Mammalian Activity-Regulated Cytoskeleton-Associated protein (Arc, and its variants, products of immediate early Arc gene) is known for its exquisite importance for synapse maturation, synaptic plasticity, learning, and memory, while the dysregulation of Arc expression can have vast consequences for normal brain function, triggering aberrant wiring of neuronal circuits (24, 25). Arc/Arg3.1 mediates activity-dependent elimination of redundant climbing fiber to Purkinje cell synapses in the developing cerebellum (26). In glutamate neurons, following activation of the NMDA receptor, Arc mRNA became significantly upregulated in the nucleus before being transported to the dendrites for translation (12, 27). At postsynaptic sites Arc mRNA localizes in the PSD-95-NMDAR complexes and binds to inactive CaMKIIβ (not bound to calmodulin) (28, 29). Activation of calcium entry via ionotropic glutamate receptors, especially the NMDA subtype, during normal synaptic transmission and in seizures results in calmodulin-dependent dissociation of Arc mRNA from CaMKIIβ and PKA-dependent induction of Arc protein expression (29, 30). Later it happens that spines on dendrites, where Arc was produced, change their morphology increasing the spine density and proportion of thin spines, together with the endocytosis of AMPA receptors leading to decreased synaptic efficacy (21). Arc hyperexpression facilitates not only AMPA endocytosis but also downregulates transcription of the GluA1 subunit of the AMPA receptor which favors synaptic downscaling (31, 32). In this way, Arc participates in synaptic long-term potentiation and the consolidation of long-term memory (33). Interestingly, Arc transcription exhibited distinct temporal kinetics depending on the activation of excitatory inputs that convey functionally distinct information (34).

Upon hyperexpression, in local dendritic compartments, Arc protein assembles into retrovirus-like capsids packing predominantly Arc mRNA. These capsids leave neurons wrapped in extracellular vesicles and can transmit mRNA to nearby cells (20, 22). Curiously, activation of metabotropic glutamate receptors (mGluR) facilitates Arc mRNA translation in capsid “infected” neurons (20, 22). Arc capsid-mediated transfection was not yet observed in the mammalian brain, but in mice, Arc expression in DRG neurons results in capsid formation modulating skin vasodilation (35). Arc mRNA transfer between cells can probably explain why after hyperactivity Arc accumulates not only in neurons but also in astrocytes, while it is originated from nearby neurons (27, 36, 37). Mammalian Arc protein lacks zinc fingers but has positively charged motifs binding polyanionic mRNA (38–40). This probably allows nonspecific packing of neuronal host mRNAs other than Arc mRNAs, because half of the RNAs encapsidated by retroviruses are host-derived RNAs (41). Extracellular vesicles are supposed to participate in the spread of different neurodegenerative pathologies over time in an activity-dependent manner via synaptic connections (42, 43). Generally, it looks probable that Arc capsids may mediate activity-dependent intercellular paracrine transfer of genetic information, which may alter neighboring cell response to network activity.

Recently it became clear that Arc protein is not only some important regulatory synaptic protein but turns out to be repurposed retrotransposon protein that mediates intercellular mRNA transfer (22, 44, 45). This mechanism involves formation (mainly by glutamatergic neurons) and expression in their synapses of pseudo-viral, retrovirus-like particles (of about 60 nm diameter) made by Arc protein multimers, which encapsulates specific neuronal mRNA and then are trafficked across synaptic boutons (20, 46). The median part of the Arc protein sequence has similarity to modern retroviruses (for example HIV) Group-specific antigen (Gag). Like HIV Gag, Arc forms capsomeres (Figure 1) which self-assemble into capsids of about 30 nM in diameter, while the multimerization of Arc is mediated by its N-terminal helical coil motif (48–50). Arc protein ensembles form multiple capsomeres with symmetric pentameric structures (Figure 1), resembling some viral ion channels (48). Arc variants are found in both birds and humans, but not fish. It is hypothesized that Arc was inserted into the ancestral genome of all tetrapods (amphibians, reptiles, birds, mammals) around 350–400 million years ago. In humans, Arc is found in the greatest amount in brain structures associated with memory, which may be associated with synaptic plasticity and consolidation of memories (44, 51).

