RXR controlled regulatory networks identified in mouse brain counteract deleterious effects of Aβ oligomers

Bexarotene, a selective agonist for Retinoid X receptors (RXR) improves cognitive deficits and amyloid-β (Aβ) clearance in mice. Here we examine if the effect of bexarotene on RXR cistrome and transcriptomes depend on APOE isoform and Aβ deposition. We found bexarotene increased RXR binding to promoter regions in cortex of APOE3 mice. Rho family GTPases and Wnt signaling pathway were highly enriched in ChIP-seq and RNA-seq datasets and members of those pathways - Lrp1, Lrp5, Sfrp5 and Sema3f were validated. The effect of APOE isoform was compared in APOE3 and APOE4 mice and we found significant overlapping in affected pathways. ChIP-seq using mouse embryonic stem cells and enrichment levels of histone marks H3K4me3 and H3K27me3 revealed that, bexarotene induced epigenetic changes, consistent with increased neuronal differentiation and in correlation with changes in transcription. Comparison of transcriptome in APOE3 and APP/APOE3 mice revealed that amyloid deposition significantly affects the response to bexarotene. In primary neurons, bexarotene ameliorated the damaged dendrite complexity and loss of neurites caused by Aβ42. Finally, we show that the disruption of actin cytoskeleton induced by Aβ42 in vitro was inhibited by bexarotene treatment. Our results suggest a mechanism to establish RXR therapeutic targets with significance in neurodegeneration.


Results
Bexarotene affects the distribution of RXR binding to genomic sites in mouse brain. In our recent study we demonstrated bexarotene increased RXR binding to genomic sites associated with genes involved in neuronal differentiation and neuronal projections in APOE4 mice 22 . To determine if RXR binding is affected by APOE isoform we applied chromatin immunoprecipitation followed by high-throughput sequencing -ChIPseq, using cortices of bexarotene treated APOE3 mice. The regions of enrichment in RXR datasets were identified at False Discovery Rate (FDR) < 0.01. RXR binding sites were connected to 562 genes in vehicle treated mice and 423 genes in bexarotene treated mice (Fig. 1A). Importantly, even though the number of total binding sites was slightly decreased by bexarotene, the treatment caused a significant increase in RXR binding to promoters. As shown on Fig. 1B there was a significant shift towards RXR binding in promoter regions of annotated genes in bexarotene treated, compared to control mice (10% vs 0.4%, Fig. 1B). In contrast, RXR binding to intergenic regions was significantly decreased (40% in bexarotene vs 55% in control). Functional annotation clustering using DAVID 26 revealed that Gene Ontologies (GO) Biological Process (BP) "transcription"; "chromatin remodeling", "neuron projections" and "axon guidance" were significantly increased in bexarotene treated mice (Fig. 1C) in agreement with our published data for APOE4 mice 22 . To validate the ChIP-seq datasets, we performed ChIP-qPCR for selected genes with enriched RXR binding using cortices of bexarotene and vehicle treated APOE3 mice (Fig. 1D). As shown, Neat1 (Nuclear paraspeckle assembly transcript 1), Eef2 (Eukaryotic elongation factor 2), Kdm6b (H3K27 demethylase 6b) and Wnt3 (wingless-type MMTV integration site family, member 3) had an increased RXR binding in Bexarotene treated samples. We used IPA web tool, as previously 22 to compare the biological function categories affected by bexarotene in APOE3 and APOE4 mice. As shown on Fig. 1E, the comparison of those categories identified by ChIP-seq showed that several pathways related to "neuronal development", "neurite extensions" and "axonal guidance" were highly enriched by bexarotene treatment in both APOE3 and APOE4 mice. These results demonstrate there is a common signature of Genome-wide changes in RXR cistrome in response to bexarotene in APOE3 and APOE4 mice.
