Abstract
Paternal stress can induce long-lasting changes in germ cells potentially propagating heritable changes across generations. To date, no studies have investigated differences in transmission patterns between stress-resilient and stress-susceptible mice. We tested the hypothesis that transcriptional alterations in sperm during chronic social defeat stress (CSDS) transmit increased susceptibility to stress phenotypes to the next generation. We demonstrate differences in offspring from stressed fathers that depend on paternal category (resilient vs susceptible) and offspring sex. Importantly, artificial insemination (AI) reveals that sperm mediates some of the behavioral phenotypes seen in offspring. Using RNA-sequencing (RNA-seq), we report substantial and distinct changes in the transcriptomic profiles of sperm following CSDS in susceptible versus resilient fathers, with alterations in long noncoding RNAs (lncRNAs) predominating especially in susceptibility. Correlation analysis revealed that these alterations were accompanied by a loss of regulation of protein-coding genes by lncRNAs in sperm of susceptible males. We also identify several co-expression gene modules that are enriched in differentially expressed genes (DEGs) in sperm from either resilient or susceptible fathers. Taken together, these studies advance our understanding of intergenerational epigenetic transmission of behavioral experience.
SIGNIFICANCE STATEMENT This manuscript contributes to the complex factors that influence the paternal transmission of stress phenotypes. By leveraging the segregation of males exposed to chronic social defeat stress (CSDS) into either resilient or susceptible categories we were able to identify the phenotypic differences in the paternal transmission of stress phenotypes across generations between the two lineages. Importantly, this work also alludes to the significance of both long noncoding RNAs (lncRNAs) and protein coding genes (PCGs) mediating the paternal transmission of stress. The knowledge gained from these data are of particular interest in understanding the risk for the development of psychiatric disorders such as anxiety and depression.
Introduction
Exposure to adverse experiences has been shown to induce long-lasting, heritable alterations in the epigenome of germ cells, potentially relaying information about the paternal environment to offspring across multiple generations (Franklin et al., 2010; Dietz et al., 2011; Bale, 2015; de Castro Barbosa et al., 2015; Cunningham et al., 2019; Manners et al., 2019; Gapp et al., 2020). Paternal exposure to a variety of stressors has been shown to influence offspring behavioral and metabolic phenotypes (Rodgers et al., 2013; Dickson et al., 2018; Manners et al., 2019). Altered offspring behavioral phenotypes are accompanied by changes in the transcriptional landscape of sperm and, importantly, experimental manipulation of small noncoding RNAs in zygotes of control offspring recapitulates altered behavioral phenotypes seen in offspring from stress lineages (Franklin et al., 2010; Gapp et al., 2014, 2020; Rodgers et al., 2015). Although long noncoding RNAs (lncRNAs) have been shown to be involved in cellular and molecular stages of development such as imprinting (Sleutels et al., 2002) and X chromosome inactivation (Brockdorff et al., 1992), only one study to date has investigated the functional role of lncRNAs in transgenerational stress transmission (Gapp et al., 2020).
Our laboratory has reported previously that paternal exposure to chronic social defeat stress (CSDS) increases baseline anxiety-like and depression-like behaviors in male and female offspring and enhanced stress sensitivity in male offspring (Dietz et al., 2011). Although there has been extensive research establishing the distinct brain circuitry (Wook Koo et al., 2016) and transcriptome (Bagot et al., 2016, 2017; Krishnan et al., 2007; Lorsch et al., 2018; Nasca et al., 2019) of resilient versus susceptible animals, how CSDS exposure differentially affects the transmission of stress phenotypes to resilient and susceptible lineages is not well understood. Here, we demonstrate that key aspects of anxiety-like behaviors can be transmitted to the offspring of socially defeated fathers, dependent in part on whether the father is resilient or susceptible. Moreover, using artificial insemination (AI), we show that sperm partly mediates the transmission patterns seen in both male and female offspring.
To further understand the molecular mechanisms underlying the intergenerational transmission of these stress phenotypes, we used RNA-sequencing (RNA-seq) to examine the transcriptomic profiles of sperm genome-wide before and after exposure to CSDS. Before exposure to CSDS, we found subtle differences in the sperm of resilient versus susceptible males. By contrast, following CSDS there are robust differences in the sperm transcriptome, with limited overlap in the biological pathways disrupted by stress exposure in the two subgroups of animals. Interestingly, after CSDS in susceptible sperm, there was a dramatic increase in the proportion of differentially expressed lncRNAs. Co-expression gene network analysis predicted lncRNAs and certain protein-coding hub regulators of these stress-specific effects on gene expression patterns in sperm.
Overall, we present novel insight into how the paternal transmission of stress phenotypes occurs in part directly via sperm. These data suggest that persistent changes in the make-up of RNAs in sperm that are transmitted to the offspring before versus after stress in stress-susceptible fathers could relay different information to offspring. Our data also suggest that lncRNAs, heretofore mostly unexplored in this context, may be an important epigenetic mechanism contributing to the intergenerational transmission of stress phenotypes.
Materials and Methods
Animals
All experimental C57BL/6J male and female mice were obtained from The Jackson Laboratory. Retired CD1 breeders used as the aggressors for the social defeat paradigm were obtained from Charles River. All mice were maintained in a temperature-controlled and humidity-controlled facility on a 12/12 h light/dark cycle (lights on at 7 A.M.) with food and water ad libitum. All procedures were conducted in accordance with the Institutional Animal Care and Use Committee guidelines of the Icahn School of Medicine at Mount Sinai.
CSDS and social interaction (SI) testing
CSDS was performed as published previously (Berton et al., 2006; Krishnan et al., 2007; Lorsch et al., 2018; Peña et al., 2019a). Briefly, F0 male C57BL/6J experimental mice were exposed to a novel aggressive CD1 male mouse for 10 min/d for 10 d. After each exposure the mice were separated for the remainder of the day by a Plexiglas barrier that allows for sensory contact without physical interaction. Non-defeated controls were run in parallel housed two animals per cage under the same conditions as their experimental counterparts but in the presence of another control mouse rather than an aggressive CD1 mouse. Twenty-four hours after the final tenth day of defeat (experimental day 11), initial SI testing was performed under red light as previously described (Bagot et al., 2016, 2017) to assess social avoidance. F0 mice were placed into an open arena (44 × 44 cm) with an empty wire cage at one side (interaction zone). Mice were given 2.5 min to explore the arena and then removed. A novel CD1 aggressor was placed in the small wire cage and the procedure was repeated for an additional 2.5 min. Time in the interaction zone was recorded automatically with video tracking software (Ethovision 10.0, Noldus). Data were analyzed as time spent in the interaction zone when the aggressor was absent (no target) compared with time spent in the interaction zone when the aggressor was present (target) to obtain a SI ratio. The segregation of defeated mice into susceptible and resilient subpopulations was performed as previously described, work which has demonstrated SI testing as a highly reliable and consistent way to differentiate susceptible versus resilient mice after CSDS (Krishnan et al., 2007; Wilkinson et al., 2009; Vialou et al., 2010). Defeated mice with SI ratio scores >1.1 were categorized as resilient. Defeated mice with SI ratio scores <0.9 were categorized as susceptible. Defeated and non-defeated males were singly housed for 28 d to allow sperm to mature and put through a second SI test (experimental day 39) to confirm their categories, as singly housing males can be stressful (Ieraci et al., 2016).
