Arc Regulates Transcription of Genes for Plasticity, Excitability and Alzheimer’s Disease

The immediate early gene Arc is a master regulator of synaptic function and a critical determinant of memory consolidation. Here, we show that Arc interacts with dynamic chromatin and closely associates with histone markers for active enhancers and transcription in cultured rat hippocampal neurons. Both these histone modifications, H3K27Ac and H3K9Ac, have recently been shown to be upregulated in late-onset Alzheimer’s disease (AD). When Arc induction by pharmacological network activation was prevented using a short hairpin RNA, the expression profile was altered for over 1900 genes, which included genes associated with synaptic function, neuronal plasticity, intrinsic excitability, and signalling pathways. Interestingly, about 100 Arc-dependent genes are associated with the pathophysiology of AD. When endogenous Arc expression was induced in HEK293T cells, the transcription of many neuronal genes was increased, suggesting that Arc can control expression in the absence of activated signalling pathways. Taken together, these data establish Arc as a master regulator of neuronal activity-dependent gene expression and suggest that it plays a significant role in the pathophysiology of AD.

Interestingly, Arc protein can also be localised in the nucleus, where it binds to a beta-spectrin IV isoform and associates with promyelocytic leukemia (PML) bodies [22][23][24][25], sites of epigenetic regulation of gene transcription [26,27]. Nuclear Arc has been reported to regulate transcription of the GluA1 AMPA receptor [28].
These newly discovered nuclear functions may point to an epigenetic role for Arc in memory consolidation. We have therefore investigated Arc's interaction with chromatin were used for gene expression quantification. Reads were assigned to genes using the expectation/maximisation (E/M) algorithm in Partek Flow [42] based on the annotation model rn6 (Ensembl transcripts release 93) for the rat samples and the annotation model hg38 (Ensembl transcripts release 94) for the HEK293 samples. To ensure only informative genes were included in the downstream analysis, noise (maximum feature counts ≤ 30) was filtered out. Read counts between samples were normalised with the Upper Quantile method [43]. As genes with very low expression might be inadequately represented and incorrectly identified as differentially expressed, a constant of 1 was added to normalised counts for rectification. Statistical analysis was performed using the gene-specific analysis (GSA) module in Partek Flow to identify differential gene expression. p-value and fold changes of differentially expressed genes were calculated based on the lognormal with shrinkage distribution. Genes with an average coverage of less than 3 were also filtered out prior to statistical analysis. Differential gene expression with a cut-off value of false discovery rate (FDR) step-up < 0.05 [44] and an absolute fold change ≥ 2 was considered for further gene ontology analysis in Partek Flow, which is based on the GO Consortium (version 2018_08_01) [45,46]. Setting up cut-off at FDR step-up < 0.05 ensures statistically significant differential gene expressions were tapped with an absolute fold change of 2 to focus on those with a higher magnitude of change [47]. Coherent biological data could then be more meaningfully interpreted through gene ontology analysis [45,46]. GO analysis was also performed on the transcriptional regulators/factors that were observed to be altered upon Arc knockdown using DAVID (version 6.8) [48,49]. The EASE score obtained from the DAVID analysis is a modified Fisher Exact p-value to indicate gene enrichment in the annotation terms. Functional analysis on the statistically significant differential gene expression (FDR step-up < 0.05; absolute fold change ≥ 2) was performed by Ingenuity Pathway Analysis (IPA) (version 01.13, Qiagen, Redwood City, CA, USA). Pathways and their associated downstream effects, diseases, regulator networks and upstream regulators were identified by IPA. Predictions on the possible activation and inhibition of pathways, downstream effects and upstream regulators were inferred from the degree of consistency in the expression of the target genes compared to the fold changes in the differentially expressed gene list. This activation or inhibition status was expressed as a z-score, with z ≥ 2 indicating activation and z ≤ 2 indicating inhibition. Inferences made were based on at least one publication or from canonical information stored in the Ingenuity Knowledge Base. Fisher's exact test was used to calculate the p-value for all analyses in IPA.

Chromatin Reorganisation in Arc-Positive Neurons
The induction of Arc protein expression by pharmacological network activation ( Figure 1) is relatively slow and reaches a maximum level between 4 and 8 h. Arc is only expressed in a subset of neurons. As shown in Figure 2, chromatin organisation is different between neurons that are positive and negative for Arc. Chromatin was visualised by labelling DNA with the fluorescent dye 4 ,6-diamidino-2-phenylindole (DAPI). Whereas chromatin in Arc-negative neurons is relatively homogenous, the nuclei of Arc-positive neurons contain many bright puncta, representing chromocenters with densely packed chromatin, in which genes are likely silenced (Figure 2A,B). The puncta are interspersed with domains of highly open chromatin, which is more supportive of efficient gene transcription. The number of puncta increased from 11.1 ± 0.8 puncta in Arc-negative nuclei to 15.9 ± 0.8 puncta in Arc-positive nuclei ( Figure 2C). However, the mean area of the puncta was not significantly different between Arc-positive and Arc-negative neurons ( Figure 2D).

Arc Associates with Dynamic Chromatin
The interaction between Arc and chromatin was studied in more detail using timelapse fluorescence microscopy of hippocampal neurons expressing Arc and histone 2B (H2B) tagged with YFP and mCherry, respectively ( Figure 3). Arc was induced in 18-day in vitro (DIV18) hippocampal neurons by a 4 h treatment with 4BF. The time-lapse movies of Arc-eYFP and H2B-mCherry revealed a highly dynamic chromatin that constantly reorganises on a time scale of seconds (Movie S1). Arc is concentrated in small puncta to which the chromatin can be seen to reach out with finger-like structures, which may represent the dynamic chromatin loops described by others [71][72][73]. The interaction between Arc and chromatin was studied in more detail using timelapse fluorescence microscopy of hippocampal neurons expressing Arc and histone 2B (H2B) tagged with YFP and mCherry, respectively ( Figure 3). Arc was induced in 18-day in vitro (DIV18) hippocampal neurons by a 4 h treatment with 4BF. The time-lapse movies of Arc-eYFP and H2B-mCherry revealed a highly dynamic chromatin that constantly reorganises on a time scale of seconds (Movie S1). Arc is concentrated in small puncta to which the chromatin can be seen to reach out with finger-like structures, which may represent the dynamic chromatin loops described by others [71][72][73].  (18 DIV) were obtained using a spinning disc confocal microscope (100×, 1.49 NA Apo TIRF objective). Z-stacks (5 images) were acquired for both YFP and mCherry channels. Three-dimensional blind deconvolution (AutoQuant) was used to remove outof-focus fluorescence. The movie is 5 min long, 3.2 s between frames, which was the time required to acquire Z-stacks from both channels. The image on the left shows a single frame of the movie in the centre of the Z-stack of a neuronal nucleus (scale bar = 1 µm). Arc (green) is seen to form puncta, while H2B (red) labels the lattice-like chromatin structure. The panels on the right show six frames of a zoomed-in section illustrating small chromatin structures transiently interact with the two Arc puncta (scale bar = 500 nm). White arrowheads indicate points of contact between Arc and chromatin. The highly dynamic interaction of chromatin with Arc puncta is most clearly seen in the Movie S1.

