Exploration of autophagy-related molecular mechanisms underlying epilepsy using multiple datasets

Abstract

Objective

To elucidate the molecular mechanisms underlying epilepsy, we investigated autophagy-related differentially expressed genes in epilepsy patients.

Methods

We analyzed GSE143272 and GSE4290 microarray datasets from the NCBI Gene Expression Omnibus database, which is established based on evaluations of peripheral blood samples. Using a bioinformatics approach, autophagy-related differentially expressed genes between epilepsy patients and healthy controls were identified. Further analyses including Least Absolute Shrinkage and Selection Operator regression, immune cell infiltration, and pathway enrichment were conducted. Experimental validation was performed using quantitative reverse transcription–polymerase chain reaction in a mouse epileptic model. Additionally, the Connectivity Map database was employed to predict potential drugs.

Results

In total, 49 autophagy-related differentially expressed genes were identified. A Least Absolute Shrinkage and Selection Operator logistic model revealed four autophagy-related differentially expressed genes, namely, CAPN2, ERN1, RELA, and SAR1A. Furthermore, a novel diagnostic model with robust validation metrics was established. Immune cell infiltration analysis underscored the significance of immune response in epilepsy, revealing distinct profiles in patients. Additionally, pathway enrichment analysis using gene set enrichment analysis and gene set variation analysis revealed that critical genes were implicated in diverse pathways, including metabolic and neurodegenerative diseases. The expression levels of these key genes were experimentally corroborated using quantitative reverse transcription–polymerase chain reaction in the hippocampus tissues of status epileptic mice. Finally, Connectivity Map analysis suggested three antiseizure drugs (cabergoline, capsazepine, and zolantidine).

Conclusions

Our results provide insights into potential biomarker candidates, thus contributing to clinical diagnosis and the development of new antiseizure drugs.

Introduction

Epilepsy, a common neurological disorder encompassing both genetic and acquired disorders, affects more than 46 million individuals worldwide. Given its intricate nature with multiple subtypes, accurate identification of biomarkers poses a persistent challenge. Individuals with epilepsy present with a diverse array of symptoms, ranging from peculiar sensations, emotional changes, and behavioral changes to convulsions, muscle spasms, and loss of consciousness, caused by aberrant brain signaling. ¹ The primary approach for the management of epilepsy entails the administration of antiepileptic drugs, and the use of this approach has increased over the years. Nevertheless, a considerable proportion (up to one-third) of epilepsy patients continue to struggle with drug-resistant epilepsy.^2,3

Temporal lobe epilepsy (TLE), emerging as the most common form of intractable epilepsy, accounts for 70% of all epilepsy cases. ⁴ TLE is frequently associated with hippocampal sclerosis, characterized by selective neuronal depletion, chronic neuroinflammation, proliferation of glial cells, and abnormal synaptic remodeling. ⁵ Epileptogenic alterations predominantly manifest during the initial stages following brain injury in animal models, ⁶ and effective preventive measures or treatments currently remain elusive.

A compelling nexus between autophagy dysfunction and epileptogenesis has emerged, substantiated by evidence showing that impaired autophagic processes exacerbate epileptic conditions. ⁷ Autophagy orchestrates cellular homeostasis through the biogenesis of double-membraned autophagosomes, which subsequently amalgamate with lysosomes to facilitate the catabolism of cytoplasmic constituents and microbial invaders. ⁸ Given the pivotal role of autophagy in modulating synaptic transmission and plasticity; excitotoxicity; neurodegeneration; astroglial apoptosis; and axonal, synaptic, and mitochondrial functionality, it is possible that dysregulated autophagy potentiates aberrant axonal plasticity and synaptic reconfiguration, culminating in the establishment of epileptogenic circuits. ⁹ Within this framework, autophagy presents itself as a promising therapeutic avenue for epilepsy management. ¹⁰

