Uncovering the role of G9a modulatory landscape promoting neuronal plasticity and neuroprotection in Alzheimer's disease


 Epigenetic alterations are a fundamental pathological hallmark of Alzheimer’s disease (AD). Herein, we uncover the unknown G9a modulation pathways involved in AD, showing the upregulation of G9a and H3K9me2 in the brains of AD patients. Likewise, treatment with a G9a inhibitor in SAMP8 mice reversed the high levels of H3K9me2 and rescued the cognitive decline. Interestingly, a transcriptional profile analysis revealed induction of neuronal plasticity and a reduction of oxidative stress and neuroinflammation; the latter being also validated in cell cultures. Furthermore, an exploratory H3K9me2 ChIP-seq analysis demonstrated that during G9a inhibition treatment, the H3K9me2 mark is enriched at the promoter of genes associated with neural functions. Lastly, we showed in Caenorhabditis elegans (C. elegans) AD transgenic strains, similar epigenetic modifications and modulated pathways were altered with increased β-amyloid levels, which were reverted by the set-25 (in C. elegans is similar to the mammalian G9a protein) knockout, including the cognitive impairment. Therefore, our findings confirm that RNAi suppression of set-25 or pharmacological G9a inhibition promotes a positive outcome in AD, being a promising therapeutic strategy.


Introduction 83
Aberrant epigenetic mechanisms represent one of the key pathophysiological 84 drivers of aging and neurodegeneration 1,2 , and are relevant to the progression 85 of age-related cognitive decline, including Alzheimer's disease (AD) 2 . 86 Transcriptionally activating and repressing global histone changes in different 87 regions have been observed in postmortem AD brains 3 . Despite enormous 88 research, none of the clinically approved drugs for AD is effective, being all of 89 them only symptomatic treatments, thereby there is an urgent need to identify 90 new targets based on epigenetic therapies for neurodegenerative diseases 4 .

91
Of note, growing evidence suggests that lysine methyltransferases (KMTs), 92 such as G9a, act as a crucial regulator in human diseases 5 . G9a is a KMT able 93 to mono-and di-methylate K9 of histone H3 (H3K9me1 and H3K9me2), which 94 are marks linked to the repression of genes implicated in synaptic plasticity 6-9 , 95 learning, and memory formation 3,10-12 . However, until now, G9a has been 96 widely explored as an anti-cancer and an anti-malarial target 13 . Recently, it has 97 been demonstrated overexpression of G9a in the brain from late-stage familial 98 AD (5XFAD) mice and AD patients 3 . Similarly, another study showed elevated 99 levels of H3K9me2 in the occipital cortex of postmortem AD brain compared to 100 the non-demented (ND) group 14 . Besides, H3K9me2 seems to participate in the 101 pathogenesis of neurodegenerative disorders, such as posttraumatic stress 102 disorder (PTSD) 15 and anxiety-like behavior 16 . In Caenorhabditis elegans (C. 103 elegans), the putative methyltransferase that targets H3K9me2 is SET-25 17 , 104 which is similar to the mammalian G9a protein (28.8% identity, 44.6% 105 similarity) 18 , and it has only been implicated in the transgenerational epigenetic 106 their direct link is not yet well described. To characterize G9a as a novel target 141 for AD, we carried out several experiments and bioinformatic analyses (Fig 1a). 142 Firstly, we examined G9a protein levels by western blot (WB) in human AD 143 patients' brains. Of relevance, we found higher levels of this protein in AD 144 patients' brains compared to the human ND patients' brains (Figs. 1b and c, 145 Supplementary Table 1). Since G9a is a KMT enzyme responsible for 146 depositing methyl groups in H3K9, we also evaluated H3K9me2, a repressive 147 mark. Higher H3K9me2 levels were statistically significant in human AD patients 148 when compared to ND group (Figs. 1d and e). Furthermore, we found that the 149 ratio of Aβ42/Aβ40 was significantly increased in AD patients in comparison with 150 the ND human group (Fig. 1f). Of relevance, we also found a strong positive 151 correlation between H3K9me2 levels and Aβ concentration in AD patients (Fig.  152 1g). Thus, the data reveal that high H3K9me2 are correlated with increased Representative WB, and quantification for G9a (EHMT2) of human patients' 173 brains. d. and e. Representative WB, and quantification for histone H3K9me2 of 174 human patients' brains. f. The ratio of Aβ42/Aβ40 by ELISA in human patients' 175 brains. Values represented are mean ± Standard error of the mean (SEM); (n = 176 14 (ND n = 6, AD n = 8)). Groups were compared by Student's t-test analysis 177 (significant at *p < 0.05; ***p < 0.001). g. Correlation between Aβ42/Aβ40 ratio 178 and H3K9me2 (slope = 0.2394). R 2 and p-values are indicated on graphs 179 (significant at *p < 0.05). Source data are provided as a Source Data file.

