Early Onset Memory Deficit of WMI Rats Compared to Their Nearly Isogenic WLIs Is Reversed by Enriched Environment in Females

ABSTRACT

The underlying mechanisms of early onset memory deficit remain poorly understood. We tested the hypothesis that environmental enrichment (EE) can attenuate early‐onset cognitive decline in a novel genetic model, the Wistar Kyoto More Immobile (WMI) inbred rat strain, which manifests the risk factors of enhanced stress reactivity and depression‐like behavior compared to its nearly isogenic control, the Wistar Kyoto Less Immobile strain (WLI). Middle‐aged (12 months) WMI females exhibited dramatically diminished fear and spatial memory in the contextual fear conditioning and Morris Water Maze paradigms, respectively, compared to young females of both strains and to middle‐aged WLI females. Middle‐aged WMI males showed a lesser, but significant, age‐induced deficit. EE from 6 to 12 months of age completely reversed the memory deficits in middle‐aged WMI females and reversed age‐induced decreases in plasma levels of estradiol. RNA sequencing from female hippocampi revealed significant strain, age, and enrichment‐induced differentially expressed genes. Among these, solute carrier family 35, member A4 (Slc35a4) and potassium inwardly rectifying channel, subfamily J, member 2 (Kcnj2) were confirmed to show hippocampal expression changes parallel to that of memory in the WMI. These genes have critical roles in the integrated stress response, cellular metabolism, and the effects of stress on neurovascular coupling, respectively. Pathway analyses revealed the involvement of oxidative phosphorylation and mitochondrial dysfunction in the hippocampal processes of aging and EE‐induced reversal. These findings underscore the critical involvement of molecular stress responses in early‐onset memory decline and suggest potential therapeutic targets for age‐related cognitive impairment.

Enriched environment housing reverses memory deficits in middle‐aged WMI females and attenuates deficits in WMI males. Plasma levels of estradiol and hippocampal gene expression followed the same pattern as memory in WMI females following environmental enrichment and may be involved in reversing the effects of aging on cognitive decline.

1. Introduction

The pathological processes underlying dementia, including Alzheimer's disease (AD), begin long before clinical onset and last approximately 15–20 years. Therefore, accessing modifiable risk factors that can delay the onset of dementia is critically important [1, 2]. Stress is among these modifiable risk factors. Stress has been associated with dementia [3, 4, 5], implicated in AD progression, and shown to lower the age of onset [3, 6, 7]. Stress‐related disorders like depression have been linked to a state of accelerated aging and subsequent pathological cognitive aging [8, 9, 10, 11, 12]. Major depressive disorder (MDD) is a risk factor for mild cognitive impairment (MCI) and dementia, with a prevalence rate of 32% in MCI and up to 37% in dementia [13]. Even moderate depression increases the risk of progression from healthy to MCI and to dementia [14]. Studying risk factors for cognitive decline in humans is challenging due to the need to account for genetic, lifestyle, physical, and mental health factors [15]. Deficits in hippocampal cellular function with aging have been shown both in humans and rodents [16]. As these deficits start before clinical signs of age‐induced cognitive decline in humans, preclinical research models can be useful for the identification of factors influencing early cognitive decline. However, few experimental models of early cognitive decline exist. Two well‐characterized models include middle‐aged Fisher 344 × Brown Norway hybrids and insulin‐resistant rats [17, 18]. Most animal studies show age‐induced memory loss in aged animals [16, 19, 20, 21, 22]. Risk factors of cognitive aging such as mid‐life or later‐life stress are also studied in aged animals [23, 24, 25, 26]. Stress has variable effects on cognitive aging [27], which may be due to genetic variation in stress reactivity and neurocognitive processes.

Genetic factors are known to contribute to the variation in the risk of AD [28], and epigenetic mechanisms are key players in dementia‐associated disorders [29, 30, 31]. The accumulation of epigenetic alterations over the lifespan is suggested to be a mechanism that regulates the transcriptome of aged brain cells, affecting learning and memory and its pathological states [30, 32, 33]. Studies of human and rodent hippocampal transcriptomes identified specific changes associated with aging‐induced cognitive decline [34, 35, 36]. The study by Pereira et al. [36] reveals that most transcriptional changes occur between middle‐aged and aged rats.

Previously, we demonstrated that female rats from a genetic model of enhanced depression‐like behavior and stress‐hyperreactivity exhibit premature cognitive decline in middle age [37]. This genetic model incorporates both stress and depression—two risk factors of cognitive decline—and was developed from the Wistar Kyoto (WKY) rat strain that demonstrates high levels of depression‐ and anxiety‐like behaviors [38, 39, 40, 41]. The WKY strain presents traits that mirror several symptoms of major depression [41, 42, 43]. Chronic treatments with antidepressants [44, 45], electroshock administration (analogous to electroconvulsive therapy) [46] and deep‐brain stimulation [46] can all reverse the depression‐like behaviors of WKYs. We exploited the fact that the WKY strain was not completely inbred, since it was distributed to different suppliers between the 12–17th generation. Bidirectional selective breeding was conducted using immobility behavior in the forced swim test (FST) as a functional selector [47]. With subsequent filial breeding, two fully inbred strains were generated: the WMI and the WLI strains [48, 49]. Compared to the WLI strain, WMI rats show enhanced responses to acute and chronic restraint stress [47], exaggerated effects of adolescent stress on adult animals [50], increases in affective behaviors [51, 52], elevated drug intake [53, 54], heightened stress‐enhanced fear learning [53, 55], and early onset memory decline [37].

One of the protective factors for cognitive decline that has been studied is environmental enrichment (EE). There is evidence that a healthy lifestyle that involves activities linked to EE can lower the risk for age‐associated memory impairment in humans [56]. Cognitive enrichment is thought to be beneficial in delaying or lessening age‐related neuropathology [57]. In animal studies, EE has been shown to reverse cognitive deficits associated with aging [58, 59, 60, 61]. EE attenuates the reinforcing effects of addictive substances and produces an antidepressant‐like effect [62, 63, 64, 65]. The effects of EE on learning and memory are thought to be caused by mitigating the effects of stress [66]. Thus, EE may act by reducing potential risk factors of cognitive decline. The WMI strain seems to accumulate potential risk factors such as enhanced depression‐like behavior and stress‐reactivity; thus, it is expected that EE could reverse or attenuate the previously found early age‐induced memory decline in this strain. Additionally, sex differences in early age‐induced cognitive deficits and the potential of hormonal and hippocampal transcriptome changes associated with aging and EE were explored in this genetically defined and naturally occurring model of early onset cognitive loss.

2. Materials and Methods

2.1. Animals

Animals were maintained at Northwestern University Feinberg School of Medicine by the Center for Comparative Medicine. All procedures were approved by the Northwestern Institutional Animal Care and Use Committee. Animals were housed in temperature‐ and humidity‐controlled cages, in a 12‐h light–dark cycle, where lights turned on at 0600 h. Food and water were readily available. The animals used in the study were inbred male and female WLI/Eer and WMI/Eer rats from the 47th–51st generations.

At weaning, animals were divided into three groups: standard group housing until 6 months of age (6 M) or 12 months of age (12 M), and animals that were housed under standard conditions up to 6 M and then placed into an EE until behavioral testing started at 12 months of age (12 M + EE). The EE groups were housed in a larger‐thanthan‐standard cage (22 × 14.5 × 8 in.), with plastic toys, chimes, tunnels, and chewing toys. All cages were cleaned and changed twice a week, and clean toys were provided. The number of animals was 8–11 per group, strain, and sex.

Animals were not handled prior to the behavioral tests as handling may act as a stressor [67]. The three groups of animals were tested in CFC at 6 months of age (6 M) or 12 months of age (12 M and 12 M + EE). Two weeks after CFC, animals were observed in the Morris Water Maze (MWM) test. Twenty‐four hours after the last component of MWM, animals were euthanized by decapitation within 5 s of being removed from their home cage.

Brains and trunk blood samples were collected from study animals. Blood samples were collected in EDTA‐coated tubes (0.3 μL/0.5 mL whole blood, 0.5 M). The samples were then centrifuged at 4°C and 4000 RPM for 10 min, and the plasma was separated for storage at −80°C. Brains were collected in RNAlater (Invitrogen, Carlsbad, CA, USA), a solution that stabilizes and protects cellular RNA, and then stored at −80°C.

2.2. Contextual Fear Conditioning

Testing was performed as previously described [53, 55]. On day 1, rats were placed into a Technical & Scientific Equipment (TSE, Bad Homburg, Germany) automated fear conditioning apparatus containing a foot‐shock grid enclosed by a clear acrylic box. Three minutes of habituation was followed by three mild shocks (0.8 mA, 1 s per min) over a three‐minute period. Between animals, the chamber was cleaned using 75% ethanol to eliminate behavioral changes caused by odor. Twenty‐four hours later (day 2), the rats were placed in the same chamber for 3 min without any shocks and examined for contextual fear memory via freeze duration, total locomotion (distance traveled), and number of rears. These measures were obtained with a computerized infrared beam system (detection rate 10 Hz) for both day 1 and day 2 of the paradigm. Day 1 freeze duration, distance traveled, and rearing data was calculated as the sum of all three one‐min intervals following each of the three one‐second foot shocks. We chose to present the sum of measures based on three mins, rather than the measures after each shock, as the summed measure has higher reliability, and to enable comparison between day 1 and day 2 of CFC. Rats that did not respond to the initial shock were excluded from the study.

