Abstract Malformations of cortical development are manifestations of mTORopathies, including tubers in context of Tuberous Sclerosis Complex (TSC) cortical tubers and Focal Cortical Dysplasia (FCD), and are associated with epilepsy, often accompanied by comorbidities such as autism spectrum disorder (ASD). This study aims to investigate the cell-type-specific transcriptional alterations and disrupted intercellular communication networks in mTORopathies, focusing on their implications for cortical network dysfunction. Using single-cell RNA sequencing, we identified 33 transcriptionally distinct cell clusters across control and pathological samples, including neuronal, glial, and endothelial populations. Our analysis revealed disease-specific changes, such as the loss of certain glutamatergic and microglial clusters in cortical tubers (TSC), MTOR_FCD and DEPDC5_FCD, and the presence of a unique endothelial cluster in pathological samples. Pathway enrichment analysis highlighted the critical role of synaptic signaling, axonogenesis, and neuroimmune regulation in these disorders. Additionally, cell–cell communication network analysis demonstrated disrupted interactions between neuron-astrocyte, astrocyte-OPC, and microglia-neuron across mTORopathies. We found that the neurexins-neuroligins (NRXN-NLGN) signaling pathway, crucial for synapse formation and stability, was altered in both glutamatergic and GABAergic neurons, reflecting dysregulated synaptic plasticity and impaired neuron-glia communication. These findings provide novel insights into the molecular underpinnings of mTORopathies and suggest potential therapeutic targets to restore cellular communication and synaptic function in these disorders. Supplementary Information The online version contains supplementary material available at 10.1186/s40478-025-02113-w. Keywords: Transcriptomic profiling, Extracellular matrix, Ligand-receptor inference, MTORopathies Introduction The mTOR pathway has gained increasing recognition as a key player in the pathogenesis of a spectrum of malformations of cortical development (MCD) collectively referred to as mTORopathies [[36]19, [37]28, [38]46]. These MCDs result from mutations in genes within the PI3K-mTOR pathway and the GATOR1 complex, leading to hyperactivation of the mechanistic target of rapamycin complex 1 (mTORC1) during cortical development [[39]20, [40]28, [41]60]. This hyperactivation causes alterations in cell size, cortical lamination, and axonal and dendritic growth, contributing to the formation of a pathological network that supports seizure development, progression, and associated comorbidities [[42]6, [43]20, [44]28, [45]49]. Among the most extensively studied mTORopathies associated with drug-resistant epilepsy are tuberous sclerosis complex (TSC) and focal cortical dysplasia (FCD) type II with mTOR related mutations [[46]20, [47]46, [48]47]. The dysregulation of mTOR activity in both conditions has led to a growing interest in understanding their overlapping mechanisms, while also unraveling the unique molecular underpinnings that distinguish them [[49]50]. TSC is a genetic disorder caused by loss-of-function mutations in either the TSC1 or TSC2 genes, which encode hamartin and tuberin, respectively [[50]20, [51]52]. These proteins form a complex that acts as a negative regulator of the mTORC1 pathway. In the absence of functional TSC1 or TSC2, mTORC1 activity is constitutively elevated, leading to excessive cellular growth and the formation of benign tumors, in multiple organs, including the brain [[52]20]. The neurological manifestations of TSC, such as epilepsy, are considered primarily driven by cortical tubers, regions of malformed cortex that exhibit disorganized neurons, astrogliosis, and loss of normal cortical architecture [[53]6, [54]20]. In contrast, FCD is a heterogeneous group of developmental brain malformations often associated with somatic mutations in genes regulating the mTOR pathway [[55]28, [56]36, [57]60]. One of the most implicated genes in FCD is DEP Domain-Containing Protein 5 (DEPDC5), which encodes a subunit of the GATOR1 complex, a key upstream inhibitor of mTORC1 [[58]7, [59]31, [60]60]. Mutations in DEPDC5, as well as in other mTOR pathway regulators such as mechanistic target of rapamycin (MTOR), lead to the hyperactivation of mTORC1 signaling, resulting in FCD with the histopathological features of FCD type IIa or IIb [[61]47]. Although both TSC- associated cortical lesions and FCD share mTOR hyperactivation as a common feature, the specific molecular consequences, including the converging and diverging mechanisms, as well as the cellular contexts in which these mutations exert their effects, remain incompletely understood, particularly regarding their roles in epileptogenesis. Cell–cell communication plays a pivotal role in maintaining the functional integrity of neural networks [[62]23]. Neurons, glial cells, and other brain-resident cell types rely on intricate signaling mechanisms