PPAR-α Hypermethylation in the Hippocampus of Mice Exposed to Social Isolation Stress Is Associated with Enhanced Neuroinflammation and Aggressive Behavior

Social behavioral changes, including social isolation or loneliness, increase the risk for stress-related disorders, such as major depressive disorder, posttraumatic stress disorder (PTSD), and suicide, which share a strong neuroinflammatory etiopathogenetic component. The peroxisome-proliferator activated receptor (PPAR)-α, a newly discovered target involved in emotional behavior regulation, is a ligand-activated nuclear receptor and a transcription factor that, following stimulation by endogenous or synthetic ligands, may induce neuroprotective effects by modulating neuroinflammation, and improve anxiety and depression-like behaviors by enhancing neurosteroid biosynthesis. How stress affects epigenetic mechanisms with downstream effects on inflammation and emotional behavior remains poorly understood. We studied the effects of 4-week social isolation, using a mouse model of PTSD/suicide-like behavior, on hippocampal PPAR-α epigenetic modification. Decreased PPAR-α expression in the hippocampus of socially isolated mice was associated with increased levels of methylated cytosines of PPAR-α gene CpG-rich fragments and deficient neurosteroid biosynthesis. This effect was associated with increased histone deacetylases (HDAC)1, methyl-cytosine binding protein (MeCP)2 and decreased ten-eleven translocator (TET)2 expression, which favor hypermethylation. These alterations were associated with increased TLR-4 and pro-inflammatory markers (e.g., TNF-α,), mediated by NF-κB signaling in the hippocampus of aggressive mice. This study contributes the first evidence of stress-induced brain PPAR-α epigenetic regulation. Social isolation stress may constitute a risk factor for inflammatory-based psychiatric disorders associated with neurosteroid deficits, and targeting epigenetic marks linked to PPAR-α downregulation may offer a valid therapeutic approach.


Introduction
The COVID-19 pandemic is currently challenging the mental health of the world's general population, demonstrated by the skyrocketing prevalence of psychological distress and the high incidence of psychiatric disorders [1]. Furthermore, individuals with pre-existing mental disorders, including anxiety, major depressive disorder (MDD), and posttraumatic stress disorder (PTSD), are at higher risk for symptom exacerbation. The emergence of suicidal thoughts and attempts, substance use disorder and addictions, and impulse control disorders is also alarming [2][3][4].
A combination of individual factors, including personal fear of contagion, history of trauma and mental disorders, and environmental factors including financial losses and an obligated social withdrawal without adequate individual-based psychological and economic support or therapeutic strategies may enhance stress-related affective instability in vulnerable populations [5][6][7]. The ability to overcome stressful events or trauma without significant psychophysiological alterations is based on the combination of genetic and

Decreased PPAR-α Expression in the Hippocampus of SI Mice Correlates with Aggressive Behavior
We previously showed a decrease in PPAR protein expression in SI mice (15). Here, we selected SI mice that showed high levels of aggressive behavior (indicated as SI-AGG) compared with low-to-non-aggressive (<5 bites/10 min) mice. These SI mice showed resilience to social isolation stress and were termed SI non-aggressive (SI-NAG) mice. We found decreased PPAR-α mRNA levels in the hippocampus of SI-AGG (one-way ANOVA: * p = 0.03; F = 3.98; N = 8), whereas no difference was observed in SI-NAG (n = 7) compared with GH mice (n = 8; Figure 1A). Moreover, the Pearson correlation analysis showed a significant negative correlation between aggressive behavior and PPAR-α mRNA expression in SI-AGG mice (Pearson r = −0.65; * p = 0.04; n = 8; Figure 1B).