Currently, the utilization of high-throughput sequencing analysis of gene expression profiles, in conjunction with advanced bioinformatics tools, holds immense promise for investigating the underlying genes involved in disease initiation and progression. ¹¹ Although previous studies have suggested that autophagy is involved in epilepsy, it remains unclear whether some key autophagy-related genes (ARGs) play a role in epilepsy development. Thus, by using a bioinformatics approach, we identified differentially expressed genes (DEGs) in the GSE143272 microarray dataset containing data from epilepsy patients. This dataset was acquired from the Gene Expression Omnibus (GEO) database established based on the evaluation of human peripheral blood samples. Our findings may offer valuable insights into potential biomarker candidates, thus contributing to clinical diagnosis and the development of therapeutic strategies for epilepsy.

Materials and methods

Data acquisition

Gene expression profiles were sourced from the NCBI GEO database, specifically GSE143272 (comprising 50 controls and 91 cases) ¹² and GSE4290 (consisting of 4 controls and 23 cases) datasets. ¹³ The annotation files GPL10558 and GPL570 were employed for these datasets, respectively. The former dataset served as the training set, while the latter was considered the validation set. Subsequent analyses focused on the intersection of DEGs and ARGs, termed as “intergenes.”

Differential expression analysis

The R package ‘limma’ (version 3.50.3) was employed to identify genes with significant differential expression between controls and cases. ¹⁴ Genes with an adjusted P-value <0.05 were considered differentially expressed. Volcano plots and heatmaps were generated to visualize these DEGs.

Gene Ontology (GO) and Metascape functional analyses of differentially expressed ARGs (DE-ARGs)

DAVID and Metascape databases served as comprehensive repositories, playing pivotal roles in gene annotation, visualization, and pathway enrichment.^15,16 GO enrichment analysis encompassed three primary domains: biological process (BP), cellular component (CC), and molecular function (MF). Functional analyses of DE-ARGs, encompassing both GO and Kyoto Encyclopedia of Genes and Genomes, were performed using the DAVID database. Concurrently, the Metascape database was used to perform enrichment analyses via DisGeNET and PaGenBase. A threshold criterion was established at an adjusted P-value <0.05, prioritizing the top 20 terms.

Establishment and evaluation of a risk score model based on DE-ARGs

In the R software environment (version 4.1.3), to streamline model parameters and mitigate overfitting, we employed the ‘glmnet’ package (version 4.1.1) to implement Least Absolute Shrinkage and Selection Operator (LASSO) regression with 10-fold cross-validation. ¹⁷ The optimal value of λ was meticulously chosen to ensure model robustness and minimize the mean cross-validated error. Genes exhibiting collinearity were excluded to further reduce model bias. The model selected the λ value that provided the best balance between model complexity and prediction accuracy, ensuring stability and avoiding overfitting. Following a multivariate logistic regression analysis of the factors derived from LASSO, only factors with a P-value <0.05 were retained as pivotal predictors in the model. The risk score was computed through a linear combination of the expression levels of each ARG (α) and their respective coefficients (β), derived using the following formula: Risk Score = α1 ×β1 + α2 × β2 + ⋯ + αn × βn. To precisely evaluate the predictive capacity of this model, the area under the receiver operating characteristic (ROC) curve (AUC) was calculated using the ‘pROC’ package (version 1.18.4).

Validation of the model and nomogram

The nomogram, a straightforward and user-friendly two-dimensional representation, is primarily designed to encapsulate specific and statistically significant parameters derived from multivariate logistic regression analyses. Within the R framework, we utilized the ‘rms’ package (version 6.7.1) to assess the probability of epilepsy onset. All independent factors from the logistic regression analysis were consolidated to construct a diagnostic nomogram for epilepsy. Furthermore, to ascertain the applicability and reliability of the model, we constructed the ROC curve of the model. In addition, we validated the efficacy of the model by constructing an ROC curve using the validation set GSE4290. To highlight the predictive power of the risk score model based on DE-ARGs, we computed the Harrell’s C-index and plotted a calibration curve using the ‘Hmisc’ (version 5.1.1) and ‘rms’ (version 6.7.1) packages, facilitating a comparison between predicted and observed outcomes.

