Creating experimental models of epilepsy in animals (mainly in rodents) is the main way to study the pathophysiological mechanisms of seizure development and to search for new targets of new antiepileptic drugs (AEDs). There are many such models available today, but no single model can fully capture all the features of epilepsy and describe all the variety of symptoms observed in humans ^[1][127].

Traditionally, the most widely used models are those in which epileptic seizures in mice or rats are induced either by repeated stimulation of various brain structures with an electric current or by administration of pharmacological drugs that decrease the seizure barrier. Such exposure has been called “kindling”. Kindling is a phenomenon in which, in response to repeated epileptogenic stimuli of subthreshold intensity, the convulsive threshold of the brain decreases, and spontaneous convulsions consistently develop. The kindling process suggests that it starts with a limited number of neural circuits and subsequently recruits additional circuits as the behavioral component of the seizure progresses to seizures ^[2][128]. The classical variant of electrical kindling is the repeated subliminal electrical stimulation of limbic structures of the brain, which results in the development of epileptic seizures. Pharmacological kindling involves repeated exposure to subthreshold doses of convulsants, such as pentylentetrazole (PTZ) or kainate, resulting in the development of spontaneous seizures. Kindling models allow the induction of spontaneously recurrent seizure development, as commonly seen in patients with temporal lobe epilepsy ^[3][129]. The historically more “old” chronic epilepsy model is cobalt-induced epilepsy ^[4][5][130,131]. The rocking of the epileptic system also occurs after a powerful single epileptogenic exposure, such as pilocarpine or kainate, inducing status epilepticus (post-status models). Post-status patterns are most consistent with refractory forms of epilepsy because they are accompanied by neuronal death, aberrant neurogenesis, development of encephalopathy, etc. ^[6][132].

The “gold standard” in the search for new potentially active (AEDs) are simple models of acute seizures: the maximal electroshock test (MES) and subcutaneous injection of PTZ in mice and rats. These models have been used for the testing and discovery of most potential AEDs ^[7][8][133,134]. The advantage of these epilepsy models is that they do not require special strains of laboratory animals. Their disadvantage is that they do not reflect such an important aspect as the contribution of genetic factors in epilepsy development.

Therefore, genetic models of seizures, i.e., animal strains with an innate predisposition to epilepsy, are widely used in laboratory practice. These models make it possible to study the contribution of individual genes, groups of genes, and related signaling cascades in epilepsy development, which is necessary to find new AEDs targets. Today, there are genetic strains derived from several animal species (baboons, chickens, rats, mice, hamsters, and felines) that are seizure-prone (both spontaneous and refractory convulsions). For example, the seizure response in Papio papio baboons and Fepi chicken strains could be induced by rhythmic flashes of light, and in cats and many rodent strains by a sudden loud sound (audiogenic epilepsy, AE) ^[9][26]. Photosensitive epilepsy is quite common in epileptic patients, whereas audiogenic seizures in humans are very rare; their closest analogs in humans are startle-induced seizures ^[9][26]. Nevertheless, it is the AE-prone rodents that are commonly used as epilepsy models in laboratory practice because their reaction to a seizure-provoking stimulus (sound) is easily reproduced, and they display the standard type of seizure fit. Standard and experimental AEDs reduce the intensity of seizures in AE-prone rodents, and thus these strains could be used for the search for new AEDs and analysis of their anticonvulsive mechanisms ^[9][10][26,135]. Therefore, rodent AE-prone strains can generally be considered to be sufficiently fit for these studies, not only for human reflex epilepsy.

Genetic abnormalities in rodent AE-prone strains and in human epilepsy may coincide, but in some cases, they demonstrate differences as well (see below). On the other hand, as mentioned earlier, even in epilepsy clinics, a huge variety of both symptomatic manifestations and their genetic causes could be found in different patients.