181
Inhibition of G9a with UNC0642 leads to a reduction in H3K9me2 and Aβ 182 levels as well as behavioral and cognitive improvements in SAMP8 mice. 183 Because aging is the main risk factor for AD 21 , we determined the G9a protein 184 in the SAMP8 strain, a well-established AD model to investigate the key SAMP8 were reduced by G9a inhibition (Fig. 2e). Notably, we found a strong 194 positive correlation between H3K9me2 levels and Aβ concentration in SAMP8 195 mice (Fig. 2f). Thus, the data suggest that high H3K9me2 is correlated with 196 increased levels of Aβ in SAMP8, suggesting the contribution of the G9a and 197 H3K9me2 in the age-related cognitive decline presented by this AD rodent 198

model. 199
Aβ levels are associated with behavioral abnormalities and cognitive decline, so 200 we next performed behavioral studies to determine whether G9a inhibition with 201 UNC0642 could revert them in SAMP8. Firstly, we used the three-chamber test 202 (TCT) to assess the general sociability behavior in mice. In the sociability 203 phase, in all mice groups, the presence of an intruder increased the time 204 significantly spent in the intruder chamber instead of the empty cup chamber 205 (Fig. 2g). Remarkably, we only found a significant increase in the time sniffing 206 the intruder mouse in the SAMP8 treated mice group, confirming the 207 improvement in social behavior after G9a inhibition by UNC0642 at 5mg/Kg 208 (Fig. 2h). To evaluate the working and spatial memories, mice were assessed in 209 the novel object recognition (NORT) and object location (OLT) tests, 210 respectively. During the familiarization phase of the NORT task (Fig. 2i), the 211 exploration time was unchanged by UNC0642 treatment. The task results 212 revealed that SAMP8 treated with UNC0642 exhibited a significant reduction in 213 cognitive deficits in both short-and long-term memories (Figs. 2j and k), 214 presented by the SAMP8 Control mice group. 215 Regarding the OLT results, the exploration during the habituation phase (Fig.  216 2l) was unaffected by the treatment. The task results showed that SAMP8 217 treated with UNC0642 exhibited an increase of the discrimination index (DI) 218 compared to the SAMP8 Control mice group (Fig. 2m), suggesting an 219 improvement in spatial memory. Finally, the polygonal graph depicts differences 220 among SAMR1 and SAMP8 Control, and SAMP8 treated mice groups by 221 graphing several TCT parameters, DI of NORT and OLT, and molecular ratios 222 of H3K9me2 and Aβ42/Aβ40 (Fig. 2n). This data suggests that the G9a inhibition 223 promotes beneficial effects on behavior and cognition. 224 and Aβ levels as well as behavioral and cognitive improvements in SAMP8 227 mice. a. and b.    To validate the RNA-seq, we performed RT-qPCR for some DEGs such as 286 Gmfb, synaptosome associated protein 25 (Snap25), T-box brain transcription 287 factor 1 (Tbr1), and calcium/Calmodulin dependent protein kinase II gamma 288 (Camk2g), which were significantly deregulated (Figs. 3e-g). Taken together, 289 these data suggest that treatment with UNC0642 induces a transcriptional 290 profile that allows beneficial effects on behavior and cognition. 291   KEGG pathway extracellular matrix organization diterpenoid metabolic process sensory perception of sound ventricular cardiac muscle tissue morphogenesis G−protein coupled receptor signaling pathway, coupled to cyclic nucleotide second messenger cardiac muscle cell contraction collagen fibril organization regulation of protein import into nucleus artery morphogenesis glycosaminoglycan catabolic process ventricular cardiac muscle cell action potential odontogenesis retinoid metabolic process adenylate cyclase−modulating G−protein coupled receptor signaling pathway positive regulation of peptidyl−tyrosine phosphorylation ventricular compact myocardium morphogenesis sensory perception of mechanical stimulus eye development cardiac ventricle morphogenesis ventricular cardiac muscle tissue development regulation of saliva secretion positive regulation of sodium ion transport adenylate cyclase−activating G−protein coupled receptor signaling pathway phospholipase C−activating G−protein coupled receptor signaling pathway regulation of dendritic spine morphogenesis