Natural mating (NM) and Artificial Insemination (AI)
On experimental day 40 (30 d after the last exposure to CSDS), F0 males were allowed to naturally mate with a naive female to produce a NM litter. Approximately 4–7 d later (experimental days 44–50), sperm from the same F0s males was used for AI of control female. We performed AI as outlined by (Stone et al., 2015). Ovulation was induced in naive females by injection with pregnant mare's serum gonadotropin (1 IU) intraperitoneal in the evening. Two days later, females were injected with human chorionic gonadotropin (1 IU) intraperitoneal and 2 d later sperm transfer was performed. Briefly, males were euthanized by cervical dislocation 1 h before the planned sperm transfer. The vas deferens and cauda epididymis were dissected and placed into 500 µl of human tubal fluid (HTF) medium (with 4 mg/ml bovine serum albumin) at 37°C in an in vitro fertilization (IVF) dish. The cauda was dissected to free the sperm. The sperm cells were allowed to swim for a few minutes, all tissue was removed, and the sperm was incubated in a CO2 incubator for ∼30 min to capacitate (the process that enables sperm to fertilize an egg). A total volume of 40 µl of sperm was delivered to the uterine horn of two females to increase the likelihood of litter production using the non-surgical embryo transfer procedure. Females were immediately paired with a vasectomized (VASEX) male overnight and copulation was confirmed by the presence of a copulation plug. After 2–3 d, the female and VASEX male were separated. VASEX males were rotated across all paternal categories to confirm that behavior of the VASEX male had no effect on maternal behavior. On embryonic day (E)12, pregnancy check was confirmed via weight gain using an 8- to 12-g increase as an indication of litter expectancy.
Offspring stress
One male and one female from each litter was exposed to stress and examined on a battery of behavioral paradigms to examine the effects of paternal exposure to CSDS on offspring stress sensitivity. The remainder of the litter was used to examine offspring behaviors at baseline (i.e., unstressed).
Subchronic variable stress (SCVS) was used to assess susceptibility to stress in male and female offspring. SCVS was performed as described previously (Bodnoff et al., 1989; Hodes et al., 2015), which consists of three different stressors with one stress given every day over 6 d. The stressors are alternated during the 6-d period to prevent stress habituation. Stressors were administered in the following order: 100 random mild foot shocks at 0.45 mA for 1 h, a tail suspension stress for 1 h and restraint stress, placed inside a 50-ml falcon tube, for 1 h within the home cage. The three stressors were given two times in the same. On day 7, F1 offspring were started on the behavioral paradigm outlined.
Modified CSDS (mCSDS) was used to assess susceptibility to a more robust stress in male offspring produced by NM and AI. Male offspring previously examined in behavioral paradigms at baseline were exposed to mCSDS. mCSDS was conducted in the same manner as standard CSDS, only performed daily for 8 d rather than the standard 10 d. Specific methods on standard CSDS can be found above.
Offspring behavioral testing
Male and female F1 offspring were examined at postnatal day (P)60 in a battery of behavioral tests in the following order: novelty-suppressed feeding (NSF), open field (OF), and SI with animals put through one behavioral test per day. Offspring were examined at baseline and following exposure to SCVS. Offspring examined at baseline were run in parallel with stressed animals. All behavioral testing was performed 1 d apart and conducted between 10 A.M. and 6 P.M.
NSF was modified as described (Peña al., 2019b). Briefly, NSF was performed after 24 h of food deprivation. On the day of testing, mice habituated to the testing room for at least 1 h. Mice were placed in a novel arena with corncob bedding and a single piece of food in the center of the arena. Time to feed was recorded manually under red light. Mice were given a maximum of 10 min to eat, after which the trial was ended, and latency of 600 s was recorded. After the mouse ate in the novel arena, the mouse was returned to the home cage where a single piece of food was located in the center, and time to eat in the home cage was recorded. Data were analyzed as latency to eat in the novel arena and latency to eat in the home cage.
Exploration of an OF arena (44 × 44 cm) was assessed during a 5-min test. A video tracking software (Ethovision 10.0, Noldus) was used to measured locomotor activity, as well as the time spent in the middle and periphery of the test arena as an index of anxiety.
SI test was performed in a similar fashion as described above on all offspring. When SI was measured in male and female animals, a same-sex C57Bl6/J mouse was used as a measure of social novelty/social avoidance. In the subset of male offspring that were exposed to mCSDS, a novel CD1 aggressor was used to measure SI after defeat exposure.
Statistics for behavior
All statistics were conducted in SPSS (version 25) or Prism GraphPad (version 8.3.1.). Outliers were removed using a Grubbs outlier test. When performing ANOVAs, if the residuals were not normally distributed, the data were transformed until a normal distribution could be achieved; where data were normally distributed (Kolmogorov–Smirnov test for normality of residuals >0.10) one-way or two-way ANOVA was used. If transformation was not successful in generating a normal distribution, a Kruskall–Wallis non-parametric test was used. If an interaction was identified, a Bonferroni post hoc analysis was conducted to determine specific differences between groups. If an effect was identified via a Kruskall–Wallis test, follow-up Dunn's multiple comparisons test were run to determine specific differences between groups. χ2 tests were used to determine differences in the distribution of RNA-seq gene biotypes. Statistical significance was set to a p ≤ 0.05 and statistically trending was set to a p value between 0.05 and 0.10.