2.3. Morris Water Maze

After a two‐week rest period following CFC, the MWM test was carried out. The hidden platform version of the test was used as an assessment of spatial learning. MWM was conducted with four consecutive days of learning trials and 6 consecutive trials/day. Trials were recorded and animals were tracked using TSE Videomot 2 software (version 5.75), automatically collecting data for distance, time, and speed in all quadrants of the maze, or near the pool wall for thigmotaxis. A water‐filled circular tank (170 cm diameter; water depth 25 cm, temperature 21–22^oC) was used. Three visual cues (i.e., different large black symbols) were placed on the walls surrounding the maze, and the fourth reference point was a curtain. The platform remained in a constant position 1 cm below water level. Rats were placed in the water in a randomly determined quadrant of the tank. Each rat was allowed to swim either until reaching the platform or until 60 s had passed, at which point it was hand‐guided to the platform. Twenty‐four hours following the last trial on day 4, the platform was removed, and a 60 s probe trial was conducted, for which animals were placed in the same quadrant opposite the missing platform. Rats were also subjected to a visual platform test on day 5 using a flag attached to the platform. All animals had to locate the visible platform to remain in the study.

Unfortunately, the TSE Videomot software stopped working after 2/3 of the animals had passed through the protocol. Therefore, distance and speed measures could not be collected for all animals, and these measures were not analyzed. The latency to reach the platform measure was scored manually to assure uniformity and compared to the computer‐collected measure. The scoring was conducted by an observer blind to strain and group. We found a very high correlation between hand scoring and computer scoring, and therefore the hand‐scored latency to reach the platform was used for all groups. Floating or immobility was also scored by an unbiased observer.

2.4. Plasma Hormone Assays

Plasma corticosterone (CORT), testosterone (T), and estradiol (E2) levels were measured by commercially available competitive ELISA kits (Corticosterone Competitive ELISA kit, ThermoFisher, USA; Testosterone ELISA kit, Biomatik, Ontario, Canada; 17‐Beta Estradiol ELISA kit, Abcam, Cambridge, United Kingdom) according to the manufacturer's protocol. The sensitivities of the assays were as follows: CORT, 18.6 pg/mL; T, 49.4 pg/mL; E2, 8.68 pg/mL. Plasma samples were diluted to a 1:1000 ratio for CORT, 1:4 for testosterone, and undiluted for estradiol. The ELISAs were performed in duplicates. ELISA plates were read on the FLUOstar Omega Microplate Reader (BMG Labtech, Ortenberg, Germany). Standard curves were generated using linear regression on log‐transformed concentration and absorbance data. The resulting equation was then used to calculate the concentration of the hormones in the samples based on absorbance measures.

2.5. Brain Dissection and RNA Extraction

Brains were thawed on ice and dissected. Hippocampi were dissected on a brain matrix and immediately stored in RNAlater (Invitrogen, Carlsbad, CA) at −80°C. Paxinos rat brain atlas coordinates were used for dorsal hippocampus (AP −2.12 to −4.16, ML 0–5.0, DV 5.40–7.60) and ventral hippocampus (AP −4.20 to −6.00, ML 0–5.00, DV 5.40–7.60). These regions were dissected separately as we observed that the precision of the dissection was increased using this method. However, both dorsal and ventral regions were combined after dissection to assure enough RNA for both RNA‐seq and qPCR experiments.

The hippocampal tissue was divided into left and right hemispheres (to increase the efficiency of RNA extraction) and homogenized using TRI Reagent (Sigma‐Aldrich, Saint Louis, MO) and a handheld tissue homogenizer (Kinetica Polytronic). Total RNA was isolated from samples using the Direct‐zol RNA MiniPrep Plus kit (Zymo Research, Irvine, CA) according to the manufacturer's instructions. RNA quality was determined using the Nanodrop instrument (Thermo Scientific). Accepted RNA qualities ranged from 1.8 to 2.2 for the 260/280 and 260/230 ratios. Total hippocampal RNA samples were stored at–80°C.

2.6. RNA Sequencing

Only female samples were included for RNA‐seq because EE completely reversed early memory loss in WMI females. Limiting the study to one sex also reduced experimental costs. Total RNA hippocampal samples (2 strains × 3 conditions × 6 biological replicates; all females) were shipped to Novogene Corporation Inc. (Sacramento, CA) for quality control assessment, library preparation, and sequencing. All samples passed the quality control test prior to library preparation. Approximately 50 million paired‐end 150 bp reads (range: 42–84 million, mean: 53.6 million) were obtained for each sample. Salmon [68] was used for the alignment of resulting fastq files to the rat transcriptome (Rattus_norvegicus.mRatBN7.2.107.gtf) and quantification of transcript abundance.

2.7. Reverse Transcription and Quantitative Polymerase Chain Reaction

Reverse transcription was done using the Super Script VILO Master Mix (Invitrogen). Hippocampal RNA (1.0 μg) was used according to the manufacturer's protocol. qPCR was performed with 5 ng cDNA, specific primer pairs, and SYBR Green Master Mix (Applied Biosystems, Foster City, CA, USA), using the QuantStudio 6 Flex Real‐Time PCR System (Applied Biosystems). Sequences of primers designed for the target genes are listed in Table S1. Triplicate reactions were performed for each cDNA sample and analyzed using QuantStudio Software (Applied Biosystems). Relative quantification, or RQ values, of target gene expression were determined relative to Gapdh and a general cDNA calibrator, acquired from a young adult WLI male, using the 2^−ΔΔCt method.

2.8. Data Analysis

2.8.1. Behavioral, Hormone and Gene Expression (qPCR) Measures

A three‐way ANOVA was conducted to analyze the contextual fear conditioning test, plasma CORT levels, and RQ in qPCR across strain (WLI, WMI), age (6 M, 12 M, 12 M + EE), and sex (male, female) using GraphPad Prism version 10.3.1 (GraphPad Software, La Jolla, CA, USA). These analyses were also applied to the probe trial and immobility in the MWM. Plasma T and E2 levels were analyzed by two‐way ANOVA in males and females, respectively. All data are presented as mean ± standard error of the mean (SEM). False Discovery Rate (FDR) post hoc analyses were conducted following significant ANOVAs using the two‐stage step‐up method of Benjamini et al. [69].

2.8.2. Morris Water Maze

Latency data from the Morris Water Maze are positively skewed (i.e., most animals find the platform quickly, while a few take longer) and exhibit greater variance as the mean latency increases. To address these properties of our data, and to evaluate the multiplicative effects of covariates in a more meaningful way, we leveraged the Gamma Generalized Linear Model (GLM) with a log link function to examine the effects of sex, group, strain, and day or trial on day 4 on the latency to reach the platform time. For the primary GLM examining average time across days, the model was fit with 411 residual degrees of freedom (df) and a sample size (N) of 426. The GLM across trials on day 4 was fit with 625 df and N of 641. In this model, exponentiated estimates (Exp_Estimate) represent the multiplicative change in mean latency per unit change in the predictor and are reported with confidence intervals (Lower_CI, Upper_CI) in the results. The primary model included all two‐ and three‐way interactions between sex, group, and strain, as well as the main effect of day, to assess the significance of primary effects and interactions. The same methodology was applied to the data for Day 4 to investigate trial‐specific effects. All analyses were performed in R [70] using the glm function for model fitting. False Discovery Rate (FDR) post hoc analyses were performed as described above for day and trial and are indicated in Figure 2. Additionally, the probe trial and floating/immobility data were analyzed by three‐way ANOVA, followed by FDR post hoc analyses.

FIGURE 2.

Daily mean latency to reach the platform, and mean latency to reach the platform during each of the six trials on Day 4 in the MWM. (A) Six daily trials were carried out over four consecutive days. 12 M WMIs took significantly longer to reach the platform especially at Day 4, compared to 12 M WLIs and both 6 M and 12 M + EE WMIs. (B) 12 M WMI females showed increased mean latency to reach the platform at trials 5 and 6 compared to both 6 M and 12 M + EE WMI females, while mean latency of 12 M WMI males did not differ from 12 M + EE WMI males. Values are mean +/−SEM for N = 7–10/sex/strain/age. Data were analyzed by Gamma Generalized Linear Model (GLM) with a log link function to examine the effects of sex, group, strain, and day or trial on day 4 on the latency to reach the platform. Post hoc analyses were conducted by FDR. Significance corrected for multiple comparison are indicated by *, a or b, while p values not corrected for multiple comparison are marked by #, a' or b'. *q < 0.05 or #p < 0.05, ##p < 0.01 shows significance between WLIs and WMIs of the same age group (6 M, 12 M or 12 M + EE) and same day or trial in the MWM; ^a q < 0.05, ^aa q < 0.01 or ^a' p < 0.05, ^a'a' p < 0.01 represent significance between 12 M vs. 6 M of the same strain (WLI or WMI) and same day or trial; ^b q < 0.05, ^bb q < 0.01 or ^b' p < 0.05, ^b'b' p < 0.01 identify significant differences between 12 M and 12 M + EE of the same WLI or WMI strain and same day or trial. Number of animals as in Figure 1.