Bexarotene significantly affects expression of genes related to nervous system development and neurites extension. Next, to determine how bexarotene affects mRNA levels in brain, we applied high-throughput RNA sequencing with RNA isolated from cortices of the same mice used for ChIP-seq. We found 1324 up-regulated genes and 482 down-regulated genes at P < 0.05 as a cutoff ( Fig. 2A). Using IPA, we identified the most significant categories affected by bexarotene. As shown on Fig. 2B, the comparison of canonical pathways identified by ChIP-seq and RNA-seq revealed several pathways related to neuronal development and differentiation that were highly enriched in both data sets derived from bexarotene treated APOE3 mice, such as Axonal guidance, Rho family GTPase, Wnt and Ephrin receptor signaling, CREB signaling and CXCR4 pathway. To validate RNA-seq data, we selected genes that are differentially affected (cut-off p < 0.05) as identified by the canonical pathways shown on Fig. 2B. Specifically we chose genes that can affect neuron development (NMDA receptors Grin1 and Grin2d 27 , Grm4, and Myo7a), neurite complexity (Lrp5, Sema3f and Akt1), neuronal death (Fas) or are related to Aβ (Lrp1 and Tfap2b). As shown on Fig. 2D,E, the effect of bexarotene on Grin2d, Lrp1, Lrp5, Sema3f, Tfap2b 28 ) and Fas as identified in RNA-seq data (Fig. 2C) was validated by qPCR in cortices of APOE3 mice (Fig. 2D) and N2a cells (Fig. 2E).
Bexarotene induces epigenetic changes in ES cells consistent with increased neuronal differentiation. As shown on Figs 1 and 2 and in agreement with the data from our previous study 22 , neuronal development and differentiation are categories significantly affected by bexarotene. To determine if bexarotene induces epigenetic changes consistent with differentiation, we performed ChIP-seq using ES cells to map the enrichment of two histone marks -H3K4me3, associated with active promoters, and H3K27me3 -a strong mark for repressed promoters, and also, associated with Polycomb repressive complex. We treated ES cells with bexarotene at the stage of embryoid bodies (EBs) for 4 days (experimental design shown on Fig. 3A) followed by ChIP-seq for H3K4me3 and H3K27me3 enrichment. As seen on Fig. 3B, the enrichment of H3K4me3 mark in promoters of pluripotency-associated genes, such as Oct4 (alternative name Pou5f1) and Nanog, was significantly decreased 4 days after bexarotene treatment compared to the expression at the start (a stage of embryoid bodies). These data are consistent with the expression of these genes only in the pluripotent progenitors and their down-regulation in cells committed to neuronal lineage 29 . In contrast, the enrichment of H3K4me3 mark in bexarotene treated cells was increased in the promoters of Pax6 (a marker for neuronal progenitors) and Syp (Synaptophysin) which is a marker for differentiating neurons. As seen from Fig. 3C, mRNA expression of Pax6 was increased 24 hours after bexarotene treatment, corresponding to its role in the transition of ES cells to the neuronal lineage. Syp mRNA  was increased at Day4 after treatment as in our previous study 22 . The results demonstrate that the expression of Pax6 and Syp correlated to the enrichment of H3K4me3 at their promoters.
Next, we correlated the enrichment of H3K4me3 and H3K27me3 datasets in each gene ontology (GO) category to the gene expression data obtained from APOE3 and APOE4 mice treated with bexarotene. The heat-map on Fig. 3D shows that in both, APOE3 and APOE4 mice, GO categories "neuron differentiation", "axonogenesis" and "neuron projection" are highly enriched in the bexarotene treated group. Importantly, the expression data highly correlated to the enrichment of H3K4me3 mark observed only in bexarotene treated ES cells. In agreement with these data, the enrichment of H3K27me3 mark was significantly changed between bexarotene and vehicle Values are mean ± SEM and statistics is by t-test; n = 3, *p < 0.05, **p < 0.01, ***p < 0.001. treated ES cells in categories such as "neurogenesis", "neuron projection", "axonogenesis" and "neuron fate commitment". To validate these results, we examined mRNA expression level of RXR targets with relevance to neuron development and neurite extensions. As visible from Fig. 3E, Eef2, Grin2, Sema3f and Lrp5 were increased 10 days after treatment corresponding to the stage of immature neurons. In contrast, Grm4 was increased 24 hour after bexarotene treatment, suggesting an early response to bexarotene. Collectively, these data indicate that bexarotene-activated RXRs affect epigenomic landscape, consistent with the transition of ES cells towards neuronal cells.