Arc Associates with a Marker of Active Enhancers
Because Arc was shown to associate with the Tip60 substrate H4K12Ac [29], we have examined interactions of Arc with other histone modifications, by comparing Arc-positive and Arc-negative neurons following pharmacological network activation. The 'histone code' [74] is complex and still incompletely understood. We have therefore focused on histone modifications whose function is best studied. In our survey, we have found  (18 DIV) were obtained using a spinning disc confocal microscope (100×, 1.49 NA Apo TIRF objective). Z-stacks (5 images) were acquired for both YFP and mCherry channels. Three-dimensional blind deconvolution (AutoQuant) was used to remove out-of-focus fluorescence. The movie is 5 min long, 3.2 s between frames, which was the time required to acquire Z-stacks from both channels. The image on the left shows a single frame of the movie in the centre of the Z-stack of a neuronal nucleus (scale bar = 1 µm). Arc (green) is seen to form puncta, while H2B (red) labels the lattice-like chromatin structure. The panels on the right show six frames of a zoomed-in section illustrating small chromatin structures transiently interact with the two Arc puncta (scale bar = 500 nm). White arrowheads indicate points of contact between Arc and chromatin. The highly dynamic interaction of chromatin with Arc puncta is most clearly seen in the Movie S1.

Arc Associates with a Marker of Active Enhancers
Because Arc was shown to associate with the Tip60 substrate H4K12Ac [29], we have examined interactions of Arc with other histone modifications, by comparing Arc-positive and Arc-negative neurons following pharmacological network activation. The 'histone code' [74] is complex and still incompletely understood. We have therefore focused on histone modifications whose function is best studied. In our survey, we have found several histone modifications for which there was a difference in nuclear organisation between Arc positive and negative neurons, including H3K9Ac, H3K4me3, and H3K14Ac (data not shown). Figure 4 illustrates the close association between Arc and H3K27Ac, which marks active enhancers [75,76]. Arc and H3K27Ac form two separate lattice-like structures that are closely inter-connected and, in some locations, appear to overlap (yellow areas in Figure 4).  Hippocampal neurons were treated with 4BF for 4 h, fixed with methanol and stained for Arc and H3K27Ac, which marks sites containing active enhancers Z-stacks of images were acquired of neuronal nuclei using a spinning disc confocal microscope (60× 1.49 NA objective). Resolution was increased using 3D blind deconvolution (AutoQuant). Scale ba is 1 µm. The enlarged section shows the close interaction between Arc and H3K27Ac. Note the elaborate interfaces between Arc (green) and H3K27Ac (red) Scale bar is 500 nm.

Arc Associates with a Marker for Active Transcription
Another histone mark that showed a strong interaction with Arc was H3K9Ac-S10P which requires the concurrent acetylation of lysine 9 of histone H3 (H3K9Ac) and phosphorylation of the neighbouring serine 10 (S10P). This dual marker indicates genomic regions undergoing active transcription [24,77,78]. Figure 5 illustrates the close interaction between Arc and this histone mark, using Stochastic Optical Reconstruction Microscopy (STORM), a form of super-resolution microscopy with a resolution of ~30 nm [79]. Both Arc and H3K9Ac-S10P are enriched at the nuclear periphery, where reorganisation o chromatin between active and inactive transcriptional states takes place [80,81]. With the increased resolution of STORM, Arc can be seen to localise to distinct puncta. H3K9Ac S10P forms an elaborate meshwork, as expected for chromatin, but also is enriched in puncta-like domains. Arrowheads in Figure 5A indicate the close apposition between these two sets of puncta. Close inspection of the interface between the two types of puncta revealed invasions of H3K9Ac-S10P into the Arc puncta (arrows in Figure 5B), resembling the finger-like chromatin structures seen in live cell imaging ( Figure 3, Movie S1). Hippocampal neurons were treated with 4BF for 4 h, fixed with methanol and stained for Arc and H3K27Ac, which marks sites containing active enhancers. Z-stacks of images were acquired of neuronal nuclei using a spinning disc confocal microscope (60×, 1.49 NA objective). Resolution was increased using 3D blind deconvolution (AutoQuant). Scale bar is 1 µm. The enlarged section shows the close interaction between Arc and H3K27Ac. Note the elaborate interfaces between Arc (green) and H3K27Ac (red) Scale bar is 500 nm.

Arc Associates with a Marker for Active Transcription
Another histone mark that showed a strong interaction with Arc was H3K9Ac-S10P, which requires the concurrent acetylation of lysine 9 of histone H3 (H3K9Ac) and phosphorylation of the neighbouring serine 10 (S10P). This dual marker indicates genomic regions undergoing active transcription [24,77,78]. Figure 5 illustrates the close interaction between Arc and this histone mark, using Stochastic Optical Reconstruction Microscopy (STORM), a form of super-resolution microscopy with a resolution of~30 nm [79]. Both Arc and H3K9Ac-S10P are enriched at the nuclear periphery, where reorganisation of chromatin between active and inactive transcriptional states takes place [80,81]. With the increased resolution of STORM, Arc can be seen to localise to distinct puncta. H3K9Ac-S10P forms an elaborate meshwork, as expected for chromatin, but also is enriched in puncta-like domains. Arrowheads in Figure 5A indicate the close apposition between these two sets of puncta. Close inspection of the interface between the two types of puncta revealed invasions of H3K9Ac-S10P into the Arc puncta (arrows in Figure 5B), resembling the finger-like chromatin structures seen in live cell imaging ( Figure 3, Movie S1).
chromatin between active and inactive transcriptional states takes place [80,81]. With the increased resolution of STORM, Arc can be seen to localise to distinct puncta. H3K9Ac-S10P forms an elaborate meshwork, as expected for chromatin, but also is enriched in puncta-like domains. Arrowheads in Figure 5A indicate the close apposition between these two sets of puncta. Close inspection of the interface between the two types of puncta revealed invasions of H3K9Ac-S10P into the Arc puncta (arrows in Figure 5B), resembling the finger-like chromatin structures seen in live cell imaging ( Figure 3, Movie S1).