to coordinate their activities, regulate synaptic plasticity, and maintain homeostasis [[63]1, [64]33]. Dysregulation of these communication networks can have profound effects on brain function, leading to aberrant synaptic connectivity [[65]39, [66]67], impaired neuronal excitability [[67]43, [68]67], and altered network dynamics [[69]14, [70]55]. In cortical tubers (TSC) and FCD, hyperactivation of the mTOR pathway likely disrupts intercellular signaling by altering the expression and secretion of key signaling molecules, including growth factors, cytokines, and extracellular matrix proteins [[71]10, [72]64, [73]71]. These alterations may not only impact cell-autonomous processes but also influence neighboring cells, thereby amplifying pathological network changes. Understanding how mTOR-driven mutations perturb cell–cell communication networks is, therefore, critical for elucidating the mechanisms underlying cortical dysplasia and epilepsy in these disorders. Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool to dissect cellular heterogeneity in complex tissues and offering unprecedented resolution to study the molecular landscapes of brain disorders [[74]56]. By capturing the transcriptional profiles of individual cells, scRNA-seq can reveal cell type-specific gene expression patterns and identify rare cell populations that may be critical drivers of disease [[75]42, [76]56, [77]61]. In the context of cortical tubers (TSC) and FCD, scRNA-seq provides an opportunity to explore how different mTORopathies have altered transcriptomic architecture in the brain and how these alterations contribute to epilepsy. Given the involvement of multiple cell types, including neurons, astrocytes, and microglia, in the pathogenesis of both disorders, an in-depth single-cell analysis is crucial for unraveling the complex molecular networks at play. In this study, we utilized scRNA-seq to map the transcriptomic landscape of brain tissue from patients with cortical tubers (TSC) and FCD type II, carrying mTOR or DEPDC5 mutations. Our goal is to compare the gene expression profiles of individual cell types across these two conditions, identify disease-specific and shared transcriptional signatures, and investigate dysregulated pathways that may drive epileptogenesis. Additionally, we aim to explore cell–cell communication and its dysregulation in cortical tubers (TSC) and FCD by inferring intercellular communication networks from ligand-receptor interactions. This approach is crucial for understanding how hyperactivation of the mTOR pathway contributes to pathological cellular interplay, potentially uncovering novel mechanisms of disease and identifying therapeutic targets to restore normal network function. Materials and methods Sample collection Study cohort Surgical and postmortem brain tissues were selected from the archives of the Department of Neuropathology of the Amsterdam UMC (Amsterdam, The Netherlands) and the UMC Utrecht (Utrecht, The Netherlands). Cortical brain samples were obtained from patients undergoing surgery for intractable epilepsy and diagnosed with TSC (n = 5 cortical tubers) or focal cortical dysplasia (MTOR mutation: n = 5, DEPDC5 mutation: n = 3). In addition, age- and tissue-matched autopsy control samples were collected (n = 3) from individuals without a history of seizures or other neurological disease. All autopsies were performed within 9 h after death. Tissue was obtained and used in accordance with the Declaration of Helsinki and the Amsterdam UMC Research Code provided by the Medical Ethics Committee and according to the Amsterdam UMC and UMC Utrecht Biobank Regulations (W21-295; 21–174). Clinical information about the brain samples is summarized in supplementary Table [78]1. Single-cell RNA sequencing Control, cortical tubers (TSC), and FCD single-cell RNA sequencing dataset Single-cell RNA sequencing was performed at Single Cell Discoveries ([79]https://www.scdiscoveries.com/) according to the 10 × genomics Chromium Single Cell Gene Expression Flex protocol. Prior to loading the samples, the frontal cortex tissue was cut into slices, which were fixed, and cells were extracted. Cells were counted to ensure quality control. For each sample, 8000 cells were loaded, and the resulting sequencing libraries were prepared following a standard 10 × Genomics protocol. Bioinformatics pipeline Processing of scRNA-seq data Sample reads were aligned to the human genome GRCh38.p14 using Cell Ranger. Filtering of empty barcodes was carried out using Cell Ranger. The data from all samples were loaded into R (version 4.3.1) and processed using the Seurat package (version 5.0.1). More specifically, cells with at least 1000 UMIs per cell and less than 5% mitochondrial gene content were retained for analysis. The data of all 10 × libraries were merged and processed together. The merged dataset was normalized for sequencing depth per cell and log transformed. After filtering, datasets were integrated using reciprocal