|Home | About | Journals | Submit | Contact Us | Français|
Neuropsychiatric disease frequently presents with an underlying hypo- or hyper- reactivity of the HPA stress axis, suggesting an exceptional vulnerability of this circuitry to external perturbations. Parental lifetime exposures to environmental challenges are associated with increased offspring neuropsychiatric disease risk, and likely contribute to stress dysregulation. While maternal influences have been extensively examined, much less is known regarding the specific role of paternal factors. To investigate the potential mechanisms by which paternal stress may contribute to offspring hypothalamic-pituitary-adrenal (HPA) axis dysregulation, we exposed mice to six weeks of chronic stress prior to breeding. As epidemiological studies support variation in paternal germ cell susceptibility to reprogramming across the lifespan, male stress exposure occurred either throughout puberty or in adulthood. Remarkably, offspring of sires from both paternal stress groups displayed significantly reduced HPA axis stress responsivity. Gene set enrichment analyses in offspring stress regulating brain regions, the paraventricular nucleus (PVN) and the bed nucleus of stria terminalis (BNST), revealed global pattern changes in transcription suggestive of epigenetic reprogramming and consistent with altered offspring stress responsivity, including increased expression of glucocorticoid-responsive genes in the PVN. In examining potential epigenetic mechanisms of germ cell transmission, we found robust changes in sperm miRNA (miR) content, where nine specific miRs were significantly increased in both paternal stress groups. Overall, these results demonstrate that paternal experience across the lifespan can induce germ cell epigenetic reprogramming and impact offspring HPA stress axis regulation, and may therefore offer novel insight into factors influencing neuropsychiatric disease risk.
Stress dysregulation is a common underlying feature in neurodevelopmental and affective disorders, with hyper- and hypo- reactivity of the HPA stress axis reported in schizophrenia, autism, and depression (Arborelius et al., 1999; Moghaddam, 2002; Nestler et al., 2002; Walker et al., 2008; Corbett et al., 2009). Parental exposures to stress, infection, malnutrition, and advanced age have been linked with increased offspring presentation of neuropsychiatric diseases (Kinney et al., 2008; Rosenfield et al., 2010; Brown, 2011; Hultman et al., 2011; Patterson, 2011). The HPA stress axis is remarkably vulnerable to environmental perturbations, and parental experiences likely influence disease risk by altering stress pathway regulation (Matthews, 2002). While maternal insults experienced during pregnancy can directly impact fetal development, the mechanisms by which lifelong maternal or paternal exposures can affect offspring neurodevelopment are less well understood (Bale et al., 2010).
Epigenetic marks in germ cells are a proposed mechanism by which parental environmental exposures shape offspring neurodevelopment (Jirtle and Skinner, 2007; Guerrero-Bosagna and Skinner, 2012). When examining mechanisms of transgenerational transmission, studies of maternal experience must consider the complex maternal:fetal/neonatal interaction, including changes in the intrauterine environment, lactation, and maternal behaviors (Champagne, 2008). However, paternal studies in rodent models, where the sire does not participate in rearing of the offspring, allow for relative exclusion of factors outside of the germ cell contribution, providing a homogenous and easily accessible tissue for examining epigenetic marks (Rando, 2012).
Research from rodent models of male exposure to dietary challenge, drugs of abuse, or social defeat support the transmission of paternal experiences to offspring through epigenetic marks in sperm (Carone et al., 2010; Guerrero-Bosagna et al., 2010; Ng et al., 2010; Dietz et al., 2011; Vassoler et al., 2013). Although mature sperm lack the machinery necessary for chromatin remodeling, spermatozoa RNA populations, histone modifications, and methylation patterns reflect heritable responses to environmental challenges that may occur during spermatogenesis (Godmann et al., 2009; Skinner and Guerrero-Bosagna, 2009). Of specific interest is the sperm microRNA (miR) content, as miRs are a mode of dynamic epigenetic regulation that rapidly respond to environmental perturbations and likely regulate gene expression post-fertilization (Giraldez et al., 2006; Rassoulzadegan et al., 2006; Grandjean et al., 2009; Leung and Sharp, 2010). Epidemiological evidence from Swedish famine years suggests that germ cell susceptibility to reprogramming, and its subsequent effects on future generations’ disease risk, is greatest during the slow growth period of preadolescence (Kaati et al., 2002; Pembrey et al., 2006; Kaati et al., 2007)
Therefore, in order to examine the effect of paternal stress exposure across the lifespan on offspring HPA axis regulation, male mice were exposed to six weeks of chronic stress, prior to breeding, either throughout puberty or only in adulthood. Offspring stress reactivity and gene array analysis of offspring stress regulatory brain regions, the PVN and BNST, were subsequently examined. Paternal sperm miR content was assessed as an indication of potential epigenetic marks associated with the reprogramming of offspring physiology or behavior.
