The ethical review boards of the University of Erlangen approved the study under agreement number 22-209-BP.

We analyzed microdissected hippocampal (n = 4), neocortical (n = 4), and white matter (n = 4) tissue obtained from patients with TLE diagnosed with hippocampal sclerosis type 1 (HS1) undergoing epilepsy surgery. Corresponding control samples were included for each region (HC: n = 4; NCx: n = 3; WM: n = 4). Control tissues were derived from post-mortem cases and underwent routine neuropathological examination at the Department of Neuropathology, which revealed no specific neurological or neuropathological disease. One neocortical control sample was excluded following outlier analysis (data not shown). Detailed information on all samples, including brain region, diagnosis, age, and sex is provided in Tables 1, 2. All tissue samples, both from TLE patients and controls, were fresh-frozen immediately after extraction and stored at −80 °C as part of the European Epilepsy Brain Bank (Blumcke et al., 2017). In accordance with hospital and legal procedures, post-mortem intervals were at least 24 h. After death, bodies were stored at 4 °C until autopsy was legally confirmed.

The mass spectrometry analysis was performed at the DTU Proteomics Core (Technical University of Denmark). 100 μl of lysis buffer (6M Guanidinium Hydrochloride, 10 mM TCEP, 40 mM CAA, 50 mM HEPES pH 8.5) was added to each sample along with a 3 mm tungsten carbide bead (Qiagen, Hilden, Germany). Samples were treated twice in the TissueLyser II homogenizer (Qiagen, Hilden, Germany) going from 3-30 Hz in 60 s. Beads were removed and samples were transferred to clean Eppendorf LoBind tubes (Eppendorf, Hamburg, Germany). Samples were boiled at 95 °C for 5 min, followed by sonication on high for 5 times 60 s on and 30 s off in a Bioruptor Pico sonication water bath (Diagenode, Liege, Belgium) at 4 °C. Samples were centrifuged at 18,000 RCF for 10 min, and supernatants were transferred to clean Eppendorf Protein LoBind tubes (Eppendorf, Hamburg, Germany). Protein concentrations were measured by BCA Rapid Gold (Thermo Scientific™, Waltham, USA) and 20 μg protein was taken forward for digestion. Sample volume was normalized between samples with additional lysis buffer and diluted 3x with digestion buffer (10% Acetonitrile in 50 mM HEPES pH 8.5) and 400 ng of LysC (MS-Grade Wako, Osaka, Japan) was added. Samples were incubated for 3 h at 37 °C while shaking at 750 RPM. After LysC digestion, samples were diluted to a final 10x in digestion buffer and digested with 200 ng Trypsin (MS-Grade, Osaka, Japan; Sigma Aldrich, St. Louis, USA) for 19 h at 37 °C while shaking at 750 RPM. Digestion was stopped by adding 2% trifluoroacetic acid (TFA) to a final concentration of 1%. The resulting peptides were desalted on a SOLAμ™ SPE plate (HRP, Thermo Scientific™, St. Louis, USA). Between each application, the solvent was spun through by centrifugation at 350 RCF. For each sample, the filters were activated with 200 μl of 100% methanol, then 200 μl of 80% acetonitrile, and 0.1% formic acid. The filters were equilibrated twice with 200 μl of 1% TFA, 3% acetonitrile, after which the samples were loaded. After washing the tips twice with 200 μl of 0.1% formic acid, the peptides were eluted into clean Eppendorf Protein LoBind tubes using 40% acetonitrile, 0.1% formic acid. The eluted peptides were concentrated in an Eppendorf Speedvac and reconstituted in 50 mM HEPES pH 8.5 for TMT labeling with 16plex tags (Thermo Scientific™, St. Louis, USA). A reference sample was prepared by mixing equal amounts of peptides from each sample and labeling that separately. Labeling was done according to manufacturer's instructions, and subsequently, labeled peptides were mixed, spiking in reference sample, resulting in two TMT pools. TFA was added to acidify and bring the acetonitrile concentration down to < 5%. Prior to mass spectrometry analysis, the peptides were desalted following the same procedure as describe previously and fractionated using an offline Thermo Scientific™ UltiMate™ 3,000 liquid chromatography system using high pH fractionation (5 mM Ammonium Bicarbonate, pH 10) at 5 μL/min flow rate (Thermo Scientific™, St. Louis, USA). Fifteen μg of each sample pool were separated over a 120-min gradient (5% to 35% acetonitrile), while collecting fractions every 130 s. The resulting 60 fractions were pooled into 30 final fractions, acidified to pH < 2 with 1% TFA and loaded onto EvoSep stage tips according to the manufacturer's protocol.