principal component analysis (RPCA). The integrated data was subsequently scaled, and dimensionality reductions were performed. Clustering was computed using the FindNeighbors function (dims = 1:20), and FindClusters at a resolution of 0.5. To identify the cell types in separate clusters, marker genes for each cell type were used (Supplementary Table [80]2). Next, to identify genes enriched in each cluster, we utilized the FindAllMarkers function (Supplementary Table [81]3). This function uses the Wilcoxon rank-sum test, to determine genes that are significantly enriched in each cluster. To perform differential expression analysis between control and cortical tubers (TSC), and FCD samples, we performed pseudo-bulk analysis. This approach involves aggregating reads from the cell-type of interest within each biological sample to create'pseudobulks'. Differential expression analysis was performed using the R package DESeq2 [[82]41]. To control the false discovery rate, we applied the Benjamini–Hochberg correction, considering gene expression changes with an adjusted p-value < 0.05 as statistically significant. Differential expression analysis Differentially expressed genes (DEGs) identified were subjected to GeneOntology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. GO enrichment was performed to classify the DEGs into functional categories, including biological processes, molecular function, and cellular components, using the R package clusterProfiler. KEGG enrichment analysis was carried out to identify enriched signaling and metabolic pathways. Adjusted p-values were calculated using the Benjamini–Hochberg procedure to control for multiple testing. Cell–cell communication analysis To infer intercellular communication networks, the CellChat R package [[83]32] was applied to the scRNA-seq data. Pre-processed scRNA-seq data, including normalized counts and cell type annotations, were used as input. CellChat uses a curated database of known ligand-receptor pairs (CellChatDB.human). Each pair consists of a ligand expressed by one cell type and a receptor expressed by another. For each cell type, CellChat identifies ligands and receptors that are significantly expressed. To determine which ligand-receptor pairs are likely to be actively involved in communication, the algorithm tests whether the ligand is overexpressed in the sender cell types. Similarly, the receptor’s overexpression is tested in the receiver cell type. Once probabilities are calculated for all ligand-receptor pairs, CellChat aggregates these probabilities to infer the activity of signaling pathways. Pathway activity scores are derived by summing the probabilities of all pairs associated with a given pathway. Additionally, differential CellChat analysis was performed to compare signaling interactions across different experimental conditions. Immunohistochemistry Human brain tissue fixed in 10% buffered formalin and embedded in paraffin was mounted on pre-coated glass slides (Star Frost, Waldemar Knittel, Braunschweig, Germany). Sections were deparaffinized in xylene, rinsed in ethanol (100%, 100%, 96%). Antigen retrieval was performed using a pressure cooker in 0.01 M sodium citrate buffer (pH 6.0) at 120 °C for 10 min. Slides were cooled in ice water for 15 min, washed with phosphate‐buffered saline (PBS, pH 7.4) and incubated overnight at 4 °C with a primary antibody against NRXN1 (1:500, Sigma #ABN161-I) or NLGN2 (1:500, SYSY #129,511). After that, sections were washed in PBS and incubated with corresponding secondary antibodies using a polymer-based peroxidase immuno cytochemistry detection kit (Brightvision plus kit, ImmunoLogic, Duiven, the Netherlands). After washing, the sections were stained with 3,3′-diaminobenzidine tetrahydrochloride (0.5 mg/ml DAB, Sigma-Aldrich, St. Louis, MO, USA) in the presence of hydrogen peroxide in Tris–HCl buffer (50 mM, pH 7.6). The sections were counterstained with hematoxylin, dehydrated in alcohol and xylene, and cover slipped. Results Classification and definition of cell types based on scRNA-seq To better define the cellular heterogeneity in human mTORopathies, we performed scRNA-seq experiments on material from control, cortical tubers (TSC) and FCD pathology with either DEPDC5 or MTOR mutations, using the 10 × Genomics flex assay platform (Fig. [84]1a). After data processing and quality controls of the scRNA-seq profiles, we obtained transcriptomic profiles for a total of 92,394 cells. Dimensional reduction and unsupervised clustering identified 33 transcriptionally distinct cell clusters, as shown by the uniform manifold approximation and projection (UMAP) (Fig. [85]1b). GABAergic neurons comprised three clusters, defined by the expression of GAD1, GAD2, PVALB, VIP, SST, LAMP5 and RELN, while glutamatergic neurons were split into nine distinct clusters, marked by SLC17A7, SATB2, SLC17A6, and GRIN1. Astrocytes were represented by seven clusters, characterized by the markers ALDH1L1, AQP4, SLC1A2, and S100B. Four