All mice bred for paternal stress studies were virgin, in-house C57BL/6:129 F1 hybrid mice. The hybrid vigor of this background strain provides a reproducible balance of stress responsivity, behavioral performance, and maternal care (Mueller and Bale, 2006; 2007; 2008; Morgan and Bale, 2011). All mice were housed in a 12:12 light:dark cycle with ambient temperature 22°C and relative humidity 42%. Food (Purina Rodent Chow; 28.1% protein, 59.8% carbohydrate, 12.1% fat) and water were provided ad libitum. All studies were performed according to experimental protocols approved by the University of Pennsylvania Institutional Animal Care and Use Committee, and all procedures were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals.
Administration of chronic variable stress was performed as previously described (Mueller and Bale, 2006). Briefly, males were randomly assigned to a control non-stressed group (n=13) or to one of two experimental groups, receiving 42 days of chronic variable stress throughout puberty (PN28-70, n=13) or in adulthood (PN56-98, n=13). Seven different stressors were randomized and administered one per day. The order of seven stressors varied across weeks. The duration of stress (42 d) was selected to encompass a complete round of spermatogenesis, exposing sperm to stress at all stages of maturation (Oakberg, 1956). Stressors, selected because they are non-habituating, do not induce pain, and do not affect food or water intake, included the following: 36 h constant light, 15 min exposure to fox odor (1:5000 2,4,5-trimethylthiazole; Acros Organics, Geel, Belgium), novel object (marbles) overnight, 15 min restraint in a 50 mL conical tube, multiple cage changes, novel 100 dB white noise (Sleep Machine; Brookstone, Merrimack, NH) overnight, and saturated bedding overnight.
Following completion of chronic variable stress, all males were housed with a single female for 14 days to remove the acute effects of stress on male behavior and mature sperm stored in the epididymis. Males were then bred with naïve females for 1–3 nights. To minimize male-female interactions, which may influence maternal investment or care and subsequently affect offspring development (Curley et al., 2011), pregnancy was established by confirmation of a copulation plug (identified each morning within 1 hr after lights on) and signaled immediate removal of the female to her own cage containing a nestlet. Breeding resulted in litters sired from control (n=8), pubertally stressed (n=10), and adult stressed (n=9) males.
Same-sex littermates were split into three groups for physiological and behavioral characterization (max 1 male and female per litter per group). Groups were examined for either HPA stress axis responsivity and prepulse inhibition (PPI), or performance on the tail suspension test (TST) and Barnes maze, or performance in the light-dark (LD) box and postmortem molecular measures. A minimum 10 d recovery period separated tests.
Plasma corticosterone was measured following an acute 15-min restraint in a 50 mL conical tube. Testing occurred 2–5 h after lights on. Tail blood from adult mice (n=6–9 per group) was collected at onset and completion of restraint (0 and 15 min, respectively) and 15 and 75 min after the end of restraint (30 min & 90 min, respectively). Tail blood collection requires <30 sec to complete. Samples were immediately mixed with 50 mM EDTA and centrifuged 10 min at 5000 rpm. Plasma was collected and stored at −80°C until analysis. Corticosterone levels were determined by 125I-corticosterone radioimmunoassay (MP Biomedicals, Orangeburg, NY) following kit instructions. The minimum detection limit of the assay was 7.7 ng/ml, and the intraassay coefficient of variation was 7.1%.
