Epilepsy is a serious chronic nervous system disease characterized by recurrent seizures that strongly affect patients’ motor, cognitive and emotional functioning. With an annual incidence rate estimated to be approximately 68 per 100,000 persons (Fiest et al., 2017) and global costs of treating patients with epilepsy surpassing 100 billion dollars per year worldwide (Begley et al., 2022), epilepsy remains one of the leading causes of human disability, significant drop in quality of life, and mortality (Beghi, 2016; Saadi et al., 2016). Therefore, epilepsy has been an area of active research in medicine for decades. Despite apparent progress in finding new molecular and neuroimaging biomarkers and the development of novel approaches to treatment, approximately 30% of patients do not appropriately respond to the pharmacological therapy being conducted (Kalilani et al., 2018). These cases of drug-resistant epilepsy are frequently considered for further neurosurgical treatment aimed at resection of the epileptogenic zone found during neuroimaging and scalp or invasive EEG recordings. Generally, neurosurgical treatment is quite successful in eliminating or at least reducing the number and intensity of seizures, as exemplified by a large body of recent evidence showing that 53 to 84% of such patients are totally seizure-free after surgery at a follow-up point of as long as 1 year (Spencer and Huh, 2008; Vakharia et al., 2018). However, the health care community is still in an urgent need of objective biomarkers able to effectively and reliably predict the outcome of surgery.

Mesial temporal lobe epilepsy (MTLE) is the most common type of focal epilepsy in the clinical population (Zhou et al., 2020). MTLE is an umbrella term for focal epilepsies caused by pathological processes within any paramedian temporal structures, including the hippocampal formation, amygdala, entorhinal and parahippocampal gyri (Vinti et al., 2021). One of the most common histopathological diagnoses among adults with drug-resistant focal epilepsy requiring surgery is hippocampal sclerosis (HS; Blumcke et al., 2017). HS is histologically described as a loss of pyramidal cells in the CA1, CA3, or CA4 regions, whereas CA2 pyramidal and dentate gyrus granule cells are seizure resistant (Blümcke et al., 2012). The loss of nerve cells is associated with neural tissue fibrosis. Despite extensive research over the past several decades, the pathophysiology of HS is still not fully understood (Takano and Coulter, 2012). It has commonly been assumed that different aberrant mechanisms, such as ion channel disturbances (Muñoz et al., 2007; Sen et al., 2007), synaptic neurotransmission abnormalities (Mathern et al., 1999; Ben-Ari and Holmes, 2005), apoptosis (Yu et al., 2021), neuroinflammation (Gales and Prayson, 2017), and brain–blood barrier dysfunction (Van Vliet et al., 2007), are involved in this process. Furthermore, alterations in the expression of different neuropeptides have also been shown to participate in the development of HS (Thom, 2014). Another type of pathological pattern in TLE is gliosis without neuronal loss, called “no hippocampal sclerosis, gliosis only” or simply “gliosis only” (Blümcke et al., 2013). Reactive astrocytes are common in HS and other epileptic foci, and it is hypothesized that they increase tissue excitability and the generation of seizures. It is also not clear whether gliosis precedes a loss of neurons that leads to HS or constitutes a different disease (Wetherington et al., 2008).

Magnetic resonance imaging (MRI) is one of the most useful diagnostic techniques, which enables an effective identification of epilepsy substrates (Linera, 2019). It is routinely used in the initial assessment of patients with epilepsy, and the diagnosis of HS depends on positive signs on MRI. In drug-resistant cases, MRI is crucial in the selection of surgical candidates and contributes to the treatment plan. It is known that the use of high magnetic field scanners (3 T and more) with high-resolution sequences (both 3D and 2D) allows the detection of even very small lesions by increasing the tissue contrast and the signal-to-noise ratio (Sone, 2021).

In addition, the development of advanced MRI sequences as well as postprocessing techniques provides valuable additional information regarding the structural, physiological and metabolic depiction of brain tissue (Tsougos et al., 2019). One of the most perspective image postprocessing tools in the field of MTLE research is automatic MR-morphometry with FreeSurfer,¹ which allows delineating the hippocampal formation with subsequent segmentation on anatomically different subfields (Iglesias et al., 2015). This technique is widely used (Sämann et al., 2022), and several recently published papers demonstrated that it could be helpful in the evaluation of patients with HS, both independently (Sone et al., 2016; Long et al., 2018; Costa et al., 2019) and as part of multimodal analysis (Rutland et al., 2018).

