation and for percent of each rDNA genotype; tissue-dependent differences in rDNA gene methylation and genotype frequencies; and associations between gene methylation and genotypes.

The present report contains a quantitative display of the effects of the paternal Cr(III) or AS treatments on molecular changes in the rDNA gene in offspring. In addition, we show that in the offspring of males treated with Cr(III) or AS there were significant changes in body and liver weights, liver/body weight ratios, serum glucose, and serum leptin. Both Cr(III) and AS had effects in directly-treated males that were consistent with evocation of a stress response. Thus, there were wide-ranging organismal and molecular consequences in offspring after treatment of males with chemical stressors.

MATERIALS AND METHODS

Animal Treatment

All animals used in this research project were cared for and used humanely according to the following policies: The U.S. Public Health Service Policy on Humane Care and Use of Animals (1996); the Guide for the Care and Use of Laboratory Animals (1996); and the U.S. Government Principles for Utilization and Care of Vertebrate Animals Used in Testing, Research, and Training (1985). All NCI-Frederick animal facilities and the animal program are accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. Male Cr:NIH Swiss mice were obtained from the NCI-Frederick Animal Production Area at 8 weeks of age and were injected intraperitoneally (i.p.) with 1 mmol/kg CrCl₃∙6H₂O(Aldrich Chemical Co., Miwaukee, WI) [Cr(III)] in 0.5 ml 0.9% NaCl adjusted to pH 4.0 with NaOH, or with only acidic saline adjusted with HCl to pH 4.0. All injections were given between 10:00 and 11:00 a.m. These treatments were compared to untreated controls (UT) maintained in parallel. For acquisition of sperm from directly-treated males, two separate series of 20 and 25 males respectively were treated, and two wks later the epididymides were frozen for sperm isolation. For the preconceptional studies, the male mice were mated with untreated females two weeks after treatments. The sperm utilized in fertilization would have been at the sensitive spermatid stage at the time of paternal treatment. The offspring of these matings are referred to here as Cr-lineage, AS-lineage or UT-lineage. In order to minimize batch-related effects, such as seasonal changes, we obtained serum and tissues (lung, liver, and epididymides) from 6-week adult offspring in three different experimental series (batches), carried out over 18 months. A total of 31 Cr-lineage, 32 AS-lineage, and 30 UT-lineage litters, and a total of 758 adult offspring, comprised this part of the study. Entire day-8 embryos, 876 total, were collected from a total of 26 Cr-lineage, 36 AS-lineage, and 36 UT-lineage litters, as two batches over a year. For the direct-treatment stress-response study, groups of 20 males per treatment were humanely euthanized and serum collected at intervals of 1 hr, 6 hrs, 24 hr, 1 wk or 2 wks for analysis of glucose, corticosterone, insulin, and leptin. These parameters were chosen because of the expected effects of Cr(III) on glucose uptake and insulin sensitivity (Anderson, 1997).

DNA Extraction and Processing

As described previously, DNA was extracted from entire day-8 embryos and 6-week lung, liver and sperm heads (Shiao et al., 2005, 2011). As described in detail in these publications, DNA methylation at 5 CpG sites in the spacer-promoter of rDNA was quantified after bisulfite treatment, polymerase chain reaction amplification (PCR), and pyrosequencing. Single-nucleotide polymorphisms (SNPs) in the spacer-promoter and in the main promoter were quantified by PCR amplification followed by pyrosequencing. Gender of the day-8 embryos was determined by real-time PCR analysis of the SRY gene on the Y chromosome and a sequence in the X-inactivation center region on the X chromosome.

Glucose and Hormone Analyses

For clinical chemistry analyses, blood samples were collected in tubes without anticoagulant and allowed to clot at room temperature. The serum was collected after centrifugation and stored at - 70^o C. Glucose was measured by the hexokinase/glucose-6-phosphate dehydrogenase method, using reagents from Beckman Coulter Inc. (Melville, NY) and an Olympus AU400e Analyzer (Beckman Coulter Inc., Irving, TX). Corticosterone and insulin-like growth factor 1 (Igf-1) were determined by radioimmunoassay with an APEX Automatic Gamma Counter, INC Micromedic Systems Inc. (Huntsville, AL). Reagents for the corticosterone assay were purchased from MP Biomedicals Diag. Div. (Solon, OH). Igf-1 assay reagents were from Diagnostic Systems Laboratories Inc. (a Beckman Coulter Company, Webster, TX). Leptin and insulin were assayed with Meso Scale Metabolic Markers kits (Meso Scale Discovery, Inc., Gaithersburg, MD).