For each fraction, peptides were analyzed using the pre-set ‘30 samples per day' method on the EvoSep One instrument. Peptides were eluted over a 44-min gradient and analyzed with an Orbitrap Eclipse™ Tribrid™ instrument (Thermo Fisher Scientific™, St. Louis, USA) with FAIMS Pro™ Interface (ThermoFisher Scientific™, St. Louis, USA) switched between CVs of −50 V and −70 V with cycle times of 1.5 s. Full MS spectra were collected at a resolution of 1,20,000 with normalized AGC target set to “standard” or maximum injection time of 50 ms and a scan range of 375-1,500 m/z. Full MS spectra were collected at a resolution of 1,20,000, with normalized AGC target set to “standard” or maximum injection time of 50 ms and a scan range of 375–1,500 m/z. MS1 precursors with an intensity of >5 × 103 and charge state of 2-7 were selected for MS2 analysis. Dynamic exclusion was set to 60 s, the exclusion list was shared between CV values (Coefficient of Variation Values) and Advanced Peak Determination was set to “off”. The precursor fit threshold was set to 70% with a fit window of 0.7 m/z for MS2. Precursors selected for MS2 were isolated in the quadrupole with a 0.7 m/z window. Ions were collected for a maximum injection time of 50 ms and normalized AGC (Automatic Gain Control) target set to “standard”. Fragmentation was performed with a CID (Collision-induced dissociation) normalized collision energy of 35% and MS2 spectra were acquired in the IT at scan rate rapid. Precursors were subsequently filtered with an isobaric tag loss exclusion of TMT (Tandem Mass Tag) and precursor mass exclusion set to 18 m/z low and 5 m/z high. Precursors were isolated for an MS3 scan using the quadrupole with a 2 m/z window, and ions were collected for a maximum injection time of 86 ms and normalized AGC target of 200%. Turbo TMT was deactivated, and the number of dependent scans set to 5. Isolated precursors were fragmented again with 63% normalized HCD (Higher-energy Collision Dissociation) collision energy, and MS3 spectra were acquired in the orbitrap at 50,000 resolution with a scan range of 100-500 m/z. MS performance was verified for consistency by running a complex cell lysate quality control standard.

Biogenity (Contract Research Biotechnology Company, Aalborg, Denmark) contributed by performing the data analysis, including figures, tables, and reports. The raw files were analyzed using Proteome Discoverer 2.4 (Thermo Fisher Scientific™, St. Louis, USA). TMT reporter ion quantitation was enabled in the processing and consensus steps, and spectra were matched against the Homo sapiens database obtained from UniProt (Universal Protein Resource). Dynamic modifications were set as oxidation (M), and acetyl on protein N-termini. Cysteine carbamidomethyl (C) and TMT 16-plex (peptide N-termini and K) were set as static modifications. All results were filtered to a 1% FDR (False Discovery Rate) using the Benjamini-Hochberg procedure. Protein quantitation was done using the built-in Minora Feature Detector. Surgical and post-mortem control samples were processed using the same standardized proteomics workflow, including identical tissue lysis, digestion, TMT labeling, fractionation, and LC-MS/MS acquisition protocols. Samples were randomized across TMT channels and batches where possible, and pooled reference channels were included to enable cross-batch normalization. Quality control measures included assessment of labeling efficiency, peptide yield, reporter ion intensity distributions, and multivariate analyses to detect potential outliers or batch-driven effects. TMT reporter ion intensities were log₂-transformed prior to downstream analysis. Sample-wise normalization was applied using global median (total signal) normalization across all TMT channels to correct for differences in protein loading and labeling efficiency. Missing values were not imputed prior to differential expression analysis. Proteins were retained only if they were quantified in the majority of samples per experimental group, in order to avoid introducing an artificial signal. The data was filtered so only proteins with 2 unique peptides and at least 60% valid values across all the samples, or at least 75% within an experimental group were used in the downstream data analysis. Statistical significance was assessed by either limma (Ritchie et al., 2015) or Wilcoxon signed-rank test depending on whether the proteins demonstrated a normal distribution or not. Filtration and statistical analysis were performed in the R programming language. Differential expression analysis was conducted using the limma package.