PPI of acoustic startle was recorded in SR-LAB startle chambers (San Diego Instruments, San Diego, CA) as previously described (Geyer and Dulawa, 2003). All test sessions were conducted 2–6 h after lights off. A PPI test session began with 5 min acclimation to background noise (65 dB) followed by five consecutive acute habituation tones (40 ms duration; 120 dB). Ten repetitions of each of the following trial types were then presented in pseudorandomized blocks: startle pulse only (40 ms; 120 dB), no stimulus (65 dB; background), & prepulses +4, +8, or +16 dB above background (20 ms; 69, 73, or 81 dB) preceding the startle pulse (40 ms; 120 dB) by 100 ms. Intertrial intervals averaged 15 sec. Acoustic startle response was defined as the peak startle magnitude recorded during 65 consecutive 1 ms readings following the startle pulse onset. Percent PPI for each prepulse intensity was calculated for each animal (n=5–6 per group) as [1 – (average response to prepulse + startle)/(average response to startle only)] x 100 before group averages were compared. Animals with average response in startle only trials of ≤ 50 mV were excluded from analysis.
The TST was performed as previously described (Steru et al., 1985). Briefly, adult mice (n=7–10 per group) were secured to a rod by adhesive tape placed 1 cm from the tip of the tail and suspended 50 cm above the bench-top. Testing occurred 2–5 h after lights on. Total immobility time, defined as the absence of movement except whisker twitches and respiration, was recorded over the 6 min test session. Testing was digitally recorded and analyzed manually using ANY-maze v4.75 software (Stoelting Co., Wood Dale, IL).
A modified Barnes maze was performed as previously described (Mueller and Bale, 2007). Briefly, adult mice (n=7–10 per group) were trained 2 trials/day for 3 days (4h separation between trials) to find a target escape box. The location of the target escape box under 1 of 24 holes evenly spaced around the perimeter of a 90 cm circular disk remained constant throughout training. Latency to identify the target escape box by nose poke or head deflection was determined for each trial. Each trial ended when the mouse identified and entered the target box or after 4 min of elapsed time. If the target box was not successfully entered within 4 min, the investigator guided the mouse into the target box and a latency of 240 s was assigned. Testing was digitally recorded and analyzed manually using ANY-maze v4.75 software (Stoelting Co., Wood Dale, IL).
The light-dark box test was performed as previously described (Bale et al., 2000; McEuen et al., 2009). Adult mice (n=6–8 per group) were placed in the light compartment at the beginning of the 10-min test session. Light intensities were set at 5 lux in the dark compartment and 300 lux in the light compartment. All testing occurred 2–5 h after lights off. Total time spent in the light compartment and the number of light-to-dark transitions were analyzed with ANY-maze v4.75 software (Stoelting Co., Wood Dale, IL).
Activation of the HPA axis by citalopram, in the absence of any additional stressor, was examined as previously described (Goel et al., 2011). Plasma corticosterone was measured in offspring (n=5–8 per group) as described above, except that animals were injected with 15 mg/kg citalopram (Sigma-Aldrich) or vehicle (0.9% saline) at time 0 min and were not restrained.
Adult male and female offspring of paternal stress or control sires (n=6–8 per group) were rapidly decapitated under isoflurane anesthesia at 22–24 weeks of age. Whole brains were removed, immediately frozen on dry ice, and stored at −80°C. The pituitary gland and left adrenal gland were removed, flash frozen in liquid nitrogen, and stored at −80°C. Approximately 5x10^6 sperm were isolated from the caudal epididymis of adult males into PBS with 1% BSA at 37°C through a swim-up assay. After settling for 30 min, sperm-containing supernatant was centrifuged for 5 min at 4000 rpm. Sperm pellets were stored −80°C.
Whole pituitary and left adrenal glands (n=6–8 per group) were bath-sonicated for 7 min (30 sec on/off cycles) in TRIzol reagent (Invitrogen, Carlsbad, CA) and total RNA was isolated according to manufacturer’s protocol. cDNA was transcribed using the High-Capacity cDNA reverse transcriptase kit (Applied Biosystems, Foster City, CA). Changes in pituitary corticotrophin-releasing factor receptor 1 (CRFr1; NM_007762.4, Mm00432670_m1), pituitary proopiomelanocortin (POMC; NM_008895.3, Mm00435874_m1), adrenal melanocortin 2 receptor (Mc2r; NM_008560.2, Mm00434865_s1), and adrenal 11-beta hydroxysteriod dehydrogenase 1 (11βHSD-1; NM_001044751.1, Mm00476182_m1) were measured by quantitative real-time PCR using TaqMan gene expression assays (Applied Biosystems, Foster City, CA). Samples were run in triplicate for the target genes and endogenous controls (pituitary, Gapdh, NM_0080842.2, Mm99999915_g1; adrenal, Actb, NM_007393.3, Mm00607939_s1) on the same 96-well plate. Analysis was performed by the comparative Ct method, and expression levels were normalized to control male offspring (Schmittgen and Livak, 2008).