Aβ deposition alters bexarotene effects on the transcriptome in APP/E3 mice. The development of amyloid phenotype can potentially affect transcription factor binding to genomic sites and therefore gene expression levels. To test this, we compared the effect of ligand activated RXR in APOE3 mice expressing either endogenous mouse App (APOE3 mice) or human APP gene (APP/E3 mice). A notable difference between the response of APP/E3 and APOE3 mice to bexarotene treatment was the significantly lower number of differentially affected genes: at p < 0.05 cut-off we identified 216 differentially affected genes in APP/E3 vs 2751 genes in APOE3 mice (Fig. 4A,B). As shown on Fig. 4C, there was a significant similarity in the categories of genes up-regulated by bexarotene between APOE3 and APP/E3 as well as with already published data for APOE4 mice 22 , and APP/PS1 mice expressing mouse APOE 24 . For example, top GO functional categories identified in up-regulated genes in APP/E3 mice included "neuron differentiation", "vasculature development", "neuron development". Common genes up-regulated in both APOE3 and APP/E3 were: Dlk1 (delta-like homologue 1) 30 , Vgat/Slc6a12 (GABAergic vesicular transporter), Apod (apolipoprotein D), Prph (peripherin); Rdm1 (involved in RNA-directed DNA methylation) among others. The up-regulation of Dlk1, an imprinted gene that acts as an atypical NOTCH ligand and regulates postnatal neurogenesis 30,31 in both data sets confirms our previous finding that activated RXR affect NOTCH signaling 22 . We found that there was a significant difference in down-regulated categories (see Fig. 4D). In APP/E3 mice 31 the top down-regulated were: "response to endogenous stimulus" (p < 10 −6 ); "lipoprotein metabolic process" (p < 10 −4 ); "acute phase response" (p < 10 −3 ) and "inflammatory response" (p = 0.0016) reminiscent of the effect of bexarotene in APP/PS1 mice expressing mouse APOE (see also the results shown in Lefterov et al. 24 ). In contrast, the top categories down-regulated by bexarotene in APOE3 mice were: "transcription" (p < 10 −9 ), "mRNA metabolic process" (p < 10 −15 ); "chromosome organization" and "chromatin modifications" (p < 10 −3 ) (Fig. 4C). Of particular interest in terms of AD, because of their role in amyloid deposition and aging, were some genes up-regulated in APP/E3 mice but not affected in APOE3 such as Klotho (Kl) 32,33 and transthyretin (Ttr) 34 . Interestingly, Ttr was also found up-regulated by bexarotene in APP mice expressing mouse Apoe 24 . In contrast, some genes with a role in Aβ clearance and phagocytosis such as Lrp1 35 , Abca7 36 and Clusterin (Clu) 37 were up-regulated by bexarotene in APOE3 mice but not in APP/E3 (Fig. 4A,B). These results indicate that amyloid deposition affects the transcriptome in mice expressing human APP and potentially can affect RXR binding to genomic sites and response to RXR agonists.
Bexarotene treatment maintains neurite complexity and prevents loss of neurites in primary neurons exposed to Aβ. Previously, we demonstrated that bexarotene-activated RXR enhanced neurite branching and complexity in mature neurons 22 . To determine the neuroprotective effects of bexarotene against Aβ , we used primary neuronal cultures established from cortices and hippocampi of E16 mouse embryos. Following Aβ treatment for 24 hours, the morphology of GFP neurons was digitally reconstructed from high resolution images using Neuromantic 22,38 . Bexarotene-treated neurons demonstrated significantly increased numbers of intersections ( Fig. 5B) and branches (Fig. 5C). Aβ 42 -treatment impaired number of intersection, branches and bifurcations in primary neurons. In contrast, pre-treatment with bexarotene before the addition of Aβ 42 prevented loss of neurites and maintained neuronal complexity (Fig. 5A). To determine bexarotene effect on mRNA expression in primary neurons, we selected few genes identified in our RNA-seq datasets as RXR targets and associated with Wnt signaling and microtubule dynamics that potentially can affect neurite complexity: secreted Frizzled-related protein 5 (Sfrp5), Wnt10a, Lrp5, Myo7a and tubulin beta 3 class III (Tubb3). As visible from Fig. 5F, bexarotene increased expression of Abca1 demonstrating target engagement. We also noted a significant up-regulation of Sfrp5 (more than 4-fold, p < 0.001) and a small but statistically significant up-regulation of Myo7a (p < 0.01) and Tubb3 (p < 0.05).

