Genomic evidence for the suitability of Göttingen Minipigs with a rare seizure phenotype as a model for human epilepsy

Epilepsy is one of the most common neurological diseases with a prevalence of almost 2% in the general human population [1]. It is characterised by two or more unprovoked seizures defined by unexpected and uncontrolled brain electrical activity. Epilepsy can result in an alteration of behaviour, movement, and sensation and typically has a significant impact on a person’s quality of life due to subsequent physical injury to the brain, emotional distress, and social isolation. While there are some treatments available, many people with epilepsy continue to experience seizures despite medical intervention, making it a serious and ongoing health challenge. Epilepsy can occur following structural abnormalities of the central nervous system, metabolic disturbances, and genetic mutations [2]. Genetics plays a prominent role in the occurrence of epilepsy, but genetic causes are still poorly defined.

Several epilepsy-associated genes have been identified, including those that encode ion channels and neurotransmitters [3]. Ion channels are important for the establishment of neuronal action potentials and in maintaining ionic homeostasis by gating ionic flux through the plasma membrane, and in cell volume regulation, which is important for the excitability of neurons [4]. Due to their function in the regulation of both neuronal excitation and inhibition, ion channels have a potential key role in epileptogenesis [5]. In particular, low-voltage activated (T-type) Ca²⁺ channels [6], which play a key role in amplifying the excitatory potentials of postsynaptic neurons [7], are of particular interest. T-type Ca²⁺ channels include three subunits of a CaV3-type pore-forming channel (CaV3.1, CaV3.2, and CaV3.3) and are encoded by the CACNA1 gene family members (CACNA1I, CACNA1G, and CACNA1H, respectively). Mutations affecting the expression of CaV3.2 channels in the hippocampus and thalamus are related to the occurrence of idiopathic generalised epilepsy [8]. Although ion channels play an important role in several types of epilepsy, other pathways and transcription factors lead to neuronal hyperexcitability [2]. However, little is known about the function of transcription factors and related pathways in the control of the CaV3.2 gene expression.

Epileptic rodent models (induced or genetic) have been used to better understand epilepsy in humans [9]. However, the value of rodent models is limited by their phenotype, particularly in models where seizures are absent. Hence, key aspects of the disease may not be faithfully recapitulated [9]. It is important to develop other animal models that more closely resemble human neurophysiology, to gain a more comprehensive understanding of epilepsy [10]. Recently, GMPs have been used extensively due to high anatomical and physiological resemblances to humans in comparison to other non-rodent species. GMPs may help bridge the gap between initial findings in rodents before human clinical trials [11‐13]. The incidence of epilepsy in a small subset of GMPs with a well-described genetic background provides a promising platform for the establishment of an animal model for characterising epilepsy and better understanding the mechanisms behind the disease in humans. Thus, this study aimed to identify genomic and transcriptomic correlates to human epilepsy in epileptic GMPs.

We identified numerous known epilepsy and seizure genes as well as novel candidate genes by comparing whole genomes of healthy with seizure-phenotype GMPs. Furthermore, we identified genes that were differentially expressed (DE) in primary skin fibroblasts isolated from epileptic and healthy GMPs and correlated these genes to mutations that were close to fixation in the genome of epileptic GMPs. Using this approach, we discovered numerous epilepsy candidate genes, e.g. CACNA1H. Furthermore, we identified the transcription factors EGR3 and HOXB6 as likely candidate regulators of CACNA1H expression. We provide the first insight into the underlying basis of the epileptic phenotype of GMPs and identified new candidate genes for validation in human epilepsy. This study is the first step toward establishing GMPs as a model system for human epilepsy.

To identify causative genetic variants for epileptic seizures in GMPs, WGS of pooled animals was performed. Variant calling in the genomic sequence data of seizure and control pools resulted in the discovery of 15.8 M SNP variants with variability within GMPs. Of these variants, 1276 showed outstanding differentiation with an \({F}_{ST}\) ≥ 0.95 between healthy and epileptic GMPs (Supplementary Information S2). Variant effect prediction revealed that 628 genes were located in close proximity (± 5 kbp) to those genes (Fig. 1A, Supplementary Information S3). Gene set enrichment analysis (GSEA) results are shown in Table 1 (complete results in Supplementary Information S4).

