Histone deacetylase‐3 regulates the expression of the amyloid precursor protein and its inhibition promotes neuroregenerative pathways in Alzheimer's disease models

HDAC3 inhibition has been shown to improve memory and reduce amyloid‐β (Aβ) in Alzheimer's disease (AD) models, but the underlying mechanisms are unclear. We investigated the molecular effects of HDAC3 inhibition on AD pathology, using in vitro and ex vivo models of AD, based on our finding that HDAC3 expression is increased in AD brains. For this purpose, N2a mouse neuroblastoma cells as well as organotypic brain cultures (OBCSs) of 5XFAD and wild‐type mice were incubated with various concentrations of the HDAC3 selective inhibitor RGFP966 (0.1–10 μM) for 24 h. Treatment with RGFP966 or HDAC3 knockdown in N2a cells was associated with an increase on amyloid precursor protein (APP) and mRNA expressions, without alterations in Aβ42 secretion. In vitro chromatin immunoprecipitation analysis revealed enriched HDAC3 binding at APP promoter regions. The increase in APP expression was also detected in OBCSs from 5XFAD mice incubated with 1 μM RGFP966, without changes in Aβ. In addition, HDAC3 inhibition resulted in a reduction of activated Iba‐1‐positive microglia and astrocytes in 5XFAD slices, which was not observed in OBCSs from wild‐type mice. mRNA sequencing analysis revealed that HDAC3 inhibition modulated neuronal regenerative pathways related to neurogenesis, differentiation, axonogenesis, and dendritic spine density in OBCSs. Our findings highlight the complexity and diversity of the effects of HDAC3 inhibition on AD models and suggest that HDAC3 may have multiple roles in the regulation of APP expression and processing, as well as in the modulation of neuroinflammatory and neuroprotective genes.

The increase in APP expression was also detected in OBCSs from 5XFAD mice incubated with 1 μM RGFP966, without changes in Aβ.In addition, HDAC3 inhibition resulted in a reduction of activated Iba-1-positive microglia and astrocytes in 5XFAD slices, which was not observed in OBCSs from wild-type mice.mRNA sequencing analysis revealed that HDAC3 inhibition modulated neuronal regenerative pathways related to neurogenesis, differentiation, axonogenesis, and dendritic spine density in OBCSs.Our findings highlight the complexity and diversity of the effects of HDAC3 inhibition on AD models and suggest that HDAC3 may have multiple roles in the regulation of APP expression and processing, as well as in the modulation of neuroinflammatory and neuroprotective genes.

| INTRODUCTION
Epigenetic-based interventions have recently garnered attention as potential therapies for Alzheimer's disease (AD) due to a greater understanding of epigenetic influences on neuropathology and neurodegeneration. 1 While specific inherited gene mutations can account for a minority of AD cases, 2 most sporadic AD cases are likely triggered by more complex interactions between genetics and environmental factors. 3Dysregulation of epigenetic mechanisms, such as the alteration of histone acetylation states, has been reported in the AD brain [4][5][6] and also in mouse models of AD, 7,8 although a consensus on the directionality of these changes is still a topic of debate.The expression of histone deacetylases (HDACs) has been found altered in AD brains, although with contrasting results depending on the specific HDACs, the brain area, and the stage of the disease. 9For example, HDAC2 was reported to be increased in the CA1 hippocampal and entorhinal cortex nuclei in AD brains as compared to controls, 10 while levels of HDAC3 and HDAC1 levels were found unchanged.On the other hand, others have shown decreases in HDAC1 and 2 in frontal cortex of AD brains, 6,11 whereas a recent study demonstrated that HDAC3 is elevated in the frontal cortex of moderate AD sufferers. 9This in agreement with larger expression of HDAC3 and HDAC4 detected in mouse models of AD. 11 In recent years, HDACs have gained attention as promising targets of therapeutic agents against AD, 12 due to their involvement in several aspects of the AD pathology, notably amyloid clearance, 13 inflammation, 14 gliosis 15 and synaptic health. 16Representing a multi-targeted approach, HDAC inhibitors have been shown to have the ability to reinstate memory even after neuronal loss. 17A small number of HDAC inhibitors are currently on trial for the treatment of AD and mild cognitive impairment (MCI). 18DAC3 is one of the most highly expressed HDACs within neurons in hippocampal and cortical brain regions, 19 and it has been closely associated with the regulation of learning and memory, 20 making it an attractive candidate to combat cognitive decline in AD.In the adult brain, HDAC3 has been involved in the regulation of behavior, learning, and memory.Focal deletion of HDAC3 in the hippocampus of mice enhanced performance in object location cognitive tests, suggesting a role for HDAC3 in the regulation of long-term memory formation. 21Alongside controlling cognitive functioning, HDAC3 appears capable of modulating AD pathology, while promoting correct neuronal function.Specific HDAC3 inhibitor, RGFP966, restored deficits in long-term potentiation (LTP) and reinstated plasticity in CA1 hippocampal neurons that had been injured by exposure to oligomeric Aβ1-42. 22Further studies using transgenic models of AD have reported reductions of Aβ following treatment with HDAC3 inhibitors 7,23 and reductions in microglia density, 7 while other HDAC inhibitors led to contrasting results. 24,25However, the molecular mechanisms whereby these effects were achieved remain unclear.
Therefore, in this study we aimed to investigate the mechanisms behind the beneficial effects of HDAC3 inhibition on AD pathology, using in vitro and ex vivo models of amyloidosis and to determine the effects on neuroinflammatory and neuroprotective pathways.Our results reveal that HDAC3 is able to regulate the transcription of APP and that the outcomes of HDAC3 inhibition on synaptogenesis or glial activation are not related to alterations in Aβ levels, but changes in neuroimmune and neuroregenerative pathways, leading to increased dendritic spine density.