2.8.3. RNAseq Analyses

Although all RNA samples passed quality testing before library preparation, group differences in RNA integrity numbers (RIN) appeared to affect RNA‐seq outcomes. Specifically, WMI and EE samples had a higher proportion of lower RIN values. Due to the length of the study, samples were collected at different times (e.g., 12 M and 12 M + EE were the last groups to have tissues harvested) and RNA was isolated from samples as they became available and included in 13 different extraction batches from 7 February through 4 August. Thus, counterbalancing of RNA extraction groups was not possible. Unfortunately, three batches in May and June included 85% of the samples with RIN values below 7. The reason for the low RIN numbers in these batches is not clear as all samples were processed by the same personnel. However, these three batches happened to contain 2/6 WMI 12 M, 5/6 WMI 12 M + EE, and 4/6 WLI 12 M samples. The batch effect impacts a moderate number of WMI and WLI 12 M samples and nearly all WMI 12 M + EE samples. The batch effect does not invalidate the results of the study but does limit the power to detect differentially expressed genes in contrasts that contain these comparisons. Therefore, RNA‐seq data analyses were performed with and without the 12 M + EE results. Two subsets of RNAseq data were analyzed that included: (1) the full data set, and (2) a data set excluding the 12 M + EE samples from both strains, which exhibited lower RIN values. For subset 1, the raw count data was normalized using the estimate SizeFactors function in DESeq2 to correct for differences in library size and contrasts (t‐tests) between 6 M, 12 M and 12 M + EE were performed for each strain. A significance threshold of p < 0.001 was applied to identify differentially expressed transcripts for each contrast. For subset 2, normalization and differential expression was performed in DESeq2 and contrasts between 6 M and 12 M were performed for each strain (Tables S2 and S3). In addition, normalized count data was also used to perform a two‐way ANOVA (Strain: two levels and Age: two levels) to identify differentially expressed transcripts with a main effect of strain, age, or an interaction effect (Tables S2 andS3). The significance threshold was set at a 10% false discovery rate with a minimum fold change of 1.2. Significant differentially expressed genes (DEGs) from subsets 1 and 2 were used for ontology enrichment analyses. Pathway and ontology enrichment analyses were performed using Qiagen Ingenuity Pathway Analysis. Please note that we entered the results of RNA‐seq into the IPA as human genes, rather than rodent, for better coverage, and that is reflected in the capitalized name of genes.

3. Results

3.1. Attenuation of Fear Memory Deficits in Aged Female WMI Following Environmental Enrichment

Contextual fear conditioning in young (6 M), middle‐aged (12 M), and middle‐aged enriched environment (12 M + EE) WLI and WMI rats confirmed and extended our previous findings [37]. While WLIs showed no age‐related differences in the freezing response on day 1 of CFC, 12 M WMIs exhibited significantly lower freezing compared to both their 6 M counterparts and 12 M WMIs that underwent 6 months of environmental enrichment (Figure 1; age, F[2110] = 13.41, p < 0.001; age × strain, F[2110] = 14.35, p < 0.001). There was also a significant sex and strain difference in the effect of aging on fear conditioning on day 1. WMI females at 12 M showed a greater decrease (approx. 70%) in freezing behavior following foot‐shock relative to 6 M females, while WMI males at 12 M had a smaller (approx. 35%) decrease compared to 6 M males (Figure 1; sex × strain, F[1110] = 5.82, p < 0.05). No interaction effect was observed in WLIs. No significant main effects of sex and strain were observed during fear conditioning.

FIGURE 1.

Fear conditioning and fear memory as measured by freeze duration during day 1 and day 2 of CFC. Day 1. The overall freeze duration after the three 1 s foot shock is significantly lower in middle aged (12 M) WMIs compared to young (6 M) WMIs, and that of middle‐aged WMIs after 6 months in EE (12 M + EE). Day 2. WMI females show decreased fear memory at 12 M that is reversed by EE, while 12 M + EE WMI males exhibit fear memory that is not different from 12 M WMI males. N = 8–13/strain/sex/age. Data as mean ± SEM. Statistical differences were determined by three‐way ANOVA. Post hoc group comparisons were carried out by two‐stage linear set‐up procedure of Benjamini, Krieger, and Yekutieli following significant ANOVA *q < 0.05; **q < 0.01 corrected for multiple comparisons.

Similar to day 1, only 12 M WMIs demonstrated significantly attenuated freezing behavior on day 2 compared to 6 M WMIs (Figure 1; age, F[2103] = 24.21, p < 0.001; strain, F[1103] = 5.57, p < 0.05; age × strain, F[2103] = 8.97, p < 0.01). This age difference between 6 M and 12 M WMI females in conditioned fear memory on day 2 disappeared after 6 months of enriched environment in 12 M + EE WMI females. In contrast, the attenuated fear memory of middle‐aged WMI males was not completely reversed by EE (age × sex, F[2103] = 3.64, p < 0.05; sex × strain, F[2103] = 7.85, p < 0.01; Figure 1). Again, the fear memory attenuation was greater in 12 M WMI females, close to 80% of fear memory of 6 M WMI females, while this was about 50% in 12 M WMI males.

Distance traveled during day 1 of CFC did not mirror the freezing behavior, and neither did it parallel distance traveled on day 2 (Figure S1). The main effect of age and sex indicated generally lower locomotion of 12 M + EE animals regardless of strain, that females traveled greater distances than males, and that the difference in age effects was also modulated by sex (age, F[2,93] = 72.65, p < 0.001; sex, F[1,93] = 9.64, p < 0.01; age × sex, F[2,93] = 8.26, p < 0.001). The difference in distance traveled by strain was also regulated by age and sex, which was shown by WLI 12 M females having decreased locomotion in response to the foot‐shocks, while 12 M males showed an increase (strain, F[1,93] = 5.49, p < 0.05; age × strain, F[2,93] = 3.20, p < 0.05; age × sex × strain, F[2,93] = 4.04, p < 0.05). Of note, on day 1 of conditioning, both WLI and WMI 12 M + EE animals showed diminished distance traveled despite differences in freeze duration between strains. For example, no change in freeze duration was observed in 12 M + EE WLIs, and an increase in freeze duration was observed in 12 M + EE WMIs.

In contrast to day 1, day 2 distance traveled showed the inverse of freezing behavior: an internal confirmation of data validity, specifically in females. WMI 12 M females traveled significantly more than either 6 M or 12 M + EE WMIs, while there was no effect of age on WMI males and WLIs of either sex (age, F[2,92] = 18.67, p < 0.001; age × strain, F[2,92] = 16.79, p < 0.001; age × sex, F[2,92] = 3.78, p < 0.05; sex × strain, F[1,92] = 9.25, p < 0.01; Figure S1).

Rearing on day 1 and day 2 of CFC showed patterns not corresponding to either freezing or distance traveled. However, the results confirmed that rearing is not simply a measure of activity (Figure S2). Female 12 M WMIs reared dramatically less than either 6 M or 12 M + EE WMI females on day 1, with similar behavior observed in WLI females, but not in males of either strain (age, F[2105] = 4.50, p = 0.01; age × sex, F[2105] = 7.38, p < 0.01). Rearing on day 2 was substantially reduced compared to day 1 and showed an opposite pattern in WLIs and WMIs. Namely, WLI 12 + EE females and males reared more than their 6 M or 12 M counterparts, respectively, while WMI 12 + EE females reared less and males showed no differences by age (sex, F[1106] = 9.81, p < 0.01; age × strain, F[2106] = 3.29, p < 0.05).