Whole brains were cryosectioned at −20°C. Using a hollow needle (Ted Pella Inc., Redding, CA), brain regions were micropunched according to the Paxinos and Franklin atlas (Paxinos and Franklin, 2001) as follows: bed nucleus of stria terminalis (BNST), 0.75mm bilateral punches from one 300 μm slice +0.25 to −0.05 relative to bregma, atlas figs. 25–32; paraventricular nucleus (PVN), 1.00 mm punch along the midline from two successive 300μm slices −0.50 to −0.80 and −0.80 to −1.10 relative to bregma, atlas figs. 36–40. Punches were bath-sonicated for 2.5 min (30 sec on/off cycles) in TRIzol reagent (Invitrogen, Carlsbad, CA). RNA was isolated by RNeasy kit (Qiagen, Valencia, CA) and suspended in RNAse-free water. Total RNA (n=5–6 per group) was sent to the University of Pennsylvania Path BioResource Molecular Profiling Core for Affymetrix GeneChip Mouse Gene 1.0 ST analysis. Gene set enrichment analysis (GSEAv2.0.7, Broad Institute, Cambridge, MA) was used to interpret patterns of gene expression (Mootha et al., 2003; Subramanian et al., 2005). Collections of c2 (curated: Kegg, Biocarta, and Reactome), c3 (motif), and c5 (gene ontology) annotated gene sets were obtained from the Molecular Signature Database (MSigDBv3.0.1, Broad Institute, Cambridge, MA) available for use with GSEA software.
Total RNA (n=4 per group) was extracted from sperm pellets using TRIzol reagent (Invitrogen, Carlsbad, CA). 100 ng of total RNA was reverse transcribed to cDNA using Megaplex RT pool A and B primers and Multiscribe reverse transcriptase, then quantified using the TaqMan Array MicroRNA card A and B set v3.0 (Applied Biosystems, Foster City, CA). Analysis was performed using the comparative Ct method (Schmittgen and Livak, 2008). For each sample, the average of the Ct values of mammu6 and sno202 was used as an endogenous loading control. Expression levels of each sample were normalized to the average expression level of male controls. miRs with expression in <50% of samples or Ct values >35 were excluded, such that analysis of significant differences in expression of specific miRs was conducted with n=2–4 per group. The miRWalk database, which identifies putative miR target sequences in mRNA 3′-UTR commonly predicted by miRWalk, miRDB.org, miRanda, and TargetScan algorithms, was used to determine predicted targets of significantly changed miRs (Dweep et al., 2011).
An investigator blind to animal treatment group conducted all studies and analyses. Total corticosterone was analyzed by multivariate ANOVA (sex x paternal stress for restraint experiment; sex x paternal stress x drug for citalopram experiment) with time as a repeated measure. PPI was analyzed by two-way ANOVA (sex x paternal stress) with prepulse intensity as a repeated measure. Baseline acoustic startle, TST immobility time, and LD box measures were analyzed by two-way ANOVA (sex x paternal stress). Latency to find target on the Barnes Maze was analyzed by two-way ANOVA (sex x paternal stress) with trial as a repeated measure. Adrenal and pituitary gene expression analysis was analyzed by two-way ANOVA (sex x paternal stress). Data greater than two standard deviations away from each group mean were considered outliers and excluded from analysis. Main effects and interactions were further analyzed with Tukey’s HSD post hoc test. Significant differences were identified at p≤0.05. All statistics above were performed using JMP8 (SAS) software, and all data are reported as mean ± standard error of the mean (SEM). Microarray data were modeled as “control” vs “paternal stress” in accordance with behavioral and physiological data. Thresholds for multiple comparisons were set at false discovery rate (FDR)≤0.05. Phenotype permutations (1000) were computed in GSEA to determine FDR, nominal p value, and normalized enrichment score (NES) of each gene set. Gene sets with FDR≤0.25, p≤0.01, and NES≥1.6 met significance thresholds. For miR analysis, heat maps and hierarchical clustering were performed using MuliExperiment Viewer (TM4.org) with Pearson’s correlation, and miRs significantly altered in both pubertal and adult stress groups were identified by one-way ANOVA with post hoc student’s t-test, p<0.05.