| Antibodies and reagents
The antibodies used for detection of proteins of interest were 6E10 (against Aβ1-16) from Covance/BioLegend (London, UK), R1(57) against the carboxy terminus of the amyloid precursor protein (APP) (a kind gift from Dr P. Mehta, NYS Institute for Basic Research in Developmental Disabilities); anti-beta-APP cleaving enzyme (BACE1) D10E5, anti-HDAC1, HDAC2, HDAC3, H3, and H3k9ac were from Cell Signaling (Leiden, The Netherlands); HDAC3 antibody from Millipore (Merck, BNP Paribas, Dublin) was used for the chromatin immunoprecipitation (ChIP) of genomic DNA isolated from wild type N2a cells; Nuclear receptor corepressor (NCoR) antibody was from Santa Cruz (Dallas, TX, USA); anti-ionized calciumbinding adapter molecule 1 (IBA1) from Wako (Fujifilm, Neuss, Germany); anti-glial fibrillary acidic protein (GFAP) (clone 2.2B10) from Invitrogen (Life Technologies Limited, Paisley, UK), anti-βactin was from Abcam (Cambridge, UK).The specificity of the antibodies against the endogenous mouse protein was assessed in cells transfected with siRNA or in tissue or cells overexpressing that protein (Figures S1 and 3E).Tissue culture reagents were purchased from Invitrogen and Millipore, and all other reagents were purchased from Sigma (Merck Life Science, Gillingham, Dorset, UK), unless stated otherwise.

| Human brains
Frozen brain tissue from human frontal cortex was obtained from the London Neurodegenerative Diseases brain bank and the Huddinge Brain Bank at the Karolinska Institutet, Stockholm, Sweden, in accordance with the laws and the permission of the ethical committees and to the provisions of the Helsinki declaration.15 sporadic AD cases were selected reporting similar pathological features associated with stages five and six of the BrainNet Europe and Braak scales of AD severity (84 years, SD ± 10 years, 6 females, 9 males).13 age-matched control samples were selected from healthy patients with no AD-related symptoms (81 years, SD ± 10 years, 5 females, 8 males).
Cells were incubated for 24 h with the HDAC3 inhibitor RGFP966 (Merck Life Science, Gillingham, Dorset, UK) diluted in dimethyl sulfoxide (DMSO, Merck Life Science, Gillingham, Dorset, UK).Concentrations used ranged from 0.5 to 10 μM.Control cells were treated with media plus vehicle (DMSO).To assess the toxicity of RGFP966 on N2a cells, a cell viability analysis was carried out using the Cell Titer 96 Aqueous One solution (Promega, Southampton, Hampshire, UK).No significant effect on cell viability was observed across the tested concentrations (Figure S2A).

| Organotypic brain cultures
Organotypic brain culture slices (OBCSs) were prepared from postnatal day (P) 7 from wild-type, 5XFAD, 26 Thy-1-GFP, 27 and double transgenics Thy-1-GFP/5XFAD mice as described previously. 28Mice were kept in individually ventilated cages and maintained on a 12/12-h light/dark cycle with controlled temperature and humidity, and food and water ad libitum.In vivo procedures (breeding) were performed in accordance with the United Kingdom Animal (Scientific Procedures) Act (1986) and approved by Imperial College London's Animal Welfare and Ethical Review Body.
The brains were sectioned in 300-μm coronal slices using a vibratome (Leica VT1200 S); following this, sections were mounted on semi-porous membrane filters (Millipore, Merck Life Science, Gillingham, Dorset, UK).Brain slices were maintained in culture for 2 weeks in nutrient media (Neurobasal-A medium (Invitrogen, Life Technologies Limited, Paisley, UK), 20% normal horse serum, 20% HBSS, 0.5% glutamine, 0.5% vitamin B27 supplement, and 1% antibiotics) and incubated at 37°C, 95%HR, 5%CO2.The nutrient medium was replaced every 2 days for 2 weeks.Slices were treated for 24 h with vehicle or with 1 μM of the HDAC3 inhibitor RGFP966 in nutrient media.After incubation, media was collected and brain slices were either homogenized or fixed in 4% PFA for 24 h and kept at 4°C in PBS and Na-azide.Tissue integrity in OBCSs was determined by propidium iodide (PI) staining, as described previously. 28o assess the toxicity of RGFP966 on neurons, double transgenic Thy-1-GFP/5xFAD mouse OBCSs were incubated with 1 and 1.5 μM HDAC3 inhibitor RGFP966 or DMSO vehicle control for 24 h.Toxicity was assessed through the appearance of neuronal branch "beading", 29 which have a "rosary bead appearance."This feature, which was apparent at the concentration of 1.5 μM of the drug (Figure S1B).Beading is considered an early step in neuronal degeneration, a step that precedes fragmentation.Therefore, 1 μM was the concentration chosen.

| Enzyme-linked immunosorbent assays
The levels of human Aβ40 and Aβ42 were determined in homogenates or in conditioned media using the High Sensitivity Human Amyloid β42 and Aβ40 enzyme-linked immunosorbent assays (ELISA) kits from Millipore (Merck Life Science, Gillingham, Dorset, UK).Conditioned cellular medium was used neat or diluted in kits' sample buffer and protein concentration was expressed as pg/mL or as percentage relative to control.