3.2. Attenuation of Spatial Hippocampal Deficits in Aged Female WMI Following Environmental Enrichment

GLM with a log link function was used to examine the effects of sex, group, strain, and day or trial on day 4 on the latency to reach the platform in the Morris Water Maze. The model revealed a significant effect of sex (Exp_Estimate = 1.23, CI = [1.10, 1.38], p < 0.001), indicating that females had a 23% longer latency to find the platform compared to males. The model also showed a significant effect of strain (Exp_Estimate = 1.17, CI = [1.03, 1.33], p < 0.05), indicating that WMI animals had 17% longer latencies compared to WLI animals. A significant effect of group was also observed (Exp_Estimate = 1.26, CI = [1.10, 1.45], p < 0.01), with 12 M animals exhibiting a 26% longer mean latency to locate the platform than 6 M animals. Furthermore, latency significantly decreased across days (Day 2: Exp_Estimate = 0.88, 95% CI = [0.81, 0.95], p < 0.001; Day 3: Exp_Estimate = 0.77, 95% CI = [0.72, 0.83], p < 0.0001; Day 4: Exp_Estimate = 0.72, 95% CI = [0.67, 0.77], p < 0.0001). These estimates indicate that mean latency on Days 2, 3, and 4 was 12%, 23%, and 28% lower than Day 1, respectively. Further post hoc analyses were conducted to explore differential responses dependent on sex, strain, and group (Figure 2). Latency to discover the hidden platform is shown for strain and group stratified by sex for all days (Figure 2A). The largest differences in latency to find the platform emerged by day 4. 12 M WLI females and males had shorter latencies relative to WMIs of both sexes at day 4. Post hoc analysis further revealed significantly longer latencies in both female and male 12 M WMIs relative to 6 M and 12 M + EE WMIs by the fourth day of testing.

A second model was used to examine changes in latency across trials in the final day. Main effects revealed that later trials had lower mean latencies compared to the first trial. Compared to Trial 1, Trial 2 had a 19% lower mean latency (Exp_Estimate = 0.81, 95% CI [0.69, 0.95], p < 0.01), Trial 3 had a 24% lower mean latency (Exp_Estimate = 0.76, 95% CI = [0.65, 0.89], p < 0.001), Trial 4 had a 30% lower latency (Exp_Estimate = 0.70, 95% CI = [0.60, 0.82], p < 0.0001), Trial 5 had a 37% lower latency (Exp_Estimate = 0.63, 95% CI = [0.54, 0.74], p < 0.0001), and Trial 6 had a 39% lower mean latency (Exp_Estimate = 0.61, 95% CI = [0.52, 0.71], p < 0.0001). Further post hoc analyses were conducted in males and females separately to explore differences between group, strain, and trial on day 4 (Figure 2B). This analysis demonstrated a more robust effect of group and strain for later trials. Specifically, WMI females at 12 M showed increased mean latency to reach the platform at trials 5 and 6 compared to 6 M and 12 M + EE WMIs. In contrast, 6 M WLI males showed decreased mean latency to reach the platform at trials 5 and 6 compared to 12 M and 12 M + EE WLIs.

Since the WLI and WMI were selectively bred based on immobility in the FST, we determined that the deficit in recognizing the platform was not related to strain differences in floating (immobility) or in general swimming ability. Floating events during the six trials of the last day of MWM were analyzed; every 5 s without movement was a floating event. If the animal floated longer than 5 s a new event was scored. Data are shown in Figure S3. Floating was high during the first trial, and it decreased precipitously during the subsequent trials (females: trial, F[5155] = 75.34; p < 0.001; males are similar). There were no significant floating/immobility differences between the groups in females (strain, F = 0.52, NS; age, F = 1.69, NS; interactions, NS). In males, the only significant difference was the decreased floating of WLI at 12 M compared to both WLI at 6 M and WLI at 12 M + EE at the first trial (strain, age, NS; trial × age, F[5,90] = 10.65, p < 0.01). Therefore, we assume that there were no age‐induced physical or overall swimming disabilities in these animals that would have resulted in increased floating in the middle‐aged compared to the young groups.

Twenty‐four hours following the last trial on day 4, the platform was removed, and a 60 s probe trial was conducted with animals placed in the same quadrant opposite to the missing platform. The probe trial results showed that 12 M WMI females searched for the hidden platform significantly longer before reaching the right quadrant (strain × age, F[2,51] = 3.95, p = 0.05; sex × age, F[2,51] = 7.04, p = 0.01; Figure S4).

3.3. Reversal of Estradiol Levels in Aged WMI Females Following Environmental Enrichment

Plasma hormone levels were measured in the trunk blood (Figure 3). Plasma CORT levels differed significantly between strains, sexes, and by age (strain, F[1101] = 59.48, p < 0.001; sex, F[1101] = 11.18, p < 0.01; age, F[2101] = 38.88, p < 0.001; sex × age, F[2101] = 2.87, p = 0.06; strain × age, F[2101] = 19.12, p < 0.001; strain × sex, F[1101] = 7.07, p < 0.01; strain × sex × age, F[2101] = 2.88, p = 0.06; Figure 3A,B). Analyzing CORT levels of males and females separately revealed that age and strain had a significant effect on plasma CORT levels in females (strain, F[1,46] = 8.50, p < 0.01; age, F[2,46] = 26.87, p < 0.001; Figure 3A). Middle‐aged 12 M WMI females had significantly lower levels of CORT than young 6 M WMI females. Notably, 6 months of environmental enrichment increased plasma CORT levels in females of both strains to levels significantly higher than those in their 6 M counterparts. This increase was greater in WMI 12 M + EE females than in WLI 12 M + EE females. In males, only age had a significant effect on CORT levels (F[2,55] = 8.94, p < 0.001; Figure 3B). WMI 6 M males showed significantly higher CORT levels compared to WLI 6 M males, and EE increased CORT levels significantly higher than those of the young males.

FIGURE 3.

Plasma corticosterone (CORT), estradiol (E2) and testosterone (T) levels from trunk blood. (A) Plasma CORT levels were significantly elevated in middle aged females of both strains who spent 6 months in enriched environment (12 M + EE). (B) In contrast, no significant differences were found in males of either strain by age. (C) E2 levels were significantly decreased in 12 M WMI females compared to 6 M WMIs, but this decrease was reversed by EE in 12 M + EE WMI females. (D) T levels were significantly lower in 6 M WMI males compared to same‐age WLI males. Statistics as described in Figure 1. N = 8–13/sex/strain/age.

E2 levels did not differ between 6 M WLI and WMI females but significantly declined at 12 M in WMI females only (strain × age, F[2,52] = 7.60, p < 0.01; Figure 3C). Six months of EE prevented the decrease in E2 in WMI females. In contrast, plasma T levels were already lower in 6 M‐old WMI males relative to WLI males, with no further change due to age or EE (strain, F[2,45] = 9.18, p < 0.01; Figure 3D).

3.4. Strain Dependent Reversal of Aging Related Expression Changes in Hippocampus Following Environmental Enrichment

Since learning/memory measures and hormone levels changed more dramatically and meaningfully in females, hippocampal RNA sequencing was carried out in female WLIs and WMIs of all three groups.

There were 65 DEGs between 6 M WLI and 6 M WMI hippocampus (Table S2). There were substantially fewer DEGs between strains for the 12 M comparison, and of the 29 DEGs, only four showed greater expression in the WMI hippocampus compared to that of WLI (Table S3). We hypothesized that EE would not only reverse memory deficits in 12 M WMI females but also restore age‐related gene expression changes in the hippocampus. To test this hypothesis, we identified DEGs with altered expression between 6 M and 12 M that were reversed by EE in each strain (Table 1). In the WLI strain, using a criterion of significance of p < 0.0001, we identified 350 DEGs whose expression was decreased by age and increased by EE and 200 DEGs whose expression was increased by age and decreased by EE. For this exploratory analysis, we reduced the significance level to p < 0.001 for the WMI strain, since the RIN issues (see Methods) were primarily present in the WMIs. In the WMI strain, the expression of only 11 and 9 DEGs was decreased by age and increased by EE or increased by age and decreased by EE, respectively.

TABLE 1.

Transcripts with significant (p < 0.001) differences between 6 M, 12 M and 12 M + EE.