Bexarotene inhibits Aβ 42 induced disruption of actin cytoskeleton in N2a cells. Previous studies
reported that disruption of actin cytoskeleton and aggregation of actin was found in hippocampal and cortical neurons of post-mortem AD brains and cultured cell following chemical or physical stress 39 . The link between Rho family GTPases and Wnt signaling, members of common pathways identified by ChIP-seq and RNA-seq in mouse brain, affect cytoskeleton remodeling through cooperation 40 . Since these pathways regulate cytoskeleton organization, we examined bexarotene effects on F-actin puncta formation in Aβ 42 -treated N2a cells. First, we found that Aβ 42 treatment increased F-actin puncta formation by 54% as compared to vehicle treated N2a cells (Fig. 6A,B). In contrast, bexarotene significantly inhibited F-actin aggregation as compared to vehicle or Aβ 42 treated cells. Our data also demonstrate that bexarotene treatment prior to the addition of Aβ 42 inhibited F-actin puncta formation (Fig. 6A,B). Next, using N2a cells, we tested the effect of bexarotene and Aβ 42 on mRNA expression of several, already identified RXR targets (see Fig. 2) involved in Wnt signaling as well as neurite extensions. Consistent with the quantification of F-actin puncta in N2a cells, the expression of genes that affect cytoskeleton dynamics as well as neuron projections was decreased after Aβ 42 treatment. In contrast, bexarotene treatment prior to Aβ application restored their expression (Fig. 6C). These data suggest that bexarotene inhibits disruption of actin cytoskeleton induced by toxic Aβ species.

Discussion
Bexarotene-activated RXRs improve behavior deficits and increase Aβ clearance from brain interstitial fluid in mice expressing human APP 6,9,41 . Previously we reported bexarotene also increased the number of neuronal progenitors in brains of APOE4 and APOE3 mice and neuronal differentiation of mouse embryonic cells 22 . To compare if the effects of bexarotene on RXR cistrome and transcriptome in mouse cortex are affected by APOE isoform, we performed high throughput massive parallel sequencing using cortices from mice expressing human APOE3.
Using ChIP-seq, we found bexarotene promoted dynamic changes of RXR binding toward promoter region and affected genes associated with "transcription", "chromatin remodeling" and "neuron projections" in APOE3 mice. To identify effects of APOE isoforms we applied IPA and compared functional categories identified by the cistrome analysis of APOE3 to those in APOE4 mice from our previous study 22 . We determined that highly enriched categories "nervous system development", "axonal guidance", "behavior" and "neurites extension" overlap in both genotypes. The analysis of the transcriptomes revealed similarity also in other functional categories of differentially affected genes related to neuronal differentiation and neuron projection that could be relevant to AD. These data were further validated in cell culture and in vivo using qPCR. Thus, the present findings Statistics is by two-way repeated measures ANOVA with number of intersections as an independent variable, treatment group as a factor, and distance from the soma as the repeated-measure factor. There is an interaction F (171,2280) = 6.23, p < 0.0001. Total neurite length (C), Number of Bifurcations (D) and Number of branches (E) were significantly increased in the neurons treated with bexarotene. Statistics for (C-E) is by one-way ANOVA. For each treatment group, images were taken from at least 4 wells and of more than 2 GFP+ neurons per well. (F) Primary neurons were treated for 6 or 24 hours with bexarotene and mRNA level for selected genes related to categories "neuron projections" and "Wnt signaling" were examined by qPCR. Values are mean ± SEM; Statistics is by t-test; n = 3, *p < 0.05, **p < 0.01 versus vehicle group.
