Epigenetic treatment of behavioral and physiological deficits in a tauopathy mouse model

Abstract Epigenetic abnormality is implicated in neurodegenerative diseases associated with cognitive deficits, such as Alzheimer's disease (AD). A common feature of AD is the accumulation of neurofibrillary tangles composed of hyperphosphorylated tau. Transgenic mice expressing mutant P301S human tau protein develop AD‐like progressive tau pathology and cognitive impairment. Here, we show that the euchromatic histone‐lysine N‐methyltransferase 2 (EHMT2) is significantly elevated in the prefrontal cortex (PFC) of P301S Tau mice (5–7 months old), leading to the increased repressive histone mark, H3K9me2, which is reversed by treatment with the selective EHMT inhibitor UNC0642. Behavioral assays show that UNC0642 treatment induces the robust rescue of spatial and recognition memory deficits in P301S Tau mice. Concomitantly, the diminished PFC neuronal excitability and glutamatergic synaptic transmission in P301S Tau mice are also normalized by UNC0642 treatment. In addition, EHMT inhibition dramatically attenuates the hyperphosphorylated tau level in PFC of P301S Tau mice. Transcriptomic analysis reveals that UNC0642 treatment of P301S Tau mice has normalized a number of dysregulated genes in PFC, which are enriched in cytoskeleton and extracellular matrix organization, ion channels and transporters, receptor signaling, and stress responses. Together, these data suggest that targeting histone methylation enzymes to adjust gene expression could be used to treat cognitive and synaptic deficits in neurodegenerative diseases linked to tauopathies.