Immune cell infiltration analysis

The CIBERSORT method, based on support vector regression, was used to deconvolve the expression matrices of immune cell subtypes. This analysis inferred the relative proportions of 22 immune cell phenotypes. Spearman correlation analysis was performed to assess the relationship between immune cell content and gene expression. ¹⁸

Gene set enrichment analysis (GSEA)

GSEA is frequently used to investigate the intricate relationship between disease categorization and its biological relevance. ¹⁹ This approach was applied to assess the variances in signaling pathways across groups with high and low gene expression. For this purpose, version 7.0 of annotated gene sets from the MsigDB database was used, specifically targeting subtype pathway gene sets. The analysis focused on the differential expression of pathways across various subtypes, leading to the sequencing of gene sets that were notably enriched (with an adjusted P-value <0.05) based on their consistency scores.

Gene set variation analysis (GSVA)

GSVA, an unsupervised, nonparametric approach, is utilized for the assessment of gene set enrichment in the transcriptome. ²⁰ GSVA transforms alterations at the gene level into modifications at the pathway level through an exhaustive scoring system, thereby facilitating the evaluation of the sample’s biological function. In our research, gene sets obtained from the Molecular Signatures Database were analyzed using the GSVA algorithm to perform extensive scoring of each gene set. This analysis aimed at assessing the prospective biological functionality alterations across various sample types.

Establishment of mouse status epilepticus (SE)

In total, 22 adult male C57BL/6 mice, with body weights ranging from 22 to 28 g, were sourced from Spebo (Beijing) Biotechnology Co., Ltd. These animals were maintained in a controlled environment (temperature of 25°C, relative humidity of 60%, and a 12-h light/dark cycle), with free access to water and food. For the establishment of an SE model, pilocarpine (300 mg/kg; Sigma, St. Louis, MO, USA) was administered to the mice via intraperitoneal injection. ²¹ The controls (n = 6) were administered an equivalent amount of normal saline solvent via intraperitoneal injection. To mitigate pilocarpine’s peripheral cholinergic effects, pretreatment with methyl scopolamine (2 mL/kg; Sigma, St. Louis, MO, USA) was performed 30 min before administering pilocarpine. The occurrence and intensity of seizure activity were monitored and assessed using the modified Racine scale. ²² SE was defined as a seizure score ≥3 and a seizure that lasted 2 h. Diazepam (4 mg/kg; Sigma, St. Louis, MO, USA) was administered intraperitoneally 2 h after the onset of SE to control the seizures. ²¹ After pilocarpine injection, six mice died and four failed to develop SE; six mice developed SE and were selected for further study. The mice were euthanized in a humane manner, and their hippocampi were extracted for further examination. All procedures adhered strictly to the ethical guidelines of the Institutional Animal Care and Use Committee (IACUC) at Guangdong Medical University.

Quantitative reverse transcription–polymerase chain reaction (qRT–PCR)

Total RNA was extracted from hippocampal tissues. Then, a PrimeScript RT kit (Vazyme, R323) was used for the following PCR experiments. A 20-μL reaction mixture was prepared and incubated at 37°C for 15 min, followed by rapid heating to 85°C for 5 s. qRT–PCR amplification was performed using the Taq Pro Universal SYBR qPCR Master Mix (Vazyme), with the amplification process conducted on the LightCycler 480 II system (Roche). The PCR was programmed as follows: initial denaturation at 95°C for 30 s, denaturation at 95°C for 10 s, annealing at 60°C for 30 s for a total of 40 cycles. In this experiment, actin served as the reference gene, with the primers detailed in Table 1. The relative expression levels of the genes were calculated using the 2^−ΔΔCt method, and the mean values were derived from three independent experiments.

