An essential part of the pathophysiology of epilepsy is played by ferroptosis, a type of programmed cell death that is dependent on iron. Iron overload has been linked to both post-traumatic23 and post-stroke24 epilepsy in several studies. It is known that iron salts or hemoglobin injections into the cortex of rats can be used to create animal models of localized persistent epilepsy25. Lipid peroxidation was shown to be considerably higher in epileptic patients compared to controls in another investigation26. Furthermore, strong evidence27,28 points to the activation of glial cells by neuroimmune pathways triggered by relevant chemical patterns (e.g., ROS) during ferroptosis, which in turn generates a variety of inflammatory factors that ultimately result in neurological illness. This implies that neuroimmune pathways regulated by ferroptosis play a significant role in the pathophysiology of epilepsy. In light of the aforementioned data, a thorough examination of immunological traits and ferroptosis-related molecular patterns is required to develop a theoretical framework for epilepsy biomarkers and a dependable research direction for upcoming experimental investigations.
Using the FerrDb database, we were able to retrieve 728 genes connected to ferroptosis for our investigation. The aforementioned genes were screened using RF and mSVM-RFE, two machine learning algorithms. To aid in the diagnosis of epilepsy, the RF algorithm and the mSVM-RFE algorithm are employed in the field of epilepsy to classify the features of EEG in patients with epilepsy29,30. After combining the study of two machine learning techniques, we were able to identify three Hub genes—RELA, TFRC, and QSOX1—that are most strongly linked to epilepsy. TFRC and QSOX1 have never been studied to confirm their relationship with epilepsy, although RELA's31 role in the pathophysiology of the condition has been experimentally validated. Another name for transferrin receptor protein 1 (TFR1) is CD71 or TFRC. The most significant membrane protein controlling iron transport in cells is TFRC, a type II transmembrane protein 32. According to recent research, TFRC plays a major role in central nervous system ferroptosis. When TFRC is strongly expressed in neurons, more iron ions can enter the cell, overloading intracellular iron and causing neurons to die33. Some studies have also shown that TFRC on astrocyte membranes can enter cells and store by transporting more iron ions, which in turn attenuates the cascade of iron ion spillage damage surrounding neurons caused by neuronal ferroptosis and protects neurons from ferroptosis damage34. These findings imply that TFRC may modulate immunological responses, which in turn influence how neuronal ferroptosis develops. Notably, our bioinformatics prediction results showed down-regulated TFRC expression, but our RT-qPCR data showed the reverse. Whereas RT-qPCR results were obtained from hippocampus tissue from mice with temporal lobe epilepsy, our bioinformatics-analyzed data came from blood samples from individuals with drug-resistant epilepsy. This discrepancy between the two warrants additional investigation. During protein folding, an enzyme called Quiescin Thiol Oxidase 1 (QSOX1) oxidizes thiols, converting molecular oxygen to hydrogen peroxide. There aren't many studies on QSOX1, and the ones that exist have mostly dealt with the topic of treatment resistance in hepatocellular carcinoma35 and non-small cell lung cancer36. According to some research, in a pig model of myocardial infarction, whole blood expression levels of QSOX1 are linked to ischemic myocardial neutrophil infiltration37. All of these findings point to QSOX1's similarly significant importance in neuroimmunity.
We used cluster analysis to further analyze the Hub gene and classified epilepsy patients into FRG Cluster A and B subtypes. It was activated in the type A FRGcluster.CD4.T.cell and turned on. The proportions of CD8.T.cell infiltration were much larger than those of type B; however, the infiltration of Fig. Immature..B.cell, Natural.killer.cell Neutrophil, Plasmacytoid.dendritic.cell, and T. follicular.helper.cell was the reverse. These findings assist us in differentiating between the severity of epilepsy patients in clinical practice since they imply that type B may be a stage of seizure deterioration due to neuroinflammation, while type A remains in a moderate inflammatory injury stage. Interestingly, Hub gene expression levels varied between types A and B, indicating the potential utility of Hub gene-based typing in determining the extent of epilepsy patients' disease development. Furthermore, we analyzed the variations in scores among subtypes by utilizing the FRGscore outcomes acquired through PCA. In the FRG Cluster, type A's FRGscore was greater than type B's, and there was a positive association between age and FRGscore but a negative correlation with treatment response. Furthermore, a correlation study between the FRGscore and immune cells revealed a strong negative connection between the FRGscore and Activated.CD4.T.cell and Activated.dendritic.cell. This suggests that dysregulation of the immune microenvironment is a critical factor in the development of epileptogenesis. This demonstrates the intricate regulatory relationship that exists between immune and ferroptosis-related biological pathways. It also implies that our Hub gene construct-based genotyping method can successfully detect epileptic patients who are still in the early phases of inflammation.
Lastly, we used the Hub gene to build a diagnostic model based on the logistic regression technique, and we tested the model's accuracy on two different datasets. Furthermore, a Nomo map was developed as a prognostic tool for the advancement of epilepsy. We find a strong correlation between the expected and actual observed values when used with the Nomo plot's calibration curve. The decision curve study also showed that Nomo plots could provide patients with more clinical benefits.
Our study is not without limits, though. To improve the clinical prognostic value of the Nomo plot, additional clinical data must first be taken into account. Second, we established and validated the diagnostic model's accuracy using data sets obtained from brain and blood tissue samples. Prospective cohort research is necessary to learn more about their diagnostic efficacy, nevertheless. The most important component of our upcoming study will be this. We intend to finish our additional research to gain a deeper understanding of the pathophysiology of epilepsy and the molecular immunological mechanisms associated with ferroptosis.