microglia clusters were identified, defined by PTPRC, ITGAM, P2RY12, and TMEM119. Oligodendrocytes were similarly grouped into four clusters, with marker genes OLIG1, CLDN11, MOG, and MBP. Oligodendrocyte progenitor cells (OPCs) consisted of two distinct clusters, identified by PDGFRA and CSPG4. Endothelial cells were also represented by three clusters, defined by the expression of CLDN5 and PECAM1. Lastly, one cluster was identified as pericytes, defined by the expression of PDGFRB and RGS5 (Fig. [86]1c and d). Fig. 1. [87]Fig. 1 [88]Open in a new tab Overview of experimental workflow and single-cell RNA sequencing analysis of cortical tubers (TSC) and FCD tissues. a Schematic representation of the experimental workflow: brain tissue samples were fixed, permeabilized, and hybridized with specific probes for transcriptomic analysis, followed by sequencing and data analysis. b UMAP visualization of single-cell RNA sequencing data showing clustering of cells into 32 distinct clusters. c Heatmap of marker gene expression used to identify major cell types. Each row represents a marker gene, and each column corresponds to the cell type, with color intensity indicating normalized expression levels. Key cell types include oligodendrocytes, microglia, OPCs (oligodendrocyte precursor cells), GABAergic neurons, glutamatergic neurons, and endothelial cells. d UMAP visualization of annotated cell type identities based on marker gene expression. e UMAP visualization of cells grouped by conditions (Control, FCD_DEPDC5, FCD_mTOR, and cortical tubers (TSC)), illustrating the distribution of clusters across different conditions. We next aimed to investigate whether there were differences in the cellular clusters between control samples and the three pathologies (cortical tubers: TSC, FCD_mTOR, and FCD_DEPDC5). We identified one cluster of glutamatergic neurons that was exclusively present in the control samples and FCD_mTOR but absent in cortical tubers (TSC) and FCD_DEPDC5. Additionally, one astrocyte cluster was detected only in control, FCD_mTOR and cortical tubers (TSC) but was missing in FCD_DEPDC5. Interestingly, a cluster of microglia was found solely in the control samples, with no representation in any of the pathological conditions. Furthermore, we uncovered a distinct cluster of endothelial cells present across all three pathologies but absent in the control samples (Fig. [89]1e). Characterization of pathology-specific cellular clusters Next, we aimed to investigate the pathology-specific cellular clusters that were previously identified, by looking at the genes that were expressed highly in these clusters compared to all other clusters. By examining these clusters that were uniquely present or absent, we aimed to identify specific molecular mechanisms underlying cortical tubers (TSC), FCD_mTOR and FCD_DEPDC5. Glutamatergic neurons – Cluster 21—Cluster 21, identified in control samples and FCD_mTOR, but absent in cortical tubers (TSC) and FCD_DEPDC5, had distinct expression of genes such as RAX2, SLC12A3, SLC22A7 and PRAP1. GO and KEGG pathway analyses revealed enrichment in processes and pathways critical for neuronal signaling, including synapse organization, calcium signaling, and glutamatergic synapse activity (Fig. [90]2A and [91]B). Fig. 2. [92]Fig. 2 [93]Open in a new tab Gene Ontology (GO) and KEGG pathway enrichment analyses for differentially expressed genes in clusters 21, 29, 22 and 31. a, b GO and GO and KEGG enrichment results for Cluster 21, a glutamatergic neuron population enriched in control samples but largely absent in cortical tubers (TSC) and FCD. GO terms highlight processes such as regulation of trans-synaptic signaling, synapse organization, and calcium ion transmembrane transport, while KEGG pathways include calcium signaling, glutamatergic synapse, and oxytocin signaling pathways. c, d GO and KEGG enrichment results for Cluster 29, a microglial population exclusively present in control samples. GO terms are associated with synapse organization, axon development, and modulation of synaptic transmission, while KEGG pathways include Rap1 signaling, cell adhesion molecules (CAMs), and GABAergic synapse pathways. e, f GO and KEGG enrichment results for Cluster 22, an astrocyte population absent in FCD_DEPDC5 but present in cortical tubers (TSC) and FCD_mTOR. Enrichment analysis reveals involvement in synaptic signaling, axonogenesis, and glutamatergic synapse pathways, along with protein processing in the endoplasmic reticulum and axon guidance. g, h GO and KEGG enrichment results for Cluster 31, an endothelial cell population present in all pathological conditions (FCD_DEPDC5, FCD_mTOR, cortical tubers (TSC)) but absent in controls. Key GO terms include modulation of synaptic signaling and axonogenesis, with KEGG pathways enriched for MAPK signaling, glutamatergic synapse, and retrograde endocannabinoid signaling Microglia – Cluster 29—Similar to cluster 21, cluster 29 was found exclusively in control samples