| Chromatin immunoprecipitation
In order to determine whether HDAC3 DNA binding was enriched in the APP promoter region, ChIP analysis was carried out to isolate DNA fragments with strong association with HDAC3.For this purpose, N2a cells were seeded into 10 cm culture dishes, with 5.5 mL DMEM, and maintained at 37°C at 5% CO2 until 90% confluence was achieved.Two dishes, corresponding to approximately 4 × 10 6 cells, were used per sample.Chromatin IP was performed using The SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads) (Cell Signaling Technology, Leiden, The Netherlands), following manufacturer instructions (https:// www.cells ignal.co.uk/ produ cts/ chip-kits-reage nts/ plus-enzym atic-chrom atin-ip-kit-magne tic-beads ).Briefly, cells were fixed with 1% formaldehyde, washed, and lysed.Nuclei were treated with 0.5 μL of micrococcal nuclease for 20 min at 37°C and digested chromatin was incubated with 1 μg of HDAC3 antibodies (Invitrogen, Life Technologies Limited, Paisley, UK) or negative control IgG (rabbit) antibody (Jackson ImmunoResearch, Cambridge, UK).After elution, cross-linking reversal, DNA was purified and quantified via qPCR, using PowerUp™ SYBR™ Green Master Mix (Thermo Fisher Scientific Ltd, Loughborough, UK) and different primers for mouse APP promoter region (see Table 1).Cycles were as follows: hot start: 15 s at 95°C, 40 amplification cycles: 20 s at 95°C, 20 s 56°C, 30 s 72°C, followed by melting analysis for specificity.

Primer
Sequence Novogene to detect differentially expressed genes (p < .05).Gene Ontology analysis was performed using Novomagic tool provided by Novogene, using all the expressed genes as background.

| Immunofluorescence staining
OBCSs slides sections were permeabilized overnight in 0.25% TBS-TX (TBS with triton X-100).Following this, sections were washed again in TBS and blocked for 60 min in 10% goat serum/1%BSA in 0.1% TBS-Triton X and incubated with the primary antibodies (anti-GFAP, 1:500, anti-IBA1 1:500) diluted in 2% goat serum, 0.2% BSA in TBS-TX 0.02% for 48 h at 4°C.Following this, sections were washed 5 times for 10 min in TBS and incubated with the secondary fluorescent antibodies (1:400 Alexa Fluor; Invitrogen) in 2% goat serum, 0.2% BSA in TBSTX 0.02% overnight at 4°C.After incubation with the secondary antibodies, sections were washed 4 times with TBS for 10 min.In the final wash, Hoechst solution (for nuclear visualization) was added (1:1000 in TBS) for 5 min.Following this, the slices were washed and mounted using ProlongTM Gold antifade reagent (Invitrogen, Life Technologies Limited, Paisley, UK).Sections were visualized with a confocal microscope (Zeiss LSM-780).Immunohistochemistry staining was quantified using the HALO software (Indico Labs, London, UK) using the area quantification FL module and represented as percentage of the total image area.

| Quantification of dendritic spines and analysis of microglia
Spine size was calculated as integrated brightness, using an adapted version of custom-written MATLAB code, as described previously. 28or microglia analysis, a fluorescence intensity profile, taken from a maximum intensity projection, was drawn through the longest dimension of the cell body and across a cross-section of the outermost tips of all associated processes.This intensity profile was then used to estimate the soma size as well as the number, area, and the perimeter of processes using custom-written code in MATLAB. 28

| Statistical analysis
The number of animals and group size to be used was determined via InVivoStat, an R-based statistical package.
The data were analyzed using GraphPad Prism 8 software (GraphPad Software Inc, Boston, MA, USA), using unpaired Student's t tests, repeated-measures 1-or 2-way analyses of variance, and followed by Tukey's or Dunnett's post hoc analysis.The Kolmogorov-Smirnov test was applied to confirm normal distribution.All quantitative data are given as mean ± SEM.Probability of less than .05was considered significant.

| HDAC3 expression is increased in the cortex of AD patients
Due to the discrepancies reported in previous studies regarding the expression of HDACs in AD, we sought to determine whether specific HDAC subtypes are altered in the brains of AD sufferers.For this purpose, HDAC expression profiles were determined in post-mortem frontal cortex homogenates of AD patients and age-matched healthy donors.The protein expression of class I HDACs (HDAC1, HDAC2, and HDAC3) was determined and normalized to constitutive βactin levels (Figure 1).No observable changes in HDAC1 and HDAC2 expression were found in AD patients (Figure 1A,B).Interestingly, a significant 28% increase was observed in HDAC3 among AD patients (t = 1.976, df = 26, p < .05,1.277 ± 0.41 SD, n = 15), relative to healthy counterparts (1 ± 0.32 SD, n = 13) (Figure 1C).