G9a inhibition treatment increases the GMFB transduction pathway, 307
restoring dendritic spine density in SAMP8 mice. 308 Pathologically, it is known that SAMP8 presents a reduction in neuronal spine 309 density 26 . In addition, GMFB is a highly conserved brain-enriched protein 310 implicated in neuroplasticity since a peak of these protein levels were correlated 311 with learning and memory formation 27 . Then, to further confirm GMFB pathway 312 activation after the G9a inhibition in SAMP8, we investigated the GMFB protein 313 levels and the downstream effector proteins using WB. Strikingly, a significant 314 increase of GMFB was observed in the SAMP8 treated with UNC0642 in 315 comparison with the SAMP8 Control mice group (Figs. 4a and b). Further, we 316 evaluated ratio p-p38/p38, and cAMP-response element binding protein (CREB) 317 protein levels. A significant increase in p-p38 protein levels were found in 318 SAMP8 treated group with UNC0642 (5 mg/Kg) (Figs. 4a and c). Likewise, the 319 ratio p-CREB/CREB was also augmented but did not reach significance 320 between groups (Figs. 4a and d). Finally, we also evaluated the protein levels of 321 BDNF and its receptor, tropomyosin-related kinase B (TrkB). Regarding the 322 ratio of protein levels of p-TrkB/TrkB, we observed a significant increase in 323 Representative WB and quantifications for the ratio of p-p38/p38. a. and d.

337
Representative WB and quantifications for the ratio of p-CREB/CREB. a. and e. 338 Representative WB and quantifications of BDNF. Values represented are the 339 mean ± SEM; (n = 12 (SAMP8 Control n = 6, SAMP8 treated with UNC0642 340 (5mg/Kg) n = 6)). Groups were compared by Student's t-test analysis 341 (significant at *p < 0.05; ***p < 0.001). f. Representative images and tracings of 342 Golgi-stained neurons from the SAMP8 Control group (top left) and the SAMP8  The inhibition of the G9a ameliorates neuroinflammation, modulating 363 molecular changes in the NF-κB pathway in SAMP8 mice. 364 We had previously shown an increased neuroinflammatory process in 365 after treatment with UNC0642. NF-κB is a transcription factor related to 371 inflammatory response and a master commander in the expression of pro-372 inflammatory genes, and its signaling is an important mediator of brain 373 inflammation in AD 29 . To investigate the activation levels of NF-κB in the 374 SAMP8 treated group, we evaluated its protein level and the expression of its 375 target genes after treatment with UNC0642. Strikingly, a significant reduction in 376 NF-κB protein levels was found in SAMP8 treated group (Figs. 5a and b). Next, 377 we assessed the gene expression of some pro-inflammatory markers, such as 378 interleukin-6 (Il-6), Cxcl10, and Tnf-α 30 . A significant decrease in the expression 379 of these pro-inflammatory genes was observed between SAMP8 treated with 380 UNC0642 and the SAMP8 control mice group (Fig. 5c). Additionally, a GSEA 381 analysis demonstrated a reduction in the enrichment of genes associated with 382 the neuroinflammatory response in SAMP8 UNC0642 (Fig. 5d). a. and b. Representative WB, and quantification of NF-κB protein levels. c. 397 Representative gene expression of pro-inflammatory markers for Il-6, Cxcl10, 398 and Tnf-α. Gene expression levels were determined by real-time PCR. Values 399 represented are the mean ± SEM; (n = 12 (SAMP8 control n = 6, SAMP8 400 UNC0642 (5mg/Kg) n = 6)). Groups were compared by Student's t-test analysis 401 (significant at **p < 0.01; ***p < 0.001; ****p < 0.0001