Arc Regulates Activity-Dependent Gene Transcription
Because Arc was found to associate with histone marks involved in transcription activation, we wanted to investigate whether network activity-induced Arc expression

Arc Regulates Activity-Dependent Gene Transcription
Because Arc was found to associate with histone marks involved in transcription activation, we wanted to investigate whether network activity-induced Arc expression alters the gene expression profile of the neurons. Four short hairpin RNAs (shRNAs) targeting the coding region of Arc were tested for their ability to suppress Arc induction following four hours of 4BF treatment. We selected the most effective shRNA to generate an adeno-associated AAV9 virus. Because AAV9 infection itself may alter the gene expression profile, we also generated a negative control consisting of AAV9 virus encoding a scrambled version of the Arc shRNA. We performed an RNA-Seq analysis of cortical neurons expressing either the Arc shRNA or its scrambled control. When 4BF-mediated Arc expression was prevented using the Arc shRNA ( Figure 6A), mRNA levels for more than 1900 genes were altered ( Figure 6B). Many gene families were affected, including those associated with plasticity (Jun, Fosb, Bdnf, Dlg4, Egr4, Npas4 and Nr4a1), synaptic proteins (syntaxin Stx12 and synaptotagmin Syt3), and neurotransmitter receptors (NMDA, AMPA, GABA, glycine, serotonin, and metabotropic glutamate receptors) ( Figure 6B). Arc also regulated the expression of genes controlling intrinsic excitability: 62 genes encoding ion channels (20 K + , 4 Na + , and 9 Ca 2+ channel subunits, 7 transient receptor potential (Trp) channels, 14 ligand-gated ion channels, 7 regulatory subunits and 1 non-selective cation channel), and 139 genes encoding transporters/pumps (for glutamate, GABA, serotonin, ADP, ATP, phosphate, glucose, inositol, alanine, cysteine, glutamine, glycine, proline, Na + , Ca 2+ , Cl − , H + and Zn 2+ ). These results suggest that Arc regulates activity-dependent gene expression relevant for synaptic function, neuronal plasticity, and intrinsic excitability. Figure 7 shows the 30 top-ranking genes sorted by absolute fold change (FC) caused by the shRNA-mediated knockdown of Arc expression. Gene names are shown together with a description of their function, their fold change, false discovery rate (FDR), and references to relevant papers. Many of the top-regulated genes are involved in synapse modulation, neurotransmission, neurogenesis and neurological disorders. Interestingly, 9 out of the top 30 genes have been implicated in the pathophysiology of AD (Fgf1, Slc30a4, Npas4, Cxcl1, Jdp2, Nts, Mmp10, Orai2 and Tomm34), while an additional 5 genes are linked to amyloid beta (Aβ) metabolism (Mmp13, Mmp12, Slc2a13, Igf1r and Apba1).

GO Analysis of Differentially Expressed Genes
A gene ontology (GO) analysis was performed on the RNA-Seq data, which aim t identify the biological processes and molecular functions altered by the reduction in Ar expression ( Figure 8). Arc knockdown altered many genes involved in the regulation o nervous system development and neuronal differentiation ( Figure 8A). In addition, man of the genes were enriched in biological processes involved in cognition, regulation of ce projection organisation and axonogenesis ( Figure 8A), processes which could modulat the structural plasticity involved in neural development, learning and memory [143,144] While the top ten regulated genes enriched for the regulation of plasma membran bounded cell projection organisation were both up-and down-regulated ( Figure 8Cii) genes enriched for cognition and the regulation of axonogenesis were mostly down regulated due to the absence of Arc (Figure 8Ci,Ciii). Many of the altered genes were als   [140][141][142]. FC indicates fold change, shown as-1/FC when <1.

GO Analysis of Differentially Expressed Genes
A gene ontology (GO) analysis was performed on the RNA-Seq data, which aim to identify the biological processes and molecular functions altered by the reduction in Arc expression ( Figure 8). Arc knockdown altered many genes involved in the regulation of nervous system development and neuronal differentiation ( Figure 8A). In addition, many of the genes were enriched in biological processes involved in cognition, regulation of cell projection organisation and axonogenesis ( Figure 8A), processes which could modulate the structural plasticity involved in neural development, learning and memory [143,144]. While the top ten regulated genes enriched for the regulation of plasma membrane bounded cell projection organisation were both up-and down-regulated (Figure 8Cii), genes enriched for cognition and the regulation of axonogenesis were mostly down-regulated due to the absence of Arc (Figure 8Ci,Ciii). Many of the altered genes were also enriched in molecular functions such as ion channel regulator activity, glutamate receptor binding and ligand-gated ion channel activity ( Figure 8B), including Sgk1 (Figure 8Di), Dlg4, which encodes PSD-95 (Figure 8Dii), and Grin2c, which encodes the NMDA receptor NR2C subunit (Figure 8Diii). These molecular functions are well-established to underlie synaptic plasticity processes crucial for formation of memory [145,146].

Arc Regulates Expression of Synaptic and Plasticity Genes
The GO results in Figure 8 indicated that the knockdown of Arc affected many gene involved in synaptic plasticity, as well as genes implicated in processes underlying learning and memory. We have therefore investigated how Arc knockdown affected genes encoding synaptic proteins by manually curating a list of differentially expressed genes whose protein products are located at the presynaptic or postsynapti compartment. A total of 232 synaptic genes were differentially expressed. Ephb3, Lrfn2
Many of these synaptic genes are also involved in neuroplasticity, cognition, learning and memory, including Syt3, Pcdh8, Pdyn, Lrfn2, Dlg4, Kcna4, Bdnf and Mapki8ip2. Figure 10 lists neuroplasticity genes and genes that are involved in cognition, learning and memory, whose activity-dependent expression is regulated by Arc. Most of these genes were downregulated when activity-dependent Arc expression was prevented.