To determine if paternal exposure to chronic stress during pubertal or adult windows altered offspring stress responsivity, we examined plasma corticosterone levels following a 15-min restraint stress. There were significant effects of paternal stress (F(2,35)=9.3, p=0.0006) and sex (F (1,35)=22.68, p<0.0001) on corticosterone levels that did not interact (F(2,35)=0.58, p=0.56) and a significant effect of time (F(3,33)=107.4, p<0.0001) (Fig. 1A,B). Similarly, there were significant effects of paternal stress (F(2,35)=5.9, p=0.006) and sex (F(1,35)=18.7, p=0.0001) on total corticosterone that did not interact (F(2,35)=0.2, p=0.86). Total corticosterone was lower in paternal stress conditions compared to controls and in males compared to females (p<0.05) (Fig. 1A,B inset). The following statistical outliers in corticosterone levels at each of the four time points were removed: 1 pubertal male & 1 pubertal female at time 0, 1 pubertal female at time 15, none at time 30, and 1 adult male & 1 adult female at time 120.
To examine the degree to which blunted stress axis responsivity was representative of global stress-sensitivity in offspring of stressed sires, we assessed offspring PPI, stress-coping behavior in the TST, spatial learning in an adverse environment on the Barnes Maze, and anxiety-like behavior in the LD box. PPI did not differ between males and females (F(1,23)=0.54, p=0.47) or between paternal stress and control groups (F(2,23)=0.28, p=0.75), though an expected within-subject effect of prepulse intensity was observed, with higher prepulses eliciting greater prepulse inhibition (F(2,22)=64.1, p<0.0001) (Fig. 2A,B). Max startle to 120dB tone was also unaffected by sex (F(1,23)=0.55, p=0.46) or paternal stress (F(2,23)=0.07, p=0.93) (Fig. 2C). In the TST, there was no effect of sex (F(1,40)=2.2, p=0.12) or paternal stress (F(2,40)=0.01, p=0.89) on total time spent immobile (Fig. 2D). Analysis of latency to find the target on the Barnes Maze similarly revealed no effect of sex (F(1,29)=0.66, p=0.78) or paternal stress (F(2,29)=0.41, p=0.66), and latency to find target decreased across trials for all groups (F(5,25)=44.83, p<0.0001) (Fig. 2E,F). In the LD box, females spent significantly more time in the light than males (F(1,32)=5.9, p=0.02), but there was no effect of paternal stress (F(2,32)=1.5, p=0.21) (Fig. 2G). To determine whether Barnes maze and LD box results were affected by locomotor ability, light-to-dark transitions were measured in the LD box. There was no effect of sex (F(1,32)=3.44, p=0.07) or paternal stress on the number of transitions (F(2,32)=0.20, p=0.81) (Fig. 2H). Statistical outliers were removed only from analysis of max startle and prepulse inhibition (1 adult female, Fig. 2 B,C) and from analysis of total time in light (1 pubertal male and 1 control female, Fig. 2G).
To investigate the potential role of serotonin in mediating changes in offspring stress axis reponsivity, we evaluated offspring corticosterone levels in response to acute SSRI administration. Citalopram administration elicited similar HPA activation in all groups, as there was a main effect of drug (F(1,52)=14.6, p=0.0004) on corticosterone level that did not interact with sex (F(1,52)=0.003, p=0.96) or paternal stress (F(2.52)=0.88, p=0.42) (Fig. 2I,J). The following statistical outliers in corticosterone levels at each of the four time points were removed: at time 0, 1 pubertal vehicle male, 1 citalopram control female, 1 pubertal vehicle female, and 1 adult vehicle female, none at time 15 or time 30, and at time 120, 1 pubertal vehicle female and 1 adult citalopram female.