Scientific RepoRts | 6:24048 | DOI: 10.1038/srep24048 demonstrate that APOE isoform by itself does not affect binding of RXR homodimers to genomic sites and consequently the effect on gene expression.
Several signaling pathways important for neuronal development and neurites extension, identified as significant in both APOE3 ChIP-seq and RNA-seq datasets also overlapped. Rho family GTPase was identified as highly enriched in ChIP-seq and up-regulated in RNA-seq dataset (Fig. 2B). Rho family GTPases are associated with cytoskeletal organization essential for neurite outgrowth and axon extension, as well as, for neurite branching and neuronal complexity 42 . They also play a role in regulating the dynamics of F-actin-rich cytoskeleton present in dendritic spines and are important for synaptic transmission 43 . Furthermore, abnormalities in cytoskeletal organization have been reported in many neurodegenerative disorders including AD 44 , and the disruption of neurite complexity and synaptic loss are closely related to neurodegenerative diseases 45,46 . Our data, shown on Fig. 6, demonstrate that oligomeric Aβ species disrupt actin cytoskeleton and bexarotene treatment counteracts this effect. Previously we have shown that bexarotene improves dendritic complexity and branching in primary neurons and in APOE3 and APOE4 mice 22 . In this study we demonstrate that bexarotene-activated RXR significantly reduced deleterious effects of oligomeric Aβ on neuronal processes, neurite length and branching in cultured primary neurons (Fig. 5).
Bexarotene treatment, as shown on Fig. 1D, increased RXR binding to Wnt3 -a key member of Wnt signaling pathway, involved in nervous system development, differentiation and plasticity 47,48 . This finding is consistent with the pathway analysis of RNA-seq data (Fig. 2B). Several members of Wnt signaling pathway such as Myo7a 49 , Lrp1 50 , Lrp5 51 , Wnt10a and Sfrp5 were identified by RNA-seq and further validated by qPCR in APOE3 cortex (Fig. 2D), N2a cells (Fig. 2) and primary neurons (Fig. 5F). SFRPs (secreted Frizzled-related proteins) are soluble ligands for the non-canonical Wnt signaling pathway involved in axon guidance 52 and anti-inflammatory response 53 . Lrp1 (LDL receptor related protein 1) is an endocytic receptor with a critical role in Aβ clearance 54 possibly via interaction with Rho GTPases 35 . Lrp5 (LDL receptor-related protein 5) is Wnt signaling receptor Values are mean ± SEM and statistics is by one-way ANOVA followed by Tukey's post-test; n = 9, *p < 0.05, ***p < 0.001. (C) N2a cells were treated as shown on A and B and mRNA level for selected RXR targets was examined by qPCR. Values are mean ± SEM; Statistics is by t-test; n = 3, *p < 0.05, **p < 0.01 versus vehicle group; ## p < 0.01, ### p < 0.001 versus Aβ group. expressed in endothelial cells, vascular smooth muscle cells and monocytes, reportedly involved in lipid internalization and motility of macrophages 55 . A recent study demonstrated that bexarotene ameliorated aging-related synapse loss depended on neuronal Lrp1 56 . The data from the current study suggest that bexarotene effects on Lrp1 and Lrp5 can account for ameliorating deleterious Aβ effects. As we demonstrate here, Lrp5 expression is up-regulated in response to bexarotene treatment, apparently due to an increased RXR binding to its proximal promoter. In a further proof that direct up-regulation of Lrp5 is associated with Aβ metabolism we show that bexarotene prevents the inhibitory effect of oligomeric Aβ on Lrp5 expression (Fig. 6C).
Other pathways related to neuronal morphology and function identified as responsive to RXR treatment were CREB and CXCR4 signaling. The transcription factor CREB regulates genes involved in neuronal survival and plays an essential role for long-term synaptic plasticity 57,58 . Transcriptional up-regulation of several members of this pathway identified by RNA-seq, namely, Akt1, Grin1, Grin2d and Grm4 was validated in APOE3 mice and in vitro (Fig. 2). CXCR4 (CXC chemokine receptor-4) is expressed in neural stem cells and plays a role in the migration of the neural progenitors from the neurogenic zones to the sites of damage after brain injury 59 .