Arc Knockdown Altered Synaptogenesis, Synaptic Plasticity and Neuroinflammation Pathways
From the GO results and the list of manually curated synaptic genes, we were interested in investigating the signalling pathways and the possible downstream effects resulting from Arc knockdown. We have analysed the differentially expressed genes and their respective fold changes using IPA. Figure 11A shows the top 15 pathways that were altered due to Arc knockdown. IPA made inferences on the activation or inhibition of the pathways based on the differential expression observed and canonical information stored in the Ingenuity Knowledge Base. The degree of activation or inhibition of each identified pathway is indicated by the z-score. The ratio is calculated as the number of differentially expressed genes for each pathway divided by the total number of genes involved in tha pathway. Many identified pathways involved cellular signalling cascades, including those mediated by CDK5, PTEN, integrin and corticotropin-releasing hormone ( Figure  11A). Pathways predicted to be responsible for the observed differential expression profile include opioid and endocannabinoid signalling, synaptogenesis, synaptic long-term depression (LTD) and neuroinflammation ( Figure 11A,B). Kcnj5, Ptgs2, Grin2c, Cacng4 and Gnaq are members of at least two of the pathways shown and are synaptic genes or associated with cognition (Figures 8-10 and 11A). Except for the neuroinflammation signalling pathway, all these pathways are associated with synaptic plasticity Knockdown of Arc-modulated neurotransmission, synaptic plasticity, spine formation/maintenance and neurite outgrowth are processes that are crucial for learning and memory ( Figure 11B) [202][203][204]. Interestingly, the two hallmarks of AD, the generation, clearance, and accumulation of amyloid beta (Aβ) and the formation o neurofibrillary tangles (NFTs), are both affected by downregulation of the neuroinflammation signalling pathway resulting from Arc knockdown ( Figure 11B) These alterations in the generation and clearance of molecular markers and triggers of AD could indicate a possible role of Arc in the pathophysiology of AD [205].

Arc Knockdown Altered Synaptogenesis, Synaptic Plasticity and Neuroinflammation Pathways
From the GO results and the list of manually curated synaptic genes, we were interested in investigating the signalling pathways and the possible downstream effects resulting from Arc knockdown. We have analysed the differentially expressed genes and their respective fold changes using IPA. Figure 11A shows the top 15 pathways that were altered due to Arc knockdown. IPA made inferences on the activation or inhibition of the pathways based on the differential expression observed and canonical information stored in the Ingenuity Knowledge Base. The degree of activation or inhibition of each identified pathway is indicated by the z-score. The ratio is calculated as the number of differentially expressed genes for each pathway divided by the total number of genes involved in that pathway. Many identified pathways involved cellular signalling cascades, including those mediated by CDK5, PTEN, integrin and corticotropin-releasing hormone ( Figure 11A). Pathways predicted to be responsible for the observed differential expression profile include opioid and endocannabinoid signalling, synaptogenesis, synaptic long-term depression (LTD) and neuroinflammation ( Figure 11A,B). Kcnj5, Ptgs2, Grin2c, Cacng4 and Gnaq are members of at least two of the pathways shown and are synaptic genes or associated with cognition (Figures 8, 9, 10 and 11A). Except for the neuroinflammation signalling pathway, all these pathways are associated with synaptic plasticity. Knockdown of Arc-modulated neurotransmission, synaptic plasticity, spine formation/maintenance and neurite outgrowth are processes that are crucial for learning and memory ( Figure 11B) [202][203][204]. Interestingly, the two hallmarks of AD, the generation, clearance, and accumulation of amyloid beta (Aβ) and the formation of neurofibrillary tangles (NFTs), are both affected by downregulation of the neuroinflammation signalling pathway resulting from Arc knockdown ( Figure 11B). These alterations in the generation and clearance of molecular markers and triggers of AD could indicate a possible role of Arc in the pathophysiology of AD [205].

Arc Knockdown Changes the Expression of Alzheimer's Disease Genes
Considering that the generation, clearance and accumulation of amyloid beta and neurofibrillary tangles was predicted to be altered due to the knockdown of Arc, we investigated whether any neurological diseases or psychological disorders were correlated with the profile of differentially expressed genes mediated by Arc knock-down. Figure 12 summarises the disease annotation and predicted activation state for two disease/disorder classes whose associated genes were significantly altered by Arc knockdown.