To determine if changes in peripheral control of the HPA axis explain offspring blunted stress responsivity following paternal stress exposure, we evaluated gene expression in offspring pituitary and adrenal glands. Neither pubertal nor adult paternal stress significantly altered the expression of genes involved in the offspring HPA axis response, including pituitary CRFr1 (F(2,27)=1.30, p=0.29), pituitary POMC (F(2,27)=0.13, p=0.90), adrenal Mc2r (F(2,26)=2.32, p=0.12), or adrenal 11βHSD-1 (F(2,26)=0.55, p=0.58). Consistent with previously reported sex differences in the expression of these genes (Goel and Bale, 2010), significant effects of sex were observed: CRFr1 was higher in males (F(1,27)=21.42, p<0.0001), POMC was higher in females (F(1,27)=14.39, p=0.001), Mc2r was higher in females (F(1,26)=24.28, p<0.0001), and 11βHSD-1 was higher in males (F(1,26)=16.92, p=0.003). There were no interactions of paternal stress and sex for CRFr1 (F(2,27)=0.24, p=0.79), POMC (F(2,27)=0.01, p=0.98), MC2R (F(2,26)=2.0, p=0.16), or 11βHSD-1 (F(2,26)=0.05, p=0.95) (Fig. 3A,B,C,D).
To examine changes in central regulators of the HPA axis following paternal stress, we evaluated patterns of gene expression in the PVN and BNST with a microarray. By multiple comparison analysis, no single gene met both threshold and significance criteria. By GSEA analyses, c3 gene sets in the PVN and c3 and c5 gene sets in the BNST were enriched following stress (Table 1). Gene sets enriched in offspring PVN following paternal stress were sorted according to the degree of enrichment (Table 2). Due to the large number of significant gene sets in the BNST, enriched gene sets were organized according to predicted function or miR family (Table 3). Members of c3 gene sets share known or likely regulatory elements in their promoters or 3′-UTRs, and c3 gene sets are named according to the transcription factor or miR that binds that conserved sequence. c3 gene sets sharing the same name represent distinct binding sites for the given transcription factor or miR. Members of c5 gene sets share a common gene ontology (GO) annotation, and c5 gene sets are named for that GO annotation.
To examine the role of paternal stress exposure in regulation of sperm miR environment, we assayed changes in sperm miR expression following breeding of stressed and control sires. Hierarchical clustering analysis using Pearson’s correlation as a metric successfully segregated pubertal and adult stress samples from control samples, as visualized in a heat map (Fig. 4A). Expression of the following miRs was significantly increased in both pubertal stress and adult stress samples: miR-193* (F(2,9)=7.3, p=0.01), miR-204 (F(2,9)=5.7, p=0.02), miR-29c (F(2,8)=5.3, p=0.03), miR-30a (F(2,9)=17.2, p=0.0008), miR-30c (F(2,8)=16.6, p=0.001), miR-32 (F(2,6)=5.3, p=0.046), miR-375 (F(2,8)=13.5, p=0.003), miR-532-3p (F(2,7)=5.4, p=0.04), and miR-698 (F(2,8)=6.3, p=0.02) (Fig. 4B). Genes predicted to be targets of at least 3 of the 9 significantly increased miRs are shown in Table 4.
Dysregulation of stress neurocircuitry is a common underlying feature in neuropsychiatric disease (Arborelius et al., 1999; Moghaddam, 2002; Nestler et al., 2002; Walker et al., 2008; Corbett et al., 2009). Adverse parental life experiences likely contribute to offspring HPA stress axis dysregulation through germ cell epigenetic modifications (Bale et al., 2010; Franklin et al., 2012). Recent studies have demonstrated that transgenerational metabolic, endocrine, and stress phenotypes can transmit through the male lineage following an initial maternal exposure, findings that emphasize the dynamic reprogramming of male germ cells (Jimenez-Chillaron et al., 2009; Franklin et al., 2010; Guerrero-Bosagna et al., 2010; Dunn and Bale, 2011; Franklin et al., 2011; Morgan and Bale, 2011). Further, identification of offspring physiological and behavioral changes following direct paternal perturbations has highlighted sperm epigenetic marks as a potential mechanism of transmission (Carone et al., 2010; Ng et al., 2010; Dietz et al., 2011; Vassoler et al., 2013). Therefore, in our study male mice received six weeks of stress prior to breeding, and offspring outcomes were evaluated. As epidemiological studies suggest that the greatest effects on offspring and grand-offspring health occur during developmental windows of increased germ cell vulnerability, we exposed males to stress either throughout puberty or only in adulthood (Kaati et al., 2002; Pembrey et al., 2006; Kaati et al., 2007). To identify potential epigenetic marks in sperm associated with the transmission of offspring phenotype, sperm miR content was assessed.