For all datasets, gene set enrichment analysis (GSEA) was applied using ranked protein lists to detect coordinated pathway-level alterations in the proteomic dataset using the clusterProfiler R package (Korotkevich et al., 2016). Enrichment analysis was performed for Reactome pathway annotations for hippocampal, white matter and neocortical TLE tissue (Milacic et al., 2024). Pathway enrichment p-values in the GSEA Reactome analyses were adjusted for multiple testing. The corresponding enrichment tables for the most enriched pathways in hippocampal TLE tissue are provided in the Supplementary Tables 2, 3, and tables for neocortical and white matter regions can be made available upon request. For visualization of Reactome pathways enrichment results, waterfall plots were generated in R programming language using the ggplot2 and associated R packages. (Korotkevich et al., 2016).

We performed mass-spectrometry-based analysis of hippocampal, temporal neocortical, and white matter samples from four TLE patients compared to location-matched controls to discover proteomic alterations under these chronic pathophysiological disease conditions. Histological evaluation confirmed TLE pathology with hippocampal sclerosis type 1 (HS type I, international consensus classification of hippocampal sclerosis in temporal lobe epilepsy, ILAE) across all pathologic samples [Table 2, (Blümcke et al., 2013)]. Control cases had no specific neuropathology findings (Table 1). Approximately 5,800 proteins were identified within each of the comparative sample groups (HC_Ctrl vs. HC_TLE; WM_Ctrl vs. WM_TLE; NCx_Ctrl vs. NCx_TLE, Supplementary Figure 1). This indicates a comparable quality of the tested samples. For exploratory analysis, proteins with a nominal p ≤ 0.05 and an absolute log2 fold change ≥±0.5 were considered differentially expressed. A more stringent log2 fold change cut-off of ≥ ± 0.9 was applied for proteins with an adjusted p-value of ≤ 0.05. In the comparison group of the hippocampal tissue, 1193 proteins showed significantly different expression with a p-value below or equal to 0.05 and a |log2 fold change| ≥0.5. One hundred and forty nine proteins were altered showing an adjusted p-value ≤ 0.05 after Benjamini-Hochberg procedure (Figure 1; Supplementary Table 1, row 1). In the comparison group of the white matter tissue, 1134 proteins showing altered expression levels with a p-value below or equal to 0.05 and a |log2 fold change| ≥ 0.5. Eighty proteins were significantly altered following adjusting for multiple comparisons with an adjusted p-value ≤ 0.05 and |log2 fold change| ≥ 0.9 (Supplementary Table 1, row 2; Supplementary Figure 2). When comparing the expression data of neocortical tissue from TLE patients and the respective control tissue, 109 proteins were altered with a p-value below or equal to 0.05 with |log2 fold change| ≥ 0.5. Thirteen proteins were still significantly altered after adjusting for multiple comparisons with an adjusted p-value ≤ 0.05 and absolute |log2 fold change| ≥ 0.9 (Supplementary Table 1, row 3; Supplementary Figure 3).

