A literature search was performed using three electronic databases, Scopus, PubMed, and Ovid MEDLINE, covering publications from database inception until 11th March 2024. The search terms applied were “Animal model*” AND “Cognit*” AND “Genetic epilep*”.

Only original studies written in English were included in this study. The inclusion criteria comprised all original studies on rodent models that examined the effect of GE on cognition. Non-original articles (i.e., systematic or literature reviews, book chapters, abstracts, etc.) and studies that did not meet the inclusion criteria were excluded during screening. Non-original studies were omitted due to insufficient datafor comparison and evaluation. Studies were included if they (i) employed rodent models with confirmed genetic or transgenic modifications linked to epilepsy phenotypes, (ii) assessed cognitive functions using validated behavioral paradigms (e.g., Morris Water Maze, Barnes Maze, Fear Conditioning, Novel Object Recognition), and (iii) reported quantitative outcomes. Studies involving solely pharmacological or environmental induction without genetic manipulation were excluded.

Two authors were involved in extracting the data. The studies obtained through database search were exported into Covidence, a tool used to streamline systematic reviews. Covidence enabled automatic removal of duplicated articles, and the remaining articles were initially reviewed through title and abstract screening. Following the elimination of studies that failed to meet the inclusion criteria, full-text articles were evaluated for eligibility (Figure 1).

The methodological quality of included studies was assessed using the Systematic Review Centre for Laboratory animal Experimentation (SYRCLE) Risk of Bias (RoB) tool. This tool consists of 10 domains covering selection, performance, detection, attrition, reporting, and other sources of bias. Each domain was rated as “Yes” (low risk of bias), “No” (high risk of bias), or “Unclear” (insufficient information). Two authors independently assessed each article, with discrepancies resolved through discussion.

A search of the three databases using the keywords mentioned in the methodology yielded 916 articles, followed by the removal of 124 duplicates. Titles and abstracts of the remaining 792 articles were screened, and 759 articles were excluded for not meeting the inclusion criteria. A total of 33 articles were included for full-text screening, of which 17 articles were omitted for the following reasons: (a) nine had wrong outcomes, (b) four involved non-rodent model, and (c) four had inappropriate study designs (Figure 1). Ultimately, 16 eligible articles were extracted and included for discussion in this systematic review, as compiled in Table 1.

The eligible studies involved rodent models, including both mice and rat models. Eight studies used mice, while another eight used rat, contributing to the total of sixteen eligible studies. Among the eight rat studies, four different strains were used. One study used the GEPR-3s strain (Aguilar et al., 2018), three studies used the GAERS strain (Marks et al., 2016a; Marks et al., 2016b; Marks et al., 2019), two employed the WAG/Rij strain (Jafarian et al., 2015; Russo et al., 2013), and two studies used genetically modified Sprague-Dawley (SD) rats with shRNA encoding downregulation of the Pcdh19 gene (Cwetsch et al., 2022) or the Scn1a gene (Bender et al., 2016). The control rats used for comparison were of the same background: transgenic studies used SD rats with or without control shRNA sequences (Bender et al., 2016; Cwetsch et al., 2022); GEPR-3s studies used SD control rats (Aguilar et al., 2018); WAG/Rij rat studies employed either Wistar rats or WAG/Rij rats injected with vehicle; and GAERS studies used the non-epileptic control (NEC) strain (Marks et al., 2016a; Marks et al., 2016b; Marks et al., 2019).

In the eight mouse studies, transgenic models were used, representing four different strains. Five studies used the C57BL/6J strain, targeting mutations in genes such as Scn1a, Scn2a, and the GABA_A receptor β3 Subunit (Dutton et al., 2017; Lena and Mantegazza, 2019; Nwosu et al., 2023; Phillips et al., 2014; Qu et al., 2020). Among these, one study employed both C57BL/6J strain and DBA/2J strains, encoding for SWD-related transcriptional downregulation of the hippocampal HCN1 channel gene, resulting in the R43Q mutation (Phillips et al., 2014). One study used the C57BL/6N strain with a Gabrb3 ^+/N110D gene knock-in (Qu et al., 2023). Two studies used mice with mixed backgrounds: one used 129P2/OlaHsd mixed C57BL/6J hybrids carrying an Scn1a mutation (Salgueiro-Pereira et al., 2019), while the other study used C3HeB/FeJ mixed C57BL/6J hybrids with a truncation mutation in Scn1a gene (Gheyara et al., 2014). Notably, four of the sixteen studies evaluated only male rodents (Lena and Mantegazza, 2019), while the remaining studies included both sexes to investigate the sex difference in cognitive function (Dutton et al., 2017; Gheyara et al., 2014; Nwosu et al., 2023; Phillips et al., 2014; Qu et al., 2020; Qu et al., 2023; Salgueiro-Pereira et al., 2019).