Historically, work on genetic models of epilepsy in rodents began in the early 1900s, when the outbred strain of albino Wistar rats (existing till now) was derived in the Wistar Institute (Philadelphia, PA, USA). A certain percentage of these animals developed epileptiform seizures in response to loud sounds ^[11][21]. Different sources indicate different percentages of Wistar rats sensitive to sound, from 15 to 50%; apparently, sensitivity varies considerably in different Wistar substrains maintained in isolation in different catteries for many years ^[12][13][121,136]. Subsequently, a similar phenomenon of AE-proneness was found in mice ^[14][15][137,138]. In general, one may conclude that many laboratory rodent strains are generally characterized by increased sensitivity to loud sounds, which rodents do not encounter in a natural habitat, and which is presumably the hypertrophied startle reaction in response to alarming stimuli ^[16][139]. Therefore, the selection of strains predisposed to AE in rodents is rather quick, as rodent CNS has presumably the “stereotyped pattern” of reaction to strong sound. Thus, mutation events, affecting different genes in such a cascade, enhance the reaction up to the pathological level. A wide set of rodent strains (including rats, mice, and hamsters) with genetically determined AE were obtained. Animals of these strains respond to loud sounds by displaying generalized seizures. In all AE-prone strains, seizures proceed according to a similar pattern. In response to 100–120 dB sound onset, the first phase of a seizure develops (it is the so-called “wild run stage”), during which the animals rush around the cage or sound chamber. In essence, it is the defense reaction as an attempt to avoid sound, although with an admixture of involuntary movements, indicating the start of convulsions proper (sometimes this stage is called the “clonic run phase”) ^[16][139]. The next stages of the seizure proceed: clonic and tonic convulsions, followed by a post-convulsive state (catalepsy, or prolonged excitation) ^[14][17][18][137,140,141].

The first strain of rats 100% predisposed to audiogenic epilepsy was obtained at Moscow State University by L.V. Krushinsky, L.N. Molodkina and D.A. Fless, which was derived in the late 1940s from the Wistar strain (Krushinsky-Molodkina, abbreviated KM strain) ^[19][142]. Independently the WAR (Wistar audiogenic rat) strain was developed at the University of São Paulo, Brazil in the 1990s ^[20][25]. In the late 1950s, two GEPR (genetically epilepsy-prone rat) strains, GEPR-3 and GEPR-9, were bred at the University of Arizona based on the Sprague Dawley outbred rat strain, maintained further as independent strains with different levels of AE intensity ^[21][143]. The strain of AE-prone rats (P77PMC) also was selected in China ^[22][23][144,145]. The AE-prone hamster strain GASH/Sal (genetic audiogenic seizure hamster, Salamanca) was obtained in the University of Salamanca ^[24][146]. There are also several AE-prone mouse strains that were obtained as a result of spontaneous mutations: Frings, DBA/2J, Black Swiss, 101/HY and BALB/c ^{[14][15][25][26]}[137,138,147,148]. It has now become a rather common practice to derive AE models by the knockout of genes, presumably associated with epilepsy—they are Lgi1 mice, Fmr1 strain, etc. ^[9][27][20,26]. Thus, there is the set of strains with a similar AE phenotype, but with different genetic backgrounds and different genetic causes of this pathology.

Analysis of gene expression profiles and localization of respective mutations as well as the comparisons with each other and with the original AE-non-prone strains could allow revealing both common genetic patterns characteristic of AE and individual strain characteristics, leading to the same result: the development of AE. Using AE-prone strains, it is also possible to compare genetic changes with already established genetic causes of human epilepsy in order to find common mutations and molecular mechanisms leading to seizure proneness. Thus, rodent AE-prone strains increase the scientists’ potential for searching for mutations leading to epilepsy development.

Studies of the biochemical peculiarities of AE-prone animals performed in many studies demonstrate numerous “deviations” from the AE-non-prone phenotype practically in all brain neurotransmitter systems tested: glutamatergic, GABAergic, monoaminergic, and purinergic ^[28][149]. Numerous studies show that, at the level of biochemical changes, the mechanisms of epileptic seizure in rodents with AE (as in other epilepsy models) and in human epilepsy, in general, are reduced to an imbalance between the glutamatergic and GABAergic systems. In AE-prone rodents, impaired mitochondrial functions, neuroinflammatory processes and increased levels of MAPK signaling cascade activity have also been demonstrated ^{[29][30][31][32]}[66,150,151,152].