Arc Knockdown Changes the Expression of Alzheimer's Disease Genes
Considering that the generation, clearance and accumulation of amyloid beta and neurofibrillary tangles was predicted to be altered due to the knockdown of Arc, we investigated whether any neurological diseases or psychological disorders were correlated with the profile of differentially expressed genes mediated by Arc knock-down. Figure 12 summarises the disease annotation and predicted activation state for two disease/disorder classes whose associated genes were significantly altered by Arc knockdown. Absence of Arc was predicted to increase damage of the cerebral cortex and its neurons and cells. In addition, Arc knockdown was also associated with psychological disorders, including Huntington's disease, basal ganglia disorder, central nervous system (CNS) amyloidosis, tauopathy and Alzheimer's disease. Of note, CNS amyloidosis and tauopathy are predictors of AD. The activation states of the five psychological disorders were not reported, possibly due to inconsistencies in the literature findings with respect to fold changes of the differentially expressed genes. However, the p-values for all five disorders were highly significant, suggesting that the progression of these disorders may be modulated by Arc function.
We next investigated how Arc knockdown could affect genes that were previously identified to increase susceptibility to AD. We have manually curated genes that were found to be genetic risk factors of AD and validated them by referencing the genome-wide association studies (GWAS) catalogue [206]. Notably, critical genetic risk factors of AD such as Picalm, Apoe, Slc24a4, and Clu were downregulated upon the knockdown of Arc [207][208][209][210][211] (Figure 13), indicating that activity-induced Arc expression is linked to enhanced transcription of these genes. Out of a total of 39 AD susceptibility genes identified, 26 were regulated by Arc ( Figure 13). Because Arc plays a role in the aetiology of AD by modulating its genetic risk factors, we investigated whether Arc regulates genes that are more broadly involved in the pathophysiology of AD. Figure 14 lists the results. While some differentially expressed genes control amyloid beta formation/accumulation through the regulation of cleavage and stabilisation of amyloid precursor protein (APP) (Mmp13, Slc2a13, Apba1, Casp8, Ptgs2, Absence of Arc was predicted to increase damage of the cerebral cortex and its neurons and cells. In addition, Arc knockdown was also associated with psychological disorders, including Huntington's disease, basal ganglia disorder, central nervous system (CNS) amyloidosis, tauopathy and Alzheimer's disease. Of note, CNS amyloidosis and tauopathy are predictors of AD. The activation states of the five psychological disorders were not reported, possibly due to inconsistencies in the literature findings with respect to fold changes of the differentially expressed genes. However, the p-values for all five disorders were highly significant, suggesting that the progression of these disorders may be modulated by Arc function.
We next investigated how Arc knockdown could affect genes that were previously identified to increase susceptibility to AD. We have manually curated genes that were found to be genetic risk factors of AD and validated them by referencing the genomewide association studies (GWAS) catalogue [206]. Notably, critical genetic risk factors of AD such as Picalm, Apoe, Slc24a4, and Clu were downregulated upon the knockdown of Arc [207][208][209][210][211] (Figure 13), indicating that activity-induced Arc expression is linked to enhanced transcription of these genes. Out of a total of 39 AD susceptibility genes identified, 26 were regulated by Arc ( Figure 13). Absence of Arc was predicted to increase damage of the cerebral cortex and its neurons and cells. In addition, Arc knockdown was also associated with psychological disorders, including Huntington's disease, basal ganglia disorder, central nervous system (CNS) amyloidosis, tauopathy and Alzheimer's disease. Of note, CNS amyloidosis and tauopathy are predictors of AD. The activation states of the five psychological disorders were not reported, possibly due to inconsistencies in the literature findings with respect to fold changes of the differentially expressed genes. However, the p-values for all five disorders were highly significant, suggesting that the progression of these disorders may be modulated by Arc function.
We next investigated how Arc knockdown could affect genes that were previously identified to increase susceptibility to AD. We have manually curated genes that were found to be genetic risk factors of AD and validated them by referencing the genome-wide association studies (GWAS) catalogue [206]. Notably, critical genetic risk factors of AD such as Picalm, Apoe, Slc24a4, and Clu were downregulated upon the knockdown of Arc [207][208][209][210][211] (Figure 13), indicating that activity-induced Arc expression is linked to enhanced transcription of these genes. Out of a total of 39 AD susceptibility genes identified, 26 were regulated by Arc ( Figure 13). Because Arc plays a role in the aetiology of AD by modulating its genetic risk factors, we investigated whether Arc regulates genes that are more broadly involved in the pathophysiology of AD. Figure 14 lists the results. While some differentially expressed genes control amyloid beta formation/accumulation through the regulation of cleavage and stabilisation of amyloid precursor protein (APP) (Mmp13, Slc2a13, Apba1, Casp8, Ptgs2, Because Arc plays a role in the aetiology of AD by modulating its genetic risk factors, we investigated whether Arc regulates genes that are more broadly involved in the pathophysiology of AD. Figure 14 lists the results. While some differentially expressed genes control amyloid beta formation/accumulation through the regulation of cleavage and stabilisation of amyloid precursor protein (APP) (Mmp13, Slc2a13, Apba1, Casp8, Ptgs2, Gpr3, Pawr, Timp3, Kcnip3, Plk2, Aplp2, Bace2, Apoe and Apba2), others are involved in the hyperphosphorylation of tau and formation of neurofibrillary tangles (Npas4, Cxcl1, Dryrk2, Tril, Pltp, Plk2 and Selenop). Arc knockdown also altered the expression of genes that are associated with the neurodegeneration and neurotoxicity observed in AD (Casp8, Bcl2l11, Alg2, Tac1, Bdnf, Hmox1, Pawr, Ccl2, Selenop and Atf6). Finally, Arc regulated genes associated with altered cognitive function, a characteristic of AD (Mmp13, Pdyn, Tac1, Bdnf, Nr4a2, Penk, Pltp and Ccl2). To date, presenilin 1 (Psen1) and glycogen synthase kinase 3 beta (Gsk3b) are the only AD mediators that have been reported to physically associate and interact with Arc [212][213][214]. Arc also interacts with endophilin 2/3 and dynamin and recruits them to early/recycling endosomes to traffic APP and beta secretase 1 (BACE1), crucial determinants of AD progression [214]. However, the observation that knocking down Arc resulted in more than 100 differentially expressed genes that are either AD susceptibility genes or genes implicated in the pathophysiology of AD (Figures 13 and 14) suggests that Arc could be mediating the expression of these genes via transcriptional regulation and not simply physical interactions. Arc has previously been reported to reside in the nucleus [22,28,50,51], and we have shown how Arc physically associates with chromatin and with markers of active transcription and enhancers (Figures 3-5). Therefore, we wanted to investigate how Arc downregulation affects transcription regulation.  [212][213][214]. Arc also interacts with endophilin 2/3 and dynamin and recruits them to early/recycling endosomes to traffic APP and beta secretase 1 (BACE1), crucial determinants of AD progression [214]. However, the observation that knocking down Arc resulted in more than 100 differentially expressed genes that are either AD susceptibility genes or genes implicated in the pathophysiology of AD (Figures 13 and 14) suggests that Arc could be mediating the expression of these genes via transcriptional regulation and not simply physical interactions. Arc has previously been reported to reside in the nucleus [22,28,50,51], and we have shown how Arc physically associates with chromatin and with markers of active transcription and enhancers (Figures 3-5). Therefore, we wanted to investigate how Arc downregulation affects transcription regulation.

Upstream Regulators Associated with Arc-Dependent Genes
Because Arc knockdown resulted in the differential expression of 1945 genes ( Figure  6), altering downstream pathways ( Figure 11) possibly leading to disease states ( Figure  12), we identified the upstream modulators that could explain the vast differential expression pattern observed. From the IPA analysis, 11 upstream regulators were predicted to critically contribute to the differential expression profile (Figure 17).

