Unlocking the epigenome: Stress and exercise induced Bdnf regulation in the prefrontal cortex

Abstract

Aversive caregiving in early life is a risk factor for aberrant brain and behavioral development. This outcome is related to epigenetic dysregulation of the brain-derived neurotrophic factor (Bdnf) gene. The Bdnf gene encodes for BDNF, a neurotrophin involved in early brain development, neural plasticity, learning, and memory. Recent work suggests that exercise may be neuroprotective in part by supporting BDNF protein and gene expression, making it an exciting target for therapeutic interventions. To our knowledge, exercise has never been studied as a therapeutic intervention in preclinical rodent models of caregiver maltreatment. To that end, the current study investigated the effect of an adult voluntary wheel running intervention on Bdnf methylation and expression in the prefrontal cortex of rats who experienced aversive caregiving in infancy. We employed a rodent model (Long Evans rats) wherein rat pups experienced intermittent caregiver-induced stress from postnatal days 1–7 and were given voluntary access to a running wheel (except in the control condition) from postnatal days 70–90 as a young adulthood treatment intervention. Our results indicate that maltreatment and exercise affect Bdnf gene methylation in an exon, CG site, and sex-specific manner. Here we add to a growing body of evidence of the ability for our experiences, including exercise, to permeate the brain.

1.1. Introduction

Positive and negative experiences during early development can stay with us throughout our life span by affecting the way the brain develops over critical and sensitive periods (Milbocker et al., 2021). Experiencing stress particularly affects brain development by interfering with neural processes, like synaptogenesis and synaptic pruning (Milbocker et al., 2021). These disruptions in early neural circuit programming set the stage for a child to be more susceptible to developing psychiatric disorders as they age, including depression (LeMoult et al., 2020), anxiety (Lähdepuro et al., 2019), schizophrenia (Barhari-Javan et al., 2017; Carr et al., 2013), and bipolar disorder (Menezes et al., 2022; Syed & Nemeroff, 2017).

One way our everyday experiences permeate the brain is through epigenetic processes (Barhari-Javan et al., 2017; Roth et al., 2009; Keller et al., 2019; Collins et al., 2022; Fachim et al., 2021). Epigenetics refers to molecular mechanisms that change the capacity for genes to be expressed in the brain and body without making changes to the underlying genomic sequence itself. DNA methylation is one epigenetic mechanism and refers to the addition of methyl groups to cytosine-guanine dinucleotides, or CG sites, on the DNA sequence (Roth et a., 2009; Fachim et al., 2021). Generally, DNA methylation occurring around the promoter region of a gene typically leads to gene silencing or gene repression (Roth et al., 2009; Roth et al., 2011; Kundakovic et al., 2014).

DNA methylation is a known mechanism by which adversity affects brain development. Numerous studies show that early life stress increases methylation of the brain-derived neurotrophic factor (Bdnf) gene and decreases Bdnf expression in numerous regions of the brain (Roth et al., 2009; Collins et al., 2022; Seo et al., 2016; Tata et al., 2021). Bdnf codes for the BDNF protein, which is not just a vital protein for proper development of the central nervous system, but also plays a significant role in brain and behavior throughout the lifespan. Indeed, many of the adverse outcomes associated with early life stress, like bipolar disorder and depression, are associated with epigenetic dysregulation of Bdnf (Brunoni et al., 2008; Rana et al., 2020; Chiou & Huang, 2019). This can have tremendous consequences for prefrontal cortex (PFC) development (Milbocker et al., 2021). Disruptions in PFC development is considered paramount when isolating risk factors for developing psychiatric disorders associated with early life stress (Chocky et al., 2013; Milbocker et al., 2021).

In contrast to stress, positive experiences, like exercise, are associated with increases in Bdnf expression and improved symptomatology in anxiety and depression (Piepmeier & Etnier, 2015; Murawska-Cialowicz et al., 2021; Jayakody et al., 2014). Exercise is thought to be neuroprotective by strengthening the brain against neurodegeneration (Marques-Aleixo et al., 2021). For example, exercise improves behavioral and neurological outcomes in rodent models of stroke by facilitating synaptogenesis in the central nervous system (Sasaki et a., 2016; Zhang et al., 2020a). Anxiety behaviors are also reduced in rodents exposed to chronic stress by remodeling of the medial PFC (mPFC)-basolateral amygdala circuit (Luo et al., 2023). In recent years, a handful of studies have investigated the efficacy of exercise to counteract maladaptive epigenetic and behavioral consequences of early life stress in animal models (Campbell et al., 2022). Modeling early life stress in the laboratory often involves experimental manipulations that either induce caregiver neglect or fragmented/aversive maternal care during infancy (Walker et al., 2016). This disruption in infant caregiving often results in maladaptive neural and behavioral development (Walker et al., 2016; Roth et al., 2009; Milbocker et al., 2021; Khorjahani et al., 2020; Sadeghi et al., 2016), with underlying abnormalities in epigenetic regulation (Jureuna et al., 2021; Torres-Berrio et al., 2022; Rahman & McGowan 2022). In our experiments here, we utilize a resource-scarce environment to induce fragmented maternal care and aversive behaviors from the rodent dam towards the pups (e.g. Roth et al., 2009; Collins et al., 2022).