Table 1.

Primers (forward and reverse) used in this study.

Primer	Sequence
CAPN2 forward	5′-GGTCGCATGAGAGAGCCATC-3′
CAPN2 reverse	5′-CCCCGAGTTTTGCTGGAGTA-3′
ERN1 forward	5′-ACACCGACCACCGTATCTCA-3′
ERN1 reverse	5′-CTCAGGATAATGGTAGCCATGTC-3′
RELA forward	5′-AGGCTTCTGGGCCTTATGTG-3′
RELA reverse	5′-TGCTTCTCTCGCCAGGAATAC-3′
SAR1A forward	5′-ATTCTTAGGATTGGACAATGCGG-3′
SAR1A reverse	5′-CACCGAGATCAAAAGTGGTGAAA-3′
Actin forward	5′-GGCTGTATTCCCCTCCATCG-3′
Actin reverse	5′-CCAGTTGGTAACAATGCCATGT-3′

Drug prediction using Connectivity Map (CMap)

CMap, developed by the Broad Institute, is a database featuring gene expression profiles centered around intervention-based gene expression levels. ²³ Its primary function is to uncover functional links between small-molecule compounds, genetic elements, and various disease states. The database encompasses microarray data capturing the pre- and post-treatment effects of 1309 small-molecule drugs across 5 human cell lines. The experimental conditions in the database are diverse, encompassing a range of drugs, concentrations, and treatment durations, among other variables. In our research, we utilized the database to identify potential therapeutic drugs for the disease by analyzing genes that showed differential expression in the diseased state.

Results

Identification of DEGs between epilepsy patients and controls

We sourced the GSE143272 dataset from the NCBI GEO public database, comprising 141 samples, with 50 in the control group and 91 in the disease group. Using the ‘limma’ package, we identified DEGs between the two groups, setting a criterion of adj. P < 0.05. This resulted in the identification of 3111 DEGs: 1473 upregulated and 1638 downregulated (Figure 1(a) and (b)). Subsequently, from the GeneCards database (https://www.genecards.org/), we identified 174 ARGs (Figure 1(c)). An intersection of these ARGs with DEGs yielded 49 shared genes, termed as intergenes (Figure 1(c)). We further performed pathway analysis of these intergenes. Our findings revealed that intergenes predominantly participated in BPs such as macroautophagy, cellular responses to external stimuli, and responses to reactive oxygen species. In terms of CC enrichment, intergenes were chiefly associated with structures such as autophagosomes, vacuolar membranes, and autophagosome membranes. The discerned MF terms encompassed protein binding, protein serine/threonine kinase activity, ubiquitin protein ligase binding, and ubiquitin-like protein ligase binding (Figure 2(a) to (c)). Concurrently, Metascape analyses indicated that intergenes were instrumental in signaling pathways related to shigellosis, neurodegeneration, and apoptosis (Figure 2(d)).

Figure 1.

Identification of differentially expressed genes in epilepsy patients. (a) Volcano plot of differentially expressed genes, with pink indicating upregulated expression and blue indicating downregulated expression. (b) Heatmap of differentially expressed genes, with red indicating high expression and green indicating low expression and (c) Venn diagram showing the intersection of epilepsy differentially expressed genes and autophagy-related genes.

Figure 2.

GO and Metascape Functional Analysis. GO/Metascape enrichment analysis of differentially expressed genes based on the DAVID and Metascape databases. GO: Gene Ontology.