Of course, virus-like transport of mRNA can be employed not only in long-term memory, and Arc can be presented better as a multifunctional signaling hub (52). For example, it was shown that Arc can transport mRNAs from mutated genes related to schizophrenia (53). Arc may also play a role in the immunity and activity-dependent β-amyloid generation (30, 54).

Evolutionary analysis shows that Arc is derived from a lineage of Ty3/gypsy retrotransposons, which are also ancestors to retroviruses, that have been repurposed during the evolution to mediate intercellular communication in the nervous system (20, 22). Thus, animals before tetrapods most probably use Ty3/gypsy variants for memory consolidation. On the other hand, the tetrapod Arc protein structure is similar to the one which has been found in Schizophora (true) flies (Dipteros), thus it may have been transferred to a common ancestor of these insects independently (22, 48). At least two Arc homologs (dArc1 and dArc2) are found in flies, which arose by genomic duplication of an ancestral dArc gene but were not detected in any other dipteran (e.g., mosquitoes) or protostome species (20, 22). In Drosophila Arc1 protein forms capsid-like structures that bind mRNA in neurons and are loaded into extracellular vesicles. These vesicles pass from motor neurons to muscles. The disruption of the transfer, in turn, blocks synaptic plasticity (20). Thus, the transsynaptic mechanism of mRNA transport involving retrovirus-like Arc- capsids and extracellular vesicles can be considered proven (55).

While Arc genes originated independently, they still share significant homology in the retroviral Gag domain, and thus the ability to form capsids. Interestingly, the protein participating in memory consolidation in fish retains some immunoreactivity to Arc (56). This domestication of proteins using transposable elements is a well-known phenomenon known as lateral gene transfer that can enrich the recipient and provide a mechanism for evolutionary flexibility (57, 58). About one-half of the mammalian genome consists of DNA with viral or transposon origin and about 8% belong to ancient retroviruses (2, 59). According to the Gene Expression Omnibus (GEO) database for gene expression profiling, the Arc gene was identified as a candidate gene involved in the pathogenesis of various neurological diseases, including epilepsy and a number of others, such as depression.

Actively regulated cytoskeletal-associated (ARC) protein, encoded by the ARC gene, was characterized at the end of the last century (12, 13). This protein is localized in the synaptic contacts of neurons, and its synthesis is dependent on the NMDA receptor's activity. This protein plays a critical role in the molecular processes associated with learning and memory and may also serve as a marker of plastic changes in the brain (15).

Impaired Arc protein synthesis is associated with various brain pathologies, including memory disorders, Alzheimer's disease, autism spectrum disorders, schizophrenia, and epilepsy (60–62). Arc cellular pathways have emerged as key regulators of synaptic plasticity, and are becoming known for being central players in genetic risk for many neural disorders (20).

Thus, in a study on a large sample of patients, it was shown that both rare mutations and epigenetic regulation of ARC contribute to the pathogenesis of schizophrenia, at least in some patients (20, 63). It has also been shown that small mutations affecting one or more nucleotides are extremely widespread among glutamatergic postsynaptic proteins, including proteins associated with ARC and NMDAR, regulated by synaptic activity. At the same time, genes affected by mutations in schizophrenia overlap with genes affected by mutations in autism spectrum disorders and some other brain pathologies (64).

Patients with epilepsy have an increased likelihood of experiencing psychotic symptoms, many of which are similar to those of depression. However, many psychiatric comorbidities, not just the symptoms of depression, are more common in patients with epilepsy (65). It remains an open question whether depression is a risk factor for the development of epilepsy.

Interesting results were obtained by measuring the level of Arc/Arg3.1 in the blood plasma of children with a diagnosed autism spectrum disorder (ASDs) (66). The average level of Arc/Arg3.1 protein in blood plasma in autism was significantly higher than in the control (healthy) group. However, no significant association was found between plasma Arc/Arg3.1 protein levels and measures of autism severity (66). This suggests that Arc/Arg3.1 can be used as an early biomarker for diagnosing autism (66).