513
Source data are provided as a Source Data file. H3K9me2 enrichment regulates pathways associated with the neuronal 517 system after G9a inhibition. We next sought to further characterize the 518 dynamics of the histone mark H3K9me2 after treatment with a G9a inhibitor. We 519 used public H3K9me2 ChIP-seq data corresponding to AML12 cells (murine 520 hepatocyte cell line) treated with UNC0638, a selective inhibitor of G9a 39 with 521 the same specificity as UNC0642. Analysis of H3K9me2 enrichment revealed 522 the presence of this histone mark in distal regions (60.78%) and promoters 523 (13.65%) in cells treated with the G9a inhibitor (Fig. 7a). Furthermore, 524 H3K9me2 has an enrichment around TSS (Fig. 7b), suggesting a role in 525 transcriptional regulation. Interestingly, functional analysis showed that 526 H3K9me2 is enriched at promoters of genes associated with nervous tissue 527 such as neuronal system, transmission across chemical synapses, neuroactive The distribution of the transcription factor motifs relative to TSS revealed 536 enrichment of binding motifs around TSS (Supplementary Fig. 6a, and Table 3). 537 To evaluate which transcription factors are associated with H3K9me2, we 538 performed a motif discovery analysis identifying motifs associated with PR/SET 539 Domain 6 (Prdm6), cytoplasmic polyadenylation element binding protein 1 540 (Cpeb1), TATA-Box binding protein associated factor 1 (Taf1), among others 541 ( Supplementary Fig. 6b). Additionally, an ENCODE, and ChEA Consensus TF 542 analysis identified an association with SUZ12 polycomb repressive complex 2 543 subunit (Suz12), enhancer of zeste 2 polycomb repressive complex 2 subunit 544 (Ezh2), RE1 silencing transcription factor (Rest) and SMAD family member 4 545 (Smad4) (Supplementary Fig. 6c), suggesting a regulatory network associated 546 with transcriptional repression where G9a could be a central regulator 547 ( Supplementary Fig. 6d). Together, these results demonstrate an H3K9me2 548 enrichment in genes associated with neural function, suggesting that after 549 treatment with a G9a inhibitor, the H3K9me2 reduction could allow the 550 expression of genes associated with the correct function of the CNS. 551 H3K9me2 peaks at TSS in AML12 with G9a Inhibitor  and more specifically in AD, we used the CL2006 strain, which expresses 573 human Aβ1-42 under the control of a muscle-specific promoter, and also presents 574 paralysis with age-worsening 40 . First, we confirm that the gene expression of 575 set-25 was higher in the transgenic AD strain, CL2006, than in N2 (WT) (Fig.  576   8a). As expected, due to RNAi suppression, set-25 gene expression was almost 577 completely reduced in our CL2006 strain (set-25 (RNAi) compared to the 578 CL2006 control group, and this finding is consistent with the observed reduction 579 in H3K9me2 levels (Figs. 8a-c). Then, we examined the effect of set-25 580 knockout on cognition, using the CL2355 C. elegans strain. This strain 581 expresses neuronal Aβ and significantly reduces the chemotaxis index (CI) 40 582 compared to the control strain CL2122. Of note, our results showed that set-25 583 knockdown fostered the restoration of impaired learning and memory of CL2355 584 strain by reaching a similar CI as control worms (Fig. 8d, Supplementary Fig. 7). 585 Besides, we showed a reduction in Aβ aggregation (Figs. 8e and f) in CL2006 586 in comparison with the group control (Fig. 8j). Taken together, these results 600 indicate that set-25 knockdown promotes similar modulatory pathway 601 modifications to confer better cognitive performance in C. elegans (Fig. 8k). from whole petri dish. Groups were compared by One-Way ANOVA and post-634 hoc Tukey's test (significant at **p < 0.01; ****p < 0.0001 vs CL2006 Control). k. 635 Representative scheme of the results found in the set-25 knockout strains. 636 Source data are provided as a Source Data file.