Tables 3, 4 display up- and downregulated proteins in the hippocampal TLE tissue. Within the significantly upregulated proteins, several proteins of the complement system or immune system processes are shown to be altered in TLE (e.g., PPBP, C8B, C4B, C6, HRG, PGLYRP2, GPRC5B, ANXA1). Moreover, proteins associated with the ECM were upregulated in the epileptic hippocampal tissue, such as ECM receptor integrins (ITGA2B, ITGB3), the ECM receptor for hyaluronan CD44, as well as the ECM proteins THBS1 and TNC. Proteins expressed by astrocytes, including GFAP, CD44, and AQP4, were found to be upregulated in TLE, suggesting astrocyte activation in the epileptic hippocampus. Notably, several proteins involved in synaptic signaling and neurotransmission were downregulated in hippocampal tissue from TLE patients compared to controls. This included the expression of both glutamatergic and GABAergic receptor subunits, such as GABRA1, GRIN1, GRIN2B, GRIA1, GRIA2, GABBR2, GRM5, suggesting impaired excitatory and inhibitory neurotransmission. Additionally, proteins associated with synaptic vesicle transport, including SV2B, STXBP5L, STXBP5, SYP, SYNGAP1, SYNPR were also significantly downregulated. Key scaffolding and post synaptic density proteins such as DLG4, DLGAP1, DLGAP3, SHANK3, HOMER2, LRRTM1, SHISA6, and SHISA7 showed reduced abundance, indicating potential structural and functional disruption of synaptic structure. Furthermore, proteins involved in ion transport and calcium signaling displayed a decreased expression in TLE HC, including several potassium channel subunits (KCND2, KCND3, KCNAB1, KCNIP4). The glutamate transporter SLC17A7 (VGLUT1) was also significantly downregulated in TLE conditions in the HC. In parallel, several proteins related to mitochondrial and metabolic function (eg., NDUFS7, NDUFA3, NDUFA4) exhibited reduced expression, pointing to impaired mitochondrial function and potential oxidative stress in the epileptic HC. To annotate the most affected pathways among the differentially expressed proteins in HC TLE tissue, gene set enrichment analyses were performed (Supplementary Tables 2, 3) using Reactome pathway analysis (Korotkevich et al., 2016; Milacic et al., 2024). Among the significantly altered proteins in the epileptic HC of TLE patients, the most enriched pathways were associated with complement cascade activation and inflammatory responses, highlighting a prominent involvement of innate immune processes. In addition, the upregulation of proteins linked to extracellular matrix (ECM) organization suggests ongoing structural remodeling, potentially driven by neuroinflammatory mechanisms (Figure 2). Conversely, the most prominently downregulated pathways included those involved in chemical synaptic transmission and neuronal system functions, indicating potential impairments in synaptic activity. Pathways related to mitochondrial function were also downregulated, suggesting compromised energy metabolism and mitochondrial activity (Figure 2). However, this downregulation of synaptic and mitochondrial pathways may, at least in part, reflect neuronal cell loss characteristic of hippocampal sclerosis.

In this study, the white matter TLE tissue displayed significantly altered proteins when comparing to control cases (Supplementary Figure 2). The most relevant and significantly altered proteins are listed in Tables 5, 6. Several upregulated proteins were directly involved in synaptic transmission and plasticity, including synaptic proteins (LRRTM2, CPLX3, SYT5, SYT6, CACNA1B, NPTX2, NPTXR) and glutamatergic receptors (GRIK3, GRM2). Additionally, synapse-associated signaling molecules were upregulated (GPRIN3, PLXNA5, SEMA4A, LINGO1) in the white matter of TLE patients. Our enrichment analysis using Reactome database revealed significant upregulation of proteins associated with the neuronal system, including transmission across chemical synapses, protein-protein interactions at synapses, and neurotransmitter signaling processes. In contrast, proteins annotated to the pathway ECM organization, including structural ECM components such as ACAN, VCAN, TNC, FBLN1, LAMA2, LAMB2, SPOCK, SPOCK1), and integrin-mediated ECM receptors (ITGB1, ITGA7), were downregulated. Moreover, complement and immune system associated proteins also displayed downregulated expression in white matter TLE tissue (eg., S100A8, S100A9, PRNP, C2, CFI, CFH, C3, C5, C6) (Figure 3). These findings suggested enhanced synaptic activity and potential synaptogenesis in the white matter of TLE patients, accompanied by ECM remodeling. Notably, these molecular alterations oppose the signatures observed in the hippocampal tissue, highlighting region-specific pathomechanisms in TLE.