Given the inverse relationship between stress, exercise, and Bdnf epigenetic regulation, we introduced voluntary exercise to determine 1) if exercise is sufficient to ameliorate maladaptive changes in Bdnf regulation associated with early adversity; 2) if exercise affects Bdnf regulation in the PFC, a brain region understudied in exercise models; 3) if differential, regional changes in Bdnf methylation is one mechanism whereby exercise affects Bdnf expression; and 4) if an exercise intervention could be efficacious during young adulthood when the PFC’s developmental window has begun to close (Zeiss, 2021). To our knowledge, this is the first study to investigate exercise in our model of disrupted infant-caregiver relationships and the first study to explore differential Bdnf methylation in the PFC as a mechanism explaining how exercise regulates Bdnf activity.

1.2. Methods

1.2.1. Subjects

This study includes data collected from 60 male and 60 female Long-Evans rats obtained by in-house breeding. Breeder males and dams used to generate subjects were obtained from Charles River Laboratories. Day of parturition was termed postnatal day (PN) 0 (Figure 1). Litters were culled to 12 pups (6 males, 6 females when possible) on PN1. All maternal behavioral manipulations and observations occurred during the light cycle. After weaning between PN21–23, pups were pair-housed with same-sex, non-littermates of the same infant condition in standard polypropylene cages with wood shavings. All procedures were approved by the University of Delaware Animal Care and Use Committee.

Figure 1. Experimental Timeline.

Dams and litters were left undisturbed on the day of parturition, termed postnatal day (PN) 0. 24-hours later, on PN1, pups within litters were randomly assigned to one of three infant conditions: normal maternal care, cross-foster care, and maltreatment. Pups were weaned between PN21–23 and pair housed with a same-sex, non-littermate of the same infant condition. Rat pairs either remained under these conditions, or were moved to a similar home cage with access to a running wheel on PN70 where they remained until brain extractions on PN90. Figure created with Biorender.com.

1.2.2. Caregiving Manipulation

Caregiving manipulations followed a validated variation of the limited bedding and nesting model; commonly termed the scarcity-adversity model (e.g. Roth et al., 2009; Collins et al., 2022; Blaze et al., 2013). Each experimental litter was born to a dam termed the normal maternal care dam. The pups in the experimental litter were randomly assigned to one of three postnatal caregiving conditions— maltreatment, normal maternal care, or cross-foster care. The experimental litter remained in the home cage with the normal maternal care dam, except for when the cross-foster and maltreatment pups were removed from the home cage for 30 minutes daily through PNs 1–7 for infant care manipulations and behavior recordings. Pups in the normal maternal care condition remained in the home cage with their biological dam, who was given copious nesting material, and the dam’s maternal behaviors exhibited towards the pups were recorded for 30 minutes in the colony room. Pups assigned to the maltreatment condition were placed into an opaque plexi chamber with a non-biological lactating foster dam that was not given time to habituate to the environmental or sufficient nesting material. Pups assigned to the cross-foster condition were placed into an opaque plexi chamber with another non-biological lactating foster dam that was given ample time to habituate to the environmental and sufficient nesting material. Maternal behaviors exhibited by the maltreatment and cross-foster dams were recorded for 30 minutes in a separate behavior room concurrently with the normal maternal care dam’s behavior observations. Maltreatment and cross-foster dams were matched in diet and postpartum age to the birth mother of the pups utilized in the experiment. The cross-foster condition served as a second control group to show that the maltreatment dam’s behavior towards the pups is altered by her stressful and impoverished environment and not simply due to caring for foster pups. The cross-foster condition also allowed us to control for increased experimenter handling, as the normal care pups are handled less by the experimenter due to staying in the homecage for behavioral recordings. The cross-foster and maltreatment dams also had their own biological litters, that were age matched to the experimental pups, to ensure they were lactating dams and capable of providing care to the experimental pups during behavioral manipulations. Their biological pups were placed in an incubator during behavioral recordings and returned to their home cages with their biological dams immediately following cessation of the behavioral observations each day. To increase genetic diversity in the experiment, 1) dams were never bred twice with the same male, 2) dams producing the experimental litters were changed out every few months, and 3) new breeder males were purchased every 8 months.

Behavior videos were scored to determine the percent of nurturing and aversive behaviors experienced by the pups in each of the infant conditions. Nurturing behaviors exhibited by the dams included nursing/hovering over litter and pup licking, and aversive behaviors included stepping on pups, rough handling pups (i.e. picking them up by a hindpaw), dropping pups during transport, actively avoiding the litter, and dragging pups across the cage while nursing. Occurrences of nurturing and aversive behaviors were hand-scored by two independent observers. Each 30-minute recording was sectioned into six 5-minute time bins and observers conducted time sample observations, with each behavior scored as “observed” or “not observed” in each time bin as previously done in our lab (Collins et al., 2022; Duffy & Roth, 2022; Doherty et al., 2019). The proportion of nurturing to aversive behaviors observed were calculated as a percent, as well as the average occurrence of each individual behavior (Figure 2). Inter-rater reliability (IRR) was calculated between each observer for every video scored; if IRR was not at least 80% agreement, the video was scored by a third independent observer, blind to the original behavior scoring observations. All videos below 80% IRR reached reliability criterion after one of the original observers was replaced by the third independent observer.