Sperm collection and CSDS for RNA-seq
Mature sperm were collected from surgically dissected left cauda epididymis from 8- to 10-week-old test mice on experimental day 1. Under ketamine (100 mg/kg, i.p.) and xylazine (10 mg/kg, i.p.) anesthesia a 1 cm vertical incision was made just lateral to the spine, and following blunt dissection, the left testis was identified. A suture was placed just below the vas deferens, and the cauda epididymis was removed with scissors and placed in HTF medium in a cell culture plate, punctured with a needle, and the sperm cells were allowed to swim for 10–15 min. Sperm suspension and tissue were transferred to a 1.5-ml microcentrifuge tube and incubated for 30 min in a 37
RNA extraction and RNA-seq library preparation
Total RNA was isolated from frozen mature sperm from adult male mice using a protocol combining TRIzol/chloroform extraction with RNeasy MiniElute Cleanup kit. In short, 200-µl TRIzol was used to homogenize the tissue. An additional 500-µl TRIzol was added along with 140-µl chloroform. The lysate was centrifuged at 12,000 × g for 15 min and the aqueous layer was removed and added to the RNeasy mini spin column. The rest of the protocol was completed as outlined by the manufacturer. Samples were treated with DNase to rid samples of genomic DNA. After extraction, purity and concentration were measured on an Agilent 2100 Bioanalyzer (Agilent) using an RNA 6000 Nano assay. To account for batch effects libraries were made in three batches with all batches containing equal samples from each group. Libraries were prepared using the TruSeq Stranded Total RNA Library Prep Gold protocol (Illumina). Briefly, cDNA was synthesized from ribo-depleted RNA, followed by end-repair and ligation with sequencing adaptors. The libraries were then size selected and purified using AMPure XP beads (Beckman Coulter). Bar code bases (8 bp) were introduced at one end of the adaptors during PCR amplification steps. Library size and concentration were measured by Bioanalyzer or Tape Station (Life Technologies) before sequencing. All libraries were pooled for multiplexing and sequenced on the Illumina HiSeq 2500 System using V3 chemistry with 150 base pair double-end reads at the sequenced on a HighSeq2500 System using V4 chemistry with 50 base pair single-end reads at GeneWiz LLC. QC revealed an average of 40 million reads per sample (minimum = 20 million; maximum = 50 million) with an average mapping rate of 71.3%. The number of independent tissue samples included in the final analysis was between six and eight per group.
RNA-seq differential expression analysis
Data were aligned to mm10 genome and filtered for low abundance transcripts by keeping only genes expressed at CPM ≥5 in at least 85% of samples before further analysis. After filtering, pair-wise differential expression comparisons using DEseq2 were performed to identify differentially expressed genes (DEGs) in comparisons of preresilient and postresilient versus control and susceptible versus control and a nominal significance threshold of fold change (FC) > 2 and p < 0.025 was applied. Functional annotation for gene ontology (GO) of biological processes and molecular functions was performed using EnrichR with gene identities of DEGs (Chen et al., 2013) with an adjusted p ≤ 0.05 and more than or equal to five genes reporting only non-redundant categories. A paired analysis was conducted to compare predefeat and postdefeat susceptible mice. To account for RNA-seq samples obtained from the same mice at predefeat and postdefeat time points, mouse ID was included as a covariate to the DESeq2 model.
Upstream regulator and pathway analysis
Predicted upstream regulators and molecular pathways were identified using Ingenuity Pathway Analysis Software (QIAGEN). These determinations were based on the log FC of DEGs in resilient or susceptible animals compared with controls.
Rank-rank hypergeometric overlap (RRHO) RNA-seq analysis
Threshold-free reads per kilobase million (RPKM) lists were ranked by their -log10(RPKM) values. RRHO was then applied to these gene expression values in the sperm of control animals predefeat versus postdefeat to examine the effect of time on the dynamic transcriptome of sperm.
Full threshold-free RNA expression lists were ranked by the −log10(p value) multiplied by the sign of the FC from the DESeq2 analysis and filtered to all genes with CPM ≥ 5 in at least 85% of samples. RRHO was used to evaluate the overall patterns of gene expression in resilient and susceptible sperm before and after defeat. To evaluate the overlap of gene expression between resilient and susceptible sperm before and after defeat, threshold-free genome-wide transcriptomic overlap analysis was conducted using RRHO (RRHO2; Seney et al., 2018). Briefly, the RRHO2 ranks gene lists by a signed p value, that is, the –log10(p value) multiplied by the sign of the FC, and generates matrices of overlapped genes between RNA-seq lists of interest independent of whether individual RNAs are differentially expressed or not.
lncRNA correlation analysis
Data filtered for low abundance transcripts (only genes expressed at CPM ≥ 5 in at least 85% of samples) were used to conduct correlation analysis between lncRNAs and protein coding genes (PCGs). Only samples from either susceptible or control groups and from the post defeat time point were used to calculate the RPKM values. The cor function was used to compute the r (significant: r
Weighted gene co-expression analysis
All animals were analyzed together through weighted gene co-expression network analysis (WGCNA) to examine the network structure of the expressed transcriptome following stress.
Using the Cytoscape software platform, we conducted a comprehensive analysis of the relationship between nodes using the maximal clique centrality (MCC) function, which has reported to be the most effective method of identifying hub nodes in a co-expression network (Chin et al., 2014).
Enrichment analysis
Fisher's exact tests (FETs) were conducted using the super exact test package in R as described previously to determine module preservation as well as enrichment of patterns and other gene lists (Wang et al., 2015)
Results
Sperm mediates a subset of behavioral phenotypes in offspring from stressed fathers
Our previous study found that offspring of CSDS-exposed fathers generated by IVF failed to show most of the phenotypes observed with NM (Dietz et al., 2011). However, it is known that dramatic epigenetic reprogramming can occur in the context of IVF, perhaps confounding the interpretation of this experiment. Therefore, to more accurately test whether sperm per se can be the active vehicle for intergenerational transmission of stress phenotypes, we used AI to investigate how sperm directly contributes to altered baseline and stress-induced phenotypes seen in litters generated from susceptible, resilient, and control fathers. Founder generation (F0) males were subjected to 10 d of CSDS followed by SI testing on day 11 to categorize them as either resilient (R) or susceptible (S; Krishnan et al., 2007; Bagot et al., 2015). Mice were singly housed for 28 d after SI testing, and SI phenotype (R or S) was again confirmed on day 39 with a matched group of controls run in parallel (Fig. 1B). These results show the consistency of the susceptible and resilient phenotypes over time after CSDS as has been demonstrated previously (Krishnan et al., 2007; Wilkinson et al., 2009). Mice were then bred with naive females the next day (at ∼13.5 weeks old; experimental day 40), and 4 d later (experimental days 44–50) sperm from the same mice was used to impregnate hormonally primed females of the same strain (Fig. 1A). Anxiety-like phenotypes were examined in offspring produced by both NM (n = 8−11 offspring/group; 7–11 litters/group) and AI (n = 8−11 offspring/group; 6–10 litters/group) using NSF (Reeb et al., 1992), SI testing, and OF (Christmas and Maxwell, 1970; Prut and Belzung, 2003; n = 7–12 offspring/group) at baseline and after subthreshold chronic variable stress (SCVS; Fig. 1C; NM: n = 7−12 offspring/group; AI: n = 7−10 offspring/group for each sex).