| HDAC3 inhibition regulates the expression of APP
Following this, we explored whether specific HDAC3 inhibition affected amyloid-β and the expression and processing of APP in vitro.The effect of HDAC3 inhibitor RGFP966 on the expression of full-length APP was assessed by Western blot using the 6E10 antibody in whole cell lysates of N2aSW cells, an in vitro AD model, which overexpresses APP with the Swedish mutation.N2aSW cells were incubated with concentrations of RGFP966 ranging from 0.5 to 10 μM or control vehicle, showing effective reduction in HDAC3 activity (Figure S1E) and no changes in Aβ1-42 in conditioned media (Figure 2A).Interestingly, elevations in the expression of full-length APP were observed with increasing concentrations of RGFP966, reaching a two-fold increase at μM (p <.0001, n = 6-14) (Figure 2B).Similar elevations in full-length APP were observed in cells where HDAC3 had been knockdown (Figure 2C).This increase in full-length APP expression paralleled with elevations in sAPPα in the conditioning media at 5 μM RGFP966 (Figure 2D).Higher sAPPα levels were consequence of changes in full-length APP rather than in the activity of the αsecretase, because normalizing values to full-length APP showed no differences between treatments (Figure 2E).
Because N2asw cells overexpress human APP, we next determined whether the levels of endogenous mouse APP were also altered by HDAC3 inhibition, by performing the incubation with RGFP966 in un-transfected N2a cells.In line with the results in N2asw cells, a significant increase in endogenous full-length APP levels was observed in N2a cells treated with 5 μM (p < .0001,1.852 ± 0.37 95% CI, n = 6) and 10 μM HDAC3 inhibitor (p < .0001,2.627 ± 0.37 95% CI, n = 6) (Figure 3A).Consistent with higher APP levels, the expression of the carboxy-terminal APP fragments (CTFs) showed a similar rise in response to HDAC3 inhibition (Figure 3B, p < .001,df = 17), as CTF levels were more than doubled at 5 μM RGFP966 (p < .001,2.194 ± 0.57 95% CI, n = 6).This was not due to differences in βsecretase, since BACE1 expression levels remained unaltered in cells treated with RGFP966 (n = 6) (Figure 3C).Knockdown of HDAC3 in N2a cells by transfection with HDAC3 siRNA resulted in increased APP expression as well (Figure 3E,F).
Next, we investigated whether the increases in APP protein expression were related to an increase in mRNA levels or a reduction in its degradation, we measured APP mRNA levels by qPCR.Incubation with 10 μM RGFP966 resulted in increased levels of APP mRNA by 2.5-fold as compared to control cells (p = .0236,n = 6) (Figure 3D).
To establish whether HDAC3 can bind APP promoter, suggesting a role in APP transcriptional regulation, we took advantage of a published HDAC3 ChIP-sequencing data in mouse hippocampus (GSE72196 30 ), showing enriched binding of HDAC3 occurred close to the transcription start site (TSS) of the APP gene (Figure 3G).This indicated that HDAC3 might mediate APP transcription, via epigenetic mechanisms.
To determine whether the increased APP mRNA after HDAC3 inhibition was underpinned by HDAC3-dependent transcriptional regulation or HDAC3, we directly explored the binding of HDAC3 on APP gene promoter in N2a cells.ChIP analysis was carried out on wild-type N2a cells using antibodies for HDAC3 versus preimmune IgG, and three pairs of primers, specific to the previously identified sites for HDAC3 binding at the APP promoter (Figure 3G,H), were tested.Chromatin IPs showed a specific HDCA3 binding at the APP promoter, with 5%-10% enrichment of over IgG negative control (Figure 3I).Since the deacetylase activity of HDAC3 mostly requires interaction with Nuclear Corepressor (NCoR), [31][32][33] HDAC3 ChIP samples were probed via WB using NCoR antibody, showing the interaction between HDAC3 and NCor in N2a cells (Figure 3J).In line with this, overexpression of NcoR in N2a cells resulted in a reduction in APP mRNA, which was restored by inhibition of HDAC3 (Figure 3D).

| HDAC3 inhibition regulates the expression of APP in organotypic brain cultures (OBCS)
Following this, we assessed whether the effects of HDAC3 inhibition were reproduced in an ex vivo model of AD, using organotypic brain cultures (OBCSs) of transgenic 5xFAD mouse pups.OBCSs provide the possibility of investigating the effects of pharmacological interventions on molecular changes in pathology in a system where all cell types are present and preserved within their original architecture.OBCSs were treated with 1 μM of HDAC3 inhibitor, RGFP966, or control vehicle for 24 h (Figure 4A).We chose this concentration as higher concentrations of this drug resulted in dendritic damage (Figure S2B).Interestingly, in line with the results obtained in vitro in N2a cells, we detected a 30% increase in the expression of APP in OBCSs treated with the HDAC3 inhibitor (p = .0367,n = 14-15) (Figure 4B) and a 50% increase in total-CTFs (p = .0164,n = 14-15) (Figure 4C).In agreement with the observations in cells, treatment of OBCSs with 1 μM of RGFP966 did not lead to changes in the expression of BACE1 protein in OBCSs as compared to controls (Figure 4D).
To investigate the overall effect of HDAC3 inhibition on the expression of amyloid-β species, we measured the levels of Aβ42 and Aβ40 in homogenates of 5xFAD OBCSs by ELISA.Incubation with 1 μM RGFP966 did not lead to overall changes in the expression of either amyloid-β species in OBCS homogenates (Figure 4E,F).To further support that HDAC3 inhibition was not involved in the generation or degradation of Aβ, qPCR analysis of mRNA extracts from transgenic slices revealed no changes in any of the enzymes involved in the generation or degradation of Aβ, in OBCSs treated with the HDAC3 inhibitor (Figure 4G-L).