The well-known similarity of audiogenic seizure fit is based on the similar pattern of brain excitation spreading from the cochlear nuclei up to the corpora quadrigemina (in the case of generalized clonic–tonic seizures) and up to the forebrain structures in the case of audiogenic kindling phenomena, developing after repetitive sound exposures. The AE seizure initiates as the abnormal acoustic impulsation arrives in the inferior colliculi (IC). The details of the IC structures’ involvement in seizure initiation have been described in detail with the GEPRs AE model ^[33][34][153,154], as well as in WARs and in mice ^{[35][36][37][38]}[155,156,157,158]. Bilateral lesions of the IC, lateral lemniscus and the connections between these structures blocked audiogenic seizures expression in rats and mice ^[36][37][39][156,157,159]. Further, the superior colliculi (SC) activated with the spread of abnormal excitation into brain stem nuclei and further into spinal projections ^[40][160]. Rybak and Morin (1995) showed ^[41][161] (using immunocytochemistry and in situ hybridization) a significant increase in GABA level and a larger number of GABAergic neurons in the central nucleus of the IC in GEPR-9 strain in comparison to Sprague Dawley AE-non-prone rats. In GEPRs IC, the number of small cells (<15 mcm) was also increased. The IC structure was also affected similarly in KM rats (derived independently from GEPRs). In general, the marked phenotypic similarity in brain structure involved during AE fit is characteristic of rat strains selected in Russia, USA, France and Brazil. The phenomenology of AE seizures as well as numerous data on so-called “priming” procedures in several mouse and rat genotypes were extensively presented in ^[11][21]. The role of the acoustic system of AE-prone mouse and rat strains was described in detail earlier ^[28][149]. Recently, the signs of acoustic system peculiarities were also demonstrated in GASH/Sal as well ^[42][162].

Rodents (rats, mice and Guinea pigs) could be made AE-prone with the injections of metaphit, the drug, which is the ligand of phencyclidine receptors ^[43][44][163,164]. This is another confirmation of the general AE-proneness pattern, and the prevalence of brain excitation circuits. There is experimental evidence suggesting that delta sleep inducing peptide (DSIP) reduced audiogenic seizures, induced by metaphit injections ^[45][165]. In general, the effects of this peptide on brain functions are numerous ^[46][47][166,167]. However, the most interesting for the problem of audiogenic seizures mechanism is the evidence of the DSIP influence on the brain antioxidant and glucocorticoid systems ^[48][49][168,169]. It is worth mentioning that the neuronal membrane lipid content in rats of KM strain changed significantly (both quantitatively and qualitatively, in striatum and brain-stem tissues) as the result of audiogenic seizure fit ^[50][170]. The data may suggest that the anticonvulsant effect of DSIP was realized via the regulation of membrane processes shifted by metaphit action.

AE in the mouse strain derived from the Fmr1 gene knockout was found as a model of human Martin–Bell syndrome (fragile X chromosome syndrome). In patients, this pathology develops as the result of CGG trinucleotide expansion in the X chromosome, which leads to hypermethylation of the promoter region of the FMR1 gene and a decrease in its expression ^[52][172]. The protein product of this gene (fragile X mental retardation protein, FMRP) is responsible for the transport of mRNAs along dendrites and the regulation of the local translation of mRNAs in synapses. The loss of function of this protein leads to the overexpression of glutamate receptors and impaired synaptic plasticity, which in turn results in intellectual disability and autism, macroorchidism, sensory hypersensitivity, and with up to 15% of male and 5% of female patients displaying seizures ^[53][173]. Mice with Fmr1 gene knockout show a phenotype with several similarities to humans: enlarged testes, hyperactivity and mild spatial learning impairment in the Morris water maze, as well as AE ^[27][54][20,174]. The manifestation of audiogenic seizures in fragile X mice is age-dependent. The peak of susceptibility to auditory stimuli in homozygous Fmr1-/- females was observed at 22 days of age, after which the intensity of seizures decreased ^[55][175].

In the Frings mice strain, AE is caused by one base deletion at nucleotide 7009 of the Vlgr1 gene (also known as GPR98, MASS1, Neurepin1 and ADGRV1), resulting from spontaneous mutation when the truncated protein variant expresses ^[56][57][58][22,23,24]. The deletion of exons 2 to 4 of the Vlgr1 gene also leads to the development of AE in mice ^[57][23]. The Vlgr1 gene encodes very large G-protein coupled receptor 1, a member of the adhesion-GPCR (G protein-coupled receptors) family, highly expressed in the embryonic central nervous system ^[59][178]. Mutations of the ADGRV1 (VLGR1) gene in humans are associated with the development of myoclonic epilepsy and Usher syndrome ^[58][24].

Another gene with Mendelian inheritance associated with multiple central nervous system pathologies including epilepsy in humans is WWOX (WW domain-containing oxidoreductase), previously described as a tumor suppressor ^[60][182]. Bi-allelic mutations in this gene cause a syndrome called WWOX-related epileptic encephalopathy (WOREE) with epilepsy, severe developmental delay, ataxia and premature mortality in childhood ^[61][62][183,184]. To date, the role of the WWOX gene in CNS development is not known in detail. A spontaneous Wwox mutation in inbred Wistar-Imamichi rats leads to dwarfism, postnatal lethality, male hypogonadism and a high incidence of epilepsy (lethal dwarfism with epilepsy, LDE rats) ^[63][185]. Homozygous Wwox-mutated rats display hippocampal region vacuolization and AE in 95% of animals. The tissue-specific deletion of the Wwox gene in mouse CNS led to a significant decrease in the transcript levels of genes involved in myelination, decreased axon myelination and reduced maturation of oligodendrocytes.