Social Isolation Stress Induces DNA Methylation of PPAR-α CpG-Rich Promoter Regions
To test whether a lower PPAR-α mRNA expression was the result of epigenetic changes in the Ppar-a CpG-rich fragments, we measured the immunoprecipitated methyl-cytosines of the three regions of the gene analyzed in the hippocampus of the same group of mice. A schematic representation of the Ppar-a gene structure and the regions amplified is shown in Figure 2. We found a significant increase in precipitated methylated cytosine (mC) levels in all the three Ppar-a gene regions amplified in the hippocampus of SI-AGG mice compared with the control (GH) mice ( Figure 2) (region 1: unpaired t test two-tailed: * p = 0.04; t = 2.23; region 2: * p = 0.03, t = 2.30; region 3: * p = 0.02, t = 2.59; n = 8). In addition, a positive correlation was observed between aggressive behavior and methylation status of the cytosines studied in the upstream region (region 2) of the Ppar-α gene (r = 0.70; * p = 0.02; n = 8; Figure 2). Methylation analysis in SI-NAG mice was not performed because no changes were found in the PPAR-α mRNA expression in these mice.  [23], we selected three different fragments of the gene rich in CG dinucleotides within the proximal promoter, 5 -UTR (within exon 1), or 3 UTR portions of the gene and amplified by specific primers using the qPCR technique. The selected region (region 1) consists of a CpG-enriched promoter fragment positioned at −50/+31 bp linking the 5 -upstream promoter region to the 5 -UTR fragment of Exon 1. Region 2 amplified the 5 -UTR and part of exon 1 after the transcription starting nucleotide (+1) positioned at +9 +149 bp, whereas region 3 covered the 3 -UTR and part of the exon 8 positioned at +1715 +1813 bp. The increase in cytosine methylation levels was measured by methylated DNA immunoprecipitation (MeDIP) of the Ppar-α gene fragments (region 1: unpaired t test: * p = 0.04, t = 2.23, df = 14, n = 8; region 2: * p = 0.03; t = 2.30; df = 14; n = 8; region 3: * p = 0.02; t = 2.59; df = 14; n = 8). In addition, a positive correlation between methylated cytosines (mC) levels of region 2 and aggressive behavior was observed (Pearson's r = 0.70; * p = 0.02; n = 8).

Increased Expression of Pro-Inflammatory Markers in the Hippocampus of SI Mice Is Associated with Aggression
To explore whether stress-induced Ppar-α gene suppression leads to the activation of neuroinflammatory processes, we measured the mRNA expression of TLR4, NF-κB, TNFα, and MCP-1, which are involved in the PPAR-α anti-inflammatory pathway mediated by NF-κB activation.

Neurosteroid Levels in Aggressive and Non-Aggressive SI Mice
Decreased Allo and pregnanolone levels were specifically observed in the hippocampus of SI-AGG mice (unpaired t test: Allo = * p = 0.04, t = 2.18; PA = * p = 0.04, t = 1.78; Figure 5). The optical density (OD) of the bands was quantified and normalized relative to the housekeeping protein GAPDH. The bar graphs were obtained using values as means ± S.E.M. of 5-6 samples for the controls (GH) and 6 samples for SI-AGG. The samples were loaded in the same immunoblot, and the values were calculated relative to the samples from GH mice. Individual values are shown in the relative graph plots. For the significance, values were analyzed using unpaired two-tailed t test.

Discussion
This study provides the first evidence of a PPAR-α epigenetic regulation induced by DNA methylation in the mouse brain. In the hippocampus of SI-AGG mice, our results show altered epigenetic marks, including (i) increased levels of methylated cytosines of the Ppar-α gene ( Figure 2); (ii) increased expression of HDAC1 and MeCP2, which belongs to the methyl-CpG-binding domain protein family ( Figure 3); and (iii) a reduced TET2 expression. Our main finding of the hypermethylated Ppar-α gene and its expression down-regulation is thus supported by the increase in HDAC1 and MeCP2 and decrease in TET mRNA expression, which favor hypermethylation of the CG fragments within the promoter regions of the Ppar-α gene. This evidence strongly supports a genomic effect induced by isolation stress and the involvement of DNA methylation mechanisms. Significantly decreased PPAR-α mRNA expression, downregulated brain Allo and pregnanolone levels, increased pro-inflammatory factors (NF-κB and TLR-4), and increased protein expression of the pIKKα/β and TNFα were associated with vulnerable SI mice expressing high aggression scores. These data support a Ppar-α gene hypermethylation as a leading mechanism to decreased hippocampus PPAR-α expression and favoring neuroinflammation mediated by the NF-κB/TLR-4 pathway. This evidence corroborates findings that deficits in PPAR-α act by promoting proinflammatory processes and dysfunctional behaviors exacerbated by stress exposure, including social isolation.
PPAR-α is deeply involved in several physiological and pathological conditions, including regulation of the mitochondrial and proteasomal function, neuroinflammation, oxidative stress, and neurodegeneration, which are considered key pathogenetic mechanisms involved in neurodegenerative diseases and stress-related disorders, including anxiety, MDD, PTSD, postpartum depression, and suicide [25][26][27]. PPAR-α is also known to exhibit anti-inflammatory effects and a neuroprotective activity by modulating the expression of inflammatory mediators, including IκB, an NF-κB inhibitor, leading to the suppression of cytokine production [28]. NF-κB resides inactively in the cytoplasm associated with inhibitory IκB proteins. Upon NF-κB pathway activation, the IκB protein is degraded and the NF-κB complex moves into the nucleus and modulates the expression of the target genes. IκB degradation is mediated by the IκB kinase (IKK) complex, which, via phosphorylation, induces IκB proteasomal degradation [29]. In our results, the pIKK protein expression in increased in SI aggressive mice, as shown in Figure 4, supporting the hypothesis of NF-κB activation.