| HDAC3 inhibition attenuates reactive inflammatory profile in 5XFAD OBCSs
It is known that Aβ in AD models promotes a neuroinflammatory response, leading to synaptic and neuronal loss.Therefore, we sought to define the effects of HDAC3 inhibition on glial cell density, on their morphological changes as well as on neuroinflammatory mediators in the OBCSs.
We first investigated whether HDAC3 inhibitor RGFP966 affected the density of microglia, by staining WT and 5XFAD transgenic OBCSs from frontal cortex and hippocampus, two areas highly affected by AD, with antibodies for Ionized calcium-binding adapter mole-cule1(IBA-1), a marker of microglia.Analysis of the percentage area stained revealed that the density of microglia of 5XFAD transgenic sections was increased compared to wild-type OBCSs in the frontal cortex (2.5-fold increase, p = <.0001,n = 13-18) (Figure 5A-E) and in the hippocampus (190% increase, p = <.0001,n = 14-24) (Figure 5F-J).Interestingly, in the hippocampus, incubation of OBCSs with RGFP966 led to a significant 60% decrease in the density of microglia in the transgenic 5xFAD OBCSs compared with vehicle-treated controls (p = .0001,n = 13-14) (Figure 5J).This finding was not observed in hippocampal WT OBCSs.A minor reduction, although not significant, was found in the frontal cortex of transgenic 5xFAD OBCSs (Figure 5E).
To examine the effects of RGFP966 on microglial activation state, we performed a morphological analysis of microglia by measuring parameters of microglial shape that are known to undergo change during glial activation.In the frontal cortex of 5XFAD OBCS, analysis of microglial morphology revealed significant elevations in process number, process area, and the size of microglial soma compared with WT OBCS (Figure 6).Interestingly, RGFP966 incubation led to a reduction in the number of processes, an increase in the total size of microglia and a growth in the size of the soma.In the frontal cortex, analysis of microglial morphology revealed significant elevations in process number (p = .0015,n = 13-24), process area (p = .0103,n = 13-24) and the size of microglial soma (p = <.0001,n = 13-24) in OBCSs from transgenic animals compared to wild-type slices (Figure 6A-G), indicative of a reactive phenotype.In the hippocampus, we did not detect differences in the morphology of microglial between WT and transgenic OBCSs (Figure 6H-N).In addition, we found that in transgenic 5xFAD OBCSs, RGFP966 incubation led to a significant decrease in the process area in microglia in cortex and hippocampus as compared with those treated with vehicle control, suggesting that the total size of the microglial cell was reduced (Figure 6M).However, this effect was not seen in the WT cultures.
To investigate the effect of HDAC3 inhibition on reactive astrogliosis, we measured the density of GFAPpositive astrocytes, by percentage area of GFAP staining, in the frontal cortex and hippocampus of WT and transgenic 5xFAD OBCSs following treatment with 1 μM of RGFP966 or vehicle control for 24 h.We observed an over 3-fold increase in the density of GFAP-positive astrocytes in the 5xFAD frontal cortex (p = <.0001,n = 12-18) (Figure 5K-O) and a 4-fold increase in the hippocampus (p = <.0001,n = 13-19) compared to wild-type OBCSs (Figure 5P-T).In 5xFAD OBCSs, incubation with HDAC3 inhibitor RGFP966 resulted in a significant decrease in the density of GFAP-positive astrocytes both in the frontal F I G U R E 3 APP expression is regulated by HDAC3 in N2a cells.(A) Representative Western blot and quantification of APP protein expression shows a significant increase by incubation with the HDAC3 inhibitor RGFP966 (5-10 μM).(B) Representative Western blot and quantification of total APP carboxy-terminus fragments (CTF) expression levels (using R1-57 antibody) in N2a cells treated with RGFP966.(C) Representative Western blot and quantification of BACE1 levels in N2a cells with RGFP966 treatment (the representative western for βactin is the same as for Figure 3B, since they were blotted in the same membrane) (D) qPCR analysis of mouse APP mRNA in N2a cells transfected with NcoR and treated with or without HDAC3 inhibitor RGFP966 (10 μM).cortex (43% reduction, p = n = 12-13) (Figure 5O) and hippocampus (47% reduction, p = .0019,n = 13-14) (Figure 5O,T).However, we did not detect changes in astrocytic density with RGFP966 treatment in wild-type OBCSs (Figure 5O,T).Taken together, these results suggest that transgenic OBCSs exhibit a reactive inflammatory profile as compared to WT OBCSs, and this effect is reversed by treatment with RGFP966.
To define whether changes in the glial activation detected by staining were associated with alterations in genes involved in neuroinflammation, mRNA sequencing analysis was conducted in mRNA extracts from 5XFAD OBCSs treated with the HDAC3 inhibitor and vehicle controls.HDAC3 inhibition resulted in 127 Upregulated and 204 downregulated genes in 5XFAD slices (p < .05)(Figure 6O,P).Interestingly, mRNA sequencing analysis showed molecular changes in markers of neuroinflammation driven by HDAC3 inhibition.Down-regulated genes were enriched for categories related to metabolism, cell adhesion, chemotaxis, proteolysis, and TGF-β production, while several genes largely expressed in activated astrocytes, such as AQP, Serpinb1, ApoD, and CryaB were also found reduced by GFP966 incubation (Figures 5U and 6Q).Interestingly, some of these HDAC3-downregulated genes, such as AQP1, Cryab, Col8a1, Plin5, and TGFbr3, were found to be overexpressed in single nuclei dataset of human astrocytes from AD patients (Figure 5U, genes in red). 34The mRNA sequencing analysis also revealed upregulations in the expression of genes involved in T-cell activation, chemotaxis, cytokine signaling, and gliogenesis, including transcription factors Sox11 and Sox4 (Figure 5V).These molecular changes possibly reflect a less activated and a more modulatory glial state in 5XFAD OBCSs treated with the HDAC3 inhibitor.

| HDAC3 inhibition activates neurotrophic pathways and synaptic function
To assess the effects of HDAC3 inhibition on synaptic connectivity, we measured the dendritic spine density and size in the CA1 and CA3 areas of the hippocampus of OBCSs from wild-type mice expressing Thy-1-GFP and double transgenic Thy-1-GFP/5xFAD mice treated with 1 μM of RGFP966 or vehicle control for 24 h (Figure 7A-C).
Quantification of spine density revealed that the density of hippocampal spines in transgenic 5xFAD OBCSs was 22% lower than in wild-type OBCSs (p = .0142)(Figure 7A-C).Incubation of 5xFAD OBCSs with RGFP966 increased spine density by 25% compared with vehicle-treated slices (p = .0077)and compared to wildtype OBCSs (Figure 7A,C).However, treatment with the HDAC3 inhibitor did not affect spine number in WT cultures (Figure 7A,C).
While average spine size did not differ between WT and transgenic OBCSs, incubation of OBCSs with RGFP966 resulted in a 22% decrease in the average spine size in dendritic spines of wild-type OBCSs (p = .0172),whereas there were no significant differences in transgenic OBCSs (Figure 7B,C).
This increased synaptic connectivity was associated with a strong enrichment of up-regulated genes in functional categories involved in CNS neuron and glia differentiation in OBCSs from 5XFAD treated with the HDAC3 inhibitor, including transcription factors and genes involved in axonogenesis, synaptic function, and calciumbinding proteins (Figure 7D,E).
Together, these results suggest that increases in the synaptic connectivity by HDAC3 inhibition in 5XFAD OBCSs are not related to reductions in Aβ pathology, but to an enhancement in neurotrophic and neuroprotective pathways.