Statistical Methods

Treatment differences among measurements made on directly-treated animals were assessed with standard parametric and nonparametric methods including correlation and regression analysis, analysis of variance (anova), t-tests and Wilcoxon tests. For offspring data, linear mixed-effects models were used to determine differences among the Cr(III), AS, and UT treatments and possible effects of covariates, including sex of offspring, litter size, and experimental batch effects on response endpoints (Milikien et al., 1992; Littell et al., 1996; Pinheiro and Bates, 2000). In this approach, the $n_{\text{i}}$-dimensional response vector $y_{\text{i}}$ for the n mice in the ith litter (for M litters) is expressed as

$$\begin{gathered} {\mathit{y}}_{\text{i}}={\mathit{X}}_{\text{i}}\mathit{\beta }+{\mathit{Z}}_{\text{i}}{\mathit{b}}_{\text{i}}+{\mathit{\epsilon }}_{\text{i}},\text{i}=1,\dots ,\text{M}, \\ {\mathit{b}}_{\text{i}}\sim \mathcal{N}(\mathbf{0},\mathbf{\psi }),{\mathbf{\epsilon }}_{\text{i}}\sim \mathcal{N}\left(\mathbf{0},{\sigma }^2\mathit{I}\right), \end{gathered}$$ (Eqn. 1)

where $\mathit{\beta }$ is the p-dimensional vector of fixed effects, ${\mathit{b}}_{\text{i}}$is the q-dimensional vector of random effects, ${\mathit{X}}_{\text{i}}$ (of size $n_{\text{i}}\times p$) and ${\mathit{Z}}_{\text{i}}$ (of size $n_{\text{i}}\times q$) are known fixed-effects and random-effects regressor matrices, and ${\epsilon }_{\text{i}}$ is the $n_{\text{i}}$-dimensional within-litter residual error vector with a spherical normal $(\mathcal{N})$ (Gaussian) distribution. The random effects ${\mathit{b}}_{\text{i}}$ were assumed to be independent for different litters, with variance-covariance matrix $\mathbf{\psi }$, and the within-litter errors ${\epsilon }_{\text{i}}$ were also assumed to be independent for different litters, with identity matrix $\mathit{I}$. This modeling approach recognizes multiple levels of random variation, including: (1) among-litter variation, from the fathers to which the treatment conditions were applied; (2) within-litter variation, from the offspring on which the response measures were observed; and (3) within-mouse replicate variation, and takes into account within-litter dependencies. Inferences regarding treatment differences were conducted using among-litter restricted maximum likelihood (REML) variance estimates. Further analyses included correlation and regression, analysis of variance, t-tests, and tests for variance homogeneity. Probability values less than 0.05 were considered significant. All probability values were two sided.

RESULTS

Preconceptional Cr(III) Treatment of Fathers Caused Tissue-specific Alterations in CpG Methylation in the Upstream Spacer-promoter Regulatory Region of the rDNA Gene in Offspring Lung.

In a previous study aimed at identifying differentially methylated genes in sperm of Cr(III)-treated males, we observed altered methylation in a regulatory region of the gene for ribosomal RNA, rDNA (Cheng et al. 2004; Shiao et al. 2005) (Fig. 1). This regulatory region is located several kb upstream from the transcription start site and is termed the spacer-promoter. There are 27 CpG sites in the spacer-promoter, which show varying degrees of methylation in concert (Shiao et al., 2005). Here, five representative CpG sites have been analyzed.

Fig. 1.

Diagram of the 45S rDNA gene in the mouse strain used in this study, showing SNP sites and the genotypes/haplotypes created by these SNPs.

NTS = non-transcribed upstream spacer-promoter region. ETS = external transcribed sequences. ITS = internal transcribed sequences. Methylation at sites 19–23, representative of original 27 CpG dinucleotides in the spacer promoter (Shiao et al., 2011), was measured in the present study. SNPs at −2214 in the spacer-promoter and at −218, −178, and −104 in the main promoter are shown.

In the lungs of male offspring of Cr(III)-treated males, there was a significant increase in average methylation at the CpG sites in the spacer-promoter regulatory region of rDNA (Fig. 2A). Results for all five CpG sites, individually and together, were of statistical significance (Supplementary Fig. 1). This may represent the first example of a specific gene methylation change in offspring as a result of paternal treatment.

Fig. 2.

Effects of paternal treatments on methylation of rDNA in offspring tissues. A. Methylation at CpG sites in male offspring lung. Compared with UT-lineage, a, P = 0.012, b, P = 0.039, c, P = 0.037, d, P = 0.023, e, P = 0.042, all sites together, p = 0.028, after correction for variables (see Methods). B. Methylation level was higher in lung than in liver, and this difference was greater in Cr-lineage offspring, b, P = 0.0047, c, P = 0.0069 compared with UT-lineage, based on within-animal comparisons (see Methods). C. Methylation level was higher in sperm than in lung, and this difference was greater in Cr-lineage offspring, d, P < 0.0001, and in AS-lineage offspring, e, P = 0.0002, compared with UT-lineage. D. Methylation was higher in sperm than in liver, a difference that was not significantly affected by paternal treatments.