Upstream Regulators Associated with Arc-Dependent Genes
Because Arc knockdown resulted in the differential expression of 1945 genes (Figure 6), altering downstream pathways ( Figure 11) possibly leading to disease states (Figure 12), we identified the upstream modulators that could explain the vast differential expression pattern observed. From the IPA analysis, 11 upstream regulators were predicted to critically contribute to the differential expression profile (Figure 17).
Except for Sox2, none of these upstream regulators were transcriptionally affected by Arc knockdown, suggesting that Arc controls their function through a different mechanism. SOX2 and HDAC4 were both activated by the absence of Arc, while the function of the remaining nine regulators was inhibited. Of note, the predicted inhibition of CREB1 (z-score = −3.5) and APP (z-score = −2.8) explains the differential expression of 100 and 94 genes, respectively ( Figure 17). The 11 upstream regulators predicted by IPA control the expression of Nr4a2, Slc6a1 and Igf1r, genes that are also involved in AD progression, neuroinflammation pathways and synaptic LTD (Figures 11A and 14). We have investigated the mechanisms by which Arc could alter the function of the identified upstream regulators, resulting in the alteration of downstream pathways and AD progression. The downstream pathways investigated are (i) opioid signalling, (ii) synaptogenesis, (iii) the endocannabinoid neuronal synapse pathway, (iv) synaptic LTD and (v) neuroinflammation ( Figure 18). These are also the pathways whose downstream effects we focused on in Figure 11.
Because Arc knockdown resulted in the differential expression of 1945 genes (Fig  6), altering downstream pathways ( Figure 11) possibly leading to disease states (Fig  12), we identified the upstream modulators that could explain the vast differen expression pattern observed. From the IPA analysis, 11 upstream regulators w predicted to critically contribute to the differential expression profile (Figure 17). Figure 17. Upstream regulators associated with differential gene expression observed up knockdown of Arc. Activation z-score indicates the predicted activity of upstream regulators Figure 17. Upstream regulators associated with differential gene expression observed upon knockdown of Arc. Activation z-score indicates the predicted activity of upstream regulators by IPA analysis. Upstream regulators that were predicted to be inhibited are highlighted in blue while those activated are highlighted in orange. Regulated genes were highlighted in green (downregulated) and red (upregulated). APP, CREB1 and TNF are three upstream regulators identified by IPA that controlled the highest number of genes involved in the downstream pathways highlighted ( Figure 18). The top five genes regulated by APP were Igf1r (synaptic LTD) [354], Ptgs2 (endocannabinoid neuronal synapse pathway; neuroinflammation, AD progression) [227,[355][356][357], Jun (neuroinflammation, AD progression) [358][359][360][361], Dlg4 (PSD95, synaptogenesis) [362,363] and Syn2 (synaptogenesis) [364] (Figure 18). In addition to Ptgs2 and Syn2, CREB1 regulated the differential expression of Slc6a1 (neuroinflammation) [365], Pdyn (opioid signalling) [366] and Fosb (opioid signalling) [367] (Figure 18). Interestingly, TNF, whose transcription was not altered upon knockdown of Arc, regulates 15 genes (Figure 17), the top five of which are Casp8 (neuroinflammation) [368], Ptgs2 (also regulated by APP and CREB1), Gabrg2 (neuroinflammation) [369,370], Bdnf (synaptogenesis, neuroinflammation, AD progression) [371][372][373][374] and Penk (opioid signalling, AD progression) [375]. While the top CREB1-regulated genes are mainly associated with the opioid signalling pathway, APP and TNF are implicated in neuroinflammation. Triggering of the neuroinflammation pathway leads to the altered expression of AD-associated genes such as Ptgs2, Jun, Bdnf, Hmox1 and Gabbr2. have investigated the mechanisms by which Arc could alter the function of the identified upstream regulators, resulting in the alteration of downstream pathways and AD progression. The downstream pathways investigated are (i) opioid signalling, (ii) synaptogenesis, (iii) the endocannabinoid neuronal synapse pathway, (iv) synaptic LTD and (v) neuroinflammation ( Figure 18). These are also the pathways whose downstream effects we focused on in Figure 11. Figure 18. Upstream regulators of differential gene expression caused by Arc knockdown. The map shows 11 upstream regulators (blue and orange boxes) predicted by IPA to mediate altered gene expression upon Arc knockdown. Genes are positioned in the extracellular space, the plasma membrane, the cytosol, or the nucleus, depending on where their associated proteins are located. Arc was positioned at the interface of nucleus and cytoplasm because it can be in either compartment. Only genes that were involved in the following pathways and disease annotations are shown: (i) opioid signalling, (ii) synaptogenesis, (iii) endocannabinoid neuronal synapse pathway, (iv) synaptic LTD, (v) neuroinflammation, (vi) CNS amyloidosis, (vii) tauopathy and (viii) AD. Genes associated with disease annotations are boxed in magenta. The respective fold changes are indicated below each gene.

Discussion
Activity-regulated cytoskeleton-associated protein (Arc) was discovered in 1995 as a neuronal activity-dependent immediate early gene [1,2], which is rapidly transcribed in response to network activation associated with novel experiences [7][8][9][10][11]. Knockdown of Arc expression interferes with the stabilisation of short-term memory, indicating that Arc plays a critical role in memory consolidation [3,4].
Arc's function has been most widely studied in excitatory synapses, where it regulates the endocytosis of AMPA receptors [14,17]. Interestingly, AMPA receptor removal also underlies Aβ-induced synaptic depression and dendritic spine loss [379], processes thought to be associated with cognitive dysfunction in Alzheimer's disease [380]. In Arc knockout mice, LTP is not stable, and dissipates within a few hours, consistent with the impaired memory consolidation observed in these mice [3][4][5][6]. However, the absence of the late form of LTP in Arc knockout mice cannot be explained by an AMPA receptor endocytosis deficit [4], indicating that Arc must have additional functions. The data presented here identify a second function for Arc: regulation of neuronal activity-dependent transcription for genes associated with synaptic plasticity, intrinsic excitability and cellular signalling. Analysis of the differentially expressed genes points to Arc's involvement in several neurological disorders, including autism, Huntington's disease and Alzheimer's disease. This newly proposed role for Arc is supported by its interaction with chromatin and histone markers reported here (Figures 2-5, Movie S1).

Arc and Chromatin
Pharmacological network stimulation induces Arc in a subset of cultured neurons (Figure 1). Whereas chromatin in cultured hippocampal neurons is relatively uniform, Arc-positive neurons are characterised by a larger number of densely packed heterochromatin puncta (chromocenters), likely harbouring silent genes, interspersed with highly open euchromatin domains, which are capable of active transcription (Figure 2). This result is consistent with what has been observed in vivo, where Arc-deficient mice were found to have decreased heterochromatin domains [51]. These significant changes in chromatin structure observed in Arc-positive neurons are likely associated with equally substantial alterations in gene expression profiles. The correlation between Arc expression and chromatin remodelling that we observed does not establish a causative relationship. It is possible that Arc expression requires an alteration in chromatin structure, or alternatively, Arc expression may cause chromatin remodelling. Additional experiments are needed to decide on the underlying mechanism. It is also not clear at this time what determines which neurons will express Arc following network activation, although it likely has to do with the degree of participation of individual neurons in the enhanced network activity, which in turn depends on their synaptic connectivity.
Arc appears to physically interact with DNA: time-lapse movies show dynamic chromatin loops that appear to invade Arc puncta (Figure 3, Movie S1). The interaction is transient, lasting only a few seconds. Because these Arc puncta likely contain the histone acetylase Tip60 [29], it is conceivable that this interaction alters chromatin accessibility, thereby facilitating transcription. This idea is further strengthened by the association of Arc puncta with a histone marker for active enhancers (Figure 4), as well as the close apposition between Arc puncta and puncta for a dual histone marker (H3K9Ac-S10P) that labels sites of active transcription ( Figure 5). A similar result has been obtained in vivo, where cocaine administration in rats results in an increase in nuclear Arc, which then associates with H3S10P [51]. Taken together, the data presented here on the interaction of Arc and chromatin may provide a mechanism for epigenetic regulation of gene transcription as the basis for memory consolidation.