The structural alterations visible on MR images represent accumulated damage caused by disease progression. A disturbance of the exquisite balance between excitatory and inhibitory neural processes may lead to the emergence of epileptogenic zones wherein epileptic activity is generated. Neuroplasticity plays a fundamental role in forming new synapses and pruning ineffective synapses, regulating the excitation to inhibition ratio (E:I ratio), and thus is of special interest to researchers investigating the cellular and molecular mechanisms of epilepsy. Among all members of the large family of neurotrophins that mediate neuroplasticity-associated tissue reorganization, brain-derived neurotrophic factor (BDNF) holds a unique place. BDNF is well established to be involved in the pathogenesis of a multitude of neurological and psychiatric disorders, such as Parkinson’s (Palasz et al., 2020) and Alzheimer’s diseases (Ng et al., 2019), epilepsy (Iughetti et al., 2018), depression (Yang et al., 2020), anxiety disorders (Suliman et al., 2013) and many others. Dysfunction of synaptic plasticity at the molecular level could be a common denominator for all of the abovementioned pathologies. The results of early studies on BDNF involvement in the emergence and spread of epileptic activity in the brain were ambiguous, with some researchers arguing for the antiepileptic nature of BDNF and others taking the opposite side (Koyama and Ikegaya, 2005; Simonato et al., 2006). Over time, the accumulated evidence has tipped the scales toward the stance shared by proponents of the proepileptic nature of hippocampal neurotrophic factors (Scharfman et al., 2005; Iughetti et al., 2018; Porcher et al., 2018).

It should be mentioned, however, that the overwhelming majority of studies are preclinical studies conducted on experimental animals in laboratory settings, whereas human data are scarce. Rare clinical trials performed in small-sized samples of patients with epilepsy undergoing surgical treatment are of great value and worth noting here. For instance, Martinez-Levy et al. reported that the expression of BDNF exon VI was increased in the hippocampus of patients with MTLE compared to the corresponding levels of hippocampal BDNF in individuals who died of causes other than psychiatric or neurological diseases (Martínez-Levy et al., 2016, 2018). In another early paper, in agreement with other studies, the authors gauged the BDNF mRNA level in the hippocampi of temporal lobe epilepsy patients and found that it was increased in a group of patients with hippocampal sclerosis (Mathern et al., 1997). Even more remarkable results were presented by Takahashi et al. (1999). They not only observed an elevated hippocampal level of BDNF in MTLE patients but also demonstrated that the same did not hold for other neurotrophins (nerve growth factor and neurotrophin 3) investigated in that study. Along the same lines, a study by Murray et al. adds to the growing list of studies evidencing the positive association between BDNF levels and the epileptogenic capacity of hippocampal neural circuits (Murray et al., 2000).

In most cases, when patients are not referred to neurosurgical treatment, one of the readily available options for clinicians to investigate, although indirectly, the role of BDNF in epileptogenesis is to measure its level in the patient’s serum. The number of such studies using BDNF serum levels in MTLE patients vastly outnumber those directly measuring BDNF in hippocampal tissue, which we have mentioned earlier. The results of these studies consistently reveal markedly decreased serum BDNF levels in patients with MTLE (Chen et al., 2016, 2017; Panina et al., 2023). However, noteworthy is a couple of published studies that did not segregate patients with focal epilepsy into subtypes and found no significant differences in serum BDNF levels between control and epilepsy groups (Hong et al., 2014; Nowroozi et al., 2021).

In this study, we set out to assess the relationship between BDNF measured in hippocampal samples and serum taken from epileptic patients during surgery and morphometric parameters of hippocampal subfields.