and represents microglia involved in synapse organization, axon development and neuroimmune regulation. GO and KEGG analyses highlighted roles in Rap1 signaling and cell adhesion molecules (CAMs), pathways essential for synaptic maintenance and neuron-glial interactions (Fig. [94]2C and [95]D). Astrocytes – Cluster 22—Cluster 22 is an astrocyte population that is absent in FCD_DEPDC5 but present in cortical tubers (TSC) and FCD_mTOR as well as controls. This cluster was enriched for processes such as transsynaptic signaling, axonogenesis, and glutamatergic synapse activity (Fig. [96]2E and [97]F). Endothelial cells – Cluster 31—Cluster 31 is the only cluster that is present in all pathological samples (FCD_DEPDC5, FCD_mTOR, and cortical tubers (TSC)) but absent in controls. This cluster is again associated with synaptic organization, axonogenesis, and synaptic signaling pathways (Fig. [98]2G and [99]H). The consistent presence of these endothelial cells across disease states highlights their potential role in maintaining pathological neural connectivity and differentiating malformed cortex from healthy controls. Differentially expressed genes reveal shared and pathology-specific mechanisms in cortical tubers and FCD Following our previous analysis of cellular clusters, this section focuses on the differential expression analysis across various cell types in cortical tubers (TSC) and FCD. We investigated the transcriptomic profiles of different major cell populations, including neurons, astrocytes, microglia and oligodendrocytes, to identify the specific molecular alterations associated with each pathology. In this analysis, we focused on both shared and distinct mechanisms for each pathology. GABAergic neurons—In GABAergic neurons (supplementary Fig. [100]1a), cortical tubers (TSC) exhibited the highest numbers of dysregulated genes with a total of 1660 differentially expressed genes (DEGs), compared to 275 DEGs in FCD_mTOR and 461 DEGs in FCD_DEPDC5. 123 genes overlapped between cortical tubers (TSC) and FCD subtypes, while the majority were uniquely dysregulated in cortical tubers (TSC). Pathway enrichment analysis of the overlapping 123 genes revealed significant involvement in protein homeostasis mechanisms, including “protein folding,” “response to unfolded protein,” and “chaperone-mediated protein folding” (supplementary Fig. [101]2a). Glutamatergic neurons—Glutamatergic neurons (supplementary Fig. [102]1b) demonstrated a similar trend, with cortical tubers (TSC) showing the largest set of dysregulated genes (1597 DEGs). FCD_mTOR and FCD_DEPDC5 exhibited smaller DEG sets (519 and 816, respectively. Overlaps between cortical tubers (TSC) and FCD subtypes consisted of 198 DEGs, indicating largely distinct transcriptomic profiles across pathologies. Enrichment analysis of the shared genes identified pathways related to developmental processes, including “kidney development” and “renal system development,” suggesting a role for pleiotropic developmental genes in excitatory neurons. Additionally, the “neuroactive ligand-receptor interaction” pathway highlighted potential alterations in receptor signaling, which may impact excitatory neurotransmission and contribute to network hyperexcitability (supplementary Fig. [103]2b). Astrocytes—Astrocytes (supplementary Fig. [104]1c) displayed the most extensive DEG set in cortical tubers (TSC) with 2919 DEGs. FCD_mTOR and FCD_DEPDC5 only showed 178 and 126 DEG, respectively. Notably, the overlap between cortical tubers (TSC) and FCD subtypes remained limited with only 30 DEGs overlapping across all pathologies, suggesting that astrocyte-specific dysregulation in cortical tubers (TSC) is significantly distinct. Pathway analysis of the shared genes revealed enrichment in “complement and coagulation cascades,” implicating immune-related pathways and neuroinflammatory responses in cortical tubers (TSC) and FCD. Additionally, disruptions in ECM organization and protein processing were identified (supplementary Fig. [105]2c). Microglia—Microglia (supplementary Fig. [106]1d) followed the same pattern, with cortical tubers (TSC) having the largest number of DEGs (2519 DEGs), far exceeding the number of DEGs identified in FCD_mTOR (211 DEGs) and FCD_DEPDC5 (569 DEGs). A total of 88 genes overlapped between the three pathologies. Enrichment analysis highlighted pathways related to protein homeostasis, including “protein folding” and “response to unfolded protein.” KEGG analysis identified additional pathways, such as “NF-kappa B signaling,” “lipid and atherosclerosis,” and “glycerolipid metabolism” (supplementary Fig. [107]2d). Oligodendrocytes—In oligodendrocytes (supplementary Fig. [108]1e), cortical tubers (TSC) again had the largest set of DEGs (1269 DEGs), whereas FCD_mTOR and FCD_DEPDC5 showed smaller DEG sets (80 and 242 DEGs, respectively). In contrast to other cell types, there was no overlap in DEGs between the three pathologies, indicating pathology-specific transcriptomic changes