| INTRODUC TI ON
Intracellular accumulation of neurofibrillary tau tangles is the common hallmark of Alzheimer's disease (AD; Dubois et al., 2014). Tau, a microtubule (MT) associated protein, is critically involved in MT assemble and stabilization. Normal phosphorylation of tau controls the dynamics of MT, establishing neuronal polarity, axonal outgrowth, and axonal transport (Caceres & Kosik, 1990;Harada et al., 1994;Wang & Mandelkow, 2016). Abnormally hyperphosphorylated tau assembles tau into tangles of filaments and breaks down MTs, which disrupts synaptic functions and leads to synapse loss correlated with cognitive impairment prior to overt neurodegeneration (Alonso et al., 1996;Hoover et al., 2010;Mielke et al., 2017). However, there are currently limited treatment options for tau pathology-linked disorders including AD.
Emerging evidence shows that epigenetic factors that integrate environmental factors (e.g., aging) into chromatin remodeling can alter neuronal functions, which contributes to the cognitive decline linked to neurodegenerative diseases Sen et al., 2016). Controlling gene expression through epigenetic regulation, including chromatin remodeling and histone modification, provides a potential avenue to restore synaptic and cognitive plasticity in AD (Li et al., 2019;Yuan et al., 2020).
A key epigenetic regulator of neuronal function, the euchromatichistone-lysine N-methyltransferase (EHMT), which represses genes through histone 3 lysine 9 (H3K9) methylation (Tachibana et al., 2005), targets most of genes involved in learning and memory in Drosophila (Kramer et al., 2011). Human databases have shown that the expression of EHMT1 in the frontal cortex increases with age and positively correlates with AD progression (Lu et al., 2004;Sharma et al., 2017;Yuan et al., 2020). Ablation of the negative regulator, H3K9 methylation, prevents Aβ-induced plasticity deficits in hippocampal neurons (Sharma et al., 2017), improves cognitive behavior in an Aβ-linked familial AD model (Zheng et al., 2019) mitochondrial function in aging mice (Yuan et al., 2020).
The therapeutic potential of targeting EHMT to treat tau-mediated neurodegenerative disorders remains unknown. To address this, we used a transgenic mouse model expressing human tau bearing P301S mutation, which develops filamentous tau lesions at 6 months of age (Yoshiyama et al., 2007;Zhang et al., 2014). Prefrontal cortex (PFC), a key target region of AD that is critical for high-level cognitive function (Maillet & Rajah, 2013;Spellman et al., 2015;Tan et al., 2021;Yan & Rein, 2021), was focused. We discovered that euchromatic histonelysine N-methyltransferase 2 (EHMT2) was significantly elevated in PFC of P301S Tau mice, and inhibition of EHMT led to the amelioration of behavioral and synaptic deficits, as well as the normalization of large-scale gene expression. It supports the potential of EHMT as a target for the treatment of tauopathies, including AD.
Given the elevation of EHMT2, we next tested whether EHMT inhibition could alleviate cognitive deficits in P301S Tau mice (5-7 months old). We first performed Barnes maze (BM), an assay to test the animal's retention and retrieval of spatial memory by recalling the location of one correct hole (where an escape box was attached in the training phases) from seven other incorrect holes on a round platform based on visual cues (Zheng et al., 2019). As shown in Figure 2a,b, compared with WT mice, P301 Tau mice spent significantly less time exploring the correct hole (T1) and more time exploring the seven incorrect holes (T2), while UNC0642-treated Tau mice had significantly increased time on the correct hole and decreased time on the incorrect holes (n = 7-8 mice/group, F 3,54(interaction) = 17.05, p < 0.0001, two-way ANOVA). Accordingly, the spatial memory index (T1/T2) was markedly lower in P301S Tau mice than WT animals, which was significantly elevated by UNC0642 treatment (Figure 2c, F 1,27(interaction) = 10.03, p = 0.0038, two-way ANOVA). Consistent improvement on spatial memory was observed in almost every examined P301S Tau mouse treated with UNC0642 ( Figure 2d, t (7) = 4.71, p = 0.0022, paired t test). The therapeutic effect of a brief UNC0642 administration on spatial memory deficits in P301S Tau mice sustained at least 4 days after the cessation of treatment (Figure 2e, n = 7-8 mice/group, F 9,81(interaction) = 3.34, p = 0.0016, two-way rmANOVA).
We next conducted the novel object recognition (NOR) task, an assay to test recognition memory by measuring the exploration time spent on a novel object versus a familiar one. As shown in Figure 2fh, WT mice exhibited significantly longer investigation time on the novel objects than the familiar objects in NOR test, whereas P301S Tau transgenic mice showed no discrimination between the novel and familiar objects, and this recognition memory impairment in P301S Tau mice was markedly rescued by UNC0642 treatment (n = 7-8 mice/group, H: F 3,52(interaction) = 9.17, p < 0.001, two-way ANOVA; I: F 1,26(interaction) = 22.66, p < 0.001, two-way ANOVA).
The elevated discrimination ratio induced by UNC0642 was consistently detected in individual P301S Tau mice (Figure 2i, t (7) = 12.76, p < 0.0001, paired t test). Moreover, the rescue effect persisted for ~4 days and weakened at 7 days after the cessation of treatment (Figure 2j, n = 7-8 mice/group, F 9,78(interaction) = 5.99, p < 0.001, twoway rmANOVA). UNC0642-treated WT mice did not change the examined behaviors (Figure 2a-j). Together, these results suggest that inhibition of the elevated EHMT2 can improve spatial and recognition memory in the tauopathy model of AD.