Predisposition to AE and spontaneous death after seizures was shown in serotonin 5-HT2c receptor mutant mice ^[64][188]. Audiogenic seizures in this strain started to develop at 60–75 days of age, and by day 120, 100% of animals tested were AE-prone. Data on the role of the serotoninergic system in epilepsy development are inconsistent and depend on the epilepsy form (in humans) or animal model ^{[65][66][67][68][69]}[189,190,191,192,193]. One more gene of interest in terms of inherited causes of epilepsy is GABRB2, which encodes the β2-subunit of the gamma-aminobutyric acid receptor—GABA_A. In humans, various mutant alleles of this gene determine the development of a wide range of epilepsy forms ^[70][71][72][194,195,196]. A Gabrb2-/- knockout mouse strain with a characteristic schizophrenia-like phenotype and AE was obtained ^[73][197]. Interestingly, this strain showed signs of neuroinflammation indicated by increased brain levels of the oxidative stress markers, malondialdehyde and proinflammatory cytokines. The dysfunction of the Lgi1 (leucine-rich repeat glioma-inactivated) protein, originally described as a suppressor of tumor growth and metastasis ^[74][198] is associated with epilepsy in humans. Unlike most proteins whose gene mutations are associated with the development of epilepsy, Lgi1 is a secreted protein, the ligand of ADAM22 and ADAM23 (disintegrin and metalloproteinase domain-containing proteins). Lgi1 binds to ADAM22 at the postsynaptic membrane and ADAM23 at the presynaptic membrane, mediating AMPA receptor and potassium channels activity, and thus participating in the regulation of synaptic activity ^[75][76][199,200]. The increased expression of Lgi1 can also inhibit ERK1/2-cascade activity ^[77][201]. Finally, Lgi1 regulates the activity of inwardly rectifying K⁺ (Kir) channels, which contain astrocyte Kir4.1 subunits (Kir4.1 channels) involved in the K⁺ homeostasis in the CNS ^[76][200]. In humans, at least 43 mutations in the LGI1 gene are known to be associated with autosomal dominant lateral temporal lobe epilepsy (ADLTE) ^[78][202]. The autoimmune reaction with the development of autoantibodies to Lgi1 causes a form of autoimmune epilepsy ^[79][80][42,43]. Finally, loss-of-function mutations in the KCNJ10 gene encoding Kir4.1 cause the epileptic disorders known as “EAST” (epilepsy, ataxia, sensorineural deafness, and tubulopathy) ^[76][200]. Mutations in the Lgi1 gene leading to epilepsy cause impaired secretion of the Lgi1 protein ^[76][81][200,203]. Rats carrying a missense mutation (L385R) in the Lgi1 gene were obtained. Homozygous Lgi1-mutant rats were predisposed to early-onset spontaneous epileptic seizures and died prematurely. Heterozygous Lgi1-mutant rats were more susceptible to audiogenic generalized tonic–clonic seizures than wild-type rats. Lgi1 knockout mice with a similar phenotype were also obtained ^[82][204]. The seizure threshold in Lgi1 heterozygotes decreased with age (13% sensitive animals at 21 days of age, and 52% at 28 days of age) ^[82][204].

3.3. Models of Pyridoxine-Dependent Epilepsy and Angelman Syndrome

Mouse knockout of three genes from the family, encoding the proline and acidic amino acid-rich basic leucine zipper (PAR bZip) transcription factor, develops spontaneous and audiogenic seizures. The expression of these genes in most tissues obeys circadian rhythmicity, but in brain tissues they are expressed at almost invariable levels. The scholars attribute the knockout effect of these transcription factors to the fact that one of their targets is the Pdxk gene encoding pyridoxal kinase, which converts vitamin B6 derivatives into pyridoxal phosphate, involved in amino acid and neurotransmitter metabolism. Human Angelman syndrome (AS) manifests as defects in intellectual and speech development, emotional retardation, general motor deficits, and epilepsy in 80% of cases. It is caused by a partial deletion of the 15q11.2-q13.3 region in the maternal chromosome, which leads to either loss or mutation of the UBE3A gene encoding the ubiquitin–protein ligase E3A (UBE3A) also known as E6AP ubiquitin–protein ligase (E6AP). This enzyme is involved in targeting proteins for degradation within the proteasome. Inheriting paternal allele mutation in the same locus leads to the development of Prader–Willi syndrome with more mild neurological manifestations. The molecular mechanisms of the pathophysiology and inheritance of this syndrome were previously described ^[83][217]. Mutations specifically affecting the UBE3A gene occur in 5 to 10% of AS. In more than 75% of cases, large-size deletions within the maternally inherited chromosome 15 appear, herewith the deletion includes not only the UBE3A gene, but surrounding sequences containing additional genes (namely GABRB3, GABRA5 and GABRG3, encoding different GABA_A receptor subunits) ^[84][85][218,219].