How Does Arc Regulate Transcription?
Preventing Arc induction during neuronal network activation affects the transcription of a very large number of genes ( Figure 6). The domain structure of Arc protein appears to rule out that it can function as a transcription factor [212]. This raises the question: how does Arc regulate transcription?
One possible mechanism, discussed above, is that Arc epigenetically controls gene transcription by regulating chromatin structure (through Tip60 or other chromatin remodellers) and modification of histones (e.g., H4K12Ac [29]). However, the differential gene expression associated with Arc knockdown is mediated through eleven upstream regulators identified by IPA (Figures 17 and 18). This suggests that Arc has additional, lessdirect ways of regulating transcription. Interestingly, to date, none of the eleven upstream regulator proteins have been shown to either directly interact with or be modulated by Arc. They are also not transcriptionally controlled by Arc (except for Sox2) ( Figure 17). How, then, does Arc regulate transcription by activating or inhibiting these upstream regulators? Using IPA and its ingenuity knowledge base, we were able to identify several known interactors of Arc that can modulate the action of the upstream regulators, which could then subsequentially alter gene transcription ( Figure 20A). Next, we will discuss the mechanisms by which four identified Arc interactors, NOTCH1, TIP60/Kat5, APP and GSK3B, could modulate the upstream regulators.
NOTCH1. NOTCH1 is a transmembrane receptor capable of signalling to the nucleus. Arc is required for the proteolytic cleavage of NOTCH1 to release its intracellular domain (NICD), which can translocate to the nucleus and alter transcription [381]. NICD regulates the expression of the transcriptional repressor BCL6 [100] and the activity of the calciumdependent kinase CAMK4 [382], which in turn alter the localisation and the nuclearcytoplasmic shuttling of the histone deacetylase HDAC4, thereby affecting its downstream interactions/modulation [383,384] (Figure 20B). NOTCH1 could regulate the stability, nuclear localisation and signalling of the transcription factor SOX2 through regulation of the protein kinase AKT1 and cell-surface glycoprotein CD44 [385][386][387][388] (Figure 20B). NOTCH1, through NICD, controls the expression of plasminogen activator inhibitor-1 (SERPINE1) [389], an inhibitor of thrombin (F2) [390] (Figure 20B). NOTCH1 regulates the transcriptional activity of T-cell factor 4 (TCF7L2) [391], through its interaction with the DNA-repair protein Ku70 (XRCC6) [392] (Figure 20B). NOTCH1 interacts with the nerve growth factor NR4A1/Nur77 [393], thereby modulating expression levels of the cytokine tumour necrosis factor alpha (TNF) [394]. Finally, NOTCH1 regulates the expression level of the inhibitor of apoptosis protein cIAP1/Birc2 [395], which also affects TNF expression [396] ( Figure 20B). modulated by Arc. They are also not transcriptionally controlled by Arc (except for Sox2) ( Figure 17). How, then, does Arc regulate transcription by activating or inhibiting these upstream regulators? Using IPA and its ingenuity knowledge base, we were able to identify several known interactors of Arc that can modulate the action of the upstream regulators, which could then subsequentially alter gene transcription ( Figure 20A). Next, we will discuss the mechanisms by which four identified Arc interactors, NOTCH1, TIP60/Kat5, APP and GSK3B, could modulate the upstream regulators. TIP60/Kat5. The Kat5 gene encodes TIP60, a member of the MYST family of histone acetyl transferases, which plays important roles in chromatin remodelling and transcription regulation [397]. In the fruit fly Drosophila, TIP60 has been implicated in epigenetic control of learning and memory [398], while it mediates APP-induced apoptosis and lethality in a fly AD model [399]. Nuclear Arc interacts with TIP60 at perichromatin regions and recruits TIP60 to PML bodies, sites of epigenetic transcription regulation [29]. Arc levels correlate with acetylation status of H4K12, a substrate of TIP60 and a memory mark that declines with aging [33], suggesting that Arc mediates activation of TIP60. TIP60/Kat5 facilitates the repressive action of HDAC4 through the formation of complexes with the zinc-finger transcription factor KLF4 [400,401], the cAMP-dependent transcription factor ATF3 [402][403][404] and the neurodegenerative disease protein ataxin-1 (ATXN1) [405,406] ( Figure 20B). Arc's interaction with TIP60/Kat5 may result in a complex being formed at the cIAP1/Birc2 promoter region [407] to mediate downstream signalling of TNF [396] ( Figure 20B). TIP60/Kat5 forms a complex with the Kaiso transcription factor ZBTB33 [408], resulting in the inhibition of the TCF7L2 transcriptional complex [409] (Figure 20B). Complexing of TIP60 with ARID1B could affect SOX2 signalling [410][411][412]. The regulation of SOX2 by TIP60/Kat5 could also have an implication on the transcriptional activity of Achaete-Scute homolog 1 (ASCL1), as SOX2 and ASCL1 regulate each other, possibly in a feedback loop [413,414].

Interactions among TIP60, NOTCH1 and APP
A delicate regulatory network exists among Arc's interactors TIP60/Kat5, NOTCH1 and APP ( Figure 20B). Arc's activation of the γ-secretase PSEN1 to promote cleavage of APP not only increases amyloid beta load, but also results in an increased level of the APP intracellular domain (AICD) [214,447]. AICD forms a complex with TIP60/Kat5 to alter transcriptional activity crucial for AD progression [421,[448][449][450][451][452][453] (Figure 20B). This AICD-TIP60 interaction is disrupted by NICD, formed when Arc activates NOTCH1 [381], thereby downregulating AICD signalling while promoting NICD signalling [454,455] (Figure 20B). The formation of NICD and AICD is competitive, as NOTCH1 and APP are both substrates of γ-secretase [456], whose activity is regulated by Arc [214]. In addition, the induction of TIP60 histone acetylation activity by Arc [29] could also increase the negative regulation of NOTCH1 [454] (Figure 20B). This highlights Arc as an important modulator of the relationship and downstream signalling mediated by NOTCH1, TIP60/Kat5 and APP. Of note, the mRNA levels of Notch1, Kat5 and App were not significantly altered upon knockdown of Arc, indicating that the transcriptional changes brought about were due to protein interaction and activation ( Figure 20B), which is upstream of transcription ( Figure 20A). However, the modulation of upstream regulators by Arc is also dependent on its subcellular localisation.