Further effects of paternal treatments emerged when the data for each individual animal were utilized to examine relationships among methylation in the several tissues (Shiao et al., 2011, Data S3). This approach provides additional statistical strength, since between-animal and between-litter variation does enter the analysis. As suggested by Fig. 2A and Supplementary Fig. 1, the within-animal comparisons clearly showed that degree of rDNA methylation was significantly different among the tissues of control animals, with sperm > lung > liver. As expected from the increases in meCpG at all sites in lungs of the Cr-lineage males, with no significant treatment-related changes in liver or sperm (Supplementary Fig. 2C and E), the lung vs liver differential was increased, and the lung vs sperm differentials decreased, for rDNA CpG methylation compared with the UT-lineage (Fig. 2B, C). Representation results for meCpG 19 are shown; findings were similar for the other CpG sites. This confirms increased methylation in male Cr-lineage offspring lung.

The results in Fig. 2A suggest that there might also be an increase in meCpG in AS-lineage male lungs. The increased statistical power of the within-animal comparisons supports this conclusion: the lung vs sperm differential was significantly increased in the AS-lineage males (Fig. 2C). Furthermore in the Cr-lineage females, compared with the UT-lineage, the lung vs liver rDNA methylation differential was increased significantly at all CpG sites (meCpG 19 illustrated in Fig. 2B), indicating that in females, as well as in males, Cr(III) treatment of fathers resulted in increased spacer-promoter methylation of rDNA in lungs. There was no significant effect of paternal treatment on the differential in methylation in sperm vs liver (Fig. 2D).

As a further confirmation of effects of paternal treatments, variances around mean meCpG values were calculated for each tissue and treatment group. Compared with the UT-lineage mice, variances for mean meCpG values were significantly decreased in AS-lineage male lung, increased in AS-lineage female liver, and increased in AS-lineage embryos of both genders (Supplementary Table 1).

Frequencies of rDNA Genotypes Differed as a Result of Paternal Treatments.

Several rDNA genotypes were observed, established by SNPs in promoter regions (see Fig. 1). There is a T/C SNP at site −2214 in the upstream spacer-promoter, and three SNPs in the main promoter at −218, −178, and −104 that are linked in such a way as to result in four haplotypes, designated CGC, CCA, ACC and CCC.

In the control UT-lineage mice, analysis by within-animal comparisons showed that relative percentages of these genotypes shifted in sperm compared with lung and/or liver in males, and in lung compared with liver in females (Shiao et al., 2011). Most notably, the CGC haplotype was more frequent in female lung compared with liver (Fig. 3B) and in male lung and liver compared with sperm (Figs. 3C, D). These observations in untreated mice imply normal, tissue-specific, site-specific instability in the genomic arrays of multiple copies and polymorphic forms of rDNA. The paternal treatments resulted in further significant changes in the frequencies of the genotypes in offspring tissues. For the CGC genotype, this could be demonstrated directly based on corrected tissue averages for lung (Fig. 3A). This genotype was significantly decreased in the lungs of female offspring, of both Cr- and AS-lineage. In the males the difference was most significant for the Cr-lineage. Within-animal comparisons of genotype profiles in lung and liver confirmed the effects of paternal treatments on the CGC genotype in offspring lung and suggested a reciprocal increase in the CCC genotype (Fig. 3B). This analytical approach revealed additional effects of paternal treatments on tissue-specific genotype frequencies (Shiao et al., 2011, Data S3). The lung-sperm differential in percent CGC was reduced after paternal treatments, and significant differences appeared in both relative percent ACC and relative percent CCA (Fig. 3C). There were similar results for the liver/sperm differential in percent CGC and percent ACC after paternal Cr(III) or AS paternal exposures (Fig. 3D). Overall averages for liver and sperm are shown in Supplementary Fig. 2.

Fig. 3.

Effects of paternal treatments on relative percentages of rDNA genotypes. A. Overall averages of percent CGC genotype were lower in lungs of offspring of treated fathers. a, P = 0.0053; b, P =0.0074; c, P = 0.013; d, P = 0.056 compared with the UT-lineage. B. Within-animal comparison of percent genotypes in lung compared with liver confirm the effect of paternal treatments on percent CGC in lung, with reciprocal changes in percent CCC. e, P = 0.0004; f, P = 0.0035; g, P < 0.0001; h, P = 0.028, i, = 0.0079 compared with UT-lineage. C. Within-animal comparison of percent genotypes in lung compared with sperm revealed a tissue differential for effects of paternal treatments on all genotypes. j, P = 0.0058; k, P < 0.0001; l, P = 0.025; m, P = 0.0001; n, P = 0.0045; o, P = 0.0073 compared with UT-lineage. D. Within-animal comparison of percent genotypes in liver compared with sperm showed a tissue differential for effects of paternal treatments on CGC and ACC genotypes. p, P = 0.0080; q, P = 0.015; r, P = 0.0098; s, P = 0.031 compared with UT-lineage.

Paternal Treatments Altered Relationships between Degree of Methylation and rDNA Genotypes.