WLI		WMI
6 M > 12 M < 12 M + EE	6 M < 12 M > 12 M + EE	6 M > 12 M < 12 M + EE	6 M < 12 M > 12 M + EE
1110065P20Rk	AABR07028902.1	Ajm1	Armc3
4833439L19Rik	AABR07064719.2	Celf3	Aspg
AABR0701536.1	Abhd18	Coro2b	Bhlhe41
AABR07034639.1	AC105531.1	Cplane2	Gpnmb
AABR07056686.1	Acan	Gak	Kcnj2
AABR07071000.1	Acbd3	Nudt13	Klhl6
Abhd17a	Acsl3	Pdk1	Ly86
AC121415.1	Actr8	Slc35a4	Smoc2
AC094643.2	Adamts3	Smarcd3	Zc3h13
Agap2	Adgre1	Tgfbi
Agap3	Aff1	Tubb2b
Agpat1	Afg3l2
Akt1	Aqr
Aldh3b1	Astn2
Ankrd63	Atg2b
Arhgdia	Atg4c
Arhgef4	Bhlhe40
Arl2bp	Brinp3
Arl3	C1qa
Arl8a	C3
Arpc4	Cacng5
Arpc5l	Camsap1
Ascc1	Casc3
Asphd1	Ccbe1
Asrgl1	Cd53
Atat1	Cfh
Atf4	Clcn6
Atf5	Clpx
Atg4d	Cnot1
Atn1	Cog3
Atp5mc2	Copb1
Atp5mk	Cpeb1
Atp5pf	Cpne1
Atp6v1g1	Csf1r
B4galt2	Ctss
B9d1	Cx3cr1
Bag1	Cyp7b1
Bap1	Dag1
Bax	Dapk1
Bcas1	Ddx21
Bcl7b	Dhx15
Bola1	Dhx33
Borcs7	Dld
Bri3	Dlg3
Brsk1	Dlst
Bspry	D0jc18
Btbd2	Dock8
Cabp7	Dop1b
Cacnb3	Dusp4
Cacng7	Dync1h1
Camta2	Dync2h1
Carm1	Ears2
Carmil2	Edem1
Cbarp	Ednrb
Cbx6	Edrf1
Ccm2	Epg5
Cdc42ep2	Ephb2
Cdk16	F3
Cdk9	Fads2
Cenpb	Fam13a
Cep170b	Fbln1
Chmp2a	Fcgr2b
Cnih2	Fgfr1
Cops6	Fzd1
Cox4i1	Gapvd1
Cox6b1	Gla
Cry2	Gle1
Csdc2	Gpr17
Ctbp1	Gpr183
Ctdnep1	Gpr34
Ctnnbip1	Grm7
Cyb5r3	Hdlbp
Dact3	Hexb
Dap	Hnrnpll
Dbi	Hs6st2
Ddit3	Iars1
Ddn	Itga7
Ddrgk1	Itgb2
Deaf1	Jak1
Desi1	Kcnc1
Diaph1	Kcnh8
Diras1	Kdm5b
Dlg4	Lamb1
Dlgap3	Lfng
D0jc8	Lhfpl2
D0l4	Lman1
Dpf1	Lpin1
Drap1	Lrrc8a
Drg2	Lrrk2
Dtx3	Maml2
Dynlrb1	Man2a2
Eef1a1	Maoa
Eef1g	Marchf7
Eef2	Mcm4
Ei24	Mfsd12
Eif1	Mgat2
Eif3f	Mipep
Eif3h	Morc3
Ell3	Mtmr4
Elob	Mtmr7
Epn1	Mtor
Evl	Myorg
Exosc2	0a25
Fahd2a	Ncan
Faim	Ncapd3
Fam131c	Ncbp1
Fbxo44	Nckap1l
Fkbp2	Ncoa3
Foxo4	Nemp1
Fth1	Nlgn3
Ftl1	Ntrk3
Fxn	Nup58
Gamt	Nup98
Gap43	Nxpe3
Git1	P2ry12
Gna12	P2ry13
Gnai2	P4ha1
Gnas	Parp1
Gnl1	Pcdhb12
Gpr137	Pde4dip
Grk2	Pdgfra
Gsta4	Pdia4
Gstm2	Pfkfb3
Gtf2h5	Pi4ka
Gtf3a	Pik3ap1
Guk1	Pik3c3
Hcfc1r1	Pik3cb
Hdgf	Pip5k1a
Hirip3	Pld4
Hnrnpl	Plx02
Hscb	Pml
Hspe1	Pnma8a
Hypk	Ppl
Ikzf4	Ppwd1
Ilrun	Prmt3
Insyn1	Prnp
Jph3	Prpf8
Jund	Qrich1
Kifc2	RGD1307443
Lhpp	Rhobtb2
Lmo1	Rimkla
Lrrc73	Rprd2
Lsm3	Rundc1
Lynx1	Rxfp1
Macrod1	Ryr2
Map3k10	Scube2
Mapk3	Sec24b
Mapk8ip1	Sec24c
Mapre3	Selplg
Marchf2	Sema3e
Marcksl1	Sfmbt1
Maz	Sfpq
Mbd3	Slc1a3
Mbp	Slc1a4
Meaf6	Slc24a4
Med25	Slc2a5
Mi0r2	Slc33a1
Mmp24	Slc38a9
Morn4	Slco2b1
Mpnd	Slit1
Mpped1	Smchd1
Mrpl21	Smpdl3a
Mrpl38	Snrk
Mrps25	Sppl2a
Mt‐atp6	Spred3
Mt‐atp8	Srgap2
Mt‐co1	Syt2
Mt‐co2	Tacc2
Mt‐co3	Tbc1d30
Mt‐cyb	Tbc1d4
Mt‐nd1	Tex10
Mt‐nd2	Tlr2
Mt‐nd3	Tmem119
Mt‐nd4	Tmem131
Mt3	Tmem161b
0ca	Tmem104
0 cc1	Tmem200a
Ndrg2	Tmem63c
Ndufa3	Tmx2
Ndufa8	Tnfrsf19
Ndufb3	Trappc11
Ndufb4	Trim9
Ndufs5	Tubgcp5
Nelfe	Txndc16
Neurl1	Uba6
Ngrn	Ubash3b
Nkd1	Ubr5
Npas1	Uhrf2
Nr1i3	Uqcrc2
Nr2f6	Uvrag
Nsmf	Vps39
Nsun5	Vps50
Nubp2	Washc2c
Oaz2	Wdr19
Ormdl2	Xpc
P3h4	Yars2
Palm	Zfp597
Panx2	Zfp869
Parp6	Zmynd8
Paxx	Zscan18
Pcbp4
Pcp4l1
Pde2a
Pdgfa
Pdzd4
Pea15
Pfdn1
Pfdn2
Pianp
Pigp
Pih1d2
Pink1
Pip5k1c
Pitpnb
Ppdpf
Ppp1r9b
Prelid3b
Prkcb
Prkcz
Prrt1
Psd
Psmc5
Ptma
Ptms
Ptpmt1
R3hcc1
R3hdm4
Rab11b
Rab11fip5
Rab15
Rab5c
Rapgefl1
Rasgef1a
Rasl10b
RGD1309036
Rnd2
Rnf139
Rnf10
Rnf220
Rnf208
Rnps1
Rpl12
Rpl14
Rpl17
Rpl18
Rpl19
Rpl10
Rpl21
Rpl22l1
Rpl23
Rpl23a
Rpl24
Rpl3
Rpl32
Rpl34
Rpl35a
Rpl36a
Rpl4
Rpl5
Rpl6
Rpl7a
Rpl8
Rpl9
Rplp2
Rplp0
Rps12
Rps15a
Rps18
Rps10
Rps2
Rps21
Rps23
Rps27a‐ps1
Rps20
Rps3a
Rps4x
Rps6
Rps8
Rps9
Rtl8a
Rtn2
Rusc1
Samd14
Samd4b
Sap30l
Scaf1
Sdccag8
Septin3
Septin5
Sh2d3c
Sirt2
Sirt6
Slc25a23
Slc39a5
Smarcc2
Smarcd1
Smarcd3
Smdt1
Smim19
Smox
Sncb
Snf8
Snrpd2
Snrpn
Sod1
Speg
Spsb2
Src
St6gal0c6
Stx1a
Stx1b
Suds3
Syf2
Tbc1d22b
Tcea2
Tcte1
Thra
Thtpa
Timm9
Tlcd3b
Tle5
Tma7
Tmcc2
Tmem256
Tmem250
Tnnc2
Tp53i11
Tpt1
Trappc2l
Trim41
Trim8
Ttc7b
Tusc2
Txn1
Ube2i
Ube2l3
Ube2m
Ube2q1
Ube2r2
Upf3a
Usf2
Vamp1
Vps37d
Vsig10l
Wdtc1
Wsb2
Zfand3
Zfp523
Zfp575
Znhit1

To verify the WMI‐specific DEGs and demonstrate that EE can indeed reverse gene expression changes caused by aging in this strain, hippocampal transcript levels of eight genes (Table 1 bolded) were quantified by qPCR in both strains. Significant DEGs were randomly selected to be investigated by qPCR. Results for Coronin 2b (Coro2b) and solute carrier family 35, member A4 (Slc35a4) confirmed the decreased expression in WMI 12 M compared to 6 M, and the reversal of this decreased expression by EE (Coro2b: strain, F[1,44] = 3.88, p = 0.05; age, F[2,44] = 14,01, p < 0.001; Slc35a4: strain × age, F[1,44] = 4.20, p < 0.05; Figure 4A,B). While Coro2b expression was also increased in WLI 12 + EE compared to 12 M hippocampus, Slc35a4 expression changes were specific to WMIs. In contrast, expression of nudix hydrolase 13 (Nudt13) showed no reversal by EE, and opposite patterns between WLIs and WMIs (strain × age, F[1,39] = 8.29, p < 0.001). Except for Zc3h13, all DEGs showed lower expression in 6 M than 12 M, and a reversal by EE (e.g., lower expression in 12 M + EE than in 12 M) were confirmed by qPCR. Specifically, hippocampal transcript levels of Kelch‐like family member 6 (Klhl6), SPARC‐related modular calcium binding 2 (Smoc2), potassium inwardly rectifying channel, subfamily J, member 2 (Kcnj2) and basic helix–loop–helix family, member e41 (Bhlhe41) were higher in 12 M compared with 6 M WMI females, but EE attenuated these increases in the 12 M + EE WMIs (Figure 4). While expressions of Klhl6, Smoc2, and Bhlhe41 were higher in WLI 12 M hippocampi, and Bhlhe41 expression also decreased in WLIs by EE, Kcnj2 transcript level changes were specific for WMIs ( Klhl6 : age, F[2,35] = 18.38, p < 0.001; strain × age, F[2,35] = 5.25, p = 0.01; Smoc2 : age, F[2,41] = 13.21, p < 0.001; strain × age, F[2,41] = 3.13, p = 0.05; Bhlhe41 : age, F[2,44] = 23.15, p < 0.001; strain × age, F[2,41] = 4.19, p < 0.05; Kcnj2 : age, F[2,41] = 4.09, p < 0.05; Figure 4).