3.4. Models with Polygenic or Unknown Inheritance

The first example of a putatively polygenic type of inheritance of AE in rodents is DBA/2J mice. Three loci presumably related to AE in this strain were identified: audiogenic seizure prone 1 (Asp1), Asp2, and Asp3 on chromosomes 12, 4, and 7, respectively ^[9][86][26,222]. Subsequently, evidence was obtained that the Asp1 and Asp2 loci were mainly responsible for the manifestations of AE ^[86][222]. Variations in the coding sequence of the Kcnj10 gene (localized on chromosome 1) leading to amino acid substitutions in the Kir4.1 potassium channel in DBA/2J relative to the C57BL/6B seizure-insensitive mice were also shown ^[87][223]. As cited above, mutations in the KCNJ10 gene cause epileptic disorders in humans ^[76][200]. It is assumed that the functions of the genes localized in the Asp1 and Asp2 regions are related to the regulation of Ca²⁺-ATPase activity, which is important for synaptic function, in particular affecting the amount of calcium within the synapse, which, in turn, influences the neurotransmitter release from synaptic vesicles ^[9][88][26,224]. Two substrains were selected for AE based on the Sprague Dowely population: GEPR-3, displaying moderate seizure intensity, and GEPR-9 with severe AE seizures ^[9][26]. Sound-induced seizures appear in GEPR-9 at the age of 25–35 days, with a further increase in intensity and decrease in fit latency ^[89][227]. Genetic studies indicate a polygenic inheritance of AE-proneness in GEPRs, while genes responsible for AE are still unknown ^[21][90][143,228]. At the same time, GEPRs’ GABAergic system anomalies were described in detail ^[9][91][26,229]. The genetics of AE development in three rodent strains obtained by classical selection with a presumably polygenic inheritance of this trait, WAR, KM and GASH/Sal, were described in detail. In KM rats, the audiogenic phenotype appears at 1 month of age, and by the age of 3 months, 100% of animals are AE-prone ^[20][25]. In the GASH/Sal strain, the age of AE expression start was not indicated. In WAR, the stable AE phenotype is present at the age of 70–78 days ^[20][25]. The cDNA microarray and RNA-seq methods identified the genetic abnormalities in connection with AE-proneness in these strains, followed by mutations in exon sequences, determining the expression of a wide range of genes and differences in the activity of key signaling pathways, demonstrated by the comparison of AE-prone and control strains ^{[12][92][93][94][95]}[121,230,231,232,233]. In WAR and GASH/Sal, the transcriptome of the IC, the structure, determining the start of AE seizure, was analyzed for both groups—animals not exposed to sound (naïve) and for animals after a seizure ^[92][96][230,234]. In KM rats, the transcriptome of the IC and SC of naïve animals was analyzed ^[12][121]. It is the profile of the transcriptome in naïve animals that is of maximal interest, as it allows to evaluate the contribution of various genes and signaling pathways to the AE-proneness. On the other hand, changes in the genes expression levels after seizures may include those involved in the compensatory mechanisms of seizure consequences ^[12][121]. However, the identification of specific genes with mutations and/or changes in expression levels that lead to seizure development in epilepsy models with polygenic or unknown nature of inheritance using whole-genome analysis methods faces a fundamental issue which complicates the whole problem. The thing is that most “classical” AE-prone strains (KM, GEPRs, WAR, and GASH/Sal) were bred by selection several decades ago, and after this, they were maintained in isolation from the original populations for a long time. This isolated breeding makes the direct comparison between AE-prone and AE-non-prone strains not very informative, as many genetic events could occur (both in initial and selected strains). The majority of such events should be neutral, i.e., not associated with the trait investigated—AE-proneness. Thus, 71 differential-expressing genes (DEGs) for the WAR model and 64 DEGs for GASH/Sal (in AE vs. non-AE comparison) were detected ^[4][130]. The search for genetic and biochemical peculiarities common to various AE-prone strains resulting from spontaneous mutations with further selection could be informative for revealing mechanisms of AE. Since, as mentioned above, some predisposition to AE is characteristic of rodents in general, it can assume the existence of some genetic pattern leading to this phenotype common to different strains and species. In this regard, it is of interest to compare different models for which RNA-seq data are available, i.e., WAR, KM, and GASH/Sal. Analysis of the IC transcriptome of the GASH/Sal hamster and WAR strains revealed several DEGs common to these strains in comparison to the original AE-non-prone strains—this list of genes includes Egr3, Rgs2, Ttr and Npy in both AE models ^[92][96][230,234]. The Egr3 gene overexpressed in WAR