Detailed numerical presentations of these results are in Shiao et al., 2011, Data S2. Selected representative findings are presented here in graphic form, using meCpG21, which is in the center of the five CpG sites measured. All five CpG sites gave similar results for these correlations (not shown). In all tissues of untreated mice, there were highly significant negative associations between % rDNA gene methylation and % T genotype (female lung and meCpG 21 illustrated in Fig. 4A). In each of the adult tissues, gene methylation was negatively associated with % CGC haplotype (lungs illustrated in Fig. 4B). The association was significantly positive with % CCC haplotype in all tissues except female embryos (not shown). In offspring whose fathers were exposed to Cr(III) or AS, there was a negative association between methylation and % T genotype, becoming more pronounced in Cr-lineage female lung but less strong in the AS-lineage female lung (Fig. 4A). Furthermore new significant associations and effects of paternal treatments were evident for all of the genotypes. Both Cr- and AS-lineage female lungs, and AS-lineage male lungs and livers, displayed a significant negative association between % ACC haplotype and degree of methylation (lungs are illustrated in Fig. 4C), in contrast to a lack of an association in UT-lineage lungs. Similarly, in UT-lineage lungs gene methylation did not correlate significantly with % CCA, but significant positive associations were found in both AS-lineage and Cr-lineage lungs (Fig. 4D). Paternal treatment also affected the relationship between % CGC and degree of rDNA methylation in female embryos (Fig. 4E). For all of these observations (Fig. 4A – 4E) there were similar significant results for the other meCpG sites (not shown).

Fig. 4.

Relationships between rDNA methylation and genotypes and effects of paternal treatments as indicated by analysis of population regression slopes from random coefficient regression models in offspring tissues. A. Significant negative associations between % T genotype and % methylation are illustrated for meCpG 21 and female lung. Line slope was less negative for AS-lineage compared with Cr-lineage offspring (P value on graph). B. There was a strong negative correlation between % methylation and % CGC genotype (P<0.001) that was not affected by paternal treatments. Male and female offspring lung combined are illustrated. C. In UT-lineage lungs % methylation was not significantly associated with % ACC genotype (P = 0.86), but this relationship was significantly negative for both Cr-lineage (P = 0.0048) and AS-lineage (P < 0.0001) offspring lungs, and these slopes were different from the UT-lineage slope (P values on graph). D. In UT-lineage lungs % methylation was not significantly associated with % CCA genotype (P = 0.53), but this relationship was significantly positive for both Cr-lineage (P = 0.028) and AS-lineage (P=0.0002) lungs, and these slopes were different from the UT-lineage slope (P values on graph). E. In UT-lineage female embryos % methylation was not significantly associated with % CGC genotype (P = 0.26), but this relationship was significantly negative for both Cr-lineage (P = 0.0031) and AS-lineage (P = 0.0078) female embryos, and these slopes were different from the UT-lineage slope (P values on graph).

Offspring of Male Mice Treated with Cr(III) or AS Exhibited Significant Organismal Changes.

Paternal treatments had significant effects on metabolism- and growth-related endpoints in the offspring (Fig. 5, Supplementary Table 2). Serum glucose, which is a tightly regulated parameter in mammals, was reduced in Cr-lineage female offspring, and in male offspring of both Cr- and AS-lineages (Fig. 5A). These differences were highly statistically significant. Serum leptin was significantly increased in AS-lineage female offspring (Fig. 5B).

Fig. 5.

Effects of paternal treatments on organismal parameters. A. Reduction in serum glucose in offspring of treated fathers compared with UT-lineage. a, P = 0.0083; b, P = 0.0013; c, P = 0.018. B. Increase in serum leptin in AS-lineage females. d, P = 0.016 compared with Cr-lineage. C. Increased body weights in offspring of treated fathers compared with UT-lineage. e, P = 0.052; f, P = 0.0081; g, P = 0.097; h, P = 0.085. D. Increased liver weights in female Cr-lineage offspring compared with UT-lineage, i, P = 0.026. E. Increased liver/body weight ratios in female Cr-lineage offspring compared with UT-lineage, j, P = 0.039.

Along with these serum changes, there were significant increases in average body weights of the female offspring (Fig. 5C), with most consistent and significant effects in those of the AS-lineage. Similar effects in males fell short of statistical significance. Note that statistical analyses took into account possible confounding by experimental batch, litter membership, and, importantly, litter size, which can affect offspring size. Average liver weights were also increased (Fig. 5D), though only the female offspring of Cr(III)-treated males showed increased liver weights significant at the P < 0.05 level. These female offspring also exhibited a significant increase in liver weight as a percent of body weight (Fig. 5E).