Autism and Angelman syndrome (AS) share many common characteristics, although Angelman syndrome is not usually included in ASD (67, 68). Angelman Syndrome is a neurological disorder caused by a mutation of the E3 ubiquitin ligase UBE3A, a gene whose mutation is associated with autism spectrum disorders. In childhood, seizures are observed in approximately 80–90% of patients with AS (69). Arc is one of the target proteins of the UBE3A gene. Since Arc is involved in learning and memory, its expression directly affects the manifestation of AS, including the epileptic seizures, typical of this syndrome. But the function of UBE3A during nervous system development and how UBE3A mutations give rise to cognitive impairment in individuals with AS and ASDs remains unclear. Nevertheless, we know that experience-driven neuronal activity induces UBE3A transcription. UBE3A then regulates excitatory synapse development by controlling the degradation of Arc (67). Disruption of UBE3A function in neurons leads to an increase in Arc expression and a concomitant decrease in the number of AMPA receptors at excitatory synapses. This deregulation of AMPA receptor expression at synapses may contribute to the brain pathologies that occur in AS and possibly other ASDs (67). Most researchers agree that the genesis of epileptic seizures in AS has a complex genesis (17). It is known that in patients with AS, there is a global decrease in the volume of a number of subcortical structures and an increase in the volume of gray matter. The degree of the abnormality correlates with the severity of seizures, suggesting that the occurrence of seizures may be directly related to morphological changes in the brain (17).

Fragile X Syndrome (FXS) is one of the most common inherited forms of developmental delay (70). Epilepsy is reported in up to 20% of individuals with fragile X syndrome (71). It was shown that the FMR1 gene, which produces the FMRP protein is responsible for FXS. One of the biological manifestations of FXS is elevated levels of metabotropic glutamate receptor (mGluR)-dependent long-term depression (LTD), (mGluR-LTD) a type of synaptic plasticity which is characterized by a reduction in the synaptic response at the excitatory synapses and overexpression of several proteins including Arc (72, 73). This abnormal overexpression of Arc leads to increased endocytosis of AMPARs and increased mGluR-LTD. In addition, altered dendritic spine morphology was observed not only in animal models of FXS but also in humans with this disorder (70, 74).

Epilepsy does not have a clearly identifiable cause in about half of patients. Also, this disease can be associated with various factors, including various genetic abnormalities. Some types of epilepsy are linked to certain genes, but generally, genetic factors can make a person more sensitive to the mechanisms that cause seizures. In addition, epilepsy has a wide comorbidity with other brain pathologies. Thus, we most likely cannot speak about the direct responsibility of the genetic pathologies of Arc for the formation of one form or another of epilepsy in humans, but Arc can certainly be associated with epilepsy.

Arc depletion may affect memory loss in the post-ictal state. It is known that memory problems experienced by people with epilepsy are characterized by difficulty in retrieving episodes or events that happened before a seizure and even general semantic information (43). Genetic interference of activity-dependent Arc protein expression in the rat hippocampus impairs the maintenance of long-term potentiation and blocks the consolidation of long-term memory (33, 75). Can it be associated with the dysregulation of Arc protein expression during a seizure? It can be assumed that new methods of in vivo molecular imaging can help answer this question to some extent (76, 77).

Arc is an activity-dependent immediate early gene, its mRNA is translocated only to dendritic spines of active neurons where it is translated to Arc protein, which then multimerizes in capsids forming viral-like particles implanted with mRNA. Then in the form of viral-like particles, Arc-mRNA is loaded into extracellular vesicles and trafficked to postsynaptic boutons, participating in regulating dendritic spine morphology and the receptor content of glutamatergic synapses controlling synaptic plasticity changes and memory (20, 21, 31, 47, 78, 79). Notably Arc expression is augmented in synapses of recently activated neurons of the epileptic seizure focal zone but rapidly declines, probably due to dysregulation during aberrant neuronal overexcitation (80). For example, temporal lobe epilepsy originates from the mesolimbic network and provokes damage to the hippocampus including the dentate gyrus, which is very vulnerable to status epilepticus (81–83).

It is generally accepted that hippocampal sclerosis provokes the hippocampus to generate seizures resulting from the loss of interneurons and pyramids accompany the formation of recurrent synaptic circuits by dentate gyrus cells (DGC) (84, 85). Arc upregulation in DGC between seizures accompanies increased spine density which enhances excitatory input from the entorhinal cortex, which precedes the formation of recurrent mossy fiber synapses (80, 86). Conversely, hippocampal regions not affected by seizures preserve normal Arc expression (80).