FIGURE 4.

Hippocampal gene expression patterns differ between WLI and WMI females in an age, strain and housing condition‐dependent manner. Hippocampal expression of Coro2b (A) and Slc35a4 (B) were significantly lower in WMI 12 M females compared to both WMI 6 M and 12 M + EE WMIs. Hippocampal transcript levels of Klhl6 (C), Smoc2 (D) and Bhlhe41 (F) were higher in both WLI and WMI 12 M females compared to 6 M levels, but their expression in the 12 M + EE hippocampus did not differ from those in 6 M females of both strains. In contrast, Kcnj2 (E) expression only increased in 12 M WMI compared to both 6 M and 12 M + EE WMIs. Transcript levels of Zc3h13 and Nudt13 (G, H) did not change by age and strain in contrast to the RNA‐seq results. Values are shown as mean ± SEM. Statistical analyses as described in Figure 1. *q < 0.05, **q < 0.01, corrected for multiple comparisons. Number of animals as in Figure 1.

The correlations between gene expression quantified by qPCR and RNA‐seq data are shown in Figure 5. Fold changes for qPCR results were derived from RQs of 6 M over RQs of 12 M, and 12 M + EE over 12 M. The correlation between the RNA‐seq and qPCR fold changes was significant, with r² = 0.756, p < 0.001 for 6 M vs. 12 M, and r ² = 0.865, p < 0.001 for 12 M + EE vs. 12 M.

FIGURE 5.

Significant correlation of fold change measures of DEGs between RNA‐seq and RT‐qPCR. (A) Expression of DEGs from RNA‐seq, shown in Figure 4, were measured by qPCR in 6 M and 12 M WLI and WMI female hippocampi. Fold change in qPCR was derived from relative quantification (RQ) of samples. Person correlation between the RNA‐seq and qPCR fold changes was highly significant. (B) Highly significant Pearson correlation between the qPCR results and the RNA‐seq results for 12 M vs. 12 M + EE WLI and WMI female hippocampi.

Significant DEGs were also identified based on age (6 M vs. 12 M) and strain (WLI vs. WMI) using a subset of the data that did not include the 12 M + EE samples with lower RIN values. Ontology enrichment was used to identify biological processes and pathways represented by these DEGs. Age and strain DEGs were both enriched for the processes of oxidoreduction‐driven active transmembrane transporter activity (Table 2). Additionally, the overall strain DEGs were enriched for the oxidative phosphorylation term.

TABLE 2.

Gene ontology enrichment.

Comparison	Term name	Adjusted p‐value	Intersection size
Age difference (no EE)	Protein binding	8.65e‐73	2322
	Catalytic activity	2.03e‐22	1491
	Molecular adaptor activity	9.05e‐9	139
	Channel regulator activity	1.42E‐06	56
	Oxidoreduction‐driven active transmembrane transporter activity	1.35E‐05	24
Strain difference (no EE)	Postsynaptic specialization	4.73e‐14	20
	Cytoplasm	8.06e‐13	92
	Oxidoreduction‐driven active transmembrane transporter activity	5.48e‐12	10
	Oxidative phosphorylation	6.96e‐11	14
	Plasma membrane bounded cell projection	3.11e‐06	32
Strain × age interaction (no EE)	Protein binding	5.22e‐15	389
	Cell adhesion	1.55e‐7	81
	Collagen‐containing extracellular matrix	6.07e‐7	27
	Nucleoplasm	2.06e‐05	140
	Cell motility	7.17e‐05	88
Overlap between strains by age	Phosphate‐containing compound metabolic process	8.19e‐05	35
	Multicellular organism development	0.00012	49
	Endomembrane system	0.00013	41
	Regulation of protein metabolic process	0.00027	32
	Regulation of amino acid transport	0.00031	6

Note: Adjusted p‐value is the probability of seeing a number of genes out of the total n genes in the list annotated to a particular GO term, given the proportion of genes in the whole genome that are annotated to that GO term. Intersection size is the number of genes in the input query that are annotated to the corresponding term.

All significant DEGs between WLI and WMI female hippocampus, regardless of age and EE status, were also entered into the Ingenuity Pathway Analysis (IPA) to find the most significant networks (Table 3). The top three IPA interactive networks for overall strain differences were characterized as cell morphology, cell signaling, cell‐to‐cell signaling, developmental disorder, hereditary disorder, metabolic disease, cardiovascular system development and gene expression (Table 3). The overall strain differences from data subset 1 that includes the 12 M + EE group resulted in the IPA network 2 (score: 39) of “developmental disorder, hereditary disorder, metabolic disease”, as shown in Figure 6. The hubs in this network are mitochondrial complex 1, mitochondrial electron transport chain, and cytochrome oxidase processes. The overall age and strain interactive network 1 was categorized into amino acid metabolism, DNA replication, recombination and repair (Table 3). Networks 2 and 4 showed the same characterizations of cell signaling, posttranslational modification and protein synthesis, while network 3 was classified by cancer, cell death, survival, organismal injury and abnormalities. The IPA canonical pathway analysis of strain × age interactive DEGs revealed the involvement in oxidative phosphorylation (64 molecules), Eif2 signaling (90 molecules) and mitochondrial dysfunction (107 molecules) (Table 3).

TABLE 3.

IPA networks and canonical pathways.

Top IPA networks
Comparison	Diseases and functions	Score	Focus molecules
WLI vs. WMI	Cell morphology, cell signaling, cell to cell signaling	42	20
	Developmental disorder, hereditary disorder, metabolic disease	39	19
	Cardiovascular system development, gene expression	37	18
Age × Strain	Amino acid metabolism, DNA replication, recombination and repair	44	36
	Cell signaling, posttranslational modification, protein synthesis	41	34
	Cancer, cell death, survival, organismal injury and abnormalities	41	34
	Cell signaling, posttranslational modification, protein synthesis	41	34

Top canonical pathways
Comparison	Canonical pathways	p	Z score
Age × Strain	Oxidative phosphorylation	1.185E‐15	−7.937
	Eif2 signaling	2.7434E‐14	−6.102
	Mitochondrial dysfunction	2.0751E‐12	−6.929

FIGURE 6.

IPA generated network of DEGs from WMI and WLI strain comparison. The IPA network of “developmental disorder, hereditary disorder, metabolic disease”, network 2 (score: 39), is shown. Colored nodes represent input genes. The predicted upstream and downstream effects of activation or inhibition are shown by color overlays and lines, and color coding of the nodes and connecting lines are indicated in the legend. The solid and dash lines denote direct and indirect interactions between two nodes based on the IPA database, respectively.

The oxidative phosphorylation network derived from overall strain differences without EE, the first and third canonical pathways representing DEGs identified from the strain × age (including EE) contrast, and the network hubs in Figure 6 confirmed the significance of oxidative phosphorylation and mitochondrial dysfunction in molecular aging in WMI and WLI. Based on this, expressions of mitochondrially encoded cytochrome c oxidase 1–3 (Mt‐co1, Mt‐co2, Mt‐co3) and mitochondrially encoded NADH dehydrogenase 3 (Mt‐nd3) were determined by quantitative RT‐PCR in WLI and WMI female hippocampi at six and 12 mos. Figure 7 shows that aging generally decreased the expression of these genes in both strains, and EE for 6 months reversed the effect. Expression of these mitochondrial genes was generally higher in WMIs than WLIs, but this difference was eliminated by the increased expression after EE (Mt‐co1; strain, F[1,42] = 5.80, p < 0.05; strain × age, F[2,42] = 13.57, p < 0.001; Mt‐co2; age, F[2,38] = 6.34, p < 0.01; Mt‐co3; strain, F[1,44] = 12.35, p < 0.01; age, F[2,44] = 15.72, p < 0.001; strain × age, F[2,44] = 3.91, p < 0.05; Mt‐nd3; strain, F[1,45] = 4.1, p < 0.05; age, F[2,45] = 13.16, p < 0.001). Mt‐co2 and Mt‐nd3 expression showed greater specificity for WMIs, as the decreased expression at 12 M was only reversed by EE in WMIs (Figure 7).