and GASH/Sal encodes a transcriptional regulator that belongs to the EGR (early growth response proteins) family of C2H2-type zinc-finger proteins and regulates NMDA receptor 1 and GABA_A receptor α4 subunit via the BDNF-PKC/MAPK signaling pathway ^[97][98][99][83,211,235]. Increased expression of GABA_A receptor α4 subunit and, conversely, decreased expression of α1 subunit of this receptor plays an important role in human temporal lobe epilepsy and in a mouse model of pilocarpine-induced seizures ^{[97][99][100]}[83,235,236]. Thus, increased expression levels of Egr3 may play a causative role in the development of AE in WAR and GASH/Sal as well. The AE-prone tremor mouse strain is also characterized by increased expression of the Egr3 gene ^[101][210]. On the other hand, it was unable to find data on the association of the Egr3 gene with human epilepsy. Mutations in the Ttr gene in humans cause family amyloid polyneuropathy but not epilepsy ^[102][103][237,238]. It is the only gene for which overexpression was found in WAR, KM and GASH/Sal, compared with the original AE-non-prone strains ^[12][121]. Transthyretin encoded by the Ttr gene regulates the activity of GABA_A receptors, which play an important role in the control of predisposition to seizures ^[104][239]. Deficiencies of GABA_A receptor-mediated neurotransmission were previously shown in GASH/Sal hamster and WAR rat models. In KM rats, the imbalance of glutamate-GABA content was noted at a neurochemical level as well ^{[19][29][92][94]}[66,142,230,232]. In addition to DEGs, sequencing of the transcriptome of GASH/Sal animals in comparison with the control AE-non-prone strain revealed several mutations in genes’ coding regions, and several of these genes were described as being associated with human epilepsy. In this list, mutations in the Cacna1a and Cacna2d3 genes encoding subunits of the calcium voltage-gated channel are of the most interest ^[94][232]. In humans, mutations in the CACNA1A induced developmental epileptic encephalopathies (DEEs) ^[105][249]. Grik1 (glutamate ionotropic receptor kainate type subunit 1) gene polymorphism was found in hamsters, while the human ortholog is associated with epilepsy development, including juvenile absence epilepsy (JAE) ^[94][106][232,250]. Cacna1a and Grick1, together with Grin2c, in which a substitution was also found in GASH/Sal, are parts of the glutamatergic synapse pathway, which is enhanced in GASH/Sal ^[94][232]. The Grin2c gene encodes the glutamate (NMDA) receptor subunit ε-3 ^[107][251]. Mutations in genes encoding glutamate receptor subunits ε-1 (GRIN2A) ^[108][252] and ε-2, i.e., genes related to previously mentioned (GRIN2B) ^[109][253], are also associated with human epilepsy. One more GASH/Sal gene Zeb2 (encoding Zinc finger E-box-binding homeobox 2) carries the substitution (in comparison to the control strain) and is possibly associated with seizure development. The encoded protein is the transcription factor, which participates in the transforming growth factor β (TGFβ) signaling pathway and is essential for early fetal development. Mutations in the ZEB2 gene in humans are associated with Mowat–Wilson syndrome, a complex disease with epilepsy and severe CNS developmental abnormalities ^{[94][110][111]}[232,254,255]. Mutations in several other genes found in GASH/Sal are probably neutral. WAR transcriptome analysis revealed mutations in Chrna4, Grin2a, Grin2b, Kcnq3, Vlgr1 and several other genes ^[112][181]. Chrna4 encodes a subunit of the nicotinic acetylcholine receptor, and its mutations are associated with nocturnal frontal lobe epilepsy. Kcnq3 encodes a subunit of the voltage-gated potassium channel with mutations associated with benign familial neonatal seizures ^[113][256]. Mutations in the Vlgr1 gene are associated with human epilepsy and with AE in the Frings mice (see above) ^[112][181]. One should note that no mutations in exon regions of any genes, associated with the development of epilepsy in humans, were found in the KM rat transcriptome analysis ^[12][121]. The comparison of signaling pathways, involved in synaptic regulation mechanisms, could reveal the common pattern of deviations that are specific to the ”epileptic brain” both in AE models and in humans. This pattern could be found despite the differences described in the mutation “list” for AE strains and for that associated with human epilepsy. From this point of view, the KM rat strain ^[12][121] was investigated in more detail. In the KM rats’ brain, increased activity of the MAPK signaling cascade was found which has direct pro-epileptogenic effects in the corpora quadrigemina, as well as proapoptotic and proinflammatory signaling pathways differences from Wistar ^[12][121]. MAPK/ERK1/2 activity presumably contributes to seizure threshold decrease via the upregulation of glutamatergic synaptic transmission and suppression of GABA_A receptors functions ^{[114][115][116][117][118][119][120]}[53,54,55,56,64,260,261]. 