Use of the LASSO model and logistic model to identify potential predictive markers of epilepsy

We designated the GSE143272 dataset as the training set and the GSE4290 dataset as the validation set. Using LASSO regression analysis, we selected features from the intergenes. The results indicated that 20 genes were identified as being characteristic of epilepsy, including CTSD, RGS19, PRKAB1, BNIP3, RAB7A, ERN1, RELA, RPS6KB1, SQSTM1, PELP1, FADD, NFE2L2, CAPN2, BAX, VEGFA, SAR1A, RB1, HSPA5, GNB2L1, and DNAJB1. Further refinement using the ‘glmnet’ package based on model coefficient P-values highlighted four DE-ARGs: CAPN2, ERN1, RELA, and SAR1A. The four genes were then used to construct a predictive model (Figure 3(a) and (b)) using the following formula: RiskScore = CAPN2 × (−0.221091258475995) + ERN1 × (−0.154716139704634) + RELA × (−0.092242622389523) + SAR1A × (−0.0386863939889303). The diagnostic efficacy of this model was demonstrated by the AUC values (Figure 3(a) and (b)). A nomogram for the diagnostic model is delineated in Figure 3(c), serving both as a visual representation and a practical tool for clinical application. Utilizing the actual expression levels of these four DE-ARGs in a patient’s blood sample, clinicians can locate these values on the corresponding axes of the nomogram. By projecting these values onto the “points” scale situated at the apex of the diagram, individual point values for each variant can be ascertained. The aggregate of these points constitutes the total point score, which is then projected onto the lowermost scale to estimate the patient’s risk probability for epilepsy. Calibration curves for the risk nomogram, which was employed for epilepsy risk prediction, exhibited strong concordance between the training and validation sets (Figure 4(a) and (b)). The resultant AUC values were 0.881 for the training set and 0.978 for the validation set, surpassing those achieved without the incorporation of these key factors.

Figure 3.

Establishment of a multipredictor nomogram and DE-ARG selection using LASSO and logistic regression models. (a) Cross-validation was used to select the most suitable tuning parameter lambda (λ); the first black dotted line represents the 49 features that were reduced to 21 non-zero coefficient features by LASSO. (b) The coefficients in the LASSO regression model for key DE-ARGs and (c) Predictive nomogram involving the expression profile of 4 DE-ARGs based on the selected features. DE-ARGs: differentially expressed autophagy-related genes; LASSO: Least Absolute Shrinkage and Selection Operator.

Figure 4.

Model discrimination and calibration curve analysis. (a) ROC curve for the prognostic model of epilepsy based on GSE14272. (b) ROC curve for the epilepsy prognostic model based on GSE4290 and (c) Calibration curve of the epilepsy nomogram prediction in the GSE143272 set. ROC: receiver operating characteristic.

Immune cell infiltration analysis

The immune microenvironment, which includes immune cells, the extracellular matrix, a range of growth factors, and mediators of inflammation and exhibits distinct physicochemical properties, plays a pivotal role in disease diagnosis and therapeutic responsiveness. ¹² We examined the association between critical genes and immune infiltration in the epilepsy dataset and performed an in-depth exploration of the potential molecular mechanisms by which these genes may impact epilepsy progression. The proportions of immune cells in each patient and the correlations between these cells are depicted in Figure 5(a). Notably, significant differences were observed in the proportions of resting CD4 memory T cells, activated CD4 memory T cells, and activated mast cells between the two groups (Figure 5(b)). Further analysis revealed strong correlations between the four DE-ARGs and immune cell infiltration, aligning with our expected results (Figure 5(c)).

Figure 5.

Immune infiltration status. (a) Proportional representation of 22 immune cell subtypes in the analyzed samples. (b) Comparative analysis of immune cell composition in control versus epileptic samples, where control samples are denoted in blue and epileptic samples in red; A P-value <0.05 was deemed to indicate statistical significance and (c) Analysis of the relationship between the expression of critical genes and the abundance of various immune cells.