Examination of the relationships among the organismal parameters was also revealing (Supplementary Table 3). Some parameters were significantly related in all groups: body weights with liver weights and with glucose, leptin, and Igf-1; and Igf-1 with liver weights. In addition, there were some paternal treatment-related differences. Male body weight decreased with serum corticosterone in the UT- and Cr-lineage mice, but not in the AS-lineage mice. Igf-1 correlated with serum glucose in all groups except AS-lineage females, and correlated with insulin only in AS-lineage males. Liver weight in males increased with serum glucose in all groups, except for the Cr-lineage mice, whereas liver weight increased with leptin and decreased with corticosterone only in the Cr-lineage males. Glucose decreased with leptin only in the AS-lineage males, and increased with insulin in UT- and AS-lineage males and females, but not in Cr-lineage mice. There were strong positive associations between serum leptin and insulin in the Cr- and AS-lineage mice, but these were absent in the UT-lineage offspring.

These results together show that paternal preconceptional exposure to Cr(III) or acidic saline had wide organismal effects in the offspring, affecting growth, glucose homeostasis, and several serum hormones, with complex consequences for the apparent relationships among these parameters.

Serum Igf-1 in Offspring Correlated with Several Molecular Parameters.

Since there were effects of paternal treatments on both molecular and organismal endpoints in the offspring, we inquired whether there were any potential relationships between these categories of phenomena. Average serum Igf-1 was higher in the Cr-lineage compared with UT-lineage males, 681.0 ± 17.6 vs 609 ± 18.0 ng/ml, but this difference was not of statistical significance (P = 0.12). Nevertheless, serum levels of Igf-1 showed two significant associations between molecular and organismal parameters, as related to paternal Cr(III) exposure. The relationship between % T variant in sperm DNA and serum Igf-1 level differed between Cr-lineage and UT-lineage males. The regression of serum Igf-1 levels on % T variant was negative for Cr-lineage males; however it was positive for UT-lineage males, with the difference in regression slopes being highly significant (Fig. 6A).

Fig. 6.

Effects of paternal treatments on correlations between serum Igf-1 and rDNA genotype and methylation profiles. A. The % T SNP in sperm was positively correlated with serum Igf-1 in the UT-lineage (P = 0.033) and negatively correlated in the Cr-lineage (P = 0.033); these slopes were significantly different (see P value on graph). B. The % CGC SNP in lungs of female offspring was significantly, positively correlated with serum Igf-1 in the AS-lineage (P = 0.012) but negatively (P = 0.15) correlated in the Cr-lineage; these slopes were significantly different (see P value on graph). C. The % meCpG 19 in female offspring lung of AS-lineage was significantly and negatively associated with serum Igf-1 (P = 0.0017), whereas the line slopes for the UT- and Cr-lineage tissues did not differ from zero. Slopes for the AS-lineage and Cr-lineage results were significantly different (see P value on graph). There were no other significant associations between organismal parameters and rDNA parameters (not shown).

In the female offspring, overall Igf-1 averages did not differ among the groups, but in AS-lineage females there was a significant positive relation (i.e., regression slope) between serum Igf-1 level and % CGC in the lungs of the same animal (Fig. 6B), and a corresponding significant negative relation with % gene methylation in the lungs (Fig. 6C). As noted above, there was a consistent negative relationship between % CGC and % gene methylation in lung (Fig. 4B). These observed relationships with serum Igf-1 suggests that the genotype frequencies and/or gene methylation frequencies may be hormonally regulated and influenced by paternal exposures.

Intraperitoneal Injection of Male Mice with Cr(III) or Acidic Saline Elicited Hormonal Responses.

A major question is how the exposures of the fathers, two weeks before mating, to Cr(III) or to an i.p. injection of acidic saline resulted in such a wide range of effects in the offspring. A hormonal mechanism was postulated and investigated by treatments of males with Cr(III), acidic saline or nothing, as had been done in the preconceptional study.

Cr(III) treatment had highly significant, prolonged effects on the four serum parameters (Fig. 7, Supplementary Table 4). These presumably reflect the bioretention of the metal (Sipowicz et al., 1997) and the dynamics of homeostasis in response to Cr(III) stimulation of glucose uptake into cells. The moderately acidic saline solution resulted in significant reductions in average serum insulin and leptin at 1 hr; there were later significant increases in serum insulin at 6 hrs and at 1 wk. There were also small but consistent increases in average serum glucose, which were significant by contrasts performed within ANOVAs (Supplementary Table 4). Acute decreases in serum insulin and leptin, seen for both Cr(III) and acidic saline, followed by increases, are typical of early stress response phases (Wallace et al., 2000; Lin et al., 2004; Jeschke et al., 2005; Kohl and Deutschman, 2006; Lago et al., 2008). Thus these two early changes were in common and of similar magnitude for both treatments.

Fig. 7.