638
Discussion 639 The present study demonstrated the relevance of the G9a and its repressive 640 histone mark H3K9me2 in AD patients, which correlates with increased levels of 641 the Aβ42/Aβ40 ratio, an important hallmark of the disease. We also have 642 observed in SAMP8 similar specific epigenetic modifications in G9a and 643 H3K9me2 to those observed in AD patients. More importantly, pharmacological 644 inhibition with UNC0642 restored cognitive status in SAMP8 in the same way as 645 previously published studies in 5XFAD mice 3,12 . Besides, aggressive behavior 646 displays a direct influence on social interaction, and in the same line, the 647 inhibition of G9a has been associated with a decrease in anxiety-like behaviors 648 in adult male mice 16 . Nevertheless, the beneficial effect of G9a inhibitor 649 treatment on social performance has not been previously described. Thus, for 650 the first time, we demonstrate that G9a inhibition by UNC0642 treatment 651 improved social behavior in SAMP8 mice. Consistently, it has been described 652 that inhibitors of KTMs lead to the remarkable restoration of cognitive behaviors, 653 such as recognition memory, spatial memory, and working memory 16,41 . 654 As aforementioned, it is thought that aberrant gene expression is associated 655 with epigenetic alterations such as DNA methylation and histone modifications 2 . and G9a has already been described during an immune response 52 , it is still 695 unknown in the context of neuroinflammation 12,53 . As expected, it was found 696 that NF-κB protein levels were decreased after UNC0642 treatment in SAMP8 697 mice. Accordingly, we observed significantly reduced Cxcl10, Il-6, and Tnf-α 698 gene expression levels in the treated group compared to the control group. By overlaying transcriptomic analyses, we identified several microRNAs (miR-724 34b-5p, miR-34c-3p, and miR-34c-3p) that were modified. In addition, the 725 KEGG enrichment analysis revealed the association of G9a with cholinergic 726 synapse, axon guidance, PI3K-Akt-signaling pathway, among others. One of 727 the most important up-regulated signaling pathways in the SAMP8 treated mice 728 was FOXO. FOXO is a multifunctional transcription factor, which regulates 729 several molecular events such as OS resistance, cell survival, apoptosis and 730 proliferation, among others, involved in the pathogenesis of AD 35 . In 731 accordance with our results, it has been reported that G9a protein levels were 732 elevated, and FOXO1 protein levels were decreased in human colon cancer 733 patients 59 . Hence, to our knowledge, herein we described for the first time the 734 up-regulation of FOXO signaling pathway in SAMP8 brain after a treatment with 735 UNC0642. Accordingly, we evaluated the gene expression of Sirt-1, which 736 deacetylates FOXO and regulates gene expression, participating in the process 737 of neuroprotection through OS-involved pathways, and found that it was 738 elevated in the brain of SAMP8 treated with UNC0642. Furthermore, the gene 739 expression evaluation of antioxidant markers Sod-1 and Hmox-1 after G9a 740 inhibition revealed that our results align with previously published studies 12,60 . 741 In supporting our idea that the reduction of H3K9me2 is a promising therapy for 742 AD, we performed a ChIP-seq reanalysis, which revealed an H3K9me2 young adults laid embryos for 24 hours before being removed from plates. 807

RNAi 808
RNAi was performed by feeding method as previously described 62 . Bacteria 809 HT115(DE3) carrying IPTG-inducible were incubated at 37°C for 7-8 hours with 810 100 mg/mL ampicillin. NGM plates were then seeded with 25 mg/mL 811 carbenicillin and 1 mM IPTG with the incubated cultures and let grow overnight 812 at room temperature. Young adults were plated onto RNAi bacteria at 16ºC. 813 Silenced adults were transferred to fresh OP50 plates to produce subsequent 814 generations seven days later. 815

Chemotaxis assay 817
Fifteen adult hermaphrodites were left to lay eggs for 24h and then removed 818 from the plates. Eggs were incubated at 16°C for 36h, and then at 23°C for 819 another 48h. Briefly, the assay was performed in 100 mm NGM plates, 10 µL of 820 control odorant (96% ethanol) was added to the "control" spot. On the opposite 821 side, 10 µL of odorant (0.025% benzaldehyde in 96% ethanol) was added to the 822 "attractant" spot. Adult worms were washed three times in M9 buffer, and at 823 least 120 worms were placed towards the center of the plate. The test plates 824 were incubated at 23°C for 1.5h, and worms were scored according to the 825 chemotaxis index (CI) as follows: CI = (number of worms at attractant−number 826 of worms at control)/total number of worms. 827

Three-Chamber Test 828
The social behavior of the mice was evaluated by the TCT following a 829 previously described protocol 63 . A box of transparent polyvinyl chloride 830 (15x15x20 cm) divided into three equally dimensioned rooms with openings 831 among them was used. First, each mouse was placed in the center of the box 832 and allowed to explore the three chambers for 5 min (habituation). Afterward, an 833 intruder (same-sex and age) was placed in a metal cage in one of the rooms, 834 and the behavior was recorded for 10 min. The time spent in each room and 835 interacting with the intruder (sniffing time) were measured manually. The TCT 836 apparatus was cleaned with 70% ethanol between the trials to eliminate 837 olfactory cues. 838