in oligodendrocytes. Across all cell types, cortical tubers (TSC) consistently showed the highest number of dysregulated genes compared to FCD_mTOR and FCD_DEPDC5. The overlap between the three pathologies was minimal, with most genes being uniquely dysregulated in each pathology. These findings suggest that while there are shared molecular mechanisms underlying the pathologies, the transcriptomic landscapes are predominantly distinct, particularly in cortical tubers (TSC). Dysregulation of cell–cell communication networks in mTORopathies Cell–cell communication plays a crucial role in the functioning of neural networks, and its dysregulation has been increasingly implicated in the pathophysiology of epilepsy. Effective intercellular signaling is essential for maintaining neural network homeostasis, synchronizing activity, and regulating excitatory-inhibitory balance. Given the pivotal role of these networks in epilepsy, we aimed to investigate the characteristics of cell–cell communication across mTORopathies. Using our scRNA-seq data and computational inference of ligand-receptor interactions, we quantified the number and strength of interactions for distinct cell populations within each pathology. Interaction number refers to the total quantity of potential ligand-receptor interactions, while interaction strength reflects the cumulative signaling intensity of these interactions. Comparative analysis revealed that FCD_DEPDC5 exhibited a higher overall level of intercellular communication compared to controls, FCD_mTOR and cortical tubers (TSC). Notably, all three mTORopathies displayed elevated levels of both interaction numbers and interaction strength relative to controls, reflecting a general upregulation of cellular signaling within pathological tissues (Fig. [109]3A and [110]B). Despite this overall increase, distinct patterns of dysregulated communication were observed in each mTORopathy, highlighting disease-specific alterations in intercellular signaling networks. In cortical tubers (TSC), the number of interactions involving microglia (sender), neurons, and oligodendrocyte precursor cells (OPCs) was reduced, suggesting impaired crosstalk between these cell types. Additionally, interaction strength between astrocytes (sender), OPCs and endothelial cells was reduced, indicating weakened signaling pathways among these populations (Fig. [111]3C). In FCD_mTOR, we observed a reduction in both the number and the strength of astrocyte interactions, including reduced signaling among astrocytes and between astrocytes (sender), OPCs, and endothelial cells (Fig. [112]3D). This pattern suggests a disruption in astrocyte-mediated communication networks, which are essential for maintaining neural homeostasis and supporting the vasculature. FCD_DEPDC5 showed a reduction in the number of interactions not only among astrocytes, but also between microglia and astrocytes, suggesting an altered neuroimmune signaling axis. Furthermore, interaction strength was decreased among astrocytes, between astrocytes and OPCs in both directions, and among OPCs (Fig. [113]3E). These findings suggest a weakening of astrocyte and OPC communication in FCD_DEPDC5, potentially affecting processes such as myelination and the response to injury. Fig. 3. [114]Fig. 3 [115]Open in a new tab Comparison of cellular interactions across control and mTORopathies. a Bar plots illustrating the total number of inferred interactions (left) and interaction strength (right) across the control, cortical tubers (TSC), FCD_mTOR (focal cortical dysplasia with MTOR mutation), and FCD_DEPDC5 (focal cortical dysplasia associated with DEPDC5 mutation) groups. b Interaction networks visualizing the connections among various cell types, including glutamatergic neurons, GABAergic neurons, astrocytes, oligodendrocytes, microglia, OPCs (oligodendrocyte precursor cells), and endothelial cells. Networks are shown separately for the control, cortical tubers (TSC), FCD_mTOR, and FCD_DEPDC5 groups, with node size and edge thickness reflecting interaction strength and number of interactions. c–e Heatmaps comparing the relative interaction differences compared to control (left panels) and differential interaction strengths (right panels) among cell types for cortical tubers (TSC) (c), FCD_mTOR (d), and FCD_DEPDC5 (e) relative to controls. Rows and columns represent cell types, and the color gradient reflects the change in number of predicted ligand-receptor interactions (left panels), and interaction strength (right panels), red reflects increase and blue reflects decrease To further investigate the cell-type-specific signaling changes driving network dysregulation in cortical tubers (TSC), FCD_mTOR, and FCD_DEPDC5, we analyzed the interaction strengths between cell types. By comparing incoming and outgoing interaction strengths across control and pathological conditions, this analysis aimed to uncover alterations in cell-specific communication patterns that contribute to the distinct