Cognition encompasses both basic and complex mental functions. Basic functions include attention, sensation, and perception, while the complex processes involve learning, memory, decision-making, and problem-solving. As the primary aim of this systematic review was to assess cognitive impairment in rodent models with GE, a variety of tests were used to assess memory functions in genetically mutated strains.

The cognitive assessments employed can be broadly divided into two categories; (1) tests of associative long-term memory in response to aversive stimuli, such as fear conditioning; (2) tests of learning and memory functions without aversive stimuli, including short-term recall, working memory, long-term recall, reference memory, adaptive memory, and recognition memory.

The Barnes Maze Test (BMT) was frequently used to assess spatial memory, spatial learning, recall memory and adaptive memory. Other tasks, including the Morris Water Maze (MWM), 8-arm radial maze, T maze rewarded alternation test, and Y-maze test evaluated similar parameters, with addition focus on long-term memory in rodents with GE. Social recognition memory was assessed using the three-chambered social interaction test, which determined whether mutant rodents spent more time with novel mice compared to familiar conspecifics. Other commonly used assessments included the novel object recognition test (NORT) and contextual fear conditioning (CFC), which evaluated recognition memory, long-term recall memory, and associative learning and memory, respectively. Associative memory for aversive stimuli was also measured with fear conditioning paradigms (low- and high-intensity) and the passive avoidance test. In some studies, recognition memory was further assessed using tactile, visual, and cross-modal object recognition (CMOR) test.

Because some GEs are associated with ASD-like phenotype, this review also evaluated the complex interplay between GE and ASD as a secondary outcome. All studies assessed behavioral and neuropsychiatric comorbidities relevant to ASD, such as anxiety, depression, intellectual disability, and hyperactivity. The most common tests for anxiety were open field test (OFT) and elevated plus maze (EPM) which also evaluated locomotion deficits. Other tasks used for similar purposes included the three-chamber test, locomotor activity chambers, looming threat test, and dark light test. Depression-related behaviors were assessed using tail suspension test (TST), sucrose consumption test (SCT), and forced swim test (FST). Repetitive behaviors, a hallmark of ASD, were examined using self-grooming and marble burying tests. Social interaction deficits were evaluated using ultrasonic vocalizations (USV), huddling test, reciprocal social interaction test, and the three-chambered social interaction test. Finally, sensory alterations were assessed using the hot plate test.

Overall, the quality assessment revealed that most studies demonstrated unclear risk of bias in all domains. While sequence generation and baseline characteristics were often described, reporting bias and attrition bias were rarely addressed. A summary of the SYRCLE RoB assessments for each study is presented in Table 2.

Two main hypotheses have been proposed to explain the cognitive dysfunction observed in genetic epilepsy models: one emphasizing seizure-related neuronal damage and the other implicating gene-driven alterations which are independent of seizure activity. According to the first hypothesis, cognitive impairment stems from persistent or recurrent seizures that lead to chronic neuronal damage and oxidative stress in brain regions responsible for memory and learning (Holmes, 2015). This was supported by studies investigating Scn1a mutant mice and in HCN1 (R43Q) mutants where prior induction of seizures before cognitive testing led to significant memory impairments (Dutton et al., 2017; Phillips et al., 2014; Salgueiro-Pereira et al., 2019). The second hypothesis was supported by a study where mice exposed to multiple SWDs presented with cognitive impairments which were not exhibited in their seizure-resistant R34Q counterpart (Phillips et al., 2014). The mutations in genes such as Scn1a and Scn2a directly disrupt cortical development, synaptic transmission and ion channel function, resulting in hyperexcitability in brain areas that compromise cognition independently of seizures. Consequently, prolonged neurological hyper-excitability will eventually lead to reduced neural function and adversely affect cognition over time (Staley, 2015). These two mechanisms are not mutually exclusive and may act in parallel, with genetic vulnerability shaping brain networks that are then further compromised by seizure activity.