| EHMT inhibition normalizes the excitability of PFC pyramidal neurons in P301S Tau mice
Next, we explored the physiological basis for the behavioral effects of EHMT inhibition. PFC is a brain region strongly linked to spatial memory retrieval and learning, as well as object recognition behaviors (Yan & Rein, 2021). The activity of PFC pyramidal neurons and recurrent excitation underlie PFC-mediated cognitive function (Goldman-Rakic, 1995), and loss of cortical network function contributes to cognitive decline in tauopathies (Menkes-Caspi et al., 2015). Thus, we carried out current-clamp recordings to measure the excitability of layer five PFC principal neurons.
We first measured the synaptic-driven spontaneous action potentials (sAP) using a low Mg 2+ condition to enhance the activity of brain slices . As shown in Figure 3a,b, compared with age-matched WT mice, sAP frequency of PFC pyramidal neurons was substantially lower in P301S mice (5-7 months old), which was significantly increased to the normal level by UNC0642 treatment (1 mg/kg, i.p., 3x) (n = 15-18 cells, 3-5 mice/group, F 2,46 = 22.98, p < 0.0001, one-way ANOVA). Resting membrane potential (RMP) and other sAP properties, including amplitude, rise time, decay time, and half-width, did not change significantly (Figure 3c-g).
We further examined neuronal excitability by measuring action potentials elicited by injecting a series of depolarizing currents (eAP). Figure 3h,i, the frequency of evoked spikes was significantly lower in P301S Tau mice than WT mice and was reversed by UNC0642 treatment (F 2,43 (group) = 10.71, p = 0.0002, two-way rmANOVA). Two other measures of excitability, the first spike latency, which plays an essential role in integrating synaptic events and influences the firing probability (Molineux et al., 2005), and rheobase, the minimum current required to elicit APs, were also compared. As shown in Figure 3j,k, both parameters were significantly increased in P301S Tau mice, and normalized by UNC0642 treatment (n = 15-16 cells, 3-4 mice/group, latency: F 2,43 = 9.57, p = 0.0004; rheobase: F 2,43 = 7.68, p = 0.0014, one-way ANOVA).

As shown in
In addition, UNC0642 restored the significantly reduced input resistance in P301S Tau mice ( Figure 3l, F 2 F I G U R E 2 Administration of EHMT inhibitor UNC0642 improves cognitive function in P301S Tau mice. (a) Heat maps depicting the topographical time distribution in BM tests of WT versus P301S Tau mice treated with UNC0642 (1 mg/kg, i.p., 3x) or saline. Locations of the correct holes are labeled with arrowheads. (b, c) Bar graphs showing the investigation time (T1: on the correct hole; T2: on the seven incorrect holes) (b) and the spatial memory index (T1/T2) (c) during BM tests of all groups (b: ++ p < 0.01, +++ p < 0.001, T1 vs. T2; *p < 0.05, **p < 0.01, ***p < 0.001, two-way ANOVA; c: **p < 0.01, ***p < 0.001, two-way ANOVA). (d) Scatter plots showing the BM spatial memory index of individual mice pre-and post-treatment of UNC0642 (**p < 0.01, paired t test). (e) Plots of BM spatial memory index in WT versus Tau mice treated with UNC0642 or saline at different days (***p < 0.001, WT + saline vs. Tau + saline; ### p < 0.001, Tau + saline vs. Tau + UNC, two-way rmANOVA). (f) Heat maps depicting the topographical time distribution in NOR tests of WT versus Tau mice treated with UNC0642 or saline. Locations of novel objects are labeled with arrowheads. (g, h) Bar graphs showing the exploration time (T Fam : on the familiar object; T Nov : on the novel object) (g) and the discrimination ratio (h) during NOR tests of all groups (g: +++ p < 0.001, T Fam vs. T Nov ; *p < 0.05, **p < 0.01, two-way ANOVA; h: ***p < 0.001, two-way ANOVA). (i) Scatter plots showing the NOR discrimination ratio in individual mice pre-and post-treatment of UNC0642 (***p < 0.001, paired t test). (j) Plots of NOR discrimination ratio in WT and Tau mice treated with UNC0642 or saline at different days (***p < 0.001, WT + saline vs. Tau + saline; # p < 0.05, ### p < 0.001, Tau + saline vs. Tau + UNC, two-way rmANOVA). Data in figures (a-d) and (f-i) were collected at day 1 after cessation of UNC0642 treatment ANOVA). Spike threshold and membrane capacitance were not significantly changed (Figure 3m,n). Overall, these data indicate that EHMT2 inhibition can rescue the diminished excitability of PFC pyramidal neurons in P301S Tau mice.