| DISCUSSION
Our study has revealed that HDAC3 inhibition plays a crucial role in AD models, not by affecting Aβ levels, but by inducing the expression of anti-inflammatory and neuroprotective genes.This is consistent with previous studies that have shown that epigenetic-based therapies can improve cognitive function in AD models, 20 but challenges the assumption that these effects are mediated by changes in amyloid pathology.
Histone modifications have been associated with the regulation of a number of Aβ-related genes. 1 Intriguingly, a recent study in post-mortem AD and elderly brains identified a significantly hypoacetylated peak located downstream of the APP gene. 1 In our study, we have identified HDAC3 enrichment at the APP promoter in N2a cells, together with transcriptional elevations in APP protein and mRNA levels following incubation with 10 μM of the HDAC3-selective inhibitor, RGFP966, or by HDAC3 knockdown.These results are in agreement with increases in APP protein expression in response to 10 μM of RGFP966 reported by Janczura et al. 23 in HEK/APPsw cells.Further support for the notion that HDAC3 regulates APP in a transcriptional manner stems from the presence of HDAC3 binding near to the APP gene transcription start site in the hippocampus of wild-type mice. 30In line with this, other studies have shown that APP promoter activity is sensitive to changes acetylation status and suggest that HDACs play an intrinsic role in the regulation of APP, both constitutively present to control expression and recruited by transcriptional repressors, such as NCoR 35 and others involved in T3 signaling. 36evertheless, the expression changes in APP following HDAC3 inhibition appear to have wide-reaching consequences for APP processing, as they were accompanied by elevations in sAPPα but small changes in amyloid-β42, in contrast with previous publications. 7,23As we did not observe changes in the activity of the αsecretase in vitro, we suggest that the elevations in these APP cleavage fragments are more closely related to an overall increase in APP expression than changes in this enzyme's activity.However, it is also plausible that HDAC3 has other posttranslational regulatory effects that govern the overall of APP.For example, studies have that epigenetic mechanisms, such as DNA methylation and histone modifications, could regulate alternative pre-mRNA splicing of APP isoforms, which could influence the generation of more pathogenic-associated APP isoforms over others. 37ontrary to some studies, our results show that HDAC3 inhibition did not lead to changes in Aβ42 expression in vitro and ex vivo.This highlights the potential variability in the effects of HDAC3 inhibition, depending on the experimental model, the concentration of the inhibitor, and the length of the treatment.For example, a study from Janczura and colleagues 23 reported that RGFP966 treatment in 293-HEK cells and 3xTg AD mice led to decreases in Aβ42, while increasing Aβ40 and the expression of PS2.Despite finding elevations in APP expression in HEK-293 cell lines treated with the HDAC3 inhibitor, this group attributed the reduction of Aβ42 in the in vivo treatment to a decrease in BACE1 expression and an increase in Aβ degradation mechanisms.In another study, lentiviral targeted knockdown of HDAC3 in hippocampus of APP/PS1 mice also led to decreases in amyloid-β, 7 and this was associated with a reduction in the expression of PS1, a component of the γsecretase.Our results show that treatment with RGFP966 in N2a cells and in organotypic cultures of 5XFAD mice did not lead to changes in the mechanisms of Aβ generation (PS1 or BACE1) or in Aβ degradation (neprilysin, IDE, and ApoE).On the other hand, Class I HDAC inhibitor, Trichostatin A (TSA), was shown to induce elevations in plasma Aβ40 and Aβ42 in APP/PS1 mice, 24 while similar Class I HDAC inhibitor treatment with Valproic acid (VPA) in APP23 mice and in primary neurons reduced amyloid, also suggesting that this was a consequence of decreases on γsecretase. 25owever, these inhibitors target a host of HDACs, and therefore, the responsible mechanisms are difficult to outline.Interestingly, we found that at high concentrations (>1 μM) RGFP966 affected dendritic spine survival; therefore, many of the previous publications showing results at 10 μM concentrations might be not physiologically relevant and indeed may result in potential neuronal damage.
Studies performed in models of AD and neurodegeneration have highlighted the effects of HDAC3 inhibition on microglia activation.Lentiviral-mediated knockdown of HDAC3 in hippocampus of APP/PS1 mice led to a decrease in IBA-1 density, while HDAC3 overexpression resulted in an increase in microglia activation. 7his anti-inflammatory effect of HDAC3 does not seem to be secondary to changes in Aβ levels, as we have also demonstrated in our present study, and evidences of similar effects have been reported in other neurodegenerative models.In line with this, RGFP966 suppressed inflammation and improved recovery following injury in a model of spinal cord injury. 38Interestingly, it was recently published that microglia-specific deletion of HDAC3 promoted inflammation resolution in a model of traumatic brain injury, 39 while HDAC3 K.O.macrophages displayed an impaired ability to respond to inflammatory stimuli. 40he molecular mechanisms underlying Hdac3's regulation of inflammatory gene expression in microglia are not well understood.A recent study suggests that Inhibition of Hdac3 enhanced histone acetylation globally and at specific gene loci, resulting in the release of gene repression at baseline and enhanced responses to LPS in microglial cultures.Therefore, this indicates that inhibition of Hdac3 facilitates the microglial response to inflammation and its subsequent resolution. 41Other previous reports also found that treatment of RGFP966 prior LPS in primary microglia reduced STAT, AP-1, and NF-κB signal transduction pathways. 42egarding other glial cells, such as astrocytes, histone acetylation has been implicated in suppressing GFAP expression previously, and in particular, siRNA knockdown of HDAC3 was found to reduce GFAP expression in vitro. 43ecently, HDAC3 inhibition was shown to modulate the reactive astrocytic phenotype, decreasing the expression of reactive markers and release of pro-inflammatory cytokines, using LPS models of neuroinflammation and traumatic injury, 44 while other studies have reported reduced expression of GFAP in an in vivo model of Huntington's disease, 45 yet the efficacy of this inhibition has not been described in AD animal models. 7On the other hand, Bailey et al. 46 described that astrocytes are hypoacetylated following traumatic brain injury (TBI), and suggested that histone deacetylation in astrocytes may give rise to chronic inflammation by up-regulating inflammatory mediators.However, whether increased histone acetylation in astrocytes curbs their induction of inflammatory responses in AD remains to be seen.
In our analysis of double transgenic 5xFAD/Thy-1GFP OBCSs, we noted an enhancement of both dendritic spine size and density in response to RGFP966 incubation, not seen in wild-type sections.A similar finding was shown in vivo in APP/PS1 mice whereby HDAC3 inhibition increased spine density, which was accompanied by increases in important synaptic proteins such as synaptophysin, BDNF, and p-CREB in the hippocampus of these animals. 7In agreement with this, HDAC3 has been reported to repress the expression of factors such as BDNF 47 and Npas4, 48 which are able to induce synaptic plasticity, 49 while HDAC3 inhibition promoted the opposite effect. 50In line with this, our analysis found enriched binding of HDAC3 close to the transcription start site (TSS) of neuronal genes Klf10, Tle1, and BDNF 30 (Figure S2C).Furthermore, knockout of HDAC3 in hippocampal neurons resulted in the elevation of neuronal genes Snap25 and neurogranin (Nrgn), 30 which have been linked to dendritic spine plasticity and function. 51,52hich was also increased by HDAC3 inhibition in our study, has neurotrophic and neuroprotective properties. 53In addition, increases in APP expression have been found associated with synaptodendritic protein synthesis. 54,55Our results obtained from RNAseq support that HDAC3 inhibition affects neuronal regenerative and trophic pathways, increasing the expression of genes involved in CNS neuron differentiation, and synaptic activity.
The main limitation of the present study was the use of only in vitro and ex vivo models.Although this methodology allows a more thorough manipulation of the conditions, to investigate molecular mechanisms, in the future it would be relevant to use in vivo models or even human models, such as induced pluripotent stem cells (IPSC) cultures or brain organoids, to have a more translational application to human disease.
In conclusion, our study provides a comprehensive analysis of the effects of HDAC3 inhibition in AD models, shedding light on its regulation of APP and other neuroimmune, neuroprotective, and pro-resolving factors.Given the increased levels of HDAC3 detected in AD brains, our findings suggest that HDAC3 inhibition could be a promising therapeutic strategy for AD and other neurodegenerative disorders.