Evidence from Scn1a R1648H mutant models further highlights the interplay between genetic vulnerability and seizure exposure in shaping cognitive outcomes. In these mutants, significant cognitive and behavioral impairments were observed only in rodents previously exposed to febrile seizures triggered by either hyperthermia or flurothyl, whereas mutants with the same genetic background, but without seizure exposure, did not exhibit comparable deficits (Salgueiro-Pereira et al., 2019). These findings suggest that the R1648H mutation alone is insufficient to cause cognitive and behavioral deficits but increases susceptibility to seizure-induced neuronal changes. These findings fortify the view specific genetic background may act as modifiers, amplifying the impact of seizure burden on cognition (Yi and Wei-Wei, 2023). However, it remains unclear whether this specific genetic mutation alone exacerbates the cognitive phenotype of genetic epilepsies or primarily acts through seizure-mediated mechanisms.

While several rodent models (e.g., Scn1a, Scn2a, and Gabrb3 mutants) represent specific genetic mutations associated with epilepsy, others such as GAERS or WAG/Rij rats display spontaneous spike–wave discharges typical of GGEs. Although full validation of human syndromes remains ongoing, these models offer crucial insights into how genetic or neurophysiological disturbances contribute to cognitive dysfunction. Pharmacological studies using these models, such as Russo et al. (2013), further inform how antiepileptic interventions may modulate cognitive outcomes. WAG/Rij rats consistently demonstrated age-dependent deficits in cognitive functions such as learning, long-term, or adaptive memory. In the MWM test, an age-dependent decline in learning and memory function is observed which is consistent with earlier findings of deficits in active and passive avoidance tasks (Leo et al., 2019; Sarkisova and van Luijtelaar, 2011). These impairments in the WAG/Rij strain may be explained by an age-dependent increase in SWD frequency and duration, consistent with previous studies and may not be related to the ontogenesis of GE (Jafarian et al., 2015; Karson et al., 2012; Leo et al., 2019; Sarkisova and van Luijtelaar, 2011). This is further supported by evidences showing that Wistar rats also demonstrated poorer memory functions in cognitive tasks at older ages which is concurrently associated with the development of SWDs (van Luijtelaar et al., 1995). The cerebral hypoperfusion experienced during SWD episodes may underlie these deficits, as older WAG/Rij rats show increased neuronal apoptotic cells, dark neurons, and caspase-3 activity in the affected brain areas (Jafarian et al., 2015). These findings suggest that in WAG/Rij rats, cognitive impairment is strongly associated with seizure-induced pathology. However, future research is needed to explore whether genetic abnormalities in the WAG/Rij strain might contribute to the cognitive deficits, as opposed to solely attributing them to seizure-induced damage.

Other studies support the hypothesis that genetic mutations themselves directly drive cognitive impairment, independent of seizure burden. For instance, both young (49-day-old) and adult (200-day-old) Gabrb3 ^+/D120N mutants exhibited poor spatial memory and learning across the BMT and MWM (Qu et al., 2020). The persistence of deficits across developmental stages suggests that the impairments arise from the underlying mutation rather than cumulative seizure effects. This pattern mirrors human syndromes such as DS, Lennox-Gastaut syndrome (LGS), and CAE, that generally manifest during the pediatric years of life and persists, regardless of seizure control (Anwar et al., 2019; Datta et al., 2023; Smith et al., 2018). Unfortunately, this poses a double whammy for those affected, as cognitive impairments are often more severe in individuals who develop epilepsy at a young age (Lenck-Santini and Scott, 2015).