Paternal exposure to CSDS alters stress sensitivity of male offspring in adulthood. A, Schematic timeline of paternal stress, mating, and offspring birth. B, Within animal comparison of F0 males used in NM and AI before (day 11) and after (day 39) CSDS in SI testing. C, Schematic of male offspring stress and behavior paradigms. D, NSF ratio was affected by offspring stress in F1 sons produced by NM (F(1,54) = 16.32, p = 0.0002) or AI (F(1,48) = 4.072, p = 0.0492) but not paternal stress (p > 0.05). Bonferroni's post hoc test revealed sons from susceptible lineages generated by NM following exposure to SVCS showed an increased NSF ratio when compared with non-stressed littermates (t(1,54) = 4.004, p = 0.0006). E, Offspring stress affected the percent time spent in the middle of an OF arena (F(1,55) = 13.23, p = 0.0006) in male offspring generated by NM. Bonferroni's post hoc test revealed F1 sons from control (t(1,55) = 2.597, p = 0.0361) and resilient (t(1,55) = 2.700, p = 0.0276) lineages produced by NM spent less time in the middle of the arena when compared with unstress littermates from the same paternal lineage. Neither paternal nor offspring stress had an effect on behaviors in OF test in AI sons. F, Schematic of male offspring CSDS. G, In sons produced by NM, there was a main effect of CSDS (F(1,24) = 19.75, p = 0.0002) in SI testing. Control (t(1,24) = 3.150, p = 0.013) and resilient (t(1,24) = 2.822, p = 0.0283) sons had a higher SI ratio after CSDS exposure, whereas susceptible sons did not show this effect. In sons produced by AI, offspring stress and paternal category affected SI ratio (interaction: F(1,19) = 7.328, p = 0.014). Similar to results with NM, AI resilient sons had a higher SI ratio (t(1,19) = 3.032, p = 0.0204) after CSDS, whereas susceptible sons did not. Error bars indicate mean ± SEM. Significant (p ≤0.05) two-way ANOVA interaction noted on graph: φ, main effect of offspring stress; δ, main effect of paternal category; σ, interaction of paternal category X offspring stress. For direct comparison of two groups: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001, #p ≤ 0.1. A table summary for all statistical test results can be found in Extended Data Figure 1-1 and Figure 1-2.
Extended Data Figure 1-1
Extended Data Figure 1. Download Figure 1-1, DOCX file.
Extended Data Figure 1-2
Extended Data Figure 1-2. Download Figure 1-2, DOCX file.
We hypothesized that, if sperm mediates aspects of the paternal transmission of stress phenotypes, we will see distinct behavioral phenotypes in offspring from stress lineages compared with controls and the patterns of behavior will be similar in NM and AI litters. Unstressed offspring from stressed fathers, whether S or R and whether generated by NM or AI, were indistinguishable from offspring from control (C) lineages in all behavioral tests. We exposed male offspring to a SCVS procedure, which typically does not produce anxiety-like phenotypes in males (Hodes et al., 2015). There was no effect of male offspring exposure to SCVS in SI testing (data not shown; Extended Data Fig. 1-1 and Fig. 1-2). However, after exposure to SCVS, NM sons from S lineages showed a higher NSF ratio compared with sons from control (C) lineages (t(54) = 4.004, p = 0.0006), indicative of increased anxiety-like behavior, an effect not apparent in AI sons (t(44) = 0.07265, p >0.9999; Fig. 1D). In the OF, there was a main effect of offspring stress (F(1,55) = 13.23, p = 0.0006), with both C and R sons from NM decreasing the percent time spent in the middle of the arena when compared with unstressed littermates (t(55) = 2.597, p = 0.0361 and t(55) = 2.700, p = 0.0276, respectively). By contrast, NM sons from S lineages showed no change in the percent time spent in the middle of the arena between unstressed and stressed offspring, suggesting a blunted stress response (t(55) = 0.9344, p > 0.9999; Fig. 1E). These data provide the first evidence that male offspring from stress lineages are sensitive to stressors besides subthreshold CSDS (Dietz et al., 2011). However, these phenotypes where not seen in sons produced by AI, which suggests that SCVS may not be a robust enough stressor to produce altered stress-induced anxiety-like behaviors in offspring produced by AI.
GO term summary
To examine whether a more robust stressor would induce altered stress-responses, we next studied F1 males produced by NM or AI in an SI test before and after exposure to a mCSDS. To do this, we took the unstressed offspring from the previous experiment and exposed them to an 8-d CSDS paradigm (mCSDS; Fig. 1F). After mCSDS, NM sons displayed a main effect of stress (F(1,24) = 19.75, p = 0.0002) with sons from C (t(24) = 3.15, p = 0.013) and R (t(24) = 2.822, p = 0.0283) lineages, but not S lineage (t(24) = 1.789, p = 0.2585), showing an increase in SI ratio. Importantly, the same distinct phenotypes were seen in R and S sons produced by AI, indicating that sperm directly contributes to the intergenerational transmission of this stress phenotype (t(19) = 3.032, p = 0.0204 and t(19) = 0.1693, p = 0.9977, respectively; Fig. 1G).
There are a limited number of intergenerational stress studies that have examined how paternal exposure to stress affects female offspring. Given that there are dramatic sex differences in stress responses between males and females (LaPlant et al., 2009; Maeng and Milad, 2015; Labonté et al., 2017), we investigated potential sex differences in stress responses in offspring from stress lineages. We examined the phenotypes of female offspring from stress lineages at baseline and after exposure to SCVS (Fig. 2A). In female offspring produced by NM, at baseline daughters from S lineages showed an increased SI ratio compared with C, suggesting increased social investigation (t(49) = 3.388, p = 0.0028; Fig. 2B). This effect was lost in AI daughters after exposure to SCVS (t(49) = 0.1623, p > 0.9999). In OF, daughters produced by NM at baseline displayed a main effect of offspring stress (F(1,54) = 5.323, p = 0.0249) and post hoc analysis revealed that S daughters were more sensitive to stress with increased anxiety-like behavior, as measured by a decreased percent time spent in the middle of the arena compared with unstressed littermates (t(54) = 2.612, p = 0.0349; Fig. 2C). The same phenotype was seen in offspring produced by AI, demonstrating that sperm directly contributes to aspects of the intergeneration transmission of these behavioral phenotypes seen in daughters of S but not R lineages (t(53) = 2.067, p = 0.0582). No effect of paternal category was seen in NSF at baseline or after SCVS in NM or AI daughters (data not shown; F(2,44) = 0.5224, p = 0.5967 and F(2,54) = 0.04,218, p = 0.9587, respectively; Extended Data Fig. 2-1 and 2-2).