Epigenetic Mechanisms Affecting PPAR-α Expression during Social Isolation
The mouse Ppar-α gene, located on chromosome 15, comprises a total of eight exons, with a 5 -untranslated region (5 UTR) encoded by exons 1 and 2, and part of exon 3, whereas the remaining part of exon 3 and exons 4-8 contribute to the coding region. The last portion of exon 8 (232 bp) contributes to the 3 untraslated region (3 UTR) [30]. Both 5 UTR and 3 UTR constitute regulatory regions for gene transcription [23]. These regions are also characterized by a high CpG density detected by MethPrimer. The altered expression of PPAR-α, which is involved in the recruitment of epigenetic factors (class III NAD +dependent HDAC Sirt1) [31], may affect the expression of the target genes, including genes encoding for neurosteroidogenic enzymes and pro-inflammatory markers.
Despite the numerous implications of PPAR-α in physiological and pathological conditions, little is known about its epigenetic regulation, especially within the brain. Few studies have investigated the epigenetic regulation of the Ppar-a gene promoter in obesity, cardiac and intestine pathologies, and liver disease, where the DNA methylation of PPAR-α has been suggested as a therapeutic approach for non-alcoholic fatty liver disease [32]. Thus, this study contributes the first demonstration that PPAR-α is susceptible to stress-induced epigenetic regulation in the brain.
Despite the fact that there was no difference in the DNMT mRNA expression in SI-AGG mice, the increased expression of MeCP2 and HDAC1, and reduced TET2, which together promote DNA methylation [31], suggest an imbalance in the dynamics between DNA methylation/demethylation in the hippocampus of SI-AGG mice. HDAC inhibitors are intensively investigated for specific clinical applications including cancer, neurodegenerative disorders, and as cognitive enhancers [31,[33][34][35][36]. Tet genes are decreased in chronic inflammatory conditions and animal models thereof, causing aberrant DNA methylation [37]. Reduced TET2 expression has also been implicated in altered neurogenesis and cognitive impairment [38,39]. These findings might favor pharmacological interventions aimed at inhibiting DNA methylation targeting the PPAR-α gene as a novel approach for stress-related neuroimmune disorders, which include neuropsychiatric disorders. Administering HDAC inhibitors may represent a valuable pharmacological approach to modulate PPAR-α expression and reinstate its functional role as an anti-inflammatory and behavioral regulator [40,41].

PPAR-α/Neurosteroid Downregulation and Increase in Pro-inflammatory Mediators in SI Mice
Preclinical and clinical studies show decreased Allo and pregnanolone levels in depression and PTSD, while the enhancement of their concentrations is associated with symptom improvement [42][43][44]. Of note, IV Allo (i.e., brexanolone) is FDA approved to treat post-partum depression after pivotal Phase 3 clinical trials, demonstrating high efficacy and safety profiles [45]. In SI mice, decreased Allo levels are inversely correlated with anxiety-like behavior, exaggerated contextual fear responses, and aggression, which is the most prominent behavioral trait to model suicide risk in humans [15,19]. Allo binds primarily at extrasynaptic GABA A receptors in corticolimbic pyramidal neurons [46,47], which is regarded as the mechanism underlying the improvement of behavioral dysfunction in human and animal studies [48,49]. These behavioral effects are also reproduced by administration with the PPAR-α endogenous modulator, PEA, and the synthetic ligand, fenofibrate, in SI mice, which upregulate neurosteroidogenic enzymes and proteins and Allo levels in corticolimbic areas [15], including the hippocampus and the amygdala, with well-established roles in emotional behavior regulation. Downregulation of Allo and its equipotent GABAergic isomer, pregnanolone, occurs specifically in SI mice that express aggressive behavior, but not in "resilient" mice that fail to express behavioral impairment. In the hippocampus of SI-AGG mice, the expression of TLR-4 and NF-κB, and of cytokines TNF-α and MCP1 increased, signifying a pro-inflammatory condition. Previous studies have provided evidence that an increased TLR-4 expression underlies a downstream activation of the NF-κB signaling cascade, promoting pro-inflammatory cytokines [50].
TLRs are immune receptors expressed in innate immune cells (dendritic cells and macrophages) and in non-immune cells (fibroblasts and epithelial cells) that are involved in proinflammatory responses and neuroinflammation [51]. In neurons, microglia, and astrocytes, Allo mediates potent anti-inflammatory effects through NF-κB signaling inhibi-tion [52]. Allo and its precursor, pregnenolone, inhibit TLR-4 signaling and the production of pro-inflammatory mediators by a mechanism independent of Allo activation of the GABA A receptor [53,54], suggesting a direct link between neurosteroidogenesis and TLR4mediated inflammation. Activation of the TLR4/MyD88 signaling pathway induces the expression of pro-inflammatory cytokines, whereas Allo prevented this effect by inhibiting TLR-4 binding to the MyD88 adaptor and to its specific endogenous ligand, lipopolysaccharide (LPS) [53]. Allo may also enhance LPS degradation [54]. Thus, decreased hippocampus Allo levels in SI mice may play a role in the following: (1) the inflammatory processes resulting from PPAR-α downregulation and enhancement of NF-κB signaling, and (2) the behavioral dysfunction arising from the downregulation of Allo-modulated GABAergic neurotransmission. The mechanisms of stress-induced Allo biosynthesis downregulation in the contest of impaired NF-κB and TLR-4 pathways remain to be clarified, as well as whether TLR-4 abnormal activation occurring in stress-induced models of affective disorders [55] is linked to the downregulation of brain Allo levels. Indeed, exposure to TLRs agonists suppresses PPAR-α expression via a NF-κB-dependent signaling in astrocytes [56], raising the question of whether this can be an additional mechanism responsible for low Allo levels in pathophysiological conditions characterized by elevated inflammation and affective symptoms.