Figure 2. Maternal Behavior.

Maternal Behavior analysis. (A) pups in the maltreatment condition were exposed to significantly more aversive behaviors from the dam than nurturing behaviors. ***p < 0.001 vs. nurturing behaviors in the same infant condition; ###p < 0.001 vs. same behaviors in nurturing care conditions. (B) One-way ANOVAs of individual aversive and (C) nurturing behaviors. *p < 0.05, **p = 0.002, ***p < 0.001 vs. maltreatment condition. (D) Pie charts depicting observed % of individual behaviors. Error bars represent SEM. N = 16 total litters.

1.2.3. Adult Wheel Running

On PN70, rat pairs were randomly assigned to wheel running exposed (WR) or standard housed (SH). SH rat pairs remained in the home cage, while WR pairs were moved to new home cages equipped with external running wheels. The WR cages were standard polypropylene cages, and identical in size and structure to the SH cages, but also had a doorway cut out on one side of the cage to allow voluntary access to an external running wheel (Med Associates). Rats remained under these conditions until brains were extracted on PN90. All rats were weighed on PN70, twice throughout WR treatment, and on PN90 before euthanasia to monitor exercise-induced weight loss.

1.2.4. Biochemistry

Upon extraction, brains were flash frozen using dry ice and 2-methylbutane and stored at −80°C. DNA and RNA were extracted from the prefrontal cortex (PFC) according to the manufacturer’s instructions (Qiagen Allprep DNA/RNA Mini Kit). Quantity and quality of the nucleic acids were determined using spectrophotometry (Nanodrop 2000). Gene Methylation Assay. DNA was bisulfite converted (Qiagen EpiTect Bisulfite Kit) and CG-site specific methylation was assessed using direct bisulfite DNA sequencing PCR (BSP) targeting Bdnf exons I, IV, and IX. For BSP, percent methylation of the CG sites within the amplified regions was determined by the ratio between peak values of G and A (G/[G+A]) on the electropherograms using Chromas software. Gene Expression Assay. Extracted RNA was reverse transcribed (Qiagen QuantiTect Reverse Transcription Kit) to produce complementary DNA (cDNA). The cDNA was analyzed by real-time-PCR, performed using Taqman probes (Fisher Life Technologies) designed to amplify total Bdnf mRNA (exon IX) and Tubulin mRNA (reference gene). See table 1 for probe and primer sequences.

Table 1. PCR Primers & Probes.

Target Gene Bisulfite DNA Sequencing	Primer Sequence (5′ – 3′)
Bdnf exon I	TTTATTTTTTGGAGTTTGTGGTATG
	ACTTCTCAAATAAAAATTAACAACCTCTAT
Bdnf exon IV	GGTAGAGGAGGTATTATATGATAGTTTA
	TACTCCTATTCTTCAACAAAAAAATTAAAT
Bdnf exon IX	GTGAATGGGTTTAGGGTAGGTT
	CCAACAAAAAAAACAAAAAAAACTC
Target Gene Expression	Probe List (Fisher Life Tech)
Tubb2b	Rn01435337_g1
Bdnf exon IX	Rn02531967_s1

1.2.5. Statistical Analysis

The relative gene expression between maltreatment, normal maternal care, and cross-foster care was assessed by the comparative Ct method (Schmittgen & Livak, 2008). CG level specific methylation (BSP) was scored by hand by measuring peak methylation at individual CG sites in the targeted region of DNA. Differences in gene expression, methylation levels, and behavior (exercise and maternal care) was analyzed by a series of multi-way ANOVAs. Statistical significance threshold was set to p < .05, as is standard in the field. Outliers were identified and removed using the ROUT method (Q = 1%) prior to each analysis. Only one male and one female rat from each litter were used per group in the analyses to avoid the confound of litter effects. All statistical analyses were conducted using GraphPad Prism (GraphPad Software Inc, version 9).

1.3. Results

1.3.1. Maternal Behavior

A two-way ANOVA conducted to measure differences in maternal and aversive behaviors between normal care, cross-foster, and maltreatment dams revealed a significant main effect of infant condition [F(1, 70) = 52.73, p < 0.001] and a significant interaction [F(2, 70) = 71.64, p < 0.001] (Figure 2A). Sidak’s post hoc multiple comparisons test revealed that pups in the maltreatment condition were exposed to significantly more aversive behaviors from the dam than nurturing behaviors (p < 0.001). In contrast, pups in the normal care and cross-foster conditions were exposed to significantly more nurturing behaviors than aversive behaviors (p < 0.001). Maltreatment condition pups were also exposed to significantly more aversive behaviors and significantly less nurturing behaviors than pups in the normal care or cross-foster conditions (p < 0.001). There were no significant differences detected between the cross-foster and normal care conditions (p > 0.99).