Since GO categories "neuron fate commitment", "nervous system development" and "neurogenesis" were significantly affected in both APOE3 and APOE4 mice we examined if bexarotene induces changes in epigenome that coincide with the transition from pluripotent cells to neuronal progenitors. We performed ChIP-seq and mapped the enrichment of two histone marks -H3K4me3, associated with active promoters, and H3K27me3 -a mark for repressed promoters, using mouse embryonic stem cells. We demonstrate that bexarotene induces epigenetic changes in these cells consistent with increased neuronal differentiation and in correlation with an effect on transcription (Fig. 3).
The effect of amyloid deposition on the transcriptome in APP/E3 mice in response to bexarotene is an important finding. Evidence from studies using APP transgenic mice suggest amyloid deposition and other neuropathological features affect transcriptome changes that potentially could affect transcription factor binding 60,61 . We compared transcriptomes of APOE3 and APP/E3 mice and some of common genes that are involved in neuroprotective effects such as Apod, Slc6a12, Prph and Rdm1 were identified in the up-regulated sets of genes in both genotypes. Slc6a12 (solute carrier family 6, member 12, also called vGat) is a vesicular GABA transporter and a marker of GABAergic terminals. Recent report demonstrated that Aβ injected in the hippocampus decreased vGluT1 (a marker of glutamatergic terminals) level without change in vGAT-immunopositive nerve terminals 62 . This suggests that the expression of vGat is less susceptible to the effect of amyloid deposition and is in agreement with our data. Apod, a gene which was found up-regulated by bexarotene in APP/PS1 mice 24 has a neuroprotective, as well as, anti-oxidant, and anti-inflammatory role 63 . Prph, a neuron-specific TypeIII intermediate filament is involved in neurite outgrowth during development and axonal regeneration 64 . Consistent with our previously published data on bexarotene treated APP transgenic mice expressing mouse Apoe 24 , "immune system development" was one of the top GO functional categories identified in up-regulated gene sets in APP/ E3 mice. Remarkably, however, most of the genes down-regulated by ligand activated RXR in those mice (Fig. 4), compared to APOE3 animals, clustered in significant, yet distinct, BP categories, associated with immune system and inflammatory response. Among them were "positive regulation of immune system", "acute inflammatory response", "acute-phase response" and "response to endogenous stimulus". Based on highly reproducible GWAS and numerous experimental studies, it is now firmly established that microglia, immune and inflammatory reactions in CNS play substantial role in AD pathogenesis. While it is not entirely clear if AD genetic predisposition is primarily associated with immune function as suggested 60 , microglia are the non-neuronal CNS cells most intimately associated with AD pathological changes in brain 65 . Therefore, we interpret the differential response to bexarotene in APP/E3 and APOE3 mice as a response of complex, mixed macrophage phenotypes formed by clustered regulatory and metabolic pathways 66,67 . The data and validation assays presented here suggest that increased inflammatory reactions in APP mice expressing human APOE isoforms can be pharmacologically mitigated at the level of transcription factor binding and ultimately by modulating transcription efficiency of genes that participate in a number of immune and inflammatory pathways, and therefore provide one possible mechanism to establish therapeutic targets in human disease with strong genetic predisposition.

Materials and Methods
All reagents, chemicals and standard additives, unless otherwise stated were purchased from Sigma Aldrich (St Louis, MO), or Thermo Fisher (Pittsburgh, PA).

Mice.
All animal experiments were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. All procedures were carried out in accordance with the approved guidelines. Mice with APOE3 +/+ targeted replacement (referred to as APOE3) were purchased from Taconic on C57BL/6 background and bred in house on the same background. APOE3 mice litters were used at 6 months of age. APP/PS1ΔE9 mice were bred to human APOE3 +/+ targeted replacement mice to generate APP/PS1ΔE9/APOE3 +/+ (referred to as APP/E3) mice expressing only human APOE3. Animals were randomly assigned to either bexarotene (100 mg/ kg/day; oral gavage; Targretin, Eisai Inc., WoodCliff Lake, NJ) or vehicle (0.2 mg/kg glycerol) treatment groups.