In the category TISSUES, the top four terms were “Central nervous system”, “Nervous system”, “Brain”, and “Head”, which verify that the discovered variants mainly affect tissues that are relevant for the seizure phenotype under investigation. Regarding Gene Ontology Processes and Functions terms related to the synapse, neurotransmitters and calcium channels were found, which also fit the phenotype. Gene clustering results for KEGG pathways are shown in Fig. 1B. The KEGG term with the highest significance detected was “calcium signalling pathway”, which included the genes ADORA2B, CACNA1E, CAMK1D, CAMK1G, CAMK2A, CHRM2, GRIN2B, ITPKB, LOC100154782, MCOLN2, MYLK, NFATC3, PDGFD, PHKB, PPP3CC, RYR2, and VDAC1. Since epilepsy is a disease associated with ion channel dysfunction, these genes have a high probability to being causative for the seizure phenotype under investigation in this study.

To validate the candidate mutations from the genomic analyses, we compared the primary skin fibroblast transcriptomes from three healthy minipigs against six animals that had developed a spontaneous seizure phenotype. Fibroblasts have a high predictive diagnostic power over other minimally invasive tissue samples [29]. For this, RNA was isolated from primary cultured fibroblasts established from each animal. The average overall alignment rate was 97.9% (SD = 0.4%), which resulted in an average exon coverage of 114 × (SD = 22) and a transcriptome coverage of 45 × (SD = 8.6). Differential expression analysis led to the discovery of 56 differentially expressed genes (DEGs) (abs. Log₂ fold change (Log₂FC) > 1, p-adj. < 0.001, Fig. 2A). The complete results of the differential expression analysis are summarised in Supplementary Information S5. Of the highly differentiated variants, 27 were in proximity (± 200 kbp) to the 56 DEGs (Table 2, Supplementary Information S6). As such, DEGs with extended boundaries cover approximately 26.2 Mbp, which is about 1% of the assembled chromosome length, only 14 highly differentiated SNPs are expected under the assumption of random distribution, leading to a 1.9-fold enrichment of highly differentiated SNPs around DEGs.

Gene cluster enrichment analysis was performed with DEGs (abs. Log₂FC > 1, p-value < 0.01) and we discovered the top KEGG terms affected in the comparison between healthy and seizure pigs were “MAPK signalling pathway”, “Neuroactive ligand-receptor interaction”, and “Ras signalling pathway” (Fig. 2B). Enriched GO molecular functions were largely related to receptor and channel activities (Fig. 2C). The complete dataset of gene clustering results is summarised in Supplementary Information S4.

The dGCR web tool [24] was utilised to discover genes, which are co-regulated with the DEGs in proximity to a fixed mutation. The output of dGCR was compared with our list of DEGs (abs. Log₂FC > 1, p-value < 0.01) and this resulted in the discovery of one co-regulated gene network with more than 3 nodes (Fig. 3A). This network included the two voltage-gated ion channels CACNA1H and CACNG4, the Ca²⁺ metabolism protein RASGRP1, and NGEF, which is involved in axon formation, and the signalling molecule VGF [30, 31]. Furthermore, the five transcription factors EGR3, HOXB5, HOXB6, HOXC10, and PRDM16 are part of the network. Transcription factor binding site enrichment analysis with Ciiider revealed that binding sites for the transcription factors EGR3 (Log₂ enrichment = 4.52, p-value = 0.004) and HOXB6 (Log₂ enrichment = 1.81, p-value = 0.012) are significantly enriched in proximity (1.5 kb upstream, 0.5 kb downstream) to the genes encoding for the proteins in that co-regulated network (Fig. 3B). CACNA1H is exceptional in this regard since it contains predicted binding sites for both transcription factors (Fig. 3C). Since CACNA1H is the most downregulated gene among DEGs (Log₂FC =  − 6.734), we conclude that the downregulated EGR3 (Log₂FC =  − 2.147) is an activator of CACNA1H expression and that the upregulated HOXB6 (Log₂FC = 1.728) is a repressor of CACNA1H. The complete transcription factor binding site enrichment results are summarised in Supplementary Information S7.