In the temporal neocortex (NCx) of TLE patients, only few proteins were significantly altered (Tables 7, 8, Supplementary Figure 3). Notably, we observed downregulation of pathways related to ECM organization (eg., MMP9, HAPLN4) and the immune system, including components of the complement cascade (such as S100A8, S100A9, C1QB), consistent with the patterns seen in the white matter tissue (Figure 4). A few synapse-associated proteins displayed nominal upregulation (e. g. NPTXR, SYT17, GRIK3, NPTX2, SYT6, DLG2). Overall, these limited protein changes provide only modest biological insight, suggesting that the NCx is less affected by epileptogenic processes and may exhibit compensatory responses.

In summary, our findings highlight the potential role of innate immune activation in the pathophysiology of TLE, as evidenced by the upregulation of complement components C8B, C4B, and C6. The data further suggest that neuroinflammatory processes may drive pathological remodeling of the ECM, including increased expression of integrins, CD44, thrombospondin-1 (THBS1), and tenascin C (TNC). These changes likely compromise the structural and functional integrity of hippocampal networks, as reflected by downregulation of key synaptic proteins such as synaptophysin (SYP), HOMER2, and neurotransmitter receptors, e. g. GABA and glutamate receptors. Proteomic profiling underscores the HC as central region involved in seizure generation, characterized by upregulation of astrocytic, complement, and ECM-related proteins alongside a reduction in synaptic and mitochondrial proteins. Collectively, these alterations point to a pathological cascade in which astrocytic activation and complement-mediated inflammation promote ECM remodeling, synaptic disruption, and impaired energy metabolism–interconnected mechanisms that may contribute to TLE pathology. Given the exploratory nature of this study, the findings should be interpreted with caution. Further experimental validation is essential to confirm the functional relevance of the identified proteomic changes and to evaluate their potential as biomarkers or therapeutic targets in TLE.

The present study provides valuable insights into the proteomic alterations associated with TLE, with a particular emphasis on neuroinflammation, ECM, and synaptic changes, all of which are likely relevant for epilepsy pathogenesis. However, several limitations must be acknowledged. First, the relatively small sample size limits the statistical power and generalizability of the findings, as access to human tissue suitable for high-sensitivity mass spectrometry analysis is limited. Second, control tissues were obtained postmortem, whereas TLE tissues were surgically resected. While efforts were made to minimize batch effects, differences in tissue source, postmortem delay, and processing could introduce variability. Third, slight age differences between control and TLE samples may introduce variability, and obtaining perfectly age-matched, neuropathology-free controls is challenging. Furthermore, the high degree of individual variability, with each sample possessing a unique molecular signature, adds complexity to direct comparisons and may influence the observed proteomic changes. Fourth, the proteomic data are correlative, and causal relationships between inflammation, ECM remodeling, and synaptic and mitochondrial alterations cannot be inferred. Downregulation of synaptic and mitochondrial proteins in the hippocampus may partly reflect neuronal loss due to hippocampal sclerosis rather than direct molecular downregulation. Finally, no validation using additional techniques was performed to confirm the proteomic findings. A limitation of pathway enrichment analyses is their reliance on curated databases that annotate proteins based on their shared molecular functions and interaction networks rather than tissue specificity or disease context. As a result, pathway enrichment analyses may yield terms that appear unexpected in brain tissue. Importantly, presence of such pathways does not imply direct involvement of these processes in epilepsy but rather reflects overlapping molecular components that also participate in neuroinflammatory signaling, or cellular stress response within the central nervous system. Therefore, our discussion focuses on biologically relevant processes in CNS pathology and epileptogenesis, including neuroinflammation, ECM, synaptic dysfunction, and mitochondrial impairment. Despite these limitations, this study offers novel perspectives into inflammatory-driven ECM and synaptic remodeling processes in epilepsy. The application of ultra-sensitive mass spectrometry enabled an in-depth proteomic analysis of human tissue, and the examination of different brain regions provided valuable information on region-specific alterations in TLE. Future studies with larger cohorts and complementary validation techniques will be essential to confirm these findings and further elucidate our understanding of molecular mechanisms underlying ECM and synaptic changes in TLE.