Significant pathways involving critical genes

We attempted to clarify the specific signaling cascades enriched by these quintessential genes, aiming to reveal the molecular underpinnings of their influence on epilepsy progression. GSEA insights revealed that ERN1 is predominantly associated with the B cell receptor signaling pathway, MAPK signaling trajectory, and oncogenic pathways (Figure 6(a)). RELA displayed enrichment in pathways such as acute myeloid leukemia, glycerophospholipid metabolism, and lysosomal functions (Figure 6(b)). CAPN2 was notably linked to cysteine and methionine metabolism, oxidative phosphorylation, and Parkinson’s disease pathways (Figure 6(c)). However, SAR1A demonstrated enrichment in pathways such as chronic myeloid leukemia, complement and coagulation cascades, and endometrial cancer (Figure 6(d)). Furthermore, GSVA results indicated that an elevated expression level of ERN1 was predominantly observed in pathways such as other glycan degradation, cytosolic DNA sensing pathway, and peroxisome processes (Figure 6(e)). Augmented RELA expression was associated with lysine degradation and small cell lung cancer pathways (Figure 6(f)). CAPN2, when overexpressed, showed a predilection for the renin–angiotensin system, starch and sucrose metabolism, and non–small cell lung cancer pathways (Figure 6(g)). Elevated SAR1A expression was notably associated with fructose and mannose metabolism, Huntington’s disease, and oxidative phosphorylation pathways (Figure 6(h)). These results suggested that these cardinal genes modulate epilepsy progression via these intricate pathways.

Figure 6.

GSEA and GSVA analyses of key genes. (a–d) KEGG signaling pathways involving ERN1, RELA, CAPN2, SAR1A as well as pathway regulation and genes involved in the pathways. (e–h) GSVA analysis of ERN1, RELA, CAPN2, and SAR1A. Red color represents signaling pathways associated with high expression of the genes, while blue color indicates signaling pathways associated with low expression of the genes. The background gene set used for analysis is “hallmark.” GSEA: gene set enrichment analysis; GSVA: gene set variation analysis; KEGG: Kyoto Encyclopedia of Genes and Genomes.

Confirmation of biomarker screening outcomes using qRT–PCR experiments

To validate the predicted expression levels of ERN1, RELA, CAPN2, and SAR1A, their expression levels were experimentally tested using qRT–PCR analysis of hippocampal tissues obtained from six SE mice. The results, as depicted in Figure 7, revealed that ERN1, CAPN2, and SAR1A showed lower expression levels in SE hippocampal tissues than in controls (n = 6) (P < 0.05, using Student’s t-test), whereas RELA exhibited higher expression levels.

Figure 7.

Evaluation of the key genes in animal models. (a–e) The expression levels of ERN1, RELA, CAPN2, and SAR1A mRNA in the hippocampus tissue of an SE mouse model (*P < 0.01,**P < 0.001,***P < 0.0001, ****P < 0.00001, n = 6). SE: status epilepticus.

Candidate small-molecule drug identification

To determine the optimal therapeutic strategy approach, we categorized the top 150 upregulated and downregulated genes into two sets. Leveraging the CMap database, we endeavored to predict drug targets for these differential genes. Intriguingly, the expression profiles altered by drugs such as cabergoline, capsazepine, and zolantidine exhibited significant negative correlations with the disease-altered expression profiles, suggesting their capability to mitigate, if not reverse, the disease (Table 2).

Table 2.

The top three compounds in CMap analysis.

Structural formula	Name	Score	Description
	Cabergoline	99.89	Dopamine receptor agonist
	Capsazepine	99.82	TRPV agonist
	Zolantidine	99.82	Histamine receptor antagonist

CMap: Connectivity Map; TRVP: transient receptor potential vanilloid.