Findings from Scn1a knockdown studies in the medial septum and diagonal band of Broca (MSDB) further illustrate how gene-specific effects can produce selective cognitive deficits. In a study by Bender et al. (2016), the knockdown of Nav1.1 in the MSDB resulted in a short-term working memory deficit in the mutant rats evidenced by the poor path efficiency in MWM and poor choice accuracy in the T-maze rewarded alternation test (Bender et al., 2016). Electrophysiological data revealed selective and targeted damage to the burst-firing GABAergic neurons, as well as loss of hippocampal theta oscillations, both of which are critical for spatial working memory (Bender et al., 2012). These findings align with prior studies using neurotoxins to selectively damage MSDB neurons which similarly resulted inworking memory but not reference memory deficits (Dwyer et al., 2007; Pang et al., 2011). Interestingly, Nav1.1 knockdown secondary to Scn1a mutation did not impair spatial reference memory, highlighting the task-specific nature of these effects (Leo et al., 2019). Together, these results suggest that functional GABAergic impairments due to SCN1A mutations disrupt hippocampal theta frequency, selective working memory impairments and not the reference memory.

Mutations in GABAA receptor subunits, which underlie syndromes such as Lennox-Gastaut syndrome (LGS) and infantile spasms syndrome (ISS), consistently produce severe cognitive impairments in rodent models. LGS models carrying Gabrb3 ^+/D120N and Gabrb3 ^+/N328D mutations as well as infantile spasms syndrome (ISS) models carrying Gabrb3 ^+/N110D mutations exhibited similar deficits in learning and memory, suggesting a cognitive deterioration phenotype typically seen in epileptic encephalopathies and progression of these genetic epilepsies (Nwosu et al., 2023; Qu et al., 2020; Qu et al., 2023). Clinically, cognitive impairment affects more than three-quarter of LGS patients, while 70%–90% of ISS patients develop intellectual disability (Zupanc, 2003). Notably, aberrant interictal epileptiform activity which interfere with cerebral function and normal brain developmental can still cause cognitive impairment even when the seizures are controlled (Aldenkamp and Arends, 2004; Stroink et al., 1998). This mechanism may explains the neurocognitive and behavioral deficits in Gabrb3 ^+/N110D mutants whose interictal brain waves differed significantly from wild-type controls, suggesting that abnormal background brain rhythms contribute to progressive cognitive regression in these models (Qu et al., 2023).

GABAergic neurons, distributed throughout multiple brain regions, play a critical role in maintaining inhibitory balance and cognitive stability. In Gabrb3 ^+/N328D mutated mice, reduced expression of Gabrb3 was observed in the thalamus, cerebellum, and hippocampus regions, postulating a decrease in the stability of the mutant β3 subunit protein (Nwosu et al., 2023). Given the hippocampus’s well-established role in memory formation and retrieval, deficits in GABAergic inhibition within the hippocampus may explain the impaired memory and learning ability observed in Gabrb3 ^+/N328D mutants. Notably, β3 subunits are highly expressed in the adult hippocampus, reinforcing their importance for normal cognitive processing (Hörtnagl et al., 2013).

Recognition memory allows animals to distinguish familiar from novel stimuli and has been linked to perirhinal cortex (Moreno-Castilla et al., 2018). In drug-naive GAERS, CMOR memory deficits were observed without aversive stimuli, minimizing the influence of anxiety on performance (Marks et al., 2016a). These impairments are likely linked to the dysfunction of perirhinal cortex, a region implicated in visual recognition memory (Winters and Reid, 2010). GAERS are reported to exhibit altered perirhinal cortex function, likely due to the mutation in CaV3.2 T-type calcium channels (Powell et al., 2009). This is notable given the high density of T-type calcium channel the perirhinal cortex (Talley et al., 2000). Hence, recognition memory impairments in GAERS may be mediated perirhinal cortex dysfunction linked to altered T-type calcium channel activity.

Not all studies have reported consistent recognition memory impairments, and several confounding factors may influence outcomes. In GEPR-3s rats, apparent memory consolidation deficits may be confounded by heightened anxiety-related behaviors (Aguilar et al., 2018). Elevated corticosterone levels in stressful settings can motivate stress-induced novelty avoidance behavior and impair NORT performance, without reflecting true memory deficits (Vargas-López et al., 2015). Consistently, elevated corticosterone levels have been reported in GEPR-3s rats, which may underlie their poor performance in recognition tasks (Aguilar et al., 2018). In contrast, young Scn2a mutant mice displayed age-dependent memory impairment which did not last till adulthood (Lena and Mantegazza, 2019). This developmental pattern may reflect the predominance of NaV1.2 voltage-gated sodium channels in the hippocampus during the initial 3 weeks of life, which is later taken over by Nav1.6 sodium channel subtype during maturation (Boiko et al., 2003). These findings underscore that recognition memory outcomes may be shaped by extrinsic confounds, such as stress and anxiety, and intrinsic developmental factors, such as ion channel maturation. Moreover, given the frequent comorbidity of ADHD with ASD, attentional dysfunction may also contribute to recognition impairments in young Scn2a mutants.