Effects of direct i.p. treatments of male Swiss mice on serum parameters. Cr(III) caused an immediate, large increase in serum glucose by 1 hr (a, P <0.001), followed by significant decreases at 6 and 24 hrs (b, c, P<0.001). Both insulin and leptin were decreased 1 hr after AS (d, i, P <0.01), with a significant decrease in leptin at this time after Cr(III) (P<0.0001). At 6 hrs both insulin and leptin increased relative to controls after Cr(III) (e, P<0.001, j, P < 0.01), and a similar response was significant for insulin after AS (f, P < 0.05). Cr(III) treatment resulted in reduction in serum insulin at 24 hrs (g, P < 0.0001). Results are presented relative to the average of controls at each time point, because of the well-known diurnal and environment-dependent variability in baselines for these parameters. Raw values are given in Supplementary Table 4.

DISCUSSION

A major strength of our study was utilization of large numbers of litters and several experimentation cohorts. This permitted rigorous use of the litter as the unit of exposure, as opposed to the individual mouse, as the primary unit of measurement. It also allowed for statistical control of potential confounders, such as batch effects, litter effects, litter sizes, etc. Most published studies on preconception effects in rodents have based analyses on relatively few individual descendants rather than on litters, leading to interpretational difficulties. Several qualitatively new findings emerge, firmly rooted, from our study.

(1) At the molecular level, the endpoint initially of interest, based on previous findings with sperm, i.e., the degree of methylation in the spacer-promoter regulatory region of the rDNA gene, did in fact include a significant change in at least one offspring tissue: an increase in lungs of those fathered by Cr(III)-treated males. Lung was one of the offspring tissues affected after paternal treatment with Cr(III) in two previous experiments (Anderson et al., 1994; Yu et al., 1999).

In addition to the changes in the Cr-lineage lungs, in the AS-lineage mice there was significantly increased variance in rDNA methylation in embryos and in female livers as well as decreased variance in male lung, indicating an effect of this paternal treatment on methylation regulation. Although global methylation changes have been shown in offspring of radiation-treated rodent fathers (Koturbash et al., 2006; Tamminga et al., 2008), ours may be the first report of methylation changes in a specific general-purpose gene in offspring as a result of preconceptional adult male treatments. While gene methylation is usually re-programmed during germ cell development and embryogenesis (Morgan et al., 2005), there are several examples of transgenerational maintenance of epigenetic methylation status, including paternal transmission, but these are thus far limited to special cases of genes regulated by retrotransposon insertions, A^vy and Axin^FU (Rakyan et al., 2003). Even in the case of maternally-inherited A^vy methylation, the epigenetic inheritance involves re-programming, there being no gene methylation in the blastocysts (Blewitt et al., 2006).

How methylation of the rDNA spacer-promoter CpG sites might affect the gene is still an open question. There was a lower degree of rDNA methylation and a higher level of 47S rRNA expression in liver compared with lung (Shiao et al., 2011). Methylation at CpG sites in the main promoter region of rDNA represses expression in mice (Santoro and Grummt, 2001), rats (Stancheva et al., 1997), and human cells (Majumder et al., 2006), and may be an important variable in cancer (Ghoshal et al., 2004) and aging (Oakes et al., 2003). Main promoter methylation status of rDNA is actively maintained (Schmitz et al., 2009). However, a transcript initiated at the spacer-promoter has a negative regulatory effect on rRNA production (Mayer et al., 2006, 2008; Santoro et al., 2010), and degree of methylation in the spacer-promoter did not relate to association of transcription factors (Nemeth et al., 2008).

The degree of methylation increase in the rDNA in the Cr(III)-lineage male offspring lung was about 11%. Comparison may be made to recent results in humans, whose parents experienced famine at around the time of conception (Heijmans et al., 2008; Tobi et al., 2009). Although the authors discussed maternal effects, fathers obviously could also have been a source of observed changes. In the white blood cells of their offspring as adults, evaluated as pair-wise sets with siblings, there was significantly altered methylation in 7 of 15 genes studied, by degrees ranging from 0.5% to 5.2%. These genes included, interestingly, leptin (LEP).

The increased methylation of rDNA in the offspring lungs in our study was in the opposite direction to the decreased methylation originally reported in the sperm of directly-treated males (Cheng et al., 2004; Shiao et al., 2005) and also found in the current study (though with borderline significance). There was no change in average degree of methylation in the embryos. Thus our results do not suggest direct inheritance of a methylation pattern from the sperm, as was also recently concluded for maternal inheritance of methylation of A^vy (Blewitt et al., 2006). rDNA may be altered in both sperm and certain offspring tissues, but as part of a large differentiation/reprogramming context. There may be many genes involved in this observed outcome, as suggested by Nomura et al. (2004) and Johannes et al. (2009) for other transgenerational systems.

(2) There was an additional, and completely unanticipated, molecular variation discovered: alterations in the frequencies of the several SNP-defined genotypes and haplotypes of the multiple copies of the rDNA gene. This genomic instability occurred in normal mice, since within-individuals comparisons for untreated controls showed consistently reduced % CGC haplotype in sperm, compared with liver and lung (see Shiao et al., 2011, Data S3). Percent CGC was higher in female lung than liver.