The search for biomarkers of disease progression, severity and resistance to therapy in patients with epilepsy is a pivotal direction in fundamental and applied medical science. In the current study, we focused on estimating the relationship between BDNF protein levels in the hippocampi or blood serum of patients with MTLE and the results of MR-based automatic segmentation of hippocampal subfields. Our results are generally in agreement with those published so far, showing that BDNF could indeed be an informative biomarker reflecting epilepsy-associated neural tissue reorganization. Specifically, we found that patients with more pronounced increases in hippocampal BDNF had smaller hippocampal volumes. In a similar vein, we discovered that different hippocampal subfields were not equally affected by the disease. Among the regions exhibiting the most statistically significant results were the CA1, subiculum, dentate gyrus and molecular layer. The CA1 and subiculum subfields have consistently been shown to exhibit a lowered threshold for the generation of epileptic activity and, therefore, are one of the first brain areas to undergo structural alterations detectable with high-resolution MRI (Stafstrom, 2005; Schoene-Bake et al., 2014; Kitaura et al., 2018; Santos et al., 2021). To our knowledge, this is the first attempt to compare serum and hippocampal BDNF levels with structural alterations in the hippocampal subfields according to MR morphometry.

We did not reveal any relationship between serum and hippocampal BDNF levels in this study, which is an interesting finding. Despite the fact that blood BDNF concentration reflects brain-tissue BDNF levels in some animals (Klein et al., 2011), the human data in this field are scarce. Notably, in humans, BDNF is widely expressed outside of the central nervous system, namely, in white blood cells, platelets, vascular endothelium, smooth muscles, and other peripheral tissues (Shvaikovskaya et al., 2020). Thus, its serum levels are likely to be influenced by peripheral sources rather than central BDNF levels. On the other hand, it is believed that serum BDNF levels are decreased in patients with MTLE compared to healthy controls (Chen et al., 2016, 2017), whereas hippocampal BDNF expression levels are increased (Murray et al., 2000). However, according to the results of the study by Iughetti and colleagues, serum BDNF can also be increased in human subjects suffering from epileptic seizures (Iughetti et al., 2018). It is therefore possible that hippocampal and serum BDNF concentrations change in different directions and with different dynamics, depending on the presence or absence of other confounding clinical variables. Thus, it seems likely that hippocampal and blood serum BDNF concentrations may be uncorrelated in patients with MTLE. We also did not find any correlations between serum BDNF levels and the results of MR morphometry of the hippocampal subfields. It is difficult to unambiguously ascertain the underlying causes of the absence of associations, but the results of a recently conducted preregistered study investigating a relationship between serum BDNF levels and hippocampal subfield volumes in healthy participants pointed in the same direction as our own. The authors did not find any significant correlations between BDNF and morphometric features of different hippocampal subfields and concluded that serum BDNF might have a limited predictive value for the morphological differences of the hippocampal structure (Puhlmann et al., 2021).

Surprisingly, we did not reveal any relationship between the levels of BDNF (both within the brain tissue and serum) and the volumes of the hippocampal subfields on the side of resection. The reason may be that all of our patients had severe drug-resistant epilepsy, a long history of disease, and obvious atrophy of the affected hippocampus. It is not clear from the available literature whether hippocampal volume loss is associated with the duration of MTLE (Kim et al., 2015; Caciagli et al., 2017; Kreilkamp et al., 2018; Wu et al., 2020). Loss of hippocampal volume was found to correlate with disease duration (Wu et al., 2020). Furthermore, Kim et al. showed an inverse relationship between disease duration and ipsilateral CA1 volume (Kim et al., 2015). Interestingly, this volume loss in CA1 contributes to the reduction in antiepileptic drug efficiency in models of chronic epilepsy. In contrast, despite also showing visible cell loss, the extent of neural degeneration unfolding in the CA4 region has been positively associated with the surgery outcome (King and Spencer, 1995; Savitr Sastri et al., 2014). One plausible explanation for this observation may be that the CA4 is strongly involved in the process of spreading epileptic activity to the contralateral hippocampus. Therefore, preventing this spreading, ipsilateral CA4 degeneration could protect the normally functioning hippocampus and, perhaps, make it possible for it to take over the functions of the affected hippocampus (King and Spencer, 1995). Meanwhile, Kreilkamp et al. reported that the volumes of hippocampal subfields were not correlated with clinical variables such as disease severity and disease duration (Kreilkamp et al., 2018). It has further been stated that HS can occur immediately after an initial precipitating injury; therefore, it may well be independent of chronicity and severity of the MTLE, since the majority of neurons might have already been malfunctioning or undergone irreversible changes by the time first seizures manifest (Pitkänen and Lukasiuk, 2011). Extending these findings, a series of papers have been published in the last decade, indicating a strong negative association between serum BDNF level and epilepsy duration (Hong et al., 2014; Chen et al., 2016). In contrast to the abovementioned statements, in our study, neither hippocampal nor blood serum BDNF correlated with disease duration. Notably, similar findings were reported in the recently published paper by Poniatowski et al. (2021).