Associative memory outcomes varied across various GE rat strains. In WAG/Rij rats, deficit in associative memory is likely due to an age-dependent increase in seizure frequency and duration resulting in neuronal disruptions (Jafarian et al., 2015; Karson et al., 2012; Leo et al., 2019). In contrast, Pcdh19 mutant rats lacked significant seizure history but showed noticeable impairments in associative memory during the CFC test. Supporting this, it was reported that Pcdh19 heterozygous female mice demonstrated consistent reduced fear response in fear conditioning tests, while hemizygous male mutants showed no alterations in fear response (Hayashi et al., 2017; Hoshina et al., 2021). Pcdh19-transfected rats also demonstrated poor long-term memory functions in the NORT. Together, these findings suggest that the associative memory impairments in WAG/Rij rats are likely seizure-driven, whereas in Pcdh19 mutants, they may arise directly from the underlying genetic defect.

In contrast to WAG/Rij and Pcdh19 models, GAERS models exhibited enhanced associative learning, with stronger cued and contextual memory in response to fear (Marks et al., 2016b; Marks et al., 2019). Similar enhancements were reported in aversive learning tests such as the two-way active avoidance tests (Getova et al., 1997; Sitnikova, 2024). However, these improvements may be confounded by the delayed extinction of conditioned fear, raising the question of whether the results were due to dysfunctional inhibitory learning, augmented fear conditioning, or heightened anxiety behavior (Marks et al., 2016b; Marks et al., 2019). Thus, the disparity in associative cognition observed across strains highlights both test-dependent variability and the likelihood that different mechanisms such as seizure burden, genetic mutation, or altered emotionality, underlie associative memory outcomes in GE models.

These findings also highlight how genetic mutations in epilepsy can influence not only cognition but also broader behavioral domains. The next section considers how GE models recapitulate behavioral phenotypes, including autism spectrum disorder (ASD)-related traits, anxiety, and depression.

Mutant rodents frequently exhibited behavioral phenotypes resembling autism spectrum disorder (ASD), including repetitive behaviors, deficits in social interaction and sensory hypersensitivity. For example, Pcdh19 transfected rats demonstrated socio-behavioral impairments from early life that persisted into adulthood as well as sensory hypersensitivity which are consistent with ASD-like traits (Cwetsch et al., 2022). Heterozygous Pcdh19 knockout female mice also showed reduced sociability in the three-chamber test, resembling autism-like behavior (Lim et al., 2019). Interestingly, sex difference have been noted, with male rodents exhibiting greater pain hypersensitivity compared to females (Hayashi et al., 2017). Similarly, young Scn2a mutant mice displayed repetitive behaviors and impaired social interaction, likely reflecting dysfunctions in cortico- striatal circuits in the brain that regulate social and repetitive behaviors (Lena and Mantegazza, 2019). These findings suggest that genetic epilepsies involving Pcdh19 and Scn2a mutations may directly contribute to ASD-like phenotypes. Across multiple studies, anxiety-like behavior were consistently reported in rodent GE models. Two-month-old WAG/Rij rats without seizure history demonstrated increased anxiety behavior and stress reactivity compared to age-matched Wistar rats, suggesting a strain effect rather than seizure-induced pathology (Fedosova et al., 2015). GAERS model similarly exhibited heightened anxiety behavior in the EPM (Jones et al., 2008; Powell et al., 2014). However, results from the OFT appeared to be task-specific and age-dependent (Marks et al., 2016b). GEPR-3s consistently demonstrated anxiety-like behavior across all paradigms, which were attributed to serotonin abnormalities in the brainstem (Aguilar et al., 2018; Dailey et al., 1992). Scn1a mutants displayed modest increase in anxiety, consistent with Nav1.1 haploinsufficiency models that increased thigmotaxis in the OF test (Ito et al., 2013). Collectively, these findings indicate that genetic background and neurotransmitter abnormalities may modulate anxiety phenotypes across GE models.