FIGURE 7.

Strain, age and housing conditions affect hippocampal mitochondrial gene expression. (A) Decreased expression of Mt‐co1 in 12 M female hippocampus in both strains compared to 6 M, that is completely or partially reversed by EE in WLIs and WMIs, respectively. (B–D) Enhanced expression of Mt‐co2, Mt‐co3 and Mt‐nd3 in 6 months old (6 M) WMIs compared to WLIs, but EE normalizes these strain differences. *q < 0.05, **q < 0.01 corrected for multiple comparison.

4. Discussion

Our study confirms that the genetically stress‐hyperresponsive WMI rat, which exhibits enhanced depression‐like behavior, experiences early‐onset hippocampus‐dependent memory loss. Long‐term environmental enrichment eliminated this cognitive decline in middle‐aged WMI females and attenuated it in WMI males. Analysis of molecular changes associated with the reversal of cognitive decline by EE identified parallel changes in estradiol levels in WMI females. Moreover, RNA sequencing combined with qPCR validation identified genes—Slc35a4 and Kcnj2—whose expression also paralleled changes in memory of WMI females. Furthermore, pathway analysis of DEGs revealed oxidative phosphorylation and mitochondrial dysfunction as key candidate processes underlying hippocampal aging and its reversal by environmental enrichment.

Cognitive changes can occur as early as middle age, 3–7 years before mild cognitive impairment is diagnosed [71]. Identifying modifiable risk factors for dementia can facilitate early interventions and reduce disease burden. Literature suggests many modifiable risk factors for dementia [4], some of which can be studied in animal models of naturally occurring cognitive decline. Among these are depression and stress. It has been proposed that early treatment of depressive symptoms may impact the course of disease in AD and affect the risk of developing dementia [72]. In a large prospective cohort study consisting of over 300,000 participants aged 50–70 years and followed for 10 years, depression was linked to a 51% increased risk of dementia [73], reinforcing its role as a significant risk factor. Even subsyndromal depression is associated with cognitive decline [74].

Stress is another major risk factor, and psychological stress in adulthood is shown to be associated with an increased risk of dementia [75]. Although stress is a very elusive construct, it is thought to be associated with a 20%–30% higher risk of dementia [76]. Generally, stress‐related neuropsychiatric disorders such as post‐traumatic stress disorder (PTSD), general anxiety disorder, and panic disorder exacerbate age‐related cognitive decline [77, 78, 79] and double the risk of developing AD and other dementias in older individuals [80, 81]. Does depression and stress accelerate cognitive decline starting in middle age? Indeed, there is evidence that stress, stress‐related disorders, and depression can accelerate age‐related cognitive decline [82, 83].

A key finding of this study is that the WMI strain, characterized by enhanced depression and stress reactivity, exhibits early‐onset cognitive decline compared to the WLI control strain. We use the term early onset cognitive aging in the context of the available literature on aging animal models. Rats usually live for 24–36 months, and most aging studies are carried out with rats over 18 months of age [84, 85]. For example, aged Sprague–Dawley rats (> 24 months old) exhibit deficits in contextual fear conditioning, whereas late middle‐aged (16–24 months old) or early middle‐aged rats do not [20].

However, the process of cognitive decline can start earlier than the above age range as suggested by the increased heterogeneity of aging‐induced cognitive decline observed in late middle‐aged (16–20 months old) rats [84]. Impaired episodic/working memory emerges around middle age (reviewed in [86] and continues to decline with advanced age [87, 88, 89, 90, 91, 92]). The WMI strain showed deficits in fear memory and in spatial memory as early as 12 months of age compared to their nearly isogenic control strain, which is strong evidence of differential genetic and epigenetic contributions to early memory deficits between strains.

Early‐onset memory loss was more pronounced in middle‐aged WMI females than in WMI males, contradicting findings from some studies. Deficits in recall after contextual fear conditioning emerge only after 8 months of age in APPswe/PS1ΔE9 female mice, in contrast to male mice where the recall deficits are already seen at 2 months of age [93]. In contrast, non‐transgenic studies suggest a male advantage in the effect of aging on spatial working and reference memory for rats across strains [94]. This sex difference may explain why most studies on aging‐induced cognitive decline focus on males [84, 85].

In contrast, our naturally occurring and non‐transgenic model of age‐induced cognitive decline mimics the sex difference that rules the human aging literature; human females are more vulnerable to cognitive aging than males [95, 96, 97]. The dramatic early onset memory loss in middle‐aged WMI females, compared to WMI males and WLI females, is demonstrated by a greater loss of fear memory and the inversely exaggerated activity on Day 2 of the CFC paradigm. The freezing response to the conditioning foot shock stimuli paralleled the fear memory in middle‐aged WMIs, suggesting that there is also an impairment in the learning process that is more pronounced in the WMI females. Similarly, middle‐aged WMI females showed impaired spatial learning in the MWM throughout the 4 days of training, manifesting in greater latency to reach the platform compared to young WMIs.

It is plausible that the decreased plasma estradiol levels could play a causative role in the attenuation of learning and memory in middle‐aged WMI females. Estrogen deficiency heightened the learning and memory deficit in the APP/PS1 triple transgenic mice [98]. Estrogen deficiency induces hippocampal apoptosis and cognitive dysfunctions [99, 100], while estrogen treatment, including brain‐specific estrogen prodrug, improves learning‐memory performances [101, 102]. Thus, reduced estrogen levels in middle‐aged WMI females, compared to young WMIs, and unchanged E2 levels in middle‐aged WLI females parallel both fear and spatial memory measures in females. Testosterone levels in males were consistently lower in WMI males than in WLI males, regardless of age; thus, age‐induced attenuated memory in WMI males does not parallel changes in peripheral testosterone levels.

Long‐term environmental enrichment reversed or attenuated the early onset memory loss in WMIs. Environmental enrichment enhances learning and memory while mitigating age‐related cognitive decline in animals [103, 104, 105]. The main mechanism by which EE affects behavioral and molecular mechanisms is that it can attenuate the negative effects of stress by either acting on the same neurological pathways concurrently or acting on different pathways in parallel [66]. Additionally, EE is beneficial for reducing anxiety and depression and increasing cognitive performance (see [105]), and stress can negatively impact the same traits [66, 106]. Although the exact mechanism of action through which EE mitigates aging effects in this study remains unclear, our behavioral data suggest several potential mechanisms of action. For example, all 12 M + EE groups showed a decrease in distance traveled on Day 1 of the CFC (Supplemental Figure S1) and CORT levels were also increased in all 12 M + EE groups (Figure 3). In addition, 12 M + EE WLI and WMI females showed increased rearing on Day 1 (Supplemental Figure S2). These data suggest that EE may be impacting responses to stress in both strains. However, only 12 M + EE female and male WMIs showed the reversal or attenuation of memory deficit as evidenced by increased freezing on Day 2 of the CFC (Figure 1). Thus, the specificity of EE in reversing or attenuating the cognitive decline of middle‐aged WMIs is not only related to counteracting stress but may involve other mechanism(s) yet to be elucidated. One possibility is that the striking reversal of cognitive decline in 12 M + EE WMI females could be related to EE‐dependent restoration of plasma E2 levels which were profoundly reduced by age in WMI females (Figure 3). How that increase in E2 levels occurs in the presence of elevated plasma CORT in the 12 M + EE WMI females is not clear. Nevertheless, this increase could affect oxidative stress and mitochondrial function, both of which are known to be involved in age‐induced cognitive decline [107, 108].

Lower serum levels of E2 are observed in female patients with AD compared to appropriately matched controls [109]. Estrogen has been shown to improve cognitive functions and alleviate depression in humans and rodent models [110, 111]. In a recent study, reduced transcript levels of estrogen receptors (ESR) were found in postmortem brain regions of female subjects with AD and major depressive disorder (MDD), compared to those with AD and no MDD [112]. Since hippocampal ESR2 expression correlated with transcript levels of antioxidant enzymes in that study, females with AD could have exaggerated accumulation of oxidative stress because of the reduced estrogen‐induced mitochondrial defect, compared to controls [113]. Thus, it is feasible that estrogen‐regulated processes contribute to the vulnerability to early onset cognitive decline in WMI females.