AI with sperm from stressed fathers increases stress sensitivity of female offspring in adulthood. A, Schematic of female offspring stress and behavior paradigm. Stressed female offspring were exposed to a 6-d subthreshold chronic variable stress. B, Neither paternal category nor offspring stress affected SI ratio in NM F1 daughters. In F1 daughters produced by AI, there was a main effect of offspring stress (F(1,49) = 6.168, p = 0.0165). AI F1 daughters from susceptible lineages showed an increased SI ratio at baseline compared with unstressed controls (t(1,49) = 3.388, p = 0.0028), an effect reversed by stress exposure (t(1,49) = 2.885, p = 0.0174). C, Offspring stress affected the percent time spent in the middle of an OF arena (F(1,118) = 6.854, p = 0.01) in female offspring. F1 stressed daughters, generated by NM or by AI, from susceptible lineages spent less time in the center of the OF arena when compared with unstressed female littermates (F(1,54) = 9.376, p = 0.0034). Error bars indicate mean ± SEM. Significant (p ≤0.05) two-way ANOVA interaction noted on graph: φ, main effect of offspring stress; δ, main effect of paternal category; σ, interaction of paternal category X offspring stress. For direct comparison of two groups: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001, #p ≤ 0.1. A table summary for all statistical test results can be found in Extended Data Figure 2-1 and Figure 2-2.
Extended Data Figure 2-1
Extended Data Figure 2-1. Download Figure 2-1, DOCX file.
Extended Data Figure 2-2
Extended Data Figure 2-2. Download Figure 2-2, DOCX file.
Distinct transcriptional patterns in sperm in resilient and susceptible mice
Having established that sperm can serve as the active vehicle to transmit certain effects of paternal stress to male and female offspring, we next used RNA-seq to investigate stress-induced changes in the transcriptome of sperm that may contribute to the behavioral phenotypes seen in F1 offspring produced by AI. Given that our group has previously shown that mating male mice before their exposure to CSDS does not result in altered phenotypes in offspring (Dietz et al., 2011), we sought to examine how the sperm transcriptome changes in response to stress. First, mature sperm was surgically collected from the left caudal epididymis of male mice before CSDS (Fig. 4A). The same surgically-manipulated males were put through 10 d of CSDS and 1 d later an SI test was used to categorize them as either R or S, with C F0 animals run in parallel to account for differences in age (Fig. 3A). The mice were then singly housed for 28 d and put through another SI test (experimental day 46) and 1 d later (experimental day 47; 30 d after the last defeat exposure) mice were euthanized, and mature sperm was collected from the right caudal epididymis of C, R, and S mice. Six to seven samples per group were used for RNA-seq analysis.
The effect of time on the transcriptome of sperm. A, Scheme of paradigm and sperm collection time points in control animals. Note that, while the early time point is denoted “predefeat” and the later time point is denoted “postdefeat,” all animals in this experiment were controls and not exposed to CSDS. B, Threshold-free comparison of gene expression by RRHO. Pixels represent the overlap between the transcriptome of each comparison as noted, with the significance of overlap (−log10(RPKM) of a hypergeometric test) color coded. C, Heatmaps show all predefeat and postdefeat RNAs ranked by RPKM values seeded by predefeat values. The third row of RNAs indicates DEGs that are expressed at higher (yellow) or lower (blue) levels at the later versus earlier time point.
Distinct transcriptional regulation in sperm of resilient and susceptible mice. A, Schematic timeline of F0 stress and sperm collection. B, Within animal comparison of F0 males used in NM and AI before (day 18) and after (day 46) CSDS in SI testing. C, Overlap of the gene list from these sequencing datasets with published gene lists from RNA-seq of epididymis versus sperm of mice and rats shows a higher overlap of genes from our gene list with the published sperm gene list than the epididymis gene list. Moreover, only ∼8% of the genes in our sperm dataset were not seen in the published sperm dataset, whereas ∼27% of our sperm genes were not seen in the published epididymis dataset. D, Threshold-free comparison of gene by RRHO. Pixels represent the overlap between the transcriptome of each comparison as noted, with the significance of overlap (−log10(p value) of a hypergeometric test) color coded. Lower left quadrant includes co-upregulated genes, upper right quadrant includes co-downregulated genes, and upper left and lower right quadrants include oppositely regulated genes (up-down and down-up, respectively). Genes along each axis are sorted in the same order. Note the dramatic overlap of transcriptional patterns between resilient versus susceptible mice predefeat and a virtually complete loss of such overlap postdefeat. E, Heatmaps show predefeat DEGs with at least a nominal p≤0.025 and a log2 FC ≥ 2. Heatmaps are seeded by either resilient or susceptible DEGs. F, Top three biological function GO terms enriched in resilient (orange) or susceptible (pink) versus control. G, Heatmaps show DEGs with at least a nominal p ≤ 0.025 and a log2 FC≥ 2. Heatmaps are seeded by either resilient (top) or susceptible (bottom) DEGs. H, Top biological function GO terms enriched in susceptible (pink) versus control. I, Upstream regulator analysis was conducted on both resilient and susceptible sperm DEGs and comparison analysis was used to examine the overlap of regulators. Canonical pathway analysis was conducted on both resilient and susceptible sperm DEGs and comparison analysis was used to look at the overlap. Activation Z scores in heatmaps: positive (yellow) = overrepresentation of targets activated by regulator; negative (blue) = overrepresentation of targets repressed by regulator; no direction (black) = no significant enrichment of activated versus repressed targets; white = not a predicted upstream regulator; * = statistically significant enrichment (p ≤ 0.05). A list of genes DEGs for the various comparisons described can be found in Extended Data Figure 4-1.
Extended Data Figure 4-1
These tables list DEGs for the various comparisons described in this study. Download Figure 4-1, XLSX file.