Processing of mouse brain tissue. Mice  Technologies, Grand Island, NY). Treatment of cell cultures to induce neuronal differentiation of ES cells was as previously described 22 . Briefly, ES cells were plated in petri dishes in medium without LIF to initiate the formation of Embryoid Bodies (EBs). EBs were treated with 5 μM bexarotene dissolved in DMSO, or DMSO only for 4 days. Four days after treatment, EBs were plated on poly-L-ornithine/laminin-coated plates in N2B27 medium and differentiated into neurons. Cells were collected at 24 hours, 4 days and 10 days after the treatment for RNA isolation and collected 4 days after the treatment for Chromatin Immunoprecipitation. Mouse neuroblastoma N2a cells were cultured in DMEM/F12 supplemented with 10% FBS, antibiotics (Life Technologies, Grand Island, NY). For validation of bexarotene effect, N2a cells were treated with 1 μM bexarotene and RNA was isolated 24 and 48 hours after the treatment.
Chromatin immunoprecipitation and sequencing (ChIP-seq). Brains from APOE3 mice treated with bexarotene for 10 days and ES cells treated with bexarotene for 4 days were used for ChIP-qPCR and ChIP-seq. Brain lysates were sonicated with 3 pulses of 15 sec at amplitude 30, a 120 sec pause and 3 pulses of 15 sec at amplitude 40 using a Model 705 Sonic Dismembrator (Fisher Scientific, Pittsburgh, PA), to obtain fragments of 200-600 bp. For immunoprecipitation, we used a rabbit polyclonal anti-RXR (ΔN 197, #sc-774, Santa Cruz, CA), rabbit polyclonal anti-H3K4me3 (#CS200580, Millipore, CA) and rabbit-polyclonal anti-H3K27me3 (#07-449, Millipore, CA). For ChIP validation we performed ChIP-qPCR with mouse Glucagon (Gcg) as a negative control and determined Fold enrichment normalized to % input of negative control (target gene % input/negative control % input). ChIP libraries were generated using TruSeq ChIP Sample Prep Kit (Illumina, San Diego, CA) following manufacturer's protocols. For each steps, samples were purified by AMPure XP beads (Beckman Coulter, Brea, CA). Adapter-ligated samples were separated on a 2% agarose gel to obtain 250-300 bp size-range of DNA fragment to remove unligated adapters. The libraries were validated by Agilent Technologies 2100 Bioanalyzer to check the size, purity, and concentration of the sample before the sequencing, and sequencing performed on Illumina HiSeq2000 instrument at the Functional Genomics Core, UPenn, Philadelphia (http://fgc.genomics.upenn.edu/). We used Bowtie2 (http://bowtie-bio.sourceforge.net/bowtie2) to align sequencing reads to the mouse genome (mm9 release) and Homer tools (http://biowhat.ucsd.edu) for peak calling, annotation and further processing of BAM/BED files.
RNA isolation, qPCR and sequencing. RNA was isolated from brain and cultured cells and purified using RNeasy mini kit and QiaShredder (Qiagen, Valencia, CA) according to the manufacturer's recommendations. The quality control of all RNA samples was performed on a 2100 Bioanalyzer instrument and samples with RIN >8 were further used for qPCR and library construction using mRNA Library Prep Reagent Set (Illumina, San Diego, CA). Libraries were generated by PCR enrichment including incorporation of barcodes to enable multiplexing. The libraries, 5 or 6 per condition (6 bexarotene and 5 vehicle treated males) were sequenced on Illumina HiSeq2000. Sequencing reads were aligned to the mouse genome (mm9 release) using Subread and differential expression analysis performed by edgeR 69 . For qPCR, first strand cDNA was synthesized from 1 μg of total RNA using EcoDry ™ Premix, Random Hexamers (Clontech, Mountain View, CA). qPCR was performed using TaqMan ® Universal Master Mix II (Life Technologies, Grand Island, NY). Transcript levels were normalized to Gapdh and amplification plots were analyzed by comparative ΔΔCt method.