While aging is linked to cognitive decline, individual variability in age‐related memory loss exists in both humans and animals [16, 84]. The hippocampus is of particular interest for aging and cognitive decline as it is known to play an important role in learning and memory consolidation. For these reasons, we quantified hippocampal gene expression in females as a first step toward identifying the underlying mechanisms through which EE can improve memory function in WMI females. We hypothesized that EE would reverse age‐related gene expression changes, which was confirmed by the restoration of hippocampal Slc35a4 and Kcnj2 expression. Interestingly, the SLC35A4 protein encoded by this gene is thought to regulate sensitivity to oxidative stress, as a knockout of SLC35A4 enhanced sensitivity to oxidative stress in a rescuable manner, indicating direct involvement of SLC35A4 in stress resistance [114]. Additionally, SLC35A4 codes for an inner mitochondrial membrane microprotein crucial for cellular respiration and ATP generation with a vital role in cellular metabolism [115]. Likewise, Kcnj2 (KIR2.1) encodes an inwardly rectifying potassium channel expressed in glial cells whose function has been found to be impaired following oxidative stress, leading to neuronal hyperexcitability, a situation that can be reversed by antioxidant treatment [116]. Additionally, the elevation of K+ in the perivascular space activates smooth muscle Kir channels to cause vasodilation and increased blood flow in response to the activity of the nearby neurons [117]. In rats, this activity‐driven vasodilation is impaired by chronic stress via glucocorticoid receptor‐dependent downregulation of Kir 2.1 [118]. In human subjects, acute psychosocial stress elicits changes in hemodynamic response in insular, temporal, and prefrontal cortices. These hemodynamic changes were associated with genetic differences in KCNJ2 expression [119]. In agreement with these findings, WMI 12 M females showed reduced corticosterone levels (Figure 3) and increased Kcnj2 expression compared to their 6 M controls (Figure 4). Both of these changes were reversed by EE; therefore, one additional potential mechanism of EE is remediating the impaired hemodynamic responses to stress in the brain.

Both Slc35a4 and Kcnj2 seem to respond to oxidative stress and its alleviation. Moreover, the expression of both genes is differentially altered (i.e., decrease in Slc35a4 and increase in Kcnj2) as a function of age in the middle‐aged WMI female hippocampus, and these changes in expression were reversed by EE. Taken together, these genes are likely to contribute to the enhanced vulnerability to aging‐induced cognitive decline via altering sensitivity to oxidative stress, which may be reversed through EE. Substantiating this hypothesis were our findings that mitochondria and oxidative stress pathways were enriched in gene lists based on age and strain interactions, which could be the result of age‐aggravated excess of oxidative stress in WMI females. We have previously reported differential vulnerability to oxidative stress and mitochondrial dysfunction between strains even at a very early age [120]; embryonic WMI neurons and astrocytes are more vulnerable to oxidative stress compared to WLI. Although the expression of many mitochondrial genes was altered by aging and reversed by EE in the WLI strain (Table 1), the low number of DEGs detected in the WMI strain is likely due to low RIN values, especially in the 12 M + EE condition. Moreover, when the expression of mitochondrial genes was quantified by qPCR in independent samples (Figure 7), we found that these genes are more highly expressed in WMI relative to WLI females at 6 M, expression is reduced at 12 M in both strains, and EE restores expression to about the same level in both strains.

Our study has several important limitations. Although we confirmed that EE reverses age‐related changes in plasma estradiol, gene expression, and cognitive decline in middle‐aged WMI females, as discussed above, the underlying causal mechanisms remain unclear. EE can act through multiple pathways that ultimately lead to changes in stress reactivity, depression, cognitive function, metabolic function, and estrogen levels. In addition, there are some potential behavioral differences between WMI and WLI strains that limit interpretation. For example, an inverse relationship between distance traveled and freezing is generally observed during CFC training on Day 1. However, in this study, all 12 M + EE groups demonstrated a decrease in distance traveled relative to 12 M, but freeze duration was only increased in the 12 M + EE WMI females and males. The decreased distance traveled with no change in freeze duration in the WLI 12 M + EE animals suggests some other movement, such as grooming, which would not be detected as locomotion. Grooming was not measured in the current study, but prior work demonstrated greater grooming behavior in adult WLIs compared to WMIs [121]. Whether or not the decrease in distance traveled in WLI 12 M + EE animals can be attributed to changes in grooming behavior is unknown, and the implications of EE enhancement of this behavior remain unclear.

Our transcriptome analysis was also limited in several ways. First, as cognitive decline and reversal by EE are more profound in females, we limited our analysis to females instead of including both sexes, which limits the generalizability of our findings. In addition, to obtain enough RNA to perform both bulk RNA‐seq and later qPCR validation, we combined dorsal and ventral hippocampi together for each animal instead of analyzing these functionally distinct structures separately. Ultimately, we were also limited by technical issues such as the unexplained lower RINs impacting some experimental groups. Due to the length of the study, samples were collected at different times, and RNA was also isolated in different batches. Thus, counterbalancing of RNA extraction groups could not always be performed. Unfortunately, a suspected batch effect disproportionately impacted the WMI 12 M + EE group, resulting in much lower RINs for this subgroup. As the exact source of the low RINs could not be established, we were reluctant to attempt to correct for a batch effect and instead opted to perform analyses of the data with and without the 12 M + EE groups. For these reasons, we were unable to detect many more exclusive changes in the WMI strain due to EE. Despite these limitations, we were able to validate many of the changes in gene expression by qPCR, including in mitochondrial genes, Slc35a4 and Kcnj2; however, our gene expression data alone cannot resolve whether some or all these potential underlying mechanisms are responsible for the cognitive protection afforded by EE treatment in the WMI strain. For this reason, future studies will be required to interpret the contribution of our observed changes in mitochondrial gene expression associated with strain, aging, metabolic burden, and EE to cognitive function. Specifically, quantification of mitochondrial abundance and function will be required, as an increase in mitochondrial gene expression could indicate either an increased number of mitochondria or an increase in oxidative stress/mitochondrial activation.

In conclusion, in this paper, we confirmed the early onset decline of hippocampal‐dependent memory in the stress and depression‐prone WMI rat relative to its nearly isogenic control strain, the WLI. Importantly, we demonstrate that an intervention during adulthood, EE, can reverse cognitive decline in midlife in the WMI strain, with more profound effects in female WMIs relative to males. To better understand the underlying molecular mechanisms that might be driving genetic background‐by‐age‐associated cognitive decline, we profiled hippocampal gene expression in females and plasma CORT, E2, and T levels in both sexes at different ages and under different environmental conditions. We also identified and validated gene expression changes associated with aging that were reversed by EE, some of which (i.e., Slc35a4 and Kcnj2) showed greater specificity for WMIs. We also noted the strong effect of EE on mitochondrial gene expression, possibly with a more beneficial effect in the WMI strain. Finally, we identified E2, but not CORT or T, as a potential mediator of EE's beneficial effect on memory in WMI females. Although we were unable to identify the causal molecular mediators underlying the profound therapeutic effect of EE on cognitive decline in our model, our work advances the field in at least two critical ways. First, we establish the nearly isogenic WMI and WLI strains as a novel model for investigating cellular and molecular interactions among aging, interventions, and genetic vulnerability to cognitive decline. Second, we propose candidate genes and testable hypotheses regarding the molecular mechanisms underlying genotype × age × EE interactions, focusing on E2 levels and hippocampal expression changes in mitochondria and oxidative response genes. Future work leveraging forward genetics approaches, such as genetic mapping to identify underlying variants in age‐associated cognitive decline, combined with epigenetic profiling and functional assays to determine vulnerable cell populations and precise alterations in mitochondrial number, function, and gene expression are expected to reveal molecular processes underlying the age‐associated decline in memory between strains. A forward genetics approach is especially useful in the context of the WMI/WLI model based on the very limited number of sequence variations between the two strains [49].

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1. Distance traveled in the CFC was not correlated between day 1 and day 2, but distance traveled on day 2 was inversely related to fear memory in WMIs. Day 1. The overall distance traveled after the three 1 s foot shock was significantly lower in middle aged (12 M) compared to young (6 M) females of both strains. This activity was further decreased by EE in females. 12 M + EE males of both strains traveled less than their young and middle‐aged counterparts. Day 2. Contrary to day 1, distance traveled was significantly and solely higher in 12 M WMIs compared to both 6 M and 12 M + EE WMI males and females. 8–13/strain/sex/age. Data as mean ± SEM. Statistical differences were determined by three‐way ANOVA. Post hoc group comparisons were carried out by two‐stage linear set‐up procedure of Benjamini, Krieger, and Yekutieli following significant ANOVA *q < 0.05; **q < 0.01 corrected for multiple comparisons.

Figure S2. Rearing during CFC. Day 1. Rearing events showed an age, strain and sex‐dependent pattern like that of distance traveled. Day 2. In general males reared more than females and EE affected rearing to the opposite direction in WLI and WMI females. Statistics as in Figure S1.

Figure S3. Floating/immobility during the last day of Morris Water Maze test. There were no significant differences by age, strain and housing condition in floating across the trials in males and females.

Figure S4. Probe trial latency to target quadrant on day 5 of Morris Water Maze test. The platform was removed on day 5 and the animals placed into the water maze at the opposite quadrant to the target quadrant. Almost all 12 M WMI females did not find the quadrant during the available 60 s.

Table S1. Primer sequences for quantitative RT‐PCR.

Table S2. Strain differences in hippocampal gene expression at 6 months of age (WLI vs. WMI; p < 0.01).

Table S3. Strain differences in hippocampal gene expression at 12 months of age (WLI vs. WMI; p < 0.001).

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.