4. The Use of Rodent Strains with AE in AEDs Screening

All AEDs currently used in clinics are characterized by a wide range of side effects, including rather severe ones, and they are not effective in all cases of epilepsy. For this reason, the search for new AEDs continues in order to find drugs with high efficacy and a smaller spectrum of side effects than those currently used. The development of new AEDs mainly includes studies of modified analogs of conventional drugs, possessing fewer side effects ^[121][264]. The “gold standard” models in AEDs testing are the maximal electroshock technique and the PTZ-induced seizure model ^{[122][123][124]}[266,267,268]. At the same time, the use of PTZ models proved to be sometimes ineffective, e.g., the absence of the anticonvulsant effect of levetiracetam using this type of seizure ^[125][269], whereas in both AE and other types of seizure induction, this drug was effective ^[126][127][270,271]. Numerous investigations proved that clinically approved AEDs reduce effectively the seizure intensity in AE-prone rodents. Numerous data, obtained in this field, confirm the applicability of AE-prone rodents for new AEDs generation testing ^{[10][128][129]}[135,272,273]. In general, there are probably too many AE-prone rodent strains available today for screening new anticonvulsants and for analyzing the mechanisms of their efficiency. Actually, several strains have rather explicit neurophysiological and biochemical characteristics to be used for this purpose. They are DBA/2 mice, WAR, GEPRs, KM rats and GASH/Sal. Appropriate genetic models described earlier are now used to develop specific treatments for some definite human pathologies, such as fragile X syndrome, Angelman syndrome, etc. ^{[130][131][132]}[177,274,275]. The DBA/2 mouse strain was used to reveal the anticonvulsant effects of clobazam ^[133][276]; benzodiazepine drugs ^[134][277]; adenosine (Ado) type 1 receptors (A1Rs) and their ligands ^[135][136][278,279]; effects of serotoninergic system enhancing substances ^[137][280]; and structurally diverse GABA_B positive allosteric modulators ^[138][281]. The DBA/2 strain was also used to study the synergism of carbenoxolone (the succinyl ester of glycyrrhetinic acid, an inhibitor of 11beta-hydroxy steroid dehydrogenase) and conventional antiepileptic drugs (carbamazepine, diazepam, felbamate, gabapentin, lamotrigine, phenytoin, phenobarbital and valproate) ^[139][282] as well as the pharmacodynamic potentiation of antiepileptic drugs’ effects by HMG-CoA reductase inhibitors ^[140][283]. In Frings mice, the effects of a new analog of topiramate ^[121][264], the anticonvulsive properties of soticlestat, a novel cholesterol 24-hydroxylase inhibitor ^[141][265], and the effects of synaptic and extrasynaptic GABA transporters inhibitors were investigated ^[142][284]. Fmr-/- mice were used to analyze the possibility of mGluRs and GABA receptors’ activity modulation in fragile X syndrome therapy ^[130][132][177,275]. The anticonvulsant properties of lovastatin were investigated in fragile X and Angelman syndrome models ^[131][274]. In GEPR-3 and GEPR-9 strains, the anticonvulsant effect of several ion-exchange transporter inhibitors ^[143][285] and the ability of liraglutide to reduce tolerance to diazepam ^[144][286] were studied. The KM strain was used in studies of the ability of vigabatrin (a GABA decay inhibitor) to reduce the clonic component of audiogenic seizures ^[145][287]. In general, one may conclude that along with standard models of seizures induced by maximal electroshock and PTZ, the AE-prone rodents could be widely used in the testing of old and new AEDs.