Epilepsy has a significant genetic component, with research suggesting that genetic factors may account for up to 80% of the risk of developing the condition. While specific genes and genetic variations associated with epilepsy have been identified, the inheritance pattern is often complex and multifactorial. Currently, available animal models for epilepsy do not recapitulate the diverse aspects of epilepsy [32]. The identification of the rare occurrence of GMPs with high recurrent spontaneous seizures offers a unique opportunity for the establishment of a pig model of epilepsy-type disease, both to better understand disease pathogenesis and in the development of new treatments. The advantages of a GMP model for epilepsy include the well-described genetic background and controlled breeding in an established breeding program. In addition, their high health status and microbiological definition decrease the risk of confounders and unwanted variables. The large and complex gyrencephalic brain of pigs has more similarity to the human brain than those of simpler model organisms and provides the potential for a better understanding of the genetic aspects of epilepsy [33]. Unpublished data based on pedigrees of affected animals suggest a moderate to high heritability and recessive mode of inheritance. Hence, the aim of this study was to elucidate if GMPs with a seizure phenotype have commonalities at the genome and transcriptome level with human epileptic disorders. We compared the genomes and transcriptomes of two experimental groups of healthy and seizure-phenotype GMPs to discover the seizure-associated genes. Variant calling from whole genome sequences led to the identification of numerous genes that were enriched in GO terms, which link them to the synapse, neurotransmitters, and calcium channel activity. Furthermore, 17 genes clustered in the KEGG “calcium signalling pathway” (Table 3). About half of those genes have been previously linked to epilepsy or seizures. This is a clear indicator that GMPs are suitable as a reliable and robust model system for human epilepsy. The remaining genes should be considered novel epilepsy candidate genes and they were mostly linked to other neurological disorders.

Among genes, which were not linked to neuropathology thus far, is the Ca²⁺-permeable cation channel MCOLN2 (synonym TRPML2). Constitutive activity of MCOLN2 induces cell degeneration [49] and might play a role in the irreversible brain damage and cognitive decline associated with frequent and recurrent epileptic seizures. LOC100154782 (5-hydroxytryptamine receptor 5B) is a 5-HT and serotonin receptor. 5-HT receptors are under discussion to be novel epilepsy treatment targets. [50]. The gene Htr5b is expressed in mice and rats, but not in humans, where the open reading frame is disrupted by stop codons [51]. Hence, it might play a role in seizures in animals but not in humans.

The remaining genes from Table 3 are involved in other neurological conditions and are therefore interesting epilepsy candidate genes. In particular the transcription factor NFATC3, since it was linked to excitotoxic and traumatic brain insults [43]. Excitotoxic damage is a hallmark in epilepsy and has been attributed to glutamatergic mechanisms [52]. Nuclear NFATC3 accumulation in pericytes depends on neuronal activity and the activation of group I metabotropic glutamate receptor [53]. In summary, this evidence makes NFATC3 a promising epilepsy candidate gene. The ITPKB gene protects against α-synuclein aggregation in Parkinson’s disease [41] and high cytoplasmic expression of α-synuclein was detected in brains of rats with induced experimental epilepsy [54]. A connection between the two observations should be further investigated.

To investigate the functional impact of the detected variants on gene expression, transcriptome analyses were performed. Since only primary fibroblasts were available as a source for RNA at the time, only DEGs with a highly differentiated genetic variant in proximity were considered relevant. Among the highest DEGs with fixed genetics variants in proximity, we found CACNA1H and SPECC1, which have previously been linked to epilepsy-type disorders in humans [8, 55, 56]. A wide range of epileptic disorders in humans have been related to dysregulation of the electrophysiological properties of ion channels and neurotransmitter systems, known as idiopathic generalised epilepsy (IGE). Ion channels are pore-forming proteins that allow ions to pass through the cell membrane [57, 58]. Voltage-gated Ca²⁺ channels lead to the conduction of inward Ca²⁺ currents after depolarization, which mediates the firing of an action potential [3]. These channels can contribute to the development of seizures by mediating the conduction of a Ca²⁺ current after depolarization. Thus, targeting these channels is a potential strategy for new treatments. Understanding the specific role of voltage-gated Ca²⁺ channels in epilepsy is important for the development of targeted therapies.