Surgery did not affect the proportion of animals that were later categorized as R or S (
Previous studies have found that small non-coding RNAs in sperm are dynamic and impacted by age in both mice (Chan et al., 2020) and humans (Morgan et al., 2020). However, no studies to date have examined such age-dependent patterns in sperm PCGs and lncRNAs. We used RPKM values as a measure of gene expression in the sperm of surgery-control predefeat and postdefeat mice to perform a within-animal comparison to examine the impact of time on the sperm transcriptome. Using rank-rank hypergeometric overlay (RRHO) to compare patterns and strength of genome-wide overlap in a threshold-free manner (Cahill et al., 2018), we observed robustly overlapping patterns of gene expression in predefeat versus postdefeat control animals (Fig. 3B). However, when we directly examined DEGs in postdefeat compared with predefeat control animals, we found 929 genes that were significantly different between the two groups (Fig. 3C). This suggests that, although the sperm transcriptome is mostly the same across age, a subset of transcripts show altered expression as a function of time, thus mirroring previous results reported for small non-coding RNAs.
In predefeat sperm, we expected to see minimal transcriptomic differences between R and S sperm, since all animals examined are essentially normal C57BL/6J mice without any perturbations. We again used RRHO to compare patterns and strength of genome-wide overlap in a threshold-free manner. RRHO analyses showed that, before defeat, there is strong co-regulation of genes in the same direction in R and S mice compared with C mice. CSDS disrupts this pattern with opposite regulation observed for many of the same genes (Fig. 4D). The latter finding indicates that CSDS produces a striking divergence in transcriptional regulation in sperm between R versus S mice.
To identify DEGs in S and R sperm both predefeat and postdefeat, the significance of DEGs was set at a nominal p < 0.025 and a log2 FC≥2. Predefeat there were 873 R-specific DEGs and 400 S-specific DEGs, with PCGs making up the majority of the biotypes (R: ∼92%, S: ∼94%) followed distantly by lncRNAs (R: ∼5%, S: ∼6%) and pseudogenes (R: ∼3%, S: ∼0.25%; Fig. 5A; Extended Data Fig. 4-1). This distribution of RNA biotypes is similar to what has been reported previously for sperm (Soumillon et al., 2013). Predefeat, heatmaps seeded by either R or S DEGs revealed that both R-specific and S-specific DEGs tend to be similarly dysregulated in the other group (Fig. 4E), confirming our RRHO analysis. To gain insight into the molecular and biological functions of these DEGs, we used Enrichr GO analysis. Many of the top GO terms in R sperm (∼66%) are related to cell cycle processes, whereas all top S sperm GO terms are related to cell migration (Fig. 4F; Table 1).
CSDS alters the biotype composition of RNAs in susceptible sperm. A, Pie chart represents the biotypes making up resilient and susceptible DEGs in sperm before and after CSDS in the same animals. There is a difference in the proportion of genes in each biotype composing DEGs in susceptible (
Extended Data Figure 5-1
Extended Data Figure 5-1. Download Figure 5-1, DOCX file.
Before turning to DEG analysis of sperm after CSDS, we examined the effect of surgery per se on the sperm transcriptome by performing RNA-seq on sperm from surgically-manipulated control mice compared with non-surgically-manipulated control mice. There were 1527 surgery-induced DEGs in sperm with biological functions related to flagellum function, ncRNA metabolic processes, and spermatogenesis (Table 2). In subsequent analyses, surgery DEGs were filtered out to examine stress-induced transcriptional differences in the sperm of R and S animals with the cut off for DEGs set at a nominal p < 0.025 and a log2 FC ≥ 2.
Postdefeat, differential expression analysis revealed only 62 R DEGs, with PCGs making up the large majority of the biotype (∼93%) followed by lncRNAs (∼7%; Fig. 5A). This finding illustrates a very limited effect of CSDS on the sperm transcriptome of R fathers. By contrast, there were 1460 S DEGs, with PCGs making up a smaller majority of the biotype (∼79%) followed by lncRNAs (∼18%) and pseudogenes (∼3%; Fig. 5A; Extended Data Fig. 4-1). Heatmaps of the R DEGs revealed that many of these genes are similarly regulated in S animals (Fig. 4G). However, heatmaps of S DEGs revealed that the vast majority appear unchanged in R animals (Fig. 4G).
GO analysis revealed no enrichment of postdefeat R DEGs for biological functions presumably because of the very small number of DEGs. The majority of the top biological function GO terms for S sperm (∼87%) were related to flagellum function (Fig. 4H; Table 2), with one term involved in cell projection functioning. Dysregulation of flagellum processes can impact fertility (Lehti and Sironen, 2017); however, there was no difference in litter size or proportion of male or female offspring between control and stress lineages (F(2,19) = 0.2637, p = 0.7710). This finding suggests that fertility may not be affected dramatically, but rather that genes involved in flagellum function might relay paternal environmental information to offspring.
We next examined predicted upstream regulators of stress-induced transcriptomic changes in sperm of R and S mice. In R mice, there were five predicted upstream regulators with predicted inactivation [estrogen receptor 1 (ESR1), nuclear factor erythroid 2 like 2 (NFE2L2), interleukin 1-family (IL1), IL1-β, and CCAAT enhancer binding protein α (CEBPA)] and one molecule with predicted activation [Kruppel-like factor 3 (KLF3); Fig. 4I]. There were three overlapping predicted upstream regulators between predefeat and postdefeat in R mice (ESR1, NFE2L2, and KLF3), however, there was no change in the activation state of these molecules following stress.
In S sperm, there were seven molecules predicted as upstream regulators: vascular endothelial growth factor (VEGF), Rab-like protein 6 (RABL6), stratifin, aryl hydrocarbon receptor (AHR), estrogen receptor (ER), prostaglandin ER2 (PTGER2), and tumor protein p53 (Fig. 4I). There were four overlapping predicted upstream regulators in predefeat and postdefeat S sperm (VEGF, SHR, ER, and TP53). Notably, only one molecule, VEGF, went from an activated to inhibited state after CSDS. VEGF is particularly interesting as a previous intergenerational study performed by our laboratory found that plasma levels of VEGF were lower in offspring from fathers after exposure to CSDS (Dietz et al., 2011), and numerous other studies have linked VEGF to depression in both rodent (Deyama et al., 2019) and human studies (Clark-Raymond and Halaris, 2013). Of note, ESR1 is a major subtype of ER, which indicates that ER signaling was a predicted upstream regulator of stress effects in both R and S sperm.
Regulation of lncRNA expression in sperm of resilient and susceptible mice
Analysis of the distributions of RNA biotype among DEGs revealed a significant difference in predefeat versus postdefeat in S but not R sperm. Post hoc tests revealed that this was driven by an increase in the proportion of lncRNA DEGs in S animals only, with 277 lncRNA DEGs in S mice postdefeat compared with only four lncRNA DEGs in R mice postdefeat and 24 and 37 lncRNAs in S and R mice predefeat, respectively (Fig. 5A,B). Heatmaps of only the lncRNAs regulated in S mice postdefeat revealed that the vast majority are unaffected in R mice (Fig. 5C).