We found a number of negative associations between hippocampal BDNF levels and volumes of different subfields of the contralateral hippocampus. We assume that it could reflect the spread of the disease, as contralateral hippocampus involvement in the MTLE process is well known and not uncommon (Pulsipher et al., 2007; Wu et al., 2020). Furthermore, a less favorable surgery outcome was demonstrated in patients with unilateral MTLE but the presence of contralateral hippocampus volume loss (Conz et al., 2014). In our study, negative correlations between BDNF levels and volumes of the contralateral CA1, subiculum, dentate gyrus and molecular layer subfields were revealed. There was also a trend toward a negative correlation between BDNF levels and the volume of the contralateral CA4 subfield. On the other hand, there were no correlations between CA2/3 volume and BDNF level. In general, these results are in line with the existing theoretical models of hippocampal epileptogenesis in MTLE. The CA1 and CA4 hippocampal subfields, which predominantly contain glutamatergic pyramidal neurons, are known to be the most susceptible to cell loss (Blümcke et al., 2013). However, CA2 pyramidal and dentate gyrus granule cells are relatively resistant to seizure damage with the involvement of these subfields only in late-stage disease (Thom, 2014). This can be explained by a direct excitotoxic effect of glutamate (Davolio and Greenamyre, 1995; Dillon et al., 2008) since CA2 is not part of the trisynaptic circuit (Takano and Coulter, 2012); additionally, a difference in damage susceptibility within the CA1 and CA3 subfields appears to be mediated by microglial reactivity (Lana et al., 2020). Recently, the subiculum was found to be a source of epileptogenic activity in patients with HS with its subsequent spreading to CA1, CA3, and CA4 areas (Kitaura et al., 2018); this may also be the reason for the observed volume reduction.

Another important part of epileptogenesis is neuroimmunity, which has been extensively studied during the last decade (Lima Giacobbo et al., 2019; Dai et al., 2021; Wang et al., 2021). Neuroinflammation is closely related to the development of epilepsy (Lima Giacobbo et al., 2019). For example, the protein of the high mobility group B1 (HMGB1) has shown elevated expression and an increased proportion of translocation from the nucleus to the cytoplasm in patients with epilepsy (Dai et al., 2021; Li et al., 2022). Therefore, HMGB1 and its signaling pathways may be a direction for antiepileptic drug therapy (Wang et al., 2021). However, BDNF and neuroinflammation signaling appear to be closely connected (Lima Giacobbo et al., 2019; Porter and O’Connor, 2022; Wang et al., 2022). For instance, BDNF has been shown to alleviate hippocampal neuroinflammation (Han et al., 2021). Therefore, another possible explanation for the increase in hippocampal BDNF expression in patients with MTLE is an attempt to compensate for the neuroinflammation burst.

Taken together, the results of this study provide additional evidence supporting the proepileptic action profile of BDNF and its predominantly negative relationship with hippocampal subfield volumes. However, after decades of investigation, given the scarcity of human data and conflicting findings, we still cannot confidently rely on BDNF as a biomarker for epilepsy severity, chronicity or treatment outcome prediction.

We found that hippocampal BDNF levels were negatively correlated with the volumes of the different subfields of the contralateral (but not ipsilateral) hippocampus in patients with drug-resistant MTLE. These findings are generally in accordance with existing data, arguing for a proepileptic nature of BDNF action.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by local Ethics Committee of the Federal Center for Neurosurgery, Novosibirsk, Russia (protocol No. 1 dated 1 March 2019). The patients/participants provided their written informed consent to participate in this study.

LA, KD, GM, and JR: conceptualization. EF, AP, SZ, AA, and MT: methodology. EF and AP: formal analysis. GM and MT: data curation. EF, AP, AT, SZ, AS, and AA: investigation. GM: resources. EF, AP, and GM: writing—original draft preparation. MT, JR, and LA: writing—review and editing. AP: visualization. LA, KD, and JR: supervision. MT: project administration and funding acquisition. All authors contributed to the article and approved the submitted version.

This study was funded by the Russian Science Foundation (grant no. 22-25-00588).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.