One-way ANOVAs were utilized to measure differences in individual types of aversive and nurturing behaviors from the dams (Figure 2B). Results revealed a significant difference in stepping [F(2, 35) = 3.73, p = 0.034] with post hoc (Sidak’s) indicating pups in the maltreatment condition were stepped on by the dam more frequently than pups in the normal care condition (p = 0.048). A significant difference in active avoiding [F(2, 35) = 109.3, p < 0.001] was also detected, with post hoc (Sidak’s) indicating pups in the maltreatment condition were actively avoided by the dam more often than pups in the normal care and cross-foster conditions (p < 0.001). One-way ANOVAs did not reveal a significant difference in dragging pups [F(2, 35) = 0.814, p = 0.45]. Differences in dropping [F(2, 35) = 3.066, p = 0.059] and rough handling [F(2, 35) = 3.04, p 0.06] pups approached significance, but ultimately were not statistically significant. Results further showed a significant difference in maternal licking [F(2, 35) = 7.033, p = 0.002], with pups in the maltreatment condition being licked significantly less than pups in the cross-foster condition (p = 0.002). A significant difference in nursing/hovering was also detected [F(2, 35) = 13.34, p < 0.001], with post hoc revealing pups in the maltreatment condition are hovered over or nursed significantly less than pups in the normal care (p < 0.001) and cross-foster (p < 0.001) conditions. No significant differences in individual behaviors were detected between the cross-foster and normal care conditions (Sidak’s p = 0.08–0.99). The percent of observed individual behaviors in each infant condition are depicted in Figure 2C.

1.3.2. Wheel Running Behavior

Wheel rotations were first converted to kilometers to investigate group differences in wheel running as a function of sex, infant condition, and time. Two-way ANOVA revealed no differences in accumulated wheel rotations (in Km) over the 20-day exposure period between the three infant conditions F(2, 33) = 1.56, p = 0.226). Two-way repeated measures ANOVA [Day x Sex] revealed a significant difference in accumulated Km between male and female pairs [F(1, 37) = 72.81, p < 0.001], a significant effect of time [F(19, 703) = 47.05, p < 0.001], and a significant time x sex interaction [F(19, 703) = 13.09, p < 0.001] (Figure 3A). Post hoc (Sidak’s correction) revealed female pairs ran more than male pairs on days 4–20 of wheel running exposure (day 4, p = 0.008; days 7–20, p < 0.001). Post hoc also revealed that both males and females ran more over time (Table 2). Overall, for the 20-day exercise exposure period, female pairs ran more than males [t(37) = 8.29, p < 0.001]; however, Levene’s test showed that the variances were not equal, [F(17,20) = 4.33, p < 0.001] so we ran the analysis using a nonparametric test (Figure 3B). Mann-Whitney test revealed a significant difference between males (mdn = 100.7, 22.06 – 225.1) and females (mdn = 264.9, 189.3 – 548; U =6; p < 0.001).

Figure 3. Wheel-running Behavior.

Average wheel rotation equivalent to kilometers (Km) accumulated over 20 days of voluntary wheel-running. (A) Females accumulated more wheel rotations every 24 hours compared to males beginning at day 4 of wheel-running exposure,*p < 0.01; **p < 0.001 vs. males on the same day. Both male and female pairs also accumulated more wheel rotations overtime (see table 2 for more detail), ^p < 0.05; $p < 0.01; #p < 0.001 vs. day 1 of same sex. (B) T-test of total average mile equivalent of wheel rotations over the 20-day exposure, **p < 0.001 versus males. Error bars represent SEM. N = 18–21 pairs per group.

Table 2. Wheel Rotation Results.

Female pairs	Sidak’s corrected p value
Wheel rotations different from Day Day 7: p = 0.019; Days 8–20: p < 0.001
1
Wheel rotations different from Day Days 6–20: p < 0.001
2
Wheel rotations different from Day Day 6: p < 0.01; Days 7–20: p < 0.001
3
Wheel rotations different from Day Day 7: p < 0.01; Days 8–20: p < 0.001
4
Wheel rotations different from Day Days 9–20: p < 0.001
5
Wheel rotations different from Day Day 9: p = 0.01; Days 10–20: p < 0.001
6
Wheel rotations different from Day Day 11: p < 0.01; Days 10, 12–20: p < 0.001
7
Wheel rotations different from Day Days 10, 12–20: p < 0.001
8
Wheel rotations different from Day Days 17–19: p < 0.001; Day 20, p < 0.01
9
Wheel rotations different from Day Days 17, 19: p < 0.05; Day 18: p < 0.001
11
Male pairs	Sidak’s corrected p value
Wheel rotations different from Day Days 17, 18: p < 0.05; Days 19, 20: p < 0.01
1
Wheel rotations different from Day Days 14, 16: p < 0.05; Days 17, 18: p < 0.01; Days 19, 20: p < 0.001
2
Wheel rotations different from Day Days 14–16: p < 0.05; Days 17, 18: p < 0.01; Days 19, 20: p < 0.001
3
Wheel rotations different from Day Days 17, 18: p < 0.05; Days 19, 20: p < 0.001
4
Wheel rotations different from Day Day 17: p < 0.01; Day 18, p < 0.05; Days 19, 20: p < 0.001
5
Wheel rotations different from Day Day 17: p < 0.05; Day 19: p < 0.01; Day 20: p < 0.001
6
Wheel rotations different from Day Day 19: p < 0.05; Day 20, p < 0.001
7
Wheel rotations different from Day Day 20: p < 0.01
8
Wheel rotations different from Day Day 20: p < 0.05
9

1.3.3. Bdnf Exon-Specific Methylation

Separate ANOVAs analyzing the effects of infant condition, adult condition, and sex were conducted for Bdnf exons I, IV, and IX.