To further assess the impact of altered expression of lncRNAs in sperm from S fathers, we performed correlation analysis to relate the expression of lncRNAs with the expression of PCGs (Issler et al., 2020). lncRNAs and PCGs that show correlated expression are potentially involved in similar biological functions (Rinn and Chang, 2012). Histogram plots of the distribution of highly correlated PCGs (significant = r
Co-expression analysis identifies gene modules in sperm of resilient and susceptible mice
We next used WGCNA (Langfelder and Horvath, 2008) to construct co-expression modules using the expression data of sperm after exposure to CSDS. WGCNA created a network that consisted of 25 discrete modules with between 33 and 5807 genes in each module (with each module assigned an arbitrary color name). We next identified gene modules differentially affected by CSDS in R and S mice by determining whether modules were enriched for DEGs under either condition. We found three modules (grey60, salmon, and medium orchid) that were enriched in R DEGs and five modules (royal blue, bisque4, dark magenta, dark orange, and dark orange2) that were enriched in S DEGs. Interestingly, no modules were enriched for both R and S DEGs, suggesting distinct regulation of genes in the sperm of resilient and susceptible animals only after stress (Fig. 6A).
Identification of co-expression gene networks in sperm differentially affected by CSDS in susceptible versus resilient fathers. A, Circos plot for all 25 WGCNA modules. Module names (an arbitrary color) and sizes are indicated on the outside of the Circos plot. Enrichment for either resilient or susceptible DEGs or enrichment for lncRNA genes are indicated by the semicircle colors within each module, with increasing warm colors indicating increasing −log10(p value). Arachne plots show the co-expression networks of (B) the susceptible-enriched and lncRNA-enriched module dark magenta, and (C) the resilient-enriched module medium orchid. Node size indicates the strength of stress-specific hub genes with increased size indicating top hub rank. Pink nodes indicate differential expression for that gene in susceptible animals. Orange nodes indicate differential expression for that gene in susceptible animals. Black halos indicate that gene is a lncRNA. Edges indicate significant co-expression between two particular genes.
Using the Cytoscape software platform, we performed comprehensive analysis of the relationship between nodes using the MCC function, reported to be the most effective method of finding hub genes in a co-expression network (Chin et al., 2014). Given the loss of correlation between lncRNA and PCG DEGs in S sperm, we looked for modules enriched in lncRNAs and R or S DEGs. Of particular interest, the S-enriched module, dark magenta, was also enriched in lncRNAs and contained a lncRNA, Gm27211, as a top hub gene. Additionally, Gm27211 was differentially expressed in S animals after CSDS (Fig. 6B). Gm27211 is expressed from E10 to E17 in the central nervous system, and in male and female reproductive tracts from P4 to adulthood; however, the specific function of this lncRNA is largely unknown (Brunskill et al., 2014; Bao et al., 2016; Kistler et al., 2015; Huntley et al., 2016).
Interestingly there was no overlap in R-enriched and lncRNA-enriched modules. The R-enriched module, medium orchid, had two hub genes that were differentially expressed, sarcospan (Sspn) and Rho GTPase activating protein 21 (Arhgap21), both of which are PCGs involved in cell adhesion and actin dynamics, respectively (Parvatiyar et al., 2015, 2019; Fig. 6C).
Discussion
In the present study, we provide several new lines of evidence concerning the transmission of paternal stress to the next generation. We show that susceptible and resilient fathers, after a course of CSDS, transmit different degrees of stress susceptibility to their offspring, with differences in phenotypes seen between male versus female offspring. We go on to show that a subset of these effects of paternal stress occur in offspring generated by AI, demonstrating that sperm per se is an active vehicle in the intergenerational transmission of stress phenotypes. Finally, we characterize the transcriptome of sperm of susceptible, resilient, and control fathers by RNA-seq and identify prominent regulation of lncRNAs in susceptibility selectively, thus implicating these molecules in next-generation consequences of paternal stress.
Paternal transmission studies typically capitalize on the fact that most male rodents are not involved in offspring rearing, which allows observed phenotypes in offspring to be attributed to male gamete contribution. We previously reported only minimal intergenerational transmission of stress vulnerability when offspring were produced via IVF, suggesting that behavioral or hormonal changes rather than epigenetic mechanisms may explain the dramatic alterations in offspring phenotypes (Dietz et al., 2011). However, the lack of robust behavioral phenotypes observed in this earlier study may be because of the fact that IVF alters the epigenome of sperm and egg cells (Khosla et al., 2001; Reik and Walter, 2001; Calle et al., 2012; Ventura-Juncá et al., 2015), which may wash out stress-induced epigenetic changes. In the present study, therefore, we used AI, which removes the need for many of the IVF factors that are believed to influence the epigenome. We report now for the first time that male and female offspring produced by AI using sperm from CSDS-exposed males show altered stress sensitivity, thereby establishing that epigenetic alterations in sperm do indeed mediate some of the intergenerational transmission of altered stress phenotypes observed with NM.
More specifically, we found that paternal CSDS alters how sons produced by NM respond to SCVS but does not affect such behavioral responses in sons produced by AI. This suggests the involvement of other factors such as maternal care or hormonal signals during mating in transmitting this phenotype rather than epigenetic changes in sperm. However, we hypothesized that a more robust offspring stressor would reveal stress phenotypes transmitted by sperm. We therefore exposed male offspring to a mCSDS procedure. Sons produced by NM or by AI and exposed to mCSDS display equivalent differences when compared with sons of control mice, thus demonstrating the intergenerational transmission of stress sensitivity by sperm.
Female offspring from stress lineages produced by NM and AII showed a similar heightened response to stress in OF testing. Surprisingly, daughters from susceptible lineages produced by AI but not NM showed an increase in social investigation. Although it may be surprising that female offspring from S lineages produced by AI and not by NM showed increased social investigation at baseline, previous studies have reported altered maternal care when dames are allowed to mate directly with stressed males (Mashoodh et al., 2018). These studies suggest that maternal interaction with an affected father may lead to maternal masking and underscore the importance of using gamete manipulation techniques for disentangling how male gametes and epigenetic changes to them contribute to the transmission of stress phenotypes across generations overall hinting at potential maternal masking of altered phenotypes in our NM experiments.