Arc's Subcellular Localisation Determines Its Function
When Arc is localised outside of the nucleus, it tends to accumulate in dendrites and spines, small membrane protrusions that harbour excitatory synapses. Here, Arc controls the removal of AMPA receptors by endocytosis, allowing it to regulate synaptic efficacy [14,17]. Synaptic Arc also associates with the synaptic scaffolding protein PSD-95/Dlg4, which complexes with the tyrosine kinase FYN [457][458][459], allowing it to regulate brain-derived neurotrophic factor (BDNF) signalling through tyrosine receptor kinase B (TrkB), a major pathway for synapse maturation, plasticity and neurodevelopmental disorders [460]. Activation of FYN could also mediate the secretion of TNF [461] (Figure 20B). A high-affinity interaction with calcium-calmodulin kinase 2 beta (CAMK2B) targets Arc to inactive synapses, where it removes GluA1 AMPA receptors from the postsynaptic membrane surface [444].
Arc has been shown to possess both a nuclear localisation signal (NLS) and a nuclear retention domain [28], allowing it to translocate to the nucleus autonomously. Once in the nucleus, Arc has access to several other potential binding partners, including a nuclear spectrin isoform (βSpectrinIV∑5) [22] and TIP60, a subunit of a chromosome remodelling complex [29]. Association with Amida, encoded by the Tfpt gene ( Figure 20B), facilitates Arc's entry into the nucleus [462]. Amida is a subunit of the INO80 chromatin remodelling complex, which contains the transcriptional regulator MCRS1 [463,464]. MCRS2, an isoform of MCRS1, is associated with the MLL chromatin remodelling complex, which also contains KMT2A (MLL1) ( Figure 20B). Arc's association with Amida and possibly the INO80 and MLL complexes may provide Arc with yet another opportunity to control gene expression by altering chromatin structure.
The ability of Arc to translocate between the synapse and the nucleus, with unique functions in each subcellular compartment, further strengthens its role in memory consolidation, which requires both alterations in synaptic function and de novo gene transcription [465].

Arc Controls Synaptic Plasticity and Intrinsic Excitability
Arc's well-studied ability to alter synaptic efficacy by endocytosis of AMPA receptors established it as a critical regulator of synaptic plasticity [14,17,459,466]. Whereas this mechanism of activity-dependent removal of glutamate receptors supports Arc's role in mediating long-term depression (LTD) [467][468][469][470], it does not explain the absence of stable LTP observed in Arc knock-out mice [4]. Because late-LTP is considered a critical cellular mechanism underlying memory consolidation, the molecular and cellular mechanism by which Arc supports memory stabilisation has remained elusive. The data presented here showing that Arc transcriptionally regulates the expression of a large number of synaptic proteins, with functions in both the pre-and post-synaptic compartment (Figure 9), provide a new mechanism by which Arc can control long-lasting changes in synaptic structure and function required for memory consolidation.
Formation of a memory trace not only requires long-term changes in the strength of the synapses connecting the neurons that constitute the engram, but also stable changes in their intrinsic excitability [471][472][473]. Because Arc controls the expression of a large number of ion channels and pumps/transporters, it appears that Arc is capable of supporting this functional aspect of memory consolidation as well.

Arc and Alzheimer's Disease
Alzheimer's disease is a devastating neurodegenerative disorder [474,475] characterised by the progressive loss of both synaptic function [476] and long-term memory formation [477]. There is currently no therapy that prevents, stabilises, or reverses the progression of this disease, which is projected to take on epidemic proportions as the world population ages [478,479]. Several previous studies have revealed an association between Arc and AD. A landmark study published in 2011 showed that Arc protein is required for the formation of amyloid (Aβ) plaques [214]. Moreover, Arc protein levels are aberrantly regulated in the hippocampus of AD patients [480] and are locally upregulated around amyloid plaques [481], whereas a polymorphism in the Arc gene confers a decreased likelihood of developing AD [482]. It has been shown that spatial memory impairment is associated with dysfunctional Arc expression in the hippocampus of an AD mouse model [483].
These published results together with the data presented here suggest that aberrant expression or dysfunction of Arc contribute to the pathophysiology of AD [476,484].

Arc and Ad Therapy
Arc's ability to transcriptionally regulate AD susceptibility and AD pathophysiologyrelated genes indicates a possibility for modifying expression and activity of Arc as a therapy for AD. Current treatments for AD are symptomatic, not effective disease-modifying cures [485,486]. Many hypotheses have been proposed to underlie the development of AD, including (i) amyloid beta aggregation, (ii) tau hyperphosphorylation, (iii) neuroinflammation, (iv) neurotransmitter dysfunction, (v) mitochondria dysfunction, (vi) glucose metabolism, (vii) vascular dysfunction and (viii) viral infection [485,[487][488][489]. These hypotheses have generated many new compounds, none of which have shown efficacy in slowing cognitive decline or improving global functioning [485,488]. Arc appears to be a good therapeutic candidate for AD, because of its involvement in amyloid beta production, tau phosphorylation, neuroinflammation and neurotransmission. Moreover, we have shown that Arc can modulate the expression of many genetic risk factors and genes associated with the pathophysiology of AD (Figures 11, 13, 14 and 19). Currently, known drugs that could increase mRNA or protein expression of Arc include antidepressant drugs [490], phencyclidine [491] and corticosterone, a memory-enhancing drug [492]. Arc expression could be altered by targeting TIP60 and PHF8, two histone modifiers that together control Arc transcription [24]. Drugs could also modulate Arc's effect via its interactors such as TIP60 and NOTCH1. Natural and synthetic drug molecules targeting TIP60 exist, but they are currently used for cancer treatment [493]. Modulation of NOTCH1 function often involves inhibitors of γ-secretase, which would also affect APP cleavage [456,494]. These pharmaceutical modifications of Arc expression and activity could present a promising starting point for the development of a more effective AD therapy.

Conclusions
The neuronal Arc gene is critically important for the stabilisation of memories. It encodes a protein that localises to dendritic spines, where it regulates endocytosis of glutamate receptors. However, Arc can also be found in the nucleus, where its function is less understood. We find that Arc tightly associates with two markers of active DNA transcription, both of which have recently been shown to be upregulated in Alzheimer's disease. Our results further show that Arc is a master regulator of activity-dependent gene expression, controlling mRNA levels of over 1900 genes involved in neuronal plasticity and intrinsic excitability, as well as over 100 genes implicated in the pathophysiology of AD. Because Arc function has previously been shown to be dysregulated in AD, these new findings identify Arc as new therapeutic target for the treatment of AD.

Conflicts of Interest:
The authors declare no conflict of interest.