| EHMT inhibition restores synaptic function and attenuates hyperphosphorylated tau in PFC of P301S Tau mice
The excitability of PFC pyramidal neurons is mainly controlled by glutamatergic transmission. Synaptic dysfunction is a primary feature of AD and the basis of cognitive impairment (Styr & Slutsky, 2018). To assess the effect of UNC0642 on synaptic function, we next measured NMDAR-and AMPAR-EPSC in layer 5 PFC pyramidal neurons from WT and P301S Tau mice (6-8 months old). As shown in Figure 4a A pathological hallmark of tauopathy is the presence of hyperphosphorylated tau (p-Tau). Next, we investigated whether EHMT inhibition could change tau hyperphosphorylation.
Functional protein classification revealed that enzymes, transcription regulators, and receptors were found among UNC0642rescued genes in both directions (Tables S3 & S4). Importantly, the down-regulated genes in P301S Tau mice that were elevated by

| DISCUSS ION
Epigenetic dysregulation of gene expression is one of the major contributing factors for cognitive decline related to aging and neurodegenerative diseases . Histone modification by G9a/GLP (EHMT1/2) is emerging as a pivotal epigenetic mechanism regulating cognitive processes (Maze et al., 2010;Sharma et al., 2017;Yuan et al., 2020). Our current study has provided evidence demonstrating that EHMT2 (G9a), which catalyzes the repressive histone mark H3K9me2 (Barski et al., 2007) at learning and memory genes (Kramer et al., 2011), is highly elevated in PFC neurons of P301S Tau transgenic mice. Moreover, treatment with the EHMT inhibitor UNC0642 ameliorates spatial and recognition memory deficits in P301S Tau mice at the age of 5-7 months old, a late stage that cognitive impairment correlating with neurofibrillary tangle (NFT) deposition has already occurred (DeVos et al., 2017;Mathys et al., 2019;Yoshiyama et al., 2007;Zhang et al., 2014), making it highly promising for clinical usage. The general safety of UNC0642 is suggested by the lack of behavioral abnormalities after treatment (Kim et al., 2017;Kramer et al., 2011;Liu et al., 2013).
Because of the powerful impact of epigenetic drugs on gene expression and the potential side effects with prolonged administration, we have chosen a short treatment paradigm in mice (i.p., once daily for 3 days), which was based on a similar regimen used in human cancer treatment with the FDA-approved HDAC inhibitor romidepsin (i.v., once a week for 1 month). Similar doses and durations have been used with UNC0642 in prior mouse studies (Kim et al., 2017;Liu et al., 2013;Wang et al., 2020;Zheng et al., 2019). A limitation is the disappearance of behavioral improvement ~7 days after the cessation of UNC0642 treatment, which may be due to the Our transcriptomic analysis has provided insights into the potential molecular mechanism for the rescuing effect of EHMT inhibition on synaptic and cognitive deficits in P301S mice. As an epigenetic enzyme mediating the repressive H3K9me2, EHMT1/2 plays an important role in silencing the expression of genes (Tachibana et al., 2005), including those involved in synaptic plasticity and cognition (Kramer et al., 2011;Rodenas-Ruano et al., 2012;Zheng et al., 2019).

| Animals and compounds
Care and experimental manipulation of animals followed the pro-
Slices were mounted onto slides with anti-fading mounting media (VECTASHIELD, Vector Laboratories). Images were taken by a confocal microscope (Leica TCS SP8). All specimens were imaged and analyzed with identical parameters.
Analyses of immunostaining signals were performed with Image  Luminescence was detected by Chemidoc XRS system (Bio-Rad), and density of blots was quantified by ImageJ software (NIH).

| Behavioral testing
Behavioral tests were performed in dimly lit rooms and scored by trained operators blind to experimental conditions. Any-maze be-  Figure S1.