5. Conclusions

There is currently a rapidly expanding set of new epilepsy models with monogenic inheritance obtained by directed mutagenesis, due to the amazing success of genetic engineering. These achievements have promoted an understanding of the molecular mechanisms of epileptogenesis. In monogenic models, it is possible to establish an unambiguous causal relationship between a known mutation and the development of epileptic events. One may assume that large collections of knockout and transgenic mice (such as Jackson Laboratory), may possess several mutant strains that are predisposed to AE, but this trait has not yet been revealed. This assumption seems legitimate, given that in rodents the AE trait (in some cases) could be determined by mutations in genes for which no association with human epilepsy has been discovered ^[146][147][215,293]. An example of such incidental finding of gene knockout association with a predisposition to AE was mentioned previously ^[146][215]. The available set of rodent AE-prone strains with a monogenic type of inheritance demonstrates a significant diversity of genes involved in various regulatory processes. These are mutations of genes that control the expression of AE: transcription factors (Egr3, Tef), enzymes (Wwox), receptors (5-HT2c, Vlgr1, Gabrb2), anchor proteins (Gipc3), regulators of mRNA translation and transport (Fmr1), etc. Thus, the models of various forms of epilepsy are available for epilepsy syndromes with different etiologies and in which definite CNS mechanisms are involved, including action potential generation and early developmental events. In several AE models, the altered function of signaling cascades was also described (e.g., MAPK in KM rats) ^[12][121] as well as the development of neuroinflammation ^[73][197]. Thus, the AE models available provide an opportunity to test a wide range of potential AEDs targeting different proteins involved in various cellular processes, the disruption of which could induce epileptogenesis. Moreover, knockout mouse strains represent an obligatory model system for the development of epilepsy gene therapy methods based on CRISPR/Cas or adenoviral systems targeting, i.e., the development of the individual (associated with epilepsy) mutations “corrections” ^[147][293]. Rodent AE models with polygenic inheritance (KM, GEPRs, WAR and GASH/Sal) were developed as the result of numerous spontaneous mutations’ accumulation. Only part of these mutations is responsible for AE-proneness, while others are neutral or even exert compensatory effects for the AE phenotype (having been involuntarily selected for during selection generations). On the one hand, the identification of the genes responsible for audiogenic seizures in these models faces rather serious problems. It is known that several mouse knockouts of genes, whose human orthologs are associated with epilepsy, do not display neither AE nor spontaneous seizures in mice ^[147][293]. As an example, Tsc1 and Tsc2 (Tuberous sclerosis complex) knockouts do not lead to the development of seizures in mice, unlike tuberous sclerosis cases in humans ^[147][148][293,294]. On the other hand, the opposite is also true, i.e., several mutations were described that induce seizures in rodents, and definite disorders in humans ^{[97][98][101]}[83,210,211]. Pathological deviations that lead to epileptogenesis in AE-prone strains obtained by selection in independent experiments obviously do not coincide with one another because they could be determined by different mutations. Nevertheless, they could lead to similar results, demonstrating violations of glutamate and GABAergic systems, which provoke seizures. Polygenic AE animal models could be indispensable for studying human epilepsy with a complex type of inheritance. Most monogenic mouse models of epilepsy were obtained by introducing mutations into the coding regions of target genes (reading frame shifts or substitutions). Many epilepsy cases in humans also describe mutations in the coding regions of certain genes, leading to the loss of functions of the protein encoded. At the same time, a lot of human epilepsy cases were described with mutations affecting not the coding, but the regulatory region of genes ^[149][150][297,298]. In such cases, functional disturbances occur due to changes in expression levels relative to the normal ones. In this aspect, the KM rat strain is indicated (in comparison to other rodent models), in which no mutations were found in candidate genes for association with epilepsy, but a wide range of candidate genes could be indicated in which the expression is significantly increased or decreased relative to the conditional norm of Wistar rats ^[12][121]. Gene expression could be also affected by the methylation status of gene promoters, as was demonstrated for epilepsy-associated genes FMR1 and BRD2 (bromodomain-containing transcriptional activator 2) in humans ^[52][151][172,299]. The expression level of many genes could also be regulated by the RNA interference mechanism involving miRNAs (small non-coding RNAs that regulate post-transcriptional gene expression) ^[152][300]. It was found that in some cases, miRNAs as well as long non-coding RNAs can act as biomarkers and therapeutic targets in human epilepsy ^{[153][154][155][156][157][158]}[301,302,303,304,305,306]. Thus, in the case of some rodent models with polygenic inheritance obtained by the selection, the search for mutations in the regulatory regions of some genes, for epigenetic changes, and changes in the expression of interfering RNAs represent a mandatory field for future research.