F
I G U R E 1 HDAC3 protein levels are increased in the frontal cortex of AD patients relative to healthy age-matched individuals.(A-C) Representative images of Western blots with quantification from densitometric analysis of homogenized frontal cortex tissue from human donors.(A) HDAC1 levels show no significant variation between AD sufferers and healthy patients (n = 13).(B) HDAC2 levels are not significantly altered in AD sufferers compared to healthy controls (n = 15).(C) Frontal cortex homogenates from AD patients show significantly elevated HDAC3 levels (n = 15).Individual data points are shown with columns representing the mean ± SEM and significance levels indicated under one-tailed unpaired student's t-tests (*p < .05).

F I G U R E 2
Effect of HDAC3 inhibition on APP expression and processing in N2aSW cells.(A) ELISA of Aβ42 in media from N2aSW cells incubated with 10 μM of HDAC3 inhibitor RGFP966 or control, n = 6.(B) Quantification and representative Western blot of full-length APP measured in cellular lysates of N2aSW cells incubated with 0.5-10 μM of HDAC3 inhibitor RGFP966 or control, normalized to βactin.(C) Quantification and representative Western blot of APP in N2asw cells transfected with control siRNA or HDAC3 siRNA (Dharmacon ID:J-043553-17-0002) (D) Quantification and representative Western blots of sAPP-α measured in conditioned cellular medium, normalized to βactin in whole cell lysates (the representative western for βactin is in cell lysates and is the same as for (B), since they were from the same experiment).(E) The activity of the αsecretase calculated as the expression of soluble-APPα in the cellular medium normalized to fulllength APP in the cell lysates in N2aSW cells.Results are represented as mean ± SEM.Statistical analysis was performed one-way ANOVA with Dunnett's multiple comparisons test or by two-tail unpaired Student's t test, (*p < .05,**p < .01,***p < .001,****p < .0001).