In our HPA stress axis assessment, we found that offspring from both paternal stress groups displayed significantly blunted corticosterone responses to acute restraint. While maternal stress has been shown to elicit sex-specific responses in stress-related outcomes, paternal stress surprisingly affected stress reactivity similarly in both sexes, suggesting distinct modes of maternal vs. paternal transmission (Mueller and Bale, 2007; 2008; Franklin et al., 2010; 2011; Kapoor and Matthews, 2011; Morgan and Bale, 2011). It was also unexpected that HPA stress axis dysregulation occurred irrespective of the timing of paternal stress exposure. Though epidemiological studies of Swedish famine years highlight preadolescence as a period of heightened germ cell susceptibility to malnutrition, our data illustrate that germ cell reprogramming by stress exposure occurs in adulthood (Kaati et al., 2002; 2007). Relative germ cell vulnerability throughout the lifespan may depend on the type of perturbation encountered, with severe caloric restriction and chronic stress eliciting distinct mechanisms.
The specificity of HPA stress axis reprogramming by paternal stress was remarkable, as is highlighted by the lack of significant differences in any of the assessed behavioral measures. These outcomes support divergent functioning and regulation of physiological and behavioral stress reactivity. Stress pathway dysregulation, either increased or decreased reactivity, reflects an organism’s inability to respond appropriately to a changing environment, and thus may characterize a disease state or a predisposition toward one. Alternatively, a reduced physiological stress response may reflect an adaptive response programmed by the paternal lineage as a protective measure, ensuring greater offspring fitness in what is expected to be a more stressful environment. Interestingly, rodents models have previously demonstrated offspring HPA axis dysregulation following maternal or paternal exposures to diverse challenges including alcohol, lead, chronic stress, or social defeat, suggesting that the HPA axis is especially sensitive to environmental manipulations—the proverbial canary in the coal mine—thus warranting greater mechanistic examination (Coe et al., 2003; Weaver et al., 2004; Virgolini et al., 2006; Hellemans et al., 2008; Kapoor and Matthews, 2008; Mueller and Bale, 2008; Dietz et al., 2011; Harris and Seckl, 2011; Morgan and Bale, 2011).
To assess the neural circuitry underlying the blunted HPA in paternal stress offspring, we first utilized a pharmacological approach of an acute administration of the SSRI, citalopram, to stimulate the HPA in the absence of external stress. The serotonin system is a potent modulator of HPA stress axis reactivity, and has been previously reported to be disrupted in offspring following maternal stress (Heisler et al., 2007; Mueller and Bale, 2008; Franklin et al., 2011; Huang et al., 2012). However, we found that citalopram evoked similar corticosterone production and recovery levels in control and paternally stressed offspring, suggesting that the blunted response detected following a restraint stress did not likely involve dysregulation of serotinergic projections to the PVN. Given intact serotonergic regulation following paternal stress, principal components of the HPA axis, the pituitary and adrenal glands, were also examined in paternally stressed offspring for altered gene expression that may explain stress pathway dysregulation. We found no expression differences between groups in the pituitary (CRF receptor-1, the CRF receptor on pituitary corticotropes, and proopiomelanocortin, the precursor to ACTH) or in the adrenal (melanocortin receptor-2, the ACTH receptor in the adrenal cortex, and 11βHSD-1, the rate limiting enzyme in corticosterone production), suggesting that the reprogramming of the offspring HPA axis by paternal stress exposure lies centrally.