The genotype frequencies were also affected by the preconceptional treatments. The most marked change, decrease in % CGC in lungs of offspring from Cr(III)- or AS- treated fathers, was evident and significant based on overall averages, even after correction for litter membership and other variables. Numerous other treatment-related specific genomic alterations in rDNA were revealed by within-animal comparisons. Male-mediated transgenerational effects of radiation have been interpreted as manifestations of nonspecific genomic instability, related to DNA strand breaks and reduction in DNA repair components (Morgan, 2003; Barber et al., 2006; Koturbashi et al., 2006; Kovalchuk and Baulch, 2008). General genomic instability has been reported in human offspring of irradiated fathers (Aghajanyan et al., 2011). Transgenerational genetic instability in mice due to paternal exposure to a genotoxic chemical, ethylnitrosourea, was also postulated to be non-specific (Dubrova et al., 2008). In contrast, our findings with rDNA indicate non-random shifts in genotype. Recombination and gene conversion are well known to occur in rDNA, with many gene copies in each cell and frequent repetitive elements in each gene copy. Mechanisms have been extensively studied, particularly in yeast (Tsang and Carr, 2008), but genotype- or tissue-specific events have received little attention.

With regard to possible functional significance of the rDNA genotypes, there is evidence that mouse rDNA variants may have differential expression (McStay and Grummt, 2008; Tseng et al., 2008; Santoro et al., 2010; Shaio et al., 2011). Our results, which indicate apparently directed, differentiation-related adjustments in frequencies of rDNA genotypes, also suggest that each of these variants has specific functional characteristics. The SNP sites at −218, −178 and −104 are all within known regulatory elements. In human keratinocytes, SNPs in the rDNA promoter were associated with specific binding of regulatory factors (Zhang et al., 2007).

The observed correlations between the degree of methylation in the spacer-promoter and the percentage of different genotypes also suggests functionality, as do the effects of paternal treatments on these relationships. There may be a mechanistic link between rDNA methylation and chromatin-based processes, coordinated by the remodelling complex NoRC (McStay and Grummt, 2008). The genotype could influence methylation, with the intermediacy of chromatin components in a three-dimensional structure. In Drosophila, rDNA contributes to global chromatin regulation (Paredes and Maggert, 2009) and gene expression (Paredes et al., 2011). In mice, pRNA, which is complementary to an rDNA promoter, enhanced binding of DNA methyltransferase (Schmitz et al. 2010). Alternatively, methylation in the spacer promoter might regulate the events that lead to altered genotype frequencies. Methylation suppresses recombination in yeast rDNA (Tsang and Carr, 2008). In plants, transgenerational stress responses include loci-specific hypomethylation that increased recombination frequency, and so increased chances for a well-adapted next generation (Molinier et al., 2006; Boyko et al. 2007; Boyko and Kovalchuk, 2008).

(3) Changes in several endpoints in the offspring indicated that there were organism-wide modulations in response to the paternal treatments. Serum glucose decreases in both Cr-lineage males and females and in AS-lineage males were highly significant in comparison with no treatment controls. Glucose is a tightly-regulated metabolic parameter in mammals, of key importance in obesity, diabetes and associated conditions (Scheen, 2010). Leptin, a central hormone for regulation of growth and metabolism, increased in AS-lineage females. Interestingly, leptin has been proposed as a candidate gene involved in an inheritable thrifty genotype (Stoger, 2008). Serum Igf-1 relationships to molecular endpoints were affected by paternal treatments. Igf-1 is a major growth regulator and has been implicated in cancer risk (Maki, 2010).

The increases in average body weights in the offspring, in both the Cr- and AS-lineages, are of particular interest. Both animal and human studies have indicated changes in body mass or composition as aspects of transgenerational effects. An F2 generation of rats derived from mothers that were protein restricted during pregnancy had significantly higher body weights and serum glucose, insulin and leptin (Pinheiro et al., 2008). Increased body weights of offspring at 10 weeks, as in our study, were observed for stress-prone females subjected to mild non-invasive stress during mid- to late pregnancy (Mueller and Bale, 2006); stress of mothers during early pregnancy resulted in hyperphagia but reduced weight in their adult offspring (Pankevich et al., 2009). A body weight change occurred in the offspring of male rats after cranial-only radiation (Tamminga et al., 2008). Transgenerational effects of stress on body size have been reported in birds (Naguib and Gil, 2005). In men, early onset of paternal smoking correlated with increased body mass index in their sons (Pembrey et al., 2006). The F2 descendants of women experiencing famine during pregnancy exhibited elevated neonatal adiposity (Painter et al., 2008).

Selective increase in liver weights, with an increase in liver/body weight ratio, occurred only in female offspring. Liver has been observed to be involved in transgenerational responses in rodents, showing, for example, altered proliferation rates, chromosomal aberrations, and responses to radiation, in both F1 and F2 descendants of irradiated male rats (Slovinska et al., 2004).