34 F I U R E 6
(E) Quantification and (F) Representative Western blot of APP in N2a cells untransfected (CTL), transfected with control siRNA or HDAC3 siRNA (Dharmacon ID: J-043553-05-0002).(G) Integrative Genomics Viewer (IGV) visualization of the HDAC3 ChIPseq (signal density across the TSS and gene body of APP gene).The green track represents the called peak (GSE72196), and the blue track represents example of the HDAC3 trace in one replicate (GSM2131238).(H) Schematic of the APP promoter region (TSS ± 600 BP).The regions amplified by the three pairs of primers are highlighted.(I) Quantification of HDAC3 binding to the APP promoter of N2a cells, as obtained by ChIP and real-time PCR analysis.Values are expressed as a percentage of the input DNA (mean ± SEM of 3 technical replicates).The number above represents the enrichment over rabbit IgG (IGG = 1).(J) HDAC3 ChIP samples were probed via WB using NCoR antibody, showing the interaction between HDAC3 and NCor in N2a cells.As controls, lysates from N2a cells overexpressing NcoR or untransfected were used, as well as IgG alone.Columns represent the mean ± SEM and significance levels indicated under one-way ANOVA with Tukey's or Dunnett's post hoc modifications or by two-tail unpaired student's t-test (*p < .05,**p < .01,***p < .001,****p < .0001).F I G U R E 4 Effect of HDAC3 inhibition on APP processing in ex vivo 5xFAD OBCSs.(A) Schematic of the timeline and treatment of organotypic brain cultures with HDAC3 inhibitor RGFP966.(B) Quantification and representative Western blot of full-length APP expression in OBCS lysates treated with control or 1 M of RGFP966, normalized to β-actin, n = 14-15.(C) Quantification and representative Western blot of APP-CTFs in OBCS treated with vehicle control or 1M of RGFP966, normalized to β-actin, n = 14-15.(D) Quantification and representative Western blot of BACE1 expression in OBCS lysates, normalized to β-actin, n = 8-9 (the representative western for β-actin is the same as for Figure 4C, since they were blotted in the same membrane) (E) Expression of Aβ-40 (n = 16) and (F) Aβ-42 (n = 14-19) in homogenates of OBCSs; qPCR analysis of mRNA extracts from 5XFAD OBCSs treated with RGFP966 for BACE1 (G), PS1 (H) ADAM-10 (I), Neprilysin (NEP) (J), Insulin degrading enzyme (IDE) (K) and ApoE (L) (n = 6-8 per group).Results are represented as mean ± SEM.Statistical analysis was performed using two-tail unpaired t-tests, (*p < .05).F I G U E 5 HDAC3 inhibition affects glial activation in transgenic OBCSs, but not in wild-type slices.(A-D) Representative images of IBA-1 staining in cortex of wild type (WT) and 5XFAD OBCSs treated with 1 μM of HDAC3 inhibitor.(E) quantification of the percentage area of IBA-1 staining in cortex of OBCSs treated with vehicle control or 1 μM of HDAC3 inhibitor for 24 h (n = 13-24).(F-I) Representative images and (J) quantification of the percentage area of IBA-1 staining in hippocampus of OBCSs treated with vehicle control or 1 μM of HDAC3 inhibitor for 24 h (n = 13-24).(K-N)Representative images and (O) quantification of the percentage area of GFAP staining in the frontal cortex of wild-type and 5xFAD OBCSs.(P-S) Representative images and (T) quantification of the percentage area of GFAP staining in the hippocampus of WT and 5xFAD OBCSs.Results are represented as mean ± SEM.Statistical analysis was performed using a One-way ANOVA, (*p < .05,**p < .01,***p < .001,****p < .0001).(U, V) Tables showing the gene name and relative function of differentially expressed genes in RNAseq analysis of 5XAD OBSCs treated with the HDAC3 inhibitor compared with control vehicle found in Gene Ontology enriched categories.U represents down-regulated genes and V Up-regulated genes and their functions.The genes shown in red have been found up-regulated in human AD astrocytes single nuclei RNAseq datasets.HDAC3 inhibition affects glial activation in transgenic OBCSs.(A-D) Representative images of IBA-1 positive microglia in the frontal cortex and (E) Quantification of process number, (F) process area, and (G) soma size of IBA-1 positive microglia in the frontal cortex of wild-type and transgenic 5xFAD OBCSs treated with 1 μM of HDAC3 inhibitor (n = 13-24).(H-K) Representative images of IBA-1 positive microglia in the hippocampus and (L) quantification of process number, (M) process area, and (N) soma size of IBA-1 positive microglia in the hippocampus of wild-type and transgenic 5xFAD OBCSs treated with 1 μM of HDAC3 inhibitor (n = 13-24).Values shown in graphs represent the mean value ± SEM.Statistical analysis included a One-way ANOVA with Tukey's multiple comparison post-test, (*p < .05,**p < .01,***p < .001,****p < .0001).(O) Volcano plot of mRNA seq analysis comparing 5XFAD treated with HDAC3 inhibitor vs 5XFAD vehicle.(P) Differential expressed (DE) genes between 5XFAD with HDAC3 inhibitor vs 5XFAD vehicle, up-regulated genes (red) and down-regulated genes (green).(Q) Gene Ontology Enrichment analysis scatter plot of the down-regulated genes (p < .05) in TG OBCs treated with HDAC3i vs vehicle.Clusters are color-coded to reflect the q-value, and the size reflects the gene number in each category.

F I G U R E 7
Effect on HDAC3 inhibition on synaptic plasticity in hippocampal neurons ex vivo.(A) Quantification of spinal density and (B) size and (C) representative images of dendritic spines in the hippocampus of OBCSs generated from wild-type thy-1-GFP and 5xFADthy-1-GFP mice incubated with 1M HDAC3 inhibitor RGFP966 or control for 24 h, n = 24-32.Values shown in graphs represent the mean value ± SEM.Statistical analysis included a One-way ANOVA with Tukey's multiple comparison post-test, (*p < .05,**p < .01,****p < .0001).(D) Gene Ontology Enrichment analysis scatter plot of the up-regulated genes (p < .05) in TG OBCs treated with HDAC3i vs vehicle.The first 22 statistically significant (p < .05)categories are shown.Clusters are color-coded to reflect the q-value, and the size reflects the gene number in each category.(E) Tables showing the gene name and relative function of differentially expressed genes in RNAseq analysis of 5XAD OBSCs treated with the HDAC3 inhibitor compared with control vehicle found in Gene Ontology enriched categories.