By microarray analyses, we examined broad changes in gene expression in key stress regulatory brain regions, the PVN and BNST. We hypothesized that paternal stress would more likely affect transcriptional patterns, rather than single candidate genes, as a complex cascade of programming events separates an initial paternal perturbation from its eventual effects in the adult offspring brain. Gene set enrichment analyses provided a characterization of these patterns by analyzing expression of a priori defined gene sets based on known biological associations, e.g. shared biological pathway, conserved cis- regulatory element, or common gene ontology category (Mootha et al., 2003; Subramanian et al., 2005). The enrichment of gene sets related to HPA stress axis regulation provided intriguing potential mechanisms that may underlie shifts in offspring stress responsivity. Most notably, enriched PVN expression of glucocorticoid receptor (GR) responsive genes could confer tighter control of CRF neurons in the PVN in paternal stressed offspring. Additionally, the majority of enriched gene sets we found in the PVN and BNST were those where member genes shared transcription factor or miR regulation. These broad differences in gene expression suggest potential involvement of broad chromatin changes and epigenetic programming. In particular, enriched BNST expression of gene sets associated with the activity of CREB, histone acetyltransferases such as CREB-binding protein, and 6 members of the miR-154 family, which are transcribed from a maternally imprinted locus, support the hypothesis that epigenetic mechanisms are likely involved in reprogramming of the offspring HPA axis circuitry (Seitz et al., 2003; 2004).
Paternal models of transgenerational epigenetics offer the advantage of examining germ cell transmission in the absence of confounding maternal factors, which can make deciphering the timing and identification of parental epigenetic transmission difficult. Among the limited set of epigenetic marks carried in sperm, miRs are poised to be products of a dynamic paternal environment and to uniquely contribute to post-fertilization gene expression. Therefore, we assessed the sperm miR content from paternal controls and stress groups. Hierarchical clustering of over 250 miR transcripts segregated all stressed sires’ miR expression profiles into a single clade, supporting recent evidence that sperm miRs are amenable to environmental perturbations (Li et al., 2012; Marczylo et al., 2012). Further, the finding that sperm miR content exhibited similar changes in both stress groups was consistent with the effects of paternal stress exposure on offspring stress responsivity, and therefore supports a continuous germ cell vulnerability to epigenetic reprogramming across the lifespan.
In examination of significant changes in specific sperm miRs, paternal stress both throughout puberty and in adulthood significantly increased the expression of nine (miR-193*, miR-204, miR-29c, miR-30a, miR-30c, miR-32, miR-375, miR-532-3p, & miR-698). As miRs are normally present at low basal levels in sperm, the observed variability in control miR expression and the directional increase in miR levels with stress were not surprising. Importantly, miRs have been shown to regulate gene expression post-fertilization and may impact offspring development by acting on oocyte stores of maternal mRNA (Giraldez et al., 2006; Rassoulzadegan et al., 2006; Grandjean et al., 2009). Assessment of the top predicted mRNA targets of the 9 significantly increased miRs revealed DNA methyltansferase 3a (DNMT3a), a critical regulator of de novo DNA methylation important for imprinted genes, and two proteins involved in miR processing, trinucleotide repeat containing 6b (Tnrc6b) and metadherin (Mtdh) (Kaneda et al., 2004; Lazzaretti et al., 2009; Yoo et al., 2011). These genes, targeted by 4 of the 9 upregulated miRs, suggest that paternal stress may affect offspring HPA axis by intervening in the epigenetic regulation of oocyte development. Alternatively, sperm miRs may have altered epigenetic programming earlier, during spermatogenesis (Amanai et al., 2006).
In summary, our findings demonstrate that male exposure to stress, either throughout puberty or in adulthood, reprograms paternal germ cells and results in transmission of an offspring HPA stress axis dysregulation phenotype. Whether such diminished stress reactivity would be detrimental or beneficial to offspring likely depends on the environment into which they were born, as well as genetic background factors. However, the finding that mild stress experience can produce such long-term changes in male germ cells provides an important and novel mechanism contributing to neuropsychiatric disease risk.
This work was supported by NIH Grants MH087597, MH091258, and MH099910. A.B.R was supported by the NIH Systems and Integrative Biology Training Grant T32-GM07517. We thank J. Fluharty for technical support, C. Howerton for statistical consultation, and K. Morrison for editorial assistance.
The authors declare no competing financial interests.