Functional Pathway analysis and Comparative analysis.
To identify biological process changes and the most relevant pathways related to the effects of RXR and epigenetic changes, we performed functional annotation clustering using the Database for Annotation, Visualization and Integrated Discovery (DAVID, https://david. ncifcrf.gov/) 26 and Ingenuity Pathway Analysis (IPA ® , QIAGEN, Redwood City, www.qiagen.com/ingenuity). For comparative analysis and graphics, R was used (https://www.r-project.org/): proportional Venn diagrams were generated using package VennDiagram; clusters that were generated by IPA were visualized as heatmaps using package gplots. Volcano and scattered plots were generated for visualizing differential gene expression.
Preparation of amyloid beta oligomers. Aβ 42 were synthesized at Keck's facility (Yale University) and stock solutions of synthetic Aβ 42 were disaggregated according to the method described previously 24,70 . In brief, 1 mg Aβ was dis-aggregated first with trifluoroacetic acid and hexafluoro-2-propanol, followed by removal of the solvents. The aqueous dissolution of the resulting disaggregated peptide film was followed by high speed centrifugation (100,000 × g) for 1 h to remove trace aggregates. This protocol yields mostly monomeric Aβ as determined by Western blotting (not shown). Aβ aggregation reactions were performed in serum-free medium at room temperature (20 °C) without shaking.
The cells were treated with vehicle or 0.5 μM bexarotene for 24 hours and then treated with 0.5 μM Aβ 42 for 24 hours. Cells were fixed with 3.7% paraformaldehyde for 10 min at room temperature and stained with Alexa 488 phalloidin (Life Technologies, Grand Island, NY) for visualizing F-actin puncta. DAPI was used for counterstaining and multi-channel GFP-DAPI images (1392 × 1040 pixels) captured with Nikon 90i microscope using 20× objective.
Primary embryonic mouse cortical neuronal cultures and treatment. Primary mouse cortical neuronal cultures were prepared from E16 embryos. Single cell suspension was used and cells plated at a density of 10 5 cells per cm 2 on 18 mm circular coverslips coated with 0.01% poly-L-lysine in Neurobasal medium containing antibiotics, L-glutamine, and B27 supplement. Cultures were maintained by replacing 50% of the conditioned medium with fresh medium, every 2 days. At the day of dissection (DIV0), cells were infected with GFP-Lentiviral particles (pLVX-IRES-ZsGreen1 Vector; Clontech, Mountain View, CA) at 0.2 multiplicity of infection. The moderate efficiency of this infection technique allows an easy identification of the processes associated with single Scientific RepoRts | 6:24048 | DOI: 10.1038/srep24048 neurons. At DIV14, infected neurons were pre-treated with 1 μM bexarotene or vehicle for 24 hours followed by 0.5 μM Aβ 42 for 24 hours. Six days later (at DIV21), the neurons were fixed with 4% paraformaldehyde for 20 min at room temperature, counterstained with DAPI, and multi-channel GFP-DAPI images (1392 × 1040 pixels) captured with Nikon 90i microscope using 20× objective.
Neuronal reconstruction and morphological analysis. For each treatment group, at least 4 wells and no less than 2 GFP+ neurons per well were imaged. DAPI staining was used to verify that only single neurons, but not cell clusters, were selected for reconstruction. The GFP channel of each captured image was then exported as a TIF file, and individual neurons were semi-automatically reconstructed using Neuromantic (Myatt, 2012). Neuronal reconstructions, saved as SWC files, were used for subsequent morphometric analyses. Sholl analysis was additionally performed using the Trees Toolbox (Cuntz, 2011) for Matlab (Mathworks).

Statistics.
All results are reported as means ± S.E.M. All statistical analyses (if not otherwise mentioned in the text) were performed in GraphPad Prism, version 6.0 (LA Jolla, CA) and differences considered significant where p < 0.05.
Dataset. The datasets in a proper format will be uploaded to a publicly available repository, and is directly available from Koldamova&Lefterov Laboratory.