The role of voltage-gated channels in the activity of neuronal excitability indicates that mutations in Ca²⁺ channel genes CACNA1A, CACNA1H, CACNA2D2, and CACNB4 might be associated with the occurrence of epilepsy [55]. CACNA1H encodes the α1 subunit of Ca_V3.2, which is a Ca_V3 subfamily member. Ca_V3 channels are mainly expressed in thalamic neurons and have a main role in the conduction of low-voltage activated T-type Ca²⁺ currents [8]. Nelson et al. showed that CACNA1H mutations disrupt the function of Ca²⁺ channels and may lead to an increase in synchronously firing neurons [59]. An increase in the synchronous firing of neurons leads to neural hyperexcitability and the occurrence of seizures. It should be pointed out that the unusual activity of neurons is referred normally to its onset origin, but during the occurrence of seizures, it could spread to other parts of the brain, leading to extensive abnormalities in the functionality of the brain [59]. Our findings are consistent with previous studies and strongly indicate the role of CACNA1H as a potential gene in IGE. Khosravani and Zamponi demonstrated that CACNA1H mutations and their dysfunctional regulations act interactively with many factors like other cation channels and transcription factors, which lead to many abnormal activities in the epileptic brain [8]. Another DEG identified in our study is SPECC1. Two probands with idiopathic generalised epilepsy carried deletions in SPECC1. Therefore, we suggest that SPECC1 could be a potential novel candidate gene for epilepsy [56]. By conducting gene cluster enrichment analysis with DEGs, we identified two key pathways: mitogen-activated protein kinases (MAPK) and neuroactive ligand-receptor interaction pathway. Our data indicate that the most significant DEGs were particularly enriched in MAPK signalling pathways, such as CACNA1H and TNF, which are involved in generalised epilepsy [60]. Notably, the mTOR/MAPK signalling pathway regulates RNA-binding proteins (RBPs), and then influences the mRNA expression encoding target markers of epilepsy. Thereby, misregulated mTOR/MAPK-RBPs interaction may lead to the excessive synthesis of ion channels and their receptors, resulting in hyperexcitability of the cells [58]. Accordingly, if the stimulation of synapses is continuous, this leads to further activation of the mTOR/MAPK pathway through RBPs expression and provides a basis for the occurrence of epilepsy [61]. In general, one important aspect of epileptogenesis in all epileptic animal models studied so far is dysregulation of synaptic function, which includes the dysfunction of ion channel expression, presynaptic and postsynaptic neurotransmitters, and their receptor expression. Although in this study, we have not investigated RBP expression, previous studies have demonstrated that ion channels and their receptors might be controlled by RBPs [62]. For instance, Ferron et al. indicated that FMRP as an RNA-binding protein regulates the localization of voltage-gated Ca²⁺ channels to the presynaptic nerve terminal. Therefore, it might be expected that any inopportune translation of synaptically synthesised proteins involved in excitability could lead to epilepsy [31].

We also identified the KEGG pathway neuroactive ligand-receptor interaction, consisting of DEGs encoding neuroreceptors, such as GRIA4, which is significantly associated with a molecular mechanism involved in temporal epilepsy [63]. Indeed, a mutation in the GRIA4 gene, which encodes an AMPA receptor (AMPA-R) subunit expressed in cortical as well as thalamic neurons, leads to the absence of seizures [64]. AMPA-R are ionotropic glutamate receptor subtypes that are coupled to ion channels and regulate fast cell excitability by controlling the Na⁺ and Ca²⁺ ion influx throughout the CNS [64, 65]. Another important finding in our study is the identification of the Ras signalling pathway as a possible component of epilepsy. The Ras pathway plays different roles in neuroplasticity. Mignot et al. indicated that the over activation of Ras is often related to neurological disease [66], which is consistent with the current study. Our finding showed Ras signalling in epilepsy includes the DEGs RASGRP3 and SYNGAP1. SYNGAP1 encodes the RAS-GTPase-activating protein. It is interesting to note that this gene has a critical function in the density regulation of NMDA and AMPA receptors at the glutamatergic synapse and mediates the activation of glutamate receptors [67]. Some studies have demonstrated that mutation in SYNGAP1 can be also associated with generalised epilepsy in humans [66, 68].