pathophysiological landscapes of these disorders. Analysis of incoming and outgoing interaction strengths across cell types revealed distinct changes in signaling patterns among cortical tubers (TSC) (Fig. [116]4b), FCD_mTOR (Fig. [117]4c), and FCD_DEPDC5 (Fig. [118]4d) compared to controls (Fig. [119]4a). Interaction strengths observed under control conditions served as a reference baseline, with oligodendrocyte precursor cells (OPCs) exhibiting the highest interaction levels, followed by astrocytes and glutamatergic neurons. In cortical tubers (TSC), astrocytes displayed increased incoming and outgoing interaction strengths, suggesting enhanced signaling activity, while microglia remained the least interactive. FCD_mTOR showed a similar pattern to cortical tubers (TSC), with glutamatergic neurons and astrocytes exhibiting elevated interaction strengths. However, in FCD_DEPDC5, astrocytes exhibited relatively higher incoming interaction strengths than outgoing, indicating a potential shift toward compensatory responses, where they primarily receive signals rather than broadcast them. Fig. 4. [120]Fig. 4 [121]Open in a new tab Comparative interaction strengths across cell types in control, cortical tubers (TSC), FCD_mTOR, and FCD_DEPDC5. a–d scatter plots display incoming interaction strength (y-axis) versus outgoing interaction strength (x-axis) for various cell types: glutamatergic neurons (green), GABAergic neurons (pink), astrocytes (purple), oligodendrocytes (red), OPCs (orange), endothelial cells (gray), and microglia (blue). The size of the dots represents interaction count. Control conditions exhibit balanced interaction strengths, while pathological conditions show distinct shifts in astrocytic and neuronal interaction patterns. Interaction strengths in cortical tubers (TSC) and FCD_mTOR reveal enhanced activity in astrocytes and neurons, while FCD_DEPDC5 shows disproportionate incoming signaling in astrocytes Cell-type specific interaction patterns and their contribution to cortical network dysregulation To better understand how altered cell–cell communication contributes to cortical network dysregulation in mTORopathies, we next investigated ligand-receptor pair-specific interaction strengths for key cell types. Specifically, we focused on astrocytes, glutamatergic neurons, and GABAergic neurons, which exhibited the most pronounced shifts in communication patterns in our previous analysis. By comparing both incoming and outgoing interaction strengths of ligand-receptor pathways across each pathology relative to controls, we aimed to identify disease-specific dysregulation of synaptic, adhesion, and extracellular matrix signaling. In cortical tubers (TSC), we found that signaling molecules essential for synaptic adhesion and communication, including NRXN, NCAM, and CADM, were upregulated in glutamatergic (Fig. [122]5d) and GABAergic neurons (Fig. [123]5g), both in incoming and outgoing interaction strengths. This upregulation suggests enhanced synaptic connectivity or hyperactivation within neuronal networks. Specifically, increased NRXN signaling likely reflects heightened synaptic activity or formation, while elevated NCAM and CADM suggest excessive plasticity and stronger synaptic adhesion, respectively. These changes are consistent with hyperexcitability and dysregulated circuits underlying epilepsy and cognitive dysfunction in cortical tubers (TSC). Conversely, astrocytes in cortical tubers (TSC) (Fig. [124]5a) exhibited reduced signaling for NRXN, NCAM, and CADM, highlighting impaired neuron-astrocyte communication and weakened astrocytic regulatory functions. Fig. 5. [125]Fig. 5 [126]Open in a new tab Signaling changes in astrocytes, glutamatergic neurons, and GABAergic neurons across mTORopathies. Differential signaling strengths for astrocytes, glutamatergic neurons, and GABAergic neurons are compared between control and pathological conditions (cortical tubers (TSC), FCD_mTOR, and FCD_DEPDC5). a–c Represents astrocytes in cortical tubers (TSC) (a), FCD_mTOR (b), and FCD_DEPDC5 (c). d–f Represents glutamatergic neurons in cortical tubers (TSC) (d), FCD_mTOR (e), and FCD_DEPDC5 (f). g–i Represents GABAergic neurons in cortical tubers (TSC) (g), FCD_mTOR (h), and FCD_DEPDC5 (i). The x-axis represents the differential outgoing interaction strength (ligand expression from the indicated cell type), and the y-axis represents the differential incoming interaction strength (receptor expression by the indicated cell type). Each dot corresponds to a specific ligand–receptor signaling pathway. Pathways are color-coded: black indicates signaling pathways shared across conditions, red indicates control-specific signaling, and blue denotes pathology-specific signaling In FCD_mTOR, NRXN, NCAM, and CADM displayed signaling changes similar to those in cortical tubers (TSC), with glutamatergic (Fig. [127]5e) and GABAergic neuron (Fig. [128]5h) upregulation and astrocytic (Fig. [129]5b) downregulation, emphasizing shared