Exon I.

A three-way ANOVA revealed no effects of infant condition [F(2, 106) = 0.5011, p = 0.607], adult condition [F(1, 106) = 1.879, p = 0.1734], sex [F(1, 106) = 0.3323, p = 0.5655], or statistical interactions (F = 0.07–0.34, p = 0.55–0.92) . Normal care and cross-foster conditions were collapsed into one nurturing care condition for power. A three-way ANOVA again revealed no effects of infant condition [F(1, 109) = 0.1522, p = 0.6972], adult condition [F(1, 109) = 2.858, p = 0.0938], sex [F(1, 109) = 0.2262, p = 0.6353], or statistical interactions (F = 0.05–0.90, p = 0.34–0.81). We next collapsed across sex to run a two-way ANOVA (Figure 4A), which revealed no effect of infant [F (1, 114) = 0.3892, p = 0.5340], adult condition [F(1, 114) = 2.045, p = 0.1555], or interaction [F(1, 114) = 0.1926, p = 0.6616].

Figure 4. Locus-specific Methylation.

Locus-specific Bdnf I (A), Bdnf IV (B), and Bdnf IX (C) methylation in the whole PFC of adult rats after exposure to ELS or control rearing and voluntary exercise or standard housing. (A) No differences were detected for Bdnf exon I. (B) Female wheel-running rats showed less Bdnf IV gene methylation compared to female standard housed rats and male wheel-running rats (B). (C) Wheel-running rats showed less Bdnf IX gene methylation compared to standard housed rats, and wheel-running rats reared in the maltreatment condition also showed less methylation compared to maltreated rats in the standard housing condition (C).*p < 0.05; error bars represent SEM. N = (A) 20–39 subjects per group; (B) 28–31 per group; (C) 20–39 per group.

Exon IV.

A three-way ANOVA revealed no main effect of infant condition [F(2, 104) = 0.4764, p = 0.6224], adult condition [F(1, 104) = 1.440, p = 0.2329], or sex [F(1, 104) = 0.5235, p = 0.4710]; however, a significant sex by adult condition interaction was detected [F(1, 104) = 6.862, p = 0.0101] (all other statistical interactions were not significant [F = 0.003–1.06, p = 0.34–0.99]). To probe this interaction, we collapsed across infant condition, as it was not significant, and ran a two-way ANOVA (Figure 4B) where we detected a significant sex by adult condition interaction [F(1, 111) = 8.970, p = 0.0034], and no main effects of sex [F(1, 111 = 0.9125, p = 0.3415) or adult condition [F(1, 111) = 2.164, p = 0.1441]. Sidak’s post hoc testing revealed female wheel runners have lower methylation compared to female standard housed rats (p = 0.013) and female wheel runners also have lower methylation levels compared to male wheel runners (p = 0.032). Given this sex difference, male and female analyses were run separately for the exon IV CG site-specific ANOVAs.

Exon IX.

A three-way ANOVA revealed no effects of infant condition [F(2, 107) = 0.1874, p = 0.8294], adult condition [F(1, 107) = 2.940, p = 0.0893], sex [F(1, 107) = 0.02920, p = 0.8646], or statistical interactions (F = 0.80–2.27, p = 0.10–0.44). We collapsed normal maternal and cross-foster care, creating a nurturing care group and re-ran the three-way ANOVA where we detected a significant main effect of adult condition [F(1, 111) = 5.465, p = 0.0212] and a significant infant by adult condition interaction [F(1, 111) = 4.871, p = 0.0294] that washed out in post hoc (Sidak’s). For a two-way ANOVA (Figure 4C), we kept normal and cross foster care conditions collapsed and sex to probe the statistical interaction, where we again detected a significant main effect of adult condition [F(1, 115) = 5.437, p = 0.0215],a significant infant by adult condition interaction [F(1, 115) = 5.150, p = 0.0251], and no main effect of infant condition [F(1, 115) = 0.0040, p = 0.9494]. Sidak’s post hoc test revealed maltreatment subjects exposed to wheel running have lower methylation compared to standard housed maltreatment subjects (p = 0.031).