The Scn2a model revealed a clear developmental contrast in anxiety phenotypes. Young Scn2a mutants demonstrated lower anxiety levels in EPM and OFT, as well as reduced immobility during the TST, in contrast to adult mutants (Lena and Mantegazza, 2019). Comparable behavioral findings have been reported in BTBR mice, where reduced anxiety-like behavior was also observed (Pobbe et al., 2011; Silverman et al., 2010). Researchers hypothesized that such behavioral findings were most likely explained by modifications of the hypothalamic-pituitary-adrenal (HPA) axis. Supporting this, reduced NaV1.2 expression in excitatory pathways of amygdala, hypothalamic, and limbic circuits have been shown to modify the HPA axis (Ogiwara et al., 2018). During early development, the amygdala primarily regulated emotional response towards aversive stimuli, but later phases of development the role is then taken over by the prefrontal cortex which exerts a top-down regulation and enhances emotional control (Gee et al., 2013). Consequently, reduced signaling from the amygdala may attenuate HPA activity during stress, explaining the lower anxiety phenotypes observed in young Scn2a mutants.

Evidence for depressive-like behaviors was also reported across GE models, though some findings may reflect confounding by anxiety. For instance, GEPR-3s exhibited reduced sucrose preference in the SCT, though this may reflect novelty-suppressed feeding due to heightened anxiety rather than depression (Aguilar et al., 2018; Bodnoff et al., 1988). Supporting this, fluoxetine, an anxiolytic agent, restored feeding behaviors in a corticosterone-induced anxiety model (Mendez-David et al., 2017). In a tail suspension test, adultScn2a mutant mice, showed increased immobility, suggestive of depressive-like symptoms (Lena and Mantegazza, 2019). Similarly, WAG/Rij rats at six and 12 months old also demonstrated increased immobility time in the FST (Leo et al., 2019). Collectively, these findings indicate that depressive-like behaviors may emerge in GE models, although careful interpretation is warranted due to overlap with anxiety-related outcomes.

Some findings diverged across models, highlighting the importance of genetic background. For example, Gabrb3 ^+/N328D mutants showed reduced locomotor activity, impaired social interaction and anxiety symptoms, whereas Gabrb3 ^+/D120N mutants did not, suggesting that motor deficits may secondarily influence behavioral outcomes (Nwosu et al., 2023). This observation highlights that even within the same syndrome, distinct mutations can result in divergent behavioral outcomes, reflecting the heterogeneity of GE models.

Several limitations should be acknowledged when interpreting the findings of this review. Genetic epilepsies are highly heterogeneous, both in their underlying pathophysiology and in the cognitive and behavioral phenotypes observed in animal models. As such, the outcomes summarized here are mutation- and syndrome-specific and should not be generalized across all forms of GE. In addition, some of the included studies incorporated treatment interventions; however, these were only considered when baseline cognitive or behavioral outcomes in untreated mutant animals were available, ensuring that the synthesis focused on the intrinsic phenotypes of GE models rather than treatment efficacy. We also recognize that our search strategy, which focused on the broad term “genetic epilepsy,” did not capture every gene- or syndrome-specific model described in the literature. Given the extensive and heterogeneous range of genes implicated in GE, it was not feasible to include all of them at this stage. Nonetheless, we believe that our approach was sufficient to provide a representative overview of cognitive impairments in rodent models of GE. Furthermore, the quality appraisal using the SYRCLE Risk of Bias tool highlighted that most included studies were rated as having an overall “unclear” risk of bias. Key methodological aspects such as sequence generation, allocation concealment, random housing, and blinding of investigators or outcome assessors were seldom reported. Even in more recent studies, including those assessing GAERS and pharmacological interventions, details on randomization and blinding were often absent or ambiguous. Therefore, while our review synthesizes available findings on cognition in genetic epilepsy models, the interpretability and generalizability of these results remain constrained by methodological shortcomings in the primary studies. Together, these considerations highlight the complexity of drawing broad translational conclusions and emphasize the importance of syndrome-specific and methodologically rigorous future studies.