Studies investigating potential molecular mechanisms that contribute to the transmission of stress phenotypes across multiple generations in rodent models have identified differences in microRNAs (Rodgers et al., 2013), lncRNAs (Gapp et al., 2020), and DNA methylation (Franklin et al., 2010). However, no study to date has identified how the RNA profiles in sperm are altered by paternal stress in a manner that distinguishes between those fathers that are susceptible versus resilient to stress exposure. Here, we provide an assessment of genome-wide transcriptomic changes in sperm before and after exposure to CSDS within the same F0 males who are susceptible or resilient to that stress.
A surprising finding of this experiment was the detection of differences in gene expression in the sperm of susceptible versus resilient mice before exposure to CSDS. This is surprising since, at this time point, the susceptible, resilient, and control mice represented essentially equivalent groups of wild-type C57BL/6J mice and we previously reported that mating males before stress exposure does not result in the transmission of stress phenotypes to offspring (Dietz et al., 2011). Indeed, RRHO analysis demonstrated broad convergence of gene expression patterns in the sperm across all three groups of animals, confirming very similar transcriptomes. Nevertheless, DEG analysis identified numerous genes that show different expression levels in mice that will become susceptible versus resilient when each was compared with the control group. Together, these analyses demonstrate that, within a group of inbred C57BL/6J with equivalent environmental experiences, there is a sizable degree of variation in sperm gene expression patterns, which correlates with the inherent vulnerability of an individual mouse to be either susceptible or resilient to social stress later in life. Understanding the factors responsible for such baseline variability in sperm gene expression is an interesting topic for future studies.
There is increasing evidence that stress alters the epigenome of sperm (Franklin et al., 2010; Gapp et al., 2014, 2020), however, there is little consistency across different paternal stressors and no studies have investigated whether similar changes are seen after CSDS. In contrast to the subtle predefeat differences in sperm gene expression, exposure to CSDS dramatically restructured the transcriptome of sperm from susceptible versus resilient fathers in very different ways. RRHO and DEG analyses revealed virtually no overlap in DEGs between the two subsets of fathers following exposure to CSDS. In susceptible fathers, there was a striking transcriptional reorganization of gene expression, with 1460 genes being differentially expressed. This contrasts dramatically with the resilient fathers, which show alterations in ∼20-fold fewer genes, demonstrating that resilient fathers escape the vast majority of the stress-induced changes in sperm gene expression seen in susceptible fathers. The finding that the sperm of inbred C57BL/6J mice display such remarkable differences in transcriptomic responses to the same stress underscores the tremendous potential for diversity even in the context of constant genetic background. Future studies are needed to understand the molecular mechanisms underlying these dramatic individual differences in stress responses in sperm.
Given that lncRNAs have been established as important for sperm functioning (Zhang et al., 2010, 2019; Wen et al., 2016), have been associated with depression GWAS studies (Roy et al., 2018; Wray et al., 2018), have been seen to be dysregulated in the sperm of diabetic mice (Jiang et al., 2016), and have also been identified as a partial mediator of the paternal transmission of stress phenotypes (Gapp et al., 2020), we examined this RNA biotype in sperm DEGs. We found a ∼10-fold increase in the number of stress induced differentially expressed lncRNAs in sperm from susceptible when compared with both resilient sperm after stress and prestress sperm independent of group. lncRNAs have been shown to serve as key drivers in cancer (Ashouri et al., 2016; Lanzós et al., 2017), and manipulation of lncRNAs in adult brain have been shown to induce depression-related behavioral phenotypes (Issler et al., 2020), suggesting that lncRNA expression can have broad impacts on a variety of phenotypes. To provide insight into the function of lncRNAs in our dataset, we used correlation analysis to relate lncRNA expression with PCG expression. We found similar patterns of lncRNA-PCG relationships in sperm of control and resilient fathers, but a dramatic loss of correlation in susceptible fathers. We next used WGCNA to characterize the connectivity of genes within our sequencing dataset and identified several gene modules that differentially enrich for either resilient or susceptible DEGs, with ∼80% of the susceptible-enriched modules also enriched for lncRNAs. These data are consistent with the view that lncRNAs contribute to the paternal transmission of stress phenotypes via regulation of PCGs. We looked for lncRNA hub genes in those modules and identified one, Gm27211, which now warrants further investigation. A key challenge for future research is to now relate altered levels of Gm27211 and presumably other regulatory genes in sperm to altered gene transcription early in development and eventually in the developing and adult brain, and ultimately to alterations in stress susceptibility observed in adult offspring.
lncRNAs like Gm27211 might also drive lasting phenotypes through interactions with microRNAs. It has been established that microRNAs regulate lncRNA expression in other tissues (da Rocha et al., 2008; Royo and Cavaillé, 2008), and studies suggest that lncRNAs may regulate miRNA function, creating the possibility of multiple feedback loops between long and short noncoding RNAs (Feng et al., 2006; Jeck and Sharpless, 2014; Ballantyne et al., 2016). This is consistent with previous work that has implicated certain microRNAs in sperm of stressed fathers as also contributing to the intergenerational transmission of stress susceptibility. However, efforts to understand the role of lncRNAs in paternal transmission of stress experience to their offspring are impaired by our still very limited knowledge of the biological function of the vast majority of lncRNAs (Bassett et al., 2014).
In summary, results from the present study show that male mice exposed to CSDS categorized as either susceptible or resilient transmit partly different behavioral responses to their offspring and that sperm is one active ingredient involved in this intergenerational transmission. Moreover, the behavioral changes observed in the offspring are accompanied by stress-induced disruption of the transcriptional state in sperm. This study thereby adds valuable insight into how stress interacts across generations and how the mechanisms of this transmission might differ between resilient and susceptible animals. This work thus suggests novel potential targets through which epigenetic mechanisms contribute to inherited phenotypes. At the same time, our findings raise two crucial questions for future analysis. The complex mechanisms underlying how behavioral experience induces long-lasting changes in sperm gene expression remain largely elusive. While epididymis extracellular vesicles have been shown to be one potential mechanism (Chan et al., 2020), many others may be at play but have not yet been investigated. As well, it will be essential to delineate the precise mechanisms by which changes in sperm gene expression lead to life-long changes in brain structure and function that mediate the altered behavioral phenotypes seen in adult offspring of stressed fathers.
Footnotes
This work was supported by National Instituties of Health Grants P50MH096890 and R01MH051399, R00 DA042100 and T32MH087004, and the Hope for Depression Research Foundation.
The authors declare no competing financial interests.
- Correspondence should be addressed to Deena M. Walker at walkerde{at}ohsu.edu or Eric J. Nestler at eric.nestler{at}mssm.edu
References
In this issue