PPAR-α/TLR4/Allo Pathway as a Potential Biomarker and Pharmacological Target in Stress-Induced Behavior
We previously reported that SI mice express aggression and altered fear responses associated with decreased corticolimbic Allo levels and lower PPAR-α protein content [15]. This is associated with decreased levels of the Allo biosynthetic enzyme, 5α-reductase type I (5α-RI) [57]. This study shows that social isolation also decreases pregnanolone levels specifically in vulnerable SI AGG mice. Given the role PPAR-α plays in neurosteroidogenesis [15,58], our study suggests that stress-induced epigenetic suppression of PPAR-α may underlay both the enhancement of proinflammatory markers, mediated by TLR4-activation, and neurosteroid biosynthesis downregulation. Whether altered PPAR-α expression directly interferes with the recruitment of epigenetic factors [59] involved in the regulation of Allo biosynthetic genes remains to be further analyzed. A significant correlation was found between mRNA PPAR-α-suppression, its methylation status, and aggression, as well as between the levels of neurosteroids and the expression of aggression. As such, our results suggest that an altered PPAR-α/TLR4/Allo pathway may offer biomarkers of mood instability and stress-induced resiliency deficits. Consistently, PPAR-α hypermethylation induced by protracted stress may provide a new neurobiological factor underlying stress-induced behavioral abnormalities in vulnerable populations. Epigenetic changes are linked to maladaptive stress responses and psychiatric disorders [60]. Hypermethylation of the glucocorticoid receptor (GR) gene (NR3C1) promoter was reported in the hippocampus of suicide victims with a history of childhood abuse compared with those without early-life trauma and with control subjects [61]. Increased methylation of the alpha-1 subunit of the GABA A receptor gene (Gabra1) promoter was also found in the prefrontal cortex of suicide victims [62]. Pharmacoepigenetic strategies aimed to increase PPAR-α expression may be useful in populations that show hypermethylation and suppression of PPAR-α caused by chronic stress conditions. It remains to be investigated if the PPAR-α hypermethylation status can be revealed in the peripheral blood mononuclear cells (PBMC) of stressed animals and subjects with psychiatric conditions. The goal to connect psychiatric symptoms with biological alterations as an objective indicator of pathology is still unmet. The discovery of blood or saliva biomarkers will represent a breakthrough in the management of PTSD and suicide, as well as in psychiatric disorders in general. Several biomarker candidates, including BDNF and HPA-axis alterations with increased levels of corticotropin-releasing hormone (CRH) and cortisol, have been proposed, but none have yet been established for clinical screening [63]. A low expression of PPAR-α and an increased expression of TLR-4 and pro-inflammatory cytokines have been reported in peripheral pathological conditions and metabolic inflammation [63,64]. Thus, establishing a correlation between the peripheral and central expression and DNA methylation status of PPAR-α/TLR-4/Allo in animal models and clinical studies should be prioritized.
PEA is a lipid mediator that exerts anti-inflammatory effects by direct activation of PPAR-α as a main molecular target [65]. The PEA stimulation of PPAR-α induces corticolimbic Allo biosynthesis, an effect that is directly associated with the improvement of PTSD/suicide-like phenotypes in SI mice [15,66]. The administration of PEA also stimulates PPAR-α protein and 5α-RI expression [15,67]. Although the underlying mechanisms are not yet fully understood, these findings suggest a possibility that PEA may reverse stress-induced epigenetic modifications, leading to an up-regulation of PPAR-α expression. While these mechanisms remain to be studied, treatment with PEA to alleviate neuroinflammatory-based neuropsychiatric disorders could provide a reasonable therapeutic strategy. Importantly, PPAR-α has been recently investigated as a target for natural and synthetic compounds that exert an anti-inflammatory and neuroprotective effect in rodent models for neurodegenerative diseases [26,68]. Ongoing clinical trials are studying the efficacy of natural micronutrients, such as genistein or synthetic PPAR-α ligands for the treatment of neuropsychiatric disorders, including Alzheimer's disease, bipolar depression, and alcoholism [26,69,70].