Discussion

The intricate landscape of epilepsy, a neurological disorder affecting millions globally, necessitates the development of precise diagnostic and therapeutic strategies. Utilizing high-throughput sequencing and sophisticated bioinformatics, we established an extensive framework to comprehend the autophagy-related molecular basis of epilepsy. The findings not only identified potential biomarkers but also offered a promising avenue for therapeutic interventions. Our analysis of the GSE143272 dataset led to the identification of 3111 DEGs, including 49 ARGs. Notably, four DE-ARGs—ERN1, RELA, CAPN2, and SAR1A—emerged as potential diagnostic markers. ERN1, commonly referred to as IRE1 (inositol-requiring enzyme 1), is a transmembrane protein predominantly localized to the membrane of the endoplasmic reticulum (ER). ²⁴ Our identification of ERN1 as a key player aligns with and extends the established understanding of ER stress pathways as critical mediators in epilepsy pathogenesis, as highlighted in foundational reviews of the disease mechanisms. ² Intriguingly, emerging research has elucidated that ERN1 can be activated by BIX, thereby restoring the functionality of N-methyl-D-aspartic acid receptors (NMDARs) containing pathogenic NR2A subunits and consequently mitigating the deleterious effects associated with epilepsy. ²⁵ Notably, IRE1 acts as a critical sensor and initiator of ER stress pathways, playing an indispensable role in the injury of hypothalamic neurons. ²⁶ RELA, also known as p65, is an integral component of the nuclear factor kappa B (NF-κB) complex. ²⁷ This protein is pivotal in orchestrating cellular responses to a myriad of stimuli, including stress, cytokines, free radicals, and ultraviolet irradiation as well as bacterial or viral antigens. ²⁸ In a rat model of SE, tumor necrosis factor-α–mediated phosphorylation of RELA leads to the upregulation of both endothelin B receptor and transient receptor potential canonical channel-3 (TRPC3). This finding is consistent with a broader body of evidence implicating neuroinflammatory cascades, particularly those involving NF-κB signaling, in the development and progression of epilepsy. ² This cascade subsequently elevates the expression level of endothelial nitric oxide synthase (eNOS) via the PI3K/AKT signaling pathway. ²⁹ CAPN2 (calpain 2) is an enzyme that belongs to the nonclassical calcium-dependent protease family, colloquially known as calpains. ³⁰ These enzymes are calcium-activated neutral proteases implicated in a plethora of biological processes, ranging from cell migration and signal transduction to cell cycle regulation and apoptosis. Remarkably, elevated levels of calpain-1 and calpain-2 have been detected in brain tissue resected from epilepsy patients. ³¹ Moreover, this upregulation is correlated with key clinicopathological changes integral to epileptogenesis, such as neurodegeneration, astrogliosis, and inflammation. ³² The association of CAPN2 dysregulation with these hallmark pathological features reinforces complementary evidence from prior studies linking calcium-dependent proteolytic activity to neuronal excitability and cell death in epilepsy. ² SAR1A is a small guanosine triphosphatase (GTPase) protein that plays a critical role in the regulation of vesicle trafficking, particularly during the early stages of the secretory pathway. ³³ Intriguingly, in neuronal N1E-115 cells, an elevation in the levels of GTP-bound SAR1A has been observed concomitant with the elongation of neuronal processes. ³⁴ These genes were implicated in various cellular processes, including synaptic plasticity, excitotoxicity, and neurodegeneration, thereby corroborating the role of autophagy in epilepsy pathogenesis. The robustness of these markers was further validated using the GSE4290 dataset, reinforcing their potential clinical utility.