1.3.4. Bdnf CG Site-Specific Methylation

Separate ANOVAs were conducted per CG site for Bdnf exons I, IV, and IX to determine CG-site specific patterns in methylation. All analyses began as three-way ANOVAs (sex x infant x adult condition) and were eventually collapsed across nurturing care and sex (with the exception of exon IV) following the same procedures described above for the whole exon level analyses. Results are described in Table 3 and Figure 5. Statistically significant interactions probed in post hoc using Sidak’s multiple comparison correction. In summary, we did not find significant uniform changes across all CG sites located within each targeted Bdnf locus, but rather a few changes at specific CG sites that appear to be driving the methylation changes observed at the whole-exon level. Unsurprisingly we noted largely no effects on exon I CG-specific methylation, matching up with the null finding at the whole-exon level. For exon IX, we tracked the whole-exon level interaction between maltreatment and nurturing care wheel-runners to significant methylation changes at CG sites 10 and 11, and also noted a methylation change by wheel-running at CG site 5. Taken together, these findings help explain the exercise main effect and interaction noted at the Bdnf IX whole exon level. For Bdnf exon IV, we noted a decrease in methylation by exercise at multiple CG sites in the females-only analysis but not the males-only, supporting the sex difference noted at this Bdnf locus.

Table 3. CpG-Site Specific Methylation Results.

Bdnf Exon I CpG-Site Specific Methylation
Two-way ANOVA	F Statistic	P value
CG Site 5: SH x WR	F(1, 108) = 4.205	*p = 0.042
Bdnf Exon IV CG-Site Specific Methylation
Two-way ANOVA	F Statistic	P value
(Females) CG site 7: SH x WR	F(1, 52) = 13.45	***p < 0.001
(Females) CG site 10: SH x WR	F(1, 52) = 5.2	*p = 0.026
(Females) CG site 10: Nurturing Care: SH x WR	F(1, 52) = 5.51, p = 0.023	*p < 0.01 (Sidak correction)
(Females) CG site 10: Nurturing Care SH x Maltreatment	F(1, 52) = 5.51, p = 0.023	*p = 0.04 (Sidak correction)
(Females) CG site 11: Nurturing Care: SH x WR	F(1, 52) = 9.159, p = 0.003	**p = 0.001 (Sidak correction)
(Females) CG site 12: SH x WR	F(1, 50) = 4.049	*p = 0.049
(Males) CG site 5: SH x WR	F(1, 55) = 4.906	*p = 0.031
(Males) CG site 5: Nurturing Care x Maltreatment	F(1, 55) = 4.742	*p = .033
(Males) CG site 7: SH x WR	F(1, 56) = 8.045	**p = 0.006
Bdnf Exon IX CG-Site Specific Methylation
Two-way ANOVA	F Statistic	P value
CG Site 5: SH x WR	F(1, 112) = 4.603	*p = 0.034
CG Site 10: Maltreatment: SH x WR	F(1, 110) = 5.53, p = 0.021	*p 0.045 (Sidak correction)
CG Site 11: Maltreatment: SH x WR	F(1, 111) = 4.693, p = 0.032	*p = 0.032 (Sidak correction)

SH = Standard housed; WR = Wheel running

Figure 5. CpG site-specific Methylation.

CpG-site specific methylation analysis of Bdnf exons I (A), IX (B), and IV (females C; males D). (A) Wheel-running reduced Bdnf I methylation at CpG site 5 compared to standard housed subjects. (B) wheel-running reduced Bdnf IX methylation at CpG site 5 compared to standard housed subjects and at sites 10 and 11 in the maltreatment wheel-running group compared to maltreatment standard housed subjects. (C) In female subjects, wheel-running reduced methylation at sites 7, 10, and 12 compared to standard housed subjects. Methylation was also reduced at site 10 compared to nurturing care standard housed rats and at site 11 in the nurturing care wheel-running rats. (D) In male subjects, wheel-running increased methylation at sites 5 and 7. Methylation was also higher in the nurturing care groups compared to maltreatment at site 5. SH = standard housed; WR = wheel-running. *p < 0.05 vs. maltreatment WR, ^p < 0.05 SH vs. WR, &p < 0.05 vs. nurturing care SH, #p < 0.05 nurturing care vs. maltreatment; error bars represent SEM. N = (A) 18–38 subjects per group; (B) 18–39 per group; (C) 9–19 per group; (D) 9–21 per group.

1.3.5. Bdnf Exon IX Expression

A three-way ANOVA revealed no main effect of infant condition [F(2, 91) = 0.3063, p = 0.7369] or sex [F(1, 91) = 1.217, p = 0.2729], but a marginally significant effect of adult condition [F(1, 91) = 3.940, p = 0.0502]. This effect became significant (Figure 6) when we collapsed across nurturing care [F(1, 96) = 4.240, p = 0.042] and as well as when we collapsed across infant condition for a two-way ANOVA [F(1, 100) = 5.965, p = 0.016], such that wheel running subjects had significantly higher Bdnf gene expression compared to standard housed subjects. We did not detect a sex difference, however the effect of wheel running visually appears to be driven more so by the male subjects [F(1, 100) = 2.770, p = 0.099]

Figure 6. Bdnf IX Gene Expression.

Total Bdnf mRNA (Bdnf IX) gene expression in the PFC. Wheel-running increased gene expression compared to standard housed rats, regardless of infant condition or sex. *p < 0.05 main effect of standard housed vs wheel running; error bars represent SEM. N = 23–28 subjects per group.