Animals
Male Swiss-Webster mice (Charles River Laboratories; 21-23 days old at arrival) were maintained under a 12-h dark/light cycle (light on at 06:00 a.m. and off at 18:00 p.m.) and were provided food and water ad libitum. SI mice were housed individually in a cage (24 × 17 × 12 cm) for 4-6 weeks, while group-housed (GH) mice were housed in groups of 5. The vivarium temperature was 24 • C and the humidity 65%. Depending on the study, the number of mice for each experimental group varied from 6 to 12 (see figure legends).

Behavioral Analysis
Aggressive behavior was analyzed using the resident-intruder test, as previously described [20]. Briefly, a naïve, sex and weight-matched "intruder" mouse was introduced into the "home cage" of the experimental "resident" mouse, which was housed in social isolation for 4 weeks, and resident-intruder interactions were videotaped for 10 min. Several behaviors were considered, such as number of attacks, bites, wrestling, and duration of attacks, as an index of aggressive behavior. The results were expressed as the total duration of aggression during the 10 min observation period.

Brain Neurosteroid Measurements
Extraction, derivatization, and quantification of Allo and pregnanolone were described previously [71]. After the addition of deuterium-labeled internal standards to tissue samples, steroids were extracted, purified, and separated by high-pressure liquid chromatography (HPLC, Spectra-Physics DSP8800, Milpitas, CA, USA). After derivatization with HFBA, gas chromatography-mass spectrometry (GC−MS 7820A/5977, Agilent Technologies, Santa Clara, CA, USA) analysis in the standard electron impact mode was performed [20].

Real-Time Polymerase Chain Reaction (RT-PCR)
Total mRNA was extracted from mouse hippocampus using the miRNeasy kit (Qiagen, Valencia, CA, USA). The extracts were treated with DNAse and RT-PCR was performed as previously described [26] using Applied Biosystems RT-PCR System (AriaMx: Agilent Technologies, Santa Clara, CA, USA) with SYBR Green master mix (Brilliant III Ultra-Fast: Agilent Technologies). Primer sequences used for the amplification of targeted cDNA are listed in Table 1. The relative mRNA expression values were calculated as the fold of controls using β-Actin as the internal control for sample normalization.

Statistical Analysis
The results were analyzed using a two-tailed unpaired t-test for the results in Figures 2, 3 and 4B; one-way ANOVA followed by Tukey's multiple comparison test for the results in Figures 1, 4A and 5B; and two-tailed Pearson's correlation test for the results in Figures 1B and 2. The F-value was determined by ANOVA using GraphPad Prism 8.2.1. Values are expressed as the means ± SEM of 6-8 mice for the biochemical analysis and 8-10 mice for the behavioral tests. Significance was set at * p < 0.05. The correlation results were obtained using two-tailed Pearson's correlation analysis (for the results in Figures 1B and 2B). For the biochemical analysis, we used eight samples per experimental group (GH and SI-AGG) and seven samples for SI-NAG. The specific number of values for each experiment is indicated in the respective results sections.  Data Availability Statement: All data are presented and available in the manuscript. The corresponding author is available to provide any additional information on data presented in this study.
Conflicts of Interest: G.P. is a paid consultant to PureTech Health (Boston, MA, USA) and Neu-roTrauma Sciences (Alpharetta, GA, USA). He has two pending patent applications, one on Npalmitoylethanolamine (PEA) and peroxisome proliferator-activated receptor alpha (PPAR-α) agonists, and one on allopregnanolone analogs in the treatment of neuropsychiatric disorders.