The immune microenvironment has gained attention for its role in various diseases, including neurological disorders. Our study revealed significant differences in T cell populations between epilepsy patients and controls. The correlation between these immune cells and the identified DE-ARGs sheds light on the disease mechanisms, requiring further investigation. The CMap database facilitated the identification of potential therapeutic agents, including cabergoline, capsazepine, and zolantidine. Cabergoline, a therapeutic agent marketed as Dostinex and Cabaser, functions primarily as an agonist at dopamine D2 receptors while also exhibiting moderate affinity toward D1 and serotonin receptors, alongside inhibitory actions on alpha-2 adrenoreceptors. ³⁵ Investigations into its pharmacodynamics within primary astrocyte cultures have demonstrated cabergoline’s capacity to elevate the levels of glial cell line-derived neurotrophic factor (GDNF), ³⁶ a molecule that has been implicated in the attenuation of seizure activity in experimental models of epilepsy. ³⁷ The potential of targeting neurotrophic pathways such as GDNF for seizure control is supported by previous research exploring neuroprotective strategies in epilepsy. ² Moreover, this compound may offer a novel therapeutic approach for drug-resistant epilepsy through the modulation of dopaminergic system activity. Capsazepine (CPZ), a synthetic antagonist, selectively targets the transient receptor potential vanilloid subtype 1 (TRPV1) channels, which have been implicated in nociceptive signaling. ³⁸ Notably, augmented TRPV1 expression has been documented in the cortical and hippocampal regions of patients with mesial TLE as well as in the dentate gyrus of mouse models subjected to pilocarpine-induced limbic SE. ³⁹ This observed overexpression of TRPV1 in epileptic foci provides mechanistic justification for exploring TRPV1 antagonists, aligning with the growing recognition of ion channel modulation as a viable therapeutic avenue in epilepsy. ² Experimental evidence suggests that CPZ exhibits anticonvulsant properties, effectively attenuating seizure activity in both in vivo and in vitro models. ⁴⁰ Despite these findings, CPZ has not been extensively adopted in clinical practice for epilepsy treatment. Zolantidine (ZOL) is characterized as a selective antagonist of the histamine H2 receptor. ⁴¹ Accumulating evidence underscores a robust association between brain histamine levels and the pathophysiology of seizures. Despite this correlation, empirical data suggest that ZOL does not confer significant ameliorative effects in animal models of epilepsy. ⁴² Consequently, the precise therapeutic targets and the efficacy of ZOL in epilepsy treatment warrant further investigation to elucidate its potential clinical utility.

These drugs exhibited significant negative correlations with the disease-altered expression profiles, suggesting their potential to mitigate the disease. Although these findings are preliminary, they open new avenues for drug repurposing in epilepsy treatment. The development of a risk score model based on DE-ARGs and its encapsulation into a user-friendly nomogram enhanced the study’s clinical applicability. High AUC values of the model in both training and validation sets indicated its predictive accuracy. Additionally, decision curve analysis indicated a superior net benefit ratio, emphasizing its clinical relevance.

However, our study has certain limitations. First, the sample size in the datasets used was relatively small, probably affecting the generalizability of the findings. Second, this study was cross-sectional in nature, lacking longitudinal data that could provide insights into epilepsy progression over time. Finally, the derived mechanistic insights were primarily computational and required further confirmation through basic and clinical studies. Future studies should aim to validate these findings in larger cohorts, conduct deeper investigations on the molecular mechanisms of these key ARGs in epilepsy, and establish their clinical utility.

In summary, our study provided a comprehensive bioinformatics and experimental framework for understanding the role of autophagy in epilepsy. The identified DE-ARGs not only serve as potential diagnostic markers but also offer a probable mechanistic basis for therapeutic intervention. With advancements in precision medicine, such multi-dimensional approaches would be indispensable in tackling the complexities of epilepsy and other neurological disorders. The limitations outlined above serve as a roadmap for future research aimed at refining and expanding upon the current findings.

Conclusions

Our comprehensive bioinformatics and experimental approach identified potential biomarkers for epilepsy, including CAPN2, ERN1, RELA, and SAR1A, offering insights into potential therapeutic strategies. The identified drugs, if validated in further studies, may pave the way for novel treatments for epilepsy.

Data availability statement

The data presented in this study are openly available in Gene Expression Omnibus (GEO) (URL: https://www.ncbi.nlm.nih.gov/geo/, accessed on 17 January 2023).

Declaration of conflicting interests

The authors declare no conflict of interest.

Institutional review board statement

Not applicable.

Informed consent statement

Not applicable.