| Barnes maze
A round platform (36-inch diameter) with eight equally spaced holes at the edge was used. One of the holes had an escape box attached during the training phases. Visual cues were placed on sidewalls to indicate the hole locations. An overhead light brightly illuminated the apparatus as an aversive stimulus to force the animal to run into the escape box. After 5-min habituation, the animal was allowed to explore the platform till entering the escape box during the two training phases (5-min interval). After a 15-min break in the home cage, the animal was positioned on the apparatus (escape box removed) to carry out the test phase (5 min) using distal visual cues. The time spent around the correct hole (T1) and all the other incorrect holes (T2) was recorded. The spatial memory index (T1/T2) was calculated as previously described (Cao et al., 2020;Wang et al., 2018). For repeated measurements, visual cues were changed before training.

| Novel object recognition
After habituation on a circular platform (24-inch diameter) for 5 min, the test animal was allowed to explore two identical objects on the platform for 5 min. After a 5-min interval, the mouse was returned to explore the platform containing one original familiar object and a novel object for 5 min. The amount of time spent interacting with each object was scored. The discrimination ratio was calculated by

| Electrophysiological recordings
The slice was transferred into a recording chamber on an upright

| RNA sequencing and bioinformatic analysis
RNA extraction from PFC samples of biological duplicates in each group used the RNAeasy Mini kit (Qiagen), combined with the RNasefree DNase step (Q iagen). The strand-specific RNA sequencing library was generated from the purified RNA (1 μg) by using TruSeq stranded total RNA and Ribo-zero kits (Illumina). The sequencing was analyzed by Illumina HiSeq 2500 platform at the Genomics and Bioinformatics Core of University at Buffalo. Reads were trimmed using Cutadapt to remove the 3′ end adapters and trailing sequences, then aligned to mouse RefSeq mRNAs using TopHat2 with default parameters. Reads count for each gene was estimated with featureCounts. Differences in gene expression levels between samples were defined with at least 1.2-Fold Change (FC) and p < 0.05. GO annotation was carried out as we previously described (Cao et al., 2020).
Cytoscape software was applied to visualize the gene interaction relationship network and identify Hub Genes and Key Pathways. We utilized the CytoHubba application in Cytoscape, employing MCC (Maximal Clique Centrality) algorithm to rank top genes. MCODE plugin was used to present dense regions of protein/gene interaction networks. STRING plugin was used for protein-protein interaction (PPI) network analysis.

| Quantitative real-time PCR
Total RNA was extracted from PFC punches with Trizol (Invitrogen), then incubated in DNase I (Invitrogen) to remove any contaminating DNA. The mRNA was converted into cDNA by using an iScript reverse transcription kit (Bio-Rad). The iCycler iQ™ Real-Time PCR Detection System and iQ™ Supermix (Bio-Rad) were used for qPCR.
Fold changes in the target genes were determined by the follow-

| Statistical analyses
Clampfit (Molecular Devices) and Mini Analysis (Synaptosoft) were used for electrophysiological data analyses. GraphPad Prism 8.0 was used for all statistical analyses. Student's t test (two-tailed paired or unpaired) was used for statistical analyses of experiments with two groups. ANOVA (one-way, two-way or two-way repeated measures) with Bonferroni post hoc test was used for statistical analyses of experiments with more than two groups. All data are presented as Mean ± SEM.

ACK N OWLED G EM ENTS
We thank Xiaoqing Chen and Kaijie Ma for their excellent technical support. We also thank Kevin Y. Feng for analyzing some immunohistochemical data. We acknowledge the support of University at Buffalo's Genomics and Bioinformatics Core and the New York State Center of Excellence in Bioinformatics and Life Sciences. This work was supported by grants from the National Institutes of Health (AG056060, AG064656, and AG067597) to Z.Y.

CO N FLI C T O F I NTE R E S T
The authors report no competing financial or other interests. designed experiments, supervised the project, and wrote the paper.

DATA AVA I A B I L I T Y S TAT E M E N T
Genomic data have been deposited in the GEO public repository (GSE182170).