Fresh-frozen brain tissue samples were obtained from the temporal neocortex of 6 adults with pharmacoresistant TLE who underwent left (language-dominant) temporal lobe resections for treatment of their epilepsy and who completed comprehensive neuropsychological evaluations, including assessment of episodic memory, prior to surgery. Patients were 41 years of age on average (SD = 10) with 14 years of education (SD = 2). All patients self-identified as White, non-Hispanic, and half the sample was female. Mean age at seizure onset was 27 years (SD = 14), and mean duration of epilepsy was 14 years (SD = 8). Tissue specimens and clinical data were obtained from IRB-approved epilepsy data registries at the Cleveland Clinic, and the methods were performed in accordance with relevant guidelines and regulations approved by the Cleveland Clinic Institutional Review Board. All patients provided written informed consent.

All patients completed measures of verbal episodic memory as part of preoperative neuropsychological evaluations a median of 6 months before surgery. Story recall was assessed with the Logical Memory subtest of the Wechsler Memory Scale – Third or Fourth Edition, and word-list learning was assessed with the Rey Auditory Verbal Learning Test. These measures were scored using demographically-corrected norms and transformed into standard scores (SS; mean = 100, SD = 15). Patients were separated into one of two memory phenotypes based on a mean composite delayed memory score (combined delayed story recall and word-list learning tasks). Specifically, mean scores <85 were classified as “impaired” memory (n = 4) and ≥85 were classified as “intact” memory (n = 2). As intended, memory scores were significantly lower in the impaired memory group (SS = 57) compared to the intact memory group (SS = 95).

Approximately 20 mg of fresh-frozen tissue (predominantly gray matter) from resected temporal neocortex was used from each patient to prepare the nuclei suspension. Our snRNA experiments were conducted using the “nuclei village” approach, a multiplex analysis of nuclei sampled from brain specimens from multiple donors simultaneously19. The nuclei were extracted, processed, and analyzed together, facilitating rigorous cross-sample comparisons. In subsequent computational analysis, combinations of hundreds of transcribed SNPs (for which alleles are ascertained in the RNA data) were used to assign each nucleus to the patient-of-origin. After treating the patient nuclei pool with myelin removal beads (Miltenyi Biotec, Bergisch Gladbach, Germany), 2X the standard number of 10X snRNA-Seq nuclei were loaded into 8 reactions (32,000 nuclei per reaction). The 8 reactions were sequenced on a NovaSeq S4 flow cell (Illumina, San Diego, CA, USA). After aligning to human reference GRCh38 with non-canonical contigs masked out, the libraries were run through CellBender to remove technical artifacts20. Then, high-quality cells were selected based on a combination of the number of unique molecular identifiers (UMIs, at least 400) and the percent of intronic reads (at least 40%). An average of 6486 nuclei per donor were identified. Nuclei from different samples were integrated and clustered using the Seurat package21. The clusters were then annotated using scPred22 and visualized with the t-SNE plots.

Differential expression analyses were conducted between subjects with impaired and intact memory per cell class, controlling for age, sex, and presence/absence of mesial temporal sclerosis, using the MAST package23. The following cell classes were examined: astrocytes, endothelia, GABAergic neurons, glutamatergic neurons, microglia, oligodendrocytes, and polydendrocytes. Pathway analyses were first performed with Kyoto Encyclopedia of Genes and Genomes (KEGG)24 using all DEGs. Then, a more stringent and comprehensive analysis, including only those DEGs with log2fold changes >1 or <−1, was performed with Ingenuity Pathway Analyses (IPA, QIAGEN). Enrichment for diseases and functions was also examined using IPA.

Baseline descriptive statistics stratified by memory group (intact versus impaired) were calculated. Independent samples t-tests or Fisher’s exact tests were used to examine group differences on demographic and disease-related variables. P values of <0.05 were considered statistically significant. For differential expression analysis, pathway analysis and enrichment of diseases and functions, adjusted P values of <0.05 were considered statistically significant.

Further information on research design is available in the Nature Research Reporting Summary linked to this article.