Many authors studied the association between Arc expression and epileptogenesis. It is known that spine loss and other dendritic abnormalities occur in epilepsy which to some extent may be also associated with Arc (87). Of course, the Arc gene affects a whole galaxy of genes, and therefore the phenotypic consequences of its mutations are very diverse (21, 37, 88). Also, some genes and their products are necessary for Arc functioning and transport, and the disruption of these genes lead also to changes in neural net excitation (74).

As part of the analysis of the role of Arc in epileptogenesis on clinical material, we analyzed the transcriptomic data from the National Center of Biotechnology Information (NCBI) public repository Gene Expression Omnibus (GEO) to evaluate the ARC expression in brain areas that are damaged in epilepsy (89). The RNA sequencing data were searched in the GEO Browser for the terms “epilepsy”, “epileptic”, and “seizure”. Afterward, we searched the GEO database for the terms “kainate”, “electroconvulsive”, and “pentylenetetrazol” to receive the data for the most common models of epilepsy. Each GEO dataset included in the analysis should meet the following criteria: (1) expression data in raw counts, fragments per kilobase per million mapped fragments (FPKM), or transcripts per million (TPM); (2) a clear explanation of sample origin (the datasets which comprise samples described like a “brain sample,” when the brain structure was not specified, were excluded); (3) at least three samples per study group; (4) the expression patterns were studied in brain structures, data for other tissues were excluded.

Four relevant datasets for the ARC expression pattern in human brain parts were mined (Table 1). In all datasets, only patients with temporal lobe epilepsy were included, and hippocampal structures were studied, except 17 neocortical temporal lobe samples in GSE134697. Further, no healthy or non-epileptic controls were included in the studies in the appropriate number (n ≥ 3) that did not allow for comparison of damaged tissues with normal tissues from healthy subjects.

No differences in ARC expression levels in mesial temporal lobe epilepsy patients with or without concomitant hippocampal sclerosis were identified in GSE71058 by an edgeR likelihood ratio test (Padj > 0.05) (93). Likewise, there were no significant differences in ARC expression levels between patients with low and high seizure frequencies in GSE127871. All data were count per million (CPM) normalized by edgeR package CPM function and demonstrated the congruent ARC expression levels (93). Applying the Expression Atlas recommendations, the expression levels may be interpreted as low expression (i.e. CPM between 0.5 and 10) or medium (CPM between 10 and 1,000) in all studied structures (94).

In mice, transcriptome modifications were studied in the hippocampus, neocortical structures, and cerebellum in genetic models of epilepsy (Table 2). No differences in Arc expression levels between control and study groups were identified in any study when likelihood ratio test (for raw counts) or empirical Bayes statistics eBayes function in limma package (for log 2 FPKM-normalized data) were applied (99). The only exception is the dramatic loss of Arc in the hippocampal transcriptome of mice lacking miR-22 during the epileptogenesis after kainate injection compared to the identically treated wild-type animals (Padj < 0.05). miR-22 loss results in an exacerbated epilepsy phenotype in kainate-induced epilepsy and is associated with the reduction of an inflammatory response at the transcriptional level (96).

In contrast, the significant downregulation of Arc expression was identified in CA1 superficial layer in response to kainate treatment (Padj < 0.0019) in rats (Table 3). As well, Arc expression was downregulated in the dorsal subiculum (Padj < 0.0024) in response to the induction of acute seizures in rats by electric stimulation. No differences in Arc expression were identified between control and epileptic model rats when the whole hippocampus or corpora quadrigemina were studied. So, the demonstrated discrepancy between results revealed in mouse and rat models may be caused by expression examination in the whole hippocampus instead of the separated parts of this complex structure.

Genetic studies of Arc expression during epileptogenesis in humans and animals are still rare to provide non-contradictory conclusions on Arc's role in seizure-related pathologies. According to available datasets for the Arc expression, only in mice and rats, but not in humans, the downregulation of Arc expression was revealed in response to seizures. Further accumulation of more detailed data on Arc expression in specific regions of the brain is likely to allow more specific conclusions to be drawn in the future.