Novel object recognition test 839
Short-and long-term recognition memory involving cortical areas and the 840 hippocampus was evaluated by NORT. The experimental apparatus used for 841 this test was a 90-degree, two-arm, 25-cm-long, 20-cm-high maze of black 842 polyvinyl chloride. Light intensity in the middle of the field was 30 lux. First, mice 843 were individually habituated to the apparatus for 10 min per day for 3 days. On 844 day 4, the animals were allowed to freely explore two identical objects (A and A 845 or B and B) placed at the end of each arm for a 10 min acquisition trial (first 846 trial-familiarization). Then, a 10-min retention trial (second trial) was carried out 847 2 hours (short-term memory) or 24 hours (long-term memory) later. During the 848 Short-term memory retention, objects A and B were placed in the maze 849

Object Location Test 859
The OLT evaluated the spontaneous tendency of rodents to spend more time 860 exploring a novel object location than a familiar object location and recognizing 861 when an object has been relocated. OLT was performed using a white plywood 862 apparatus (50 × 50 × 25 cm), in which three walls were white and one was 863 black. On the first day, animals just habituated to the empty open field arena for 864 10 minutes. On the second day, two objects were placed in front of the black 865 wall, equidistant from each other and the wall. The objects were 10 cm high and 866 identical. The animals were placed into the open field arena and allowed to 867 explore both objects and surroundings for 10 minutes. Afterward, animals were 868 returned to their home cages, and the OLT apparatus was cleaned with 70% 869 ethanol. On the third day, one object was moved in front of the opposite white 870 wall to test the spatial memory. Trials were recorded using a camera mounted 871 above the open field area, and the total exploration time was determined by 872 scoring the amount of time (seconds) spent sniffing the object in the new 873 location (TN) and the object in the old location (TO). DI was calculated, which is 874 defined as (TN-TO)/(TN+TO). 875

Biochemical experiments 876
Brain processing 877 SAMP8 and SAMR1 mice were euthanized 3 days after the behavioral test 878 completion by cervical dislocation. Brains were immediately removed from the 879 skull. Cortex and hippocampus were then isolated and frozen on powdered dry 880 ice. They were maintained at -80 °C for biochemical experiments. For Golgi 881 stain protocol, see the procedure in the section "Spine density and Golgi stain 882 protocol". 883

Human cases 884
Tissue samples were obtained from the Institute of Neuropathology-IDIBELL 885 Brain Bank, Hospitalet de Llobregat, following the guidelines of Spanish 886 legislation on this matter and the approval of the local ethics committee (Fig.  887   S1). The postmortem interval between death and tissue processing was 1 to 10 888 hours and was processed to minimize postmortem delay artifacts. The brain 889 tissue was immediately frozen on metal plates over dry ice, placed in individual 890 air-tight plastic bags, and maintained at -80 °C for biochemical experiments. To determine the altered microRNAs after treatment with UNC0642 we used the 958 QIAseqmiRNA platform. TMM normalization and differential expression analysis 959 were carried out using the edgeR program 67 . The cutoff of differential miRNA 960 was a fold-change threshold of 1.9 and a maximum p-value <0.05. The 961 miRTarVis tool was used to predict miRNA target 33 . 962 KEGG, Gene Ontology, and GSEA 68 were used to perform the enrichment and 963 pathway analysis. Raw data were deposited at the Gene Expression Omnibus 964 (accession GSE189250), raw fastq files for RNA-seq on mouse hippocampus 965 SAMP8 (accession GSE189249), and raw fastq files for miRNA-seq on mouse 966 hippocampus SAMP8 (accession GSE 189248). 967

ChiP-seq analysis 968
Publicly available H3K9me2 ChIP-seq data corresponding to AML12 cells 969 ChIP-X tool 76 . The detection of transcription factor binding motifs at H3K9me2 983 peaks was performed with the MEME-ChIP database 77 . Motifs with an E-984 value<0.05 were considered statistically significant. Protein-protein interaction 985 network between G9a (EHMT2) and transcription factors was performed in 986 STRING database. 987

Dendritic length, spine density and Golgi stain protocol 988
Mice were sacrificed by cervical dislocation and brains were removed from the 989 skull (n=6 whole brain per experimental group). Then, Golgi stain protocol was 990 of live cells, cortical and striatal neurons were gently detached and mixed with 1016 an equal volume of trypan blue (0.4%). Neurons (%) were counted with a 1017 TC20™ Automated Cell Counter (Biorad, 1450102).