Several of the relationships among serum parameters were significantly affected by paternal treatment, including for example an association between liver weight and serum leptin in the Cr-lineage male mice only, and between serum glucose and leptin only in the AS-lineage male mice. These results suggest some shifts in central regulatory pathways for metabolism, as part of the preconceptional effect.

Furthermore the demonstrated organismal and molecular phenomena may be connected in some way. Serum Igf-1 showed significant relationships to some of the measured molecular endpoints, suggestive of coordinating events. rRNA expression and cellular metabolism are intimately integrated through a complexity of mechanisms (Murayama et al., 2008; Grummt and Ladurner, 2008; Salminen and Kaarniranta, 2009; Zhou et al., 2009; Grummt and Voit, 2010), and at least in yeast recombination in rDNA is also linked to metabolism via the SIR2 protein (Tsang and Carr, 2008; Lin et al., 2000). Rat paternal dietary stress led to altered gene expression and metabolic dysfunction in their female offspring (Ng et al., 2010).

(4) There were changes in male mice directly-treated with Cr(III) or AS that were indicative of a stress response. This effect was strong, as expected, in the Cr(III)-treated males, where the metal caused a precipitous drop in serum glucose and an increase in corticosterone. In addition, the relatively mild insult to male mice of intraperitoneal injection of a small volume of a moderately acidic saline solution resulted in changes in serum insulin and leptin similar to those seen with Cr(III) and consistent with a stress response. We hypothesize that a hormonal stress response in the fathers underlies the preconceptional effect of these chemical treatments. Developing male rat germ cells, including spermatids, expressed a wide selection of stress response genes (Aguilar-Mahecha et al., 2001), and the expression of these was changed by treatment with cyclophosphamide (Hales et al., 2005). Furthermore, numerous hormone and cytokine receptors have been identified on sperm (Naz and Sellamuthu, 2006). Leptin and insulin can act directly on testicular cells, as well as via the neuroendocrine axis (Caprio et al., 2001). It is even possible that sperm might in theory carry hormonal signals into the zygote, since mature human sperm express both leptin and insulin (Ando and Aquila, 2006).

Once controversial, involvement of hormones in heritable epigenetic variations is becoming increasingly likely (Jablonka and Raz, 2009). In a classical experiment, exposure of male rats to high temperature, a noninvasive stressor, resulted in proliferation of testicular interstitial cells in both those directly exposed and their offspring (Steinach and Kammerer, 1920). An epidemiological study linked paternal preconceptional heat exposure (saunas, electric blankets) with increased risk of brain tumors in their children (Bunin et al., 2006). Also of possible relevance is our finding that limited food deprivation of male mice, another non-invasive stressor, led to significant decrease in serum glucose in their offspring (Anderson et al., 2006), an outcome also in our current study. Most recently, paternal cranial irradiation of male rats, with the rest of the body shielded, resulted in DNA strand breaks and loss of methylation damage in the testes of the treated males, and in their offspring reduced DNA methylation in bone marrow, spleen, thymus and liver, along with reduced DNA methyl transferase and methyl binding proteins (Tammina et al., 2008). Hormonal alterations are a possible explanation for the latter findings.

The putative information transfer from fathers to offspring could be accomplished through any of several mechanisms suggested for establishment and transfer of paternal signal(s) via the sperm, involving gene methylation, paramutation, microRNA (Cuzin et al., 2008; Filkowski et al., 2009), remodelling of chromatin and nucleus structure (de Boer et al., 2010), histones (Ruden and Lu, 2008), and exogenous DNA or RNA (Sciamanna et al., 2009). Both DNA methylation and chromatin remodeling were directly implicated in a study of epigenetic reprogramming in mice (Chong et al., 2007). Paternal hormones could modulate any of these. In addition, at present, we cannot rule out direct effects of the acidic saline or Cr(III) treatments on the testes or sperm, or even behavioral or pheromonal clues to the mothers (Curley et al., 2011).

Finally, the broader significance of the array of preconceptional effects, revealed in our study, should be considered in the context of evolution and of human disease. These effects could be manifestations of a capacity for adaptive developmental plasticity, wherein experiences of a parent results in offspring that have physiologically pre-adjusted for current conditions (Gluckman et al., 2007, 2008, 2009; Nadeau, 2009; Ho and Burrgren, 2010). This may be viewed as anticipatory Beneficial Acclimation (Huey et al., 1999). This phenomenon, while beneficial over the course of evolution, could have negative consequences in some contexts and is suggested to contribute to current major public health problems such as cardiovascular disease and diabetes (Jirtle and Skinner, 2007; Hanson et al., 2011). It could be interesting to explore the impact of paternal stress on beneficial or deleterious physiological parameters of descendants.

Supplementary Material

1

Supplementary Fig. 1. Methylation at 5 CpG sites in the spacer-promoter of rDNA in tissues of offspring and in sperm from directly-treated males.

Supplementary Fig. 2. Relative levels of rDNA genotypes in offspring liver and sperm. a, P < 0.05 vs UT-lineage.