1.4. Discussion

1.4.1. Summary

Here we replicated previous results from our lab showing that dams in an impoverished environment engage in more aversive behaviors towards the pups in their care than they do nurturing behaviors compared to dams in resource-rich environments (Roth et al., 2009; Keller et al., 2019; Collins et al., 2022; Duffy & Roth, 2020; Blaze & Roth, 2017; Doherty et al., 2019). We further showed that exercise alters the epigenetic landscape of the Bdnf gene in an exon- and sex-specific manner within the adult PFC. Specifically, we showed exercise largely has no effect on exon I methylation, suggesting this locus may not be involved in exercise-induced Bdnf gene regulation, but that exercise does decrease methylation at exons IV and IX, with changes at exon IV and IX being particularly influenced by sex and the early life environment. We also noted a sex difference in wheel running behavior that is a replication in the literature, further indicating that female rodents are more active than males (Bartling et al., 2017; Eikelboom & Mills, 1988; Milbocker et al., 2022). This sex difference may explain significant differences in our methylation data, in that female wheel-runners showed less methylation at Bdnf exon IV compared to male wheel-runners and female controls.

1.4.2. Effects of Maltreatment on Bdnf

Interestingly, we did not observe a main effect of maltreatment on Bdnf gene expression or methylation as previously described (Roth et al., 2009; Duffy & Roth, 2020; Collins et al., 2022). This could be explained by differences in specific behaviors exhibited by the maltreatment dams. In the present study, dams in the maltreatment condition exhibited more active avoidance towards the pups than previously observed (Table 4; Roth et al., 2009; Collins et al., 2022; Doherty et al., 2019; Doherty et al., 2017).

Table 4. Active Avoidance Comparison.

Active Avoid F(5, 73) = 8.98, p < 0.001 (average occurrence per cohort)
Current project vs.   Collins et al., 2022	Mean diff = 1.90 95% CI (0.92, 2.88)	***p < 0.001
Current project vs.   Duffy & Roth, 2022	Mean diff = 1.69 95% CI (0.27, 3.11)	*p = 0.01
Current project vs.   Doherty et al., 2019	Mean diff = 1.59 95% CI (0.37, 2.81)	**p = 0.003
Current project vs.   Phillips, 2019	Mean diff = 2.02 95% CI (0.76, 3.28)	***p < 0.001
Current project vs.   Blaze et al., 2015	Mean diff = 2.42 95% CI (1.28, 3.57)	***p < 0.001

Results of one-way ANOVA of the average occurrence of active avoidance in the maltreatment group compared to previous work.

This could mean that pups in the maltreatment condition were exposed to early life adversity more similar to animal models of neglect as opposed to fragmented or harsh infant-caregiver interactions. Typically in models of early life neglect, the experimental manipulation occurs for longer periods of time, often 3 hours daily for around 13 days total, suggesting that the stress exposure in our study may have been particularly mild (Campbell et al., 2022). This is important to note as the outcomes of early life adversity exposure are known to differ depending on the type of stress experienced. For example, Zhang and colleagues (2020) compared the effects of mild and severe maternal separation stress (MS) on BDNF protein expression in the PFC by exposing rats to 15 or 180 minutes of MS daily from PNs 1–21. They found that rats exposed to just 15 minutes of daily separation did not show a decrease in BDNF protein expression in the PFC, however, rats exposed to 180 minutes daily had lower BDNF expression in the PFC compared to both control subjects and subjects exposed to the mild MS. Similar findings have also been reported by labs comparing 15 and 180 minute daily MS on BDNF protein (Lippmann et al., 2007) and gene expression (Bian et al., 2015) in the hippocampus. When looking at behavioral outcomes, these labs have also shown that MS180-exposed subjects exhibit anxiety-(Zhang et al., 2020b; Lippman et al., 2007) and depressive- (Bian et al., 2015) phenotypes, whereas MS15-exposed subjects do not. Additionally, a recent study by Demaestri and colleagues (2022) also showed exposure to maternal separation stress (MS) or limited bedding and nesting (LBN) had differing effects on expression of corticotropin releasing hormone (Crh)-related genes in the amygdala. Specifically, they showed MS-exposed males had increased CrhR1 expression at PN8 compared to controls and at PN16 compared to both LBN-exposed and control males, whereas LBN-exposed females showed increased CrhR1 compared to control females at PN16. Additionally, LBN-exposed females showed increased CrhBP compared to controls with no change in MS females at PN16 and MS-exposed males showed increased CrhBP compared to LBN-exposed males and at PN21 compared to both control males and LBN-exposed males. Previous studies have also reported depressive-like behaviors on the forced-swim test in rats reared in the scarcity-adversity environment but not the LBN environment (Walker et al., 2017). Taken together, these findings suggest the type of stress experienced during early development can differentially affect gene expression in a sex and time-course specific manner, and should be taken into account when interpreting and comparing results from other studies.