Most of the GO molecular function enrichment analyses in the current study indicated epilepsy-related genes in cation channel activity and signalling ligand receptors. Among the possible target candidates, we found the DEG KCNAB1 which belongs to the voltage gate channel group associated with epilepsy [69, 70]. Previous studies demonstrated the involvement of potassium channels in some familial epileptic disorders [71]. Moreover, potassium voltage gate channels are mainly located in the presynaptic area and include two pore-forming α-subunits and one β-subunit (Kvβ1.1), which occludes the channel pore after depolarization and acts as a regulator in channel deactivation. Our findings are consistent with other studies where intronic variants in KCNAB1, which encodes Kvβ1.1, were associated with lateral temporal epilepsy [69, 72].

We discovered one co-regulated gene network (Fig. 3A) which includes the Ca²⁺ voltage-gated ion channels CACNA1H and CACNG4 and some transcription factors: HOXB5, HOXB6, HOXC10, and EGR3. Among the discovered transcription factors, binding sites for EGR3 and HOXB6 are significantly enriched near CACNA1H (Fig. 3C). Our results suggest that the downregulated EGR3 acts as an activator of CACNA1H expression. EGR3 encodes a transcriptional regulatory factor belonging to the EGR family [73], which participates in the regulation of dendritic morphology in sympathetic neurons and branch formation of terminal axons. This process is crucial for the normal development of the sympathetic nervous system [30]. Procedures that lead to neuronal hyperexcitability, such as seizures, lead to EGR gene overexpression [74]. EGR3 expression is elevated in the hippocampus in human and animal models with temporal lobe epilepsy [75]. It has been previously reported that EGR3 upregulates GABRA4 promoter activity and endogenous expression level of GABA_A receptor α4 subunits in response to a seizure-induced status [75]. Transcription factor binding site enrichment analysis revealed EGR3 binding sites upstream of CACNA1H. EGR1 binding sites are significantly overrepresented in the Ca_V3.2 promoter region and regulate transcription of the CACNA1H gene, which encodes T-type calcium channel Ca_V3.2, contributing to the development of neuronal hyperexcitability in epileptogenesis [6]. Although we have found significant enrichment of transcription factor binding sites for EGR3, sequence-specific DNA binding, transcription activities, protein structure aspects, and EGR3 regulation have many similarities to EGR1 [76]. However, comprehending the key mechanisms of controlling the Ca_V3.2 gene expression is still limited. Therefore, further experiments, such as expression of transcription factors in a specific region of the brain such as the cortex, will help to elucidate the roles of particular genes involved in epilepsy. Furthermore, we assume that upregulated HOXB6 acts as a repressor of CACNA1H expression. Transcriptomic studies demonstrate that HOX family transcription factors are mainly needed for synapse formation and also neuronal network maturation in the central nervous system (CNS) [77, 78]. To our knowledge, HOX genes were not linked to epilepsy yet.

The genomic and transcriptomic analyses of seizure-phenotype GMPs in comparison to healthy GMPs validated numerous known epilepsy and seizure genes and revealed novel candidate genes. The study presented here is the first step toward a robust and reliable non-rodent model system for human epilepsy and warrants further experimental characterisation. Transcriptomic analyses of cultured fibroblasts in three healthy and six epileptic seizure minipigs have revealed a potential role for differential expression of CACNA1H in epilepsy. Previous studies have shown that mutations in CACNA1H disrupt Ca²⁺ voltage-gated channels, leading to neural hyperexcitability and seizure occurrence. We also found the MAPK signalling pathway, neuroactive ligand-receptor interaction, and cation signalling pathways previously described in human epilepsy are dysregulated in GMPs experiencing spontaneous, recurrent seizures. Furthermore, our results pointed to the involvement of two transcription factors, EGR3 and HOXB6, as potential key regulators of CACNA1H and CACNAG4 expression, as well as the involvement of HOX genes, which may point to an early developmental disorder in GMPs.