Practically all rodent epilepsy seizure induction models affect Arc expression. In rodent models, Arc gene expression can be stimulated by practically any method which induces seizure activity in the brain, at least in the hippocampus and cortex, but Arc expression is time-dependent and the stimulation later turns to depression (80, 103, 104).

Initial hyperproduction of Arc can be seen in epilepsy models with kainate, D1-receptor agonists, 4-Aminopyridine (4AP), Bicuculline (Bic), and Forskolin, pilocarpine, pentylenetetrazole, kindling, activation of mGluR, electroconvulsive stimulation (Also Arc mRNA and proteins are rapidly induced in the striatum after acute cocaine administration (80, 105–111). Details on epileptic models inducing Arc expression are presented in Table 4.

Arc expression demonstrates complex behavior in seizure generation. The threshold of stimulation intensity (depolarization threshold) required for induction of Arc expression varies between brain regions (23). Particularly in the case of most intensive electroconvulsive stimulation, the proportion of Arc-positive neurons following seizures was highest in the dentate gyrus, intermediate in the CA3 region of the hippocampus, and lowest in the perirhinal cortex (23). In contrast, low-intensity seizure-inducing electrostimulation caused an opposite Arc expression profile (lowest in the dentate gyrus and highest in the perirhinal cortex), which indicates for Arc expression may be serving as a transcriptional threshold mechanism in CNS (23). Accordingly, Arc mRNA upregulation is positively correlated to seizure burst ratio, burst amplitudes, and length of paroxysmal episodes (80).

Peebles et al. studied spine morphology and the general stability of the glutamatergic neuronal network and found both dependent on Arc expression, employing a mouse model (21). They have confirmed that Arc expression leads to an increase in spine density, but generally decreases synaptic efficacy by reducing surface GluR1. Authors have shown that regulated synaptic strength in neuronal networks determined by Arc is very important for network stability. Studying kainite-elicited seizure activity in WT and Arc–/– mice, authors have shown that Arc–/– mice are more susceptible to kainite-elicited seizures and neuronal changes associated with epilepsy. Also, Arc–/– mice had aberrant spontaneous cortical network discharge activity, highly associated with epilepsy (21). On the other side, prenatal or perinatal deletion of Arc/Arg3.1 alters cortical network activity without excessive disruption of the balance of excitation and inhibition in the brain (88). Furthermore, Arc knockouts elevate AMPA receptor level expression in some brain regions including the nucleus accumbens which reduces the symptoms of epilepsy-associated pathologies (109). Pilocarpine-induced temporal lobe epilepsy causes a time-dependent decrease in Arc expression in hippocampal tissue (121).

Epilepsy is prevalent and often medically intractable in Angelman syndrome (AS). There are different models of AS. AS mouse models associated with UBE3A gene-deficient function (UBE3Am-/p+) shows reduced excitatory neurotransmission but a lower seizure threshold. Genetically decreased Arc expression additionally reduces abnormal EEGs and seizures in mice with Angelman syndrome associated with UBE3A gene-deficient function, because both Arc and UBE3A regulate surface expression of AMPA receptors. In another AS model, the so-called fragile X syndrome (FXS) mouse model, on the contrary, increased Arc is responsible for seizure phenotype (68, 114).

Comorbidity of schizophrenia and epilepsy are relatively common in clinical practice and animal models. Two different models of schizophrenia with seizures were developed, conditional KO (late-cKO) mice, in which Arc/Arg3.1 was deleted during late postnatal development, to investigate the causal relationship between Arc/Arg3.1 deletion and schizophrenia-linked neurophysiological and behavioral phenotypes. Nevertheless, in an animal study genetic deletion of Arc/Arg3.1 per se did not cause schizophrenia-like behavior, and a significantly higher dosage of kainic acid was required to elicit epileptic seizures in the KO mice (88). This completely contradicts the results obtained in another model of Arc/Arg3.1 knockout mice by Managò et al. showing genetic mutation disrupting Arc produced a hyperactive phenotype and amphetamine supersensitivity consistent with rodents' correlates of schizophrenia-like symptoms (120). The authors also describe important differences in the dopaminergic system in KO and wild-type mice. Most likely, this inconsistency can be explained by the different strategies used to create the two knockout mouse models.