Variations in outcomes have also been noted between studies, even when the same model is utilized. For example, some studies report an increase in basal corticosterone levels in infancy following LBN rearing while others indicate no change or even a significant reduction (Walker et al., 2017). These discrepancies could be attributed to subtle differences in behavior exhibited either by the control or stressed dams, given that natural variations in maternal licking and grooming influences play behaviors in juvenile rats (Parent & Meaney, 2008), as well as social and anxiety-like behaviors in adult rats (Starr-Phillips & Beery, 2014).

1.4.3. Exercise x Stress Interactions on Bdnf

Exercise differently affected Bdnf methylation based on early life experiences. At Bdnf exon IX, we noted that exercise decreased methylation within the maltreatment group at the whole exon and CG-site specific levels. Previous work indicates that the maltreated brain is more vulnerable to the negative effects of experiencing more stress later in life (Deng et al., 2023; Pena et al., 2017; Mayo et al., 2017), and while we agree with this sentiment, our data may indicate the maltreated brain is also more receptive to interventions meant to be neuroprotective. In fact this idea that the maltreated and normally developing brain responds differently to therapeutic interventions has support from other data. For example, A recent study by González-Pardo et al. (2019) analyzed regional brain volume in young adult male rats exposed to maternal separation (MS; or normal rearing) and environmental enrichment (or standard housing), which included running wheels, during adolescence. They showed MS subjects exposed to environmental enrichment had a larger ventral hippocampal volume whereas normal reared subjects exposed to environmental enrichment had significantly smaller ventral hippocampal volumes. They further described larger right dorsal hippocampal volume in MS rats exposed to environmental enrichment compared to normal reared subjects also exposed to environmental enrichment. Work from our lab has revealed a similar phenomenon in behavior data when studying the transgenerational effects of early life stress. Previously, our lab showed that female pups reared in our model of early adversity engage in more aversive behaviors towards their own pups later in life (Keller et al., 2019). This maladaptive behavioral outcome can be prevented by treating the dams with a methylation inhibitor, however, dams who had been exposed to normal rearing instead showed a marginally significant increase in aversive behaviors towards their pups (p = 0.051; Keller et al., 2019). Together, these results indicate that the normal reared and stressed brain respond differently to interventions later in life, specifically suggesting that interventions may be more effective for the more vulnerable versus resilient brain.

1.4.4. Early Life Stress & Exercise: An Emerging Field

To our knowledge, this is the first study investigating the use of an adult-only exercise intervention in a model of early life stress (Campbell et al., 2020). Consistent with previous literature in the early life stress and exercise field, we found that exercise significantly upregulated Bdnf expression (Maniam & Morris, 2010; Marais et al., 2009; Hakansson et al., 2017) and further related this finding to decreased exon-specific methylation (Boschen et al., 2017; Voisey et al., 2019) as one molecular mechanism facilitating exercise-induced changes in Bdnf expression. However, none of the previous studies in the field implemented voluntary exercise solely during young adulthood, as in these studies exercise encompassed adolescence into early adulthood. This adds important novelty to our study, in that we showed the brain is receptive to an intervention known to be neuroprotective after the developmental plasticity window has closed (Milbocker et al., 2021). Previous studies also did not measure Bdnf regulation in the PFC, and in fact, the PFC is often left out of the exercise literature as a whole (Campbell et al., 2020; de Sousa Fernandes et al., 2020; Liu et al., 2019; da Costa Daniele et a., 2020). This study provides unique insights into how exercise may affect molecular activity into the PFC in a way that may have implications for behavior (Cefis et al., 2019). Future studies should also include behavioral measures to determine how changes in Bdnf epigenetic regulation may relate to potential behavioral outcomes.

1.4.5. Interpreting & Relating Bdnf Methylation to Gene Expression

Though we noted sex differences and infant care by exercise interactions in Bdnf exon- and CG-specific methylation, we did not note similar effects in Bdnf exon IX gene expression. It is possible that additional changes in methylation levels existed earlier in development due to adversity that affected gene expression, but that these changes washed out over development. It is also important to note that methylation could still render an effect on activity-dependent gene expression, which was not directly tested here. Nevertheless, in the current study we did ultimately find an inverse relationship in activity-dependent Bdnf regulation, in that exercise led to an increase in total Bdnf expression that coincides with decreases in methylation.

1.4.6. Takeaways & Future Directions

Notably, this study showed that 20 days of voluntary exercise is sufficient to alter methylation and expression of the Bdnf gene in the PFC of young adult rats. The effect of exercise on Bdnf regulation was reliant on early life experiences, as exercise altered methylation based on the early life caregiving experience. Future work should target additional brain regions, such as the cerebellum and hippocampus, and different exercise exposure timepoints to shed light on how exercise may interact with the developing brain. The current study did not investigate how exercise-induced upregulation of Bdnf in the PFC may affect behavior but future studies may consider relating changes in Bdnf to performance on tests of cognition, memory, and social behavior.

Highlights:

• Exercise increases Bdnf gene expression in the rodent Prefrontal Cortex (PFC)

• Bdnf gene methylation after exercise is influenced by early life maternal care in the PFC

• Exercise affects Bdnf gene methylation in an exon-specific manner