synaptic adhesion and ECM deficits. However, astrocytes in FCD_mTOR also showed upregulated PTN, CNTN, and CD99 signaling (Fig. [130]5b), suggesting an active astrocytic response to the pathological microenvironment. In FCD_DEPDC5, NRXN, NCAM, and CADM signaling mirrored trends seen in cortical tubers (TSC) and FCD_mTOR. However, FCD_DEPDC5 displayed distinct astrocytic responses (Fig. [131]5c), with collagen signaling strongly downregulated in astrocytic incoming interactions, suggesting impaired ECM maintenance and weakened structural support. In contrast, NECTIN signaling was uniquely upregulated in astrocytes, indicating a compensatory effort to strengthen cell–cell adhesion and stabilize disrupted networks. Unlike in FCD_mTOR, laminin and collagen upregulation were not observed in glutamatergic (Fig. [132]5f) and GABAergic (Fig. [133]5i) neurons, underscoring a more astrocyte-specific ECM dysregulation in FCD_DEPDC5. These findings reveal shared features of neuronal hyperexcitability and ECM disruption across disorders, driven by NRXN, NCAM, and CADM upregulation in neurons and laminin and collagen downregulation in astrocytes in cortical tubers (TSC) and FCD_mTOR. However, disorder-specific differences highlight contrasting astrocytic roles, with FCD_DEPDC5 astrocytes exhibiting stronger compensatory mechanisms through NECTIN upregulation. Building on this analysis of interaction strengths, we aggregated incoming and outgoing signaling strengths to allow us to assess how distinct signaling pathways further contribute to cellular communication within each condition (Supplementary Fig. [134]3). Differential NRXN signaling in TSC and FCD subtypes To investigate the impact of NRXN and NLGN signaling in cortical tubers (TSC) and FCD subtypes, we analyzed the communication probabilities between these key ligand-receptor pairs across different cell types. Neurons exhibited a general trend of increased signaling, whereas astrocytes demonstrated predominantly decreased signaling (Fig. [135]6). Notably, no increased or decreased signaling was observed between microglia and glutamatergic neurons, GABAergic neurons, or astrocytes, indicating that microglial involvement is minimal in this context. Among the conditions studied, FCD_DEPDC5 displayed the least decreased NRXN-NLGN signaling compared to cortical tubers (TSC) and FCD_mTOR. Additionally, signaling from neurons to astrocytes was less pronounced in FCD_mTOR compared to cortical tubers (TSC) and FCD_DEPDC5. These findings suggest that the extent and nature of astrocytic involvement in NRXN-NLGN signaling networks vary significantly across these conditions, with FCD_DEPDC5 retaining more robust neuronal-astrocytic communication than cortical tubers (TSC) and FCD_mTOR. Fig. 6. [136]Fig. 6 [137]Open in a new tab Differential NRXN-NLGN signaling in cortical tubers (TSC)and FCD subtypes. Dot plots represent communication probabilities between NRXN (Neurexin) and NLGN (Neuroligin) signaling pairs across different cell types in (a, b) cortical tubers (TSC), c, d FCD_mTOR, and e, f FCD_DEPDC5. a Increased signaling in cortical tubers (TSC). b Decreased signaling in cortical tubers (TSC). c Increased signaling in FCD_mTOR. d Decreased signaling in FCD_mTOR. e Increased signaling in FCD_DEPDC5. f Decreased signaling in FCD_DEPDC5. Dot sizes represent statistical significance (p < 0.05 and p < 0.01), while color intensity reflects communication probability (red = maximum, blue = minimum). Cell types are grouped by control (red) and pathology (blue) on the x-axis Interestingly, in cortical tubers (TSC), NRXN-NLGN signaling was specifically identified between glutamatergic neurons and endothelial cells, a connection absent in control samples. Similarly, NCAM1-NCAM1 signaling between glutamatergic and GABAergic neurons was observed in cortical tubers (TSC), FCD_mTOR, and FCD_DEPDC5 but was absent in controls. The relative contribution of these ligand-receptor (L-R) pairs across conditions is illustrated in supplementary Fig. [138]4. These findings highlight aberrant intercellular signaling pathways unique to mTORopathies, implicating altered neuronal adhesion and communication as key pathological features. Glutamatergic signaling was more prevalent in FCD_DEPDC5 compared to cortical tubers (TSC) and FCD_mTOR across all NRXN-NLGN ligand-receptor pairs. Additionally, NCAM signaling was more pronounced in astrocytes of cortical tubers (TSC) and FCD_DEPDC5 but not in FCD_mTOR, indicating disease-specific patterns of astrocytic involvement in cell adhesion and synaptic regulation. While GABAergic neurons showed greater involvement in cortical tubers (TSC), FCD_mTOR displayed a broader range of outgoing interactions originating from glutamatergic neurons. These observations underscore distinct alterations in the cellular contributions to NRXN-NLGN signaling across mTORopathies, with each condition exhibiting unique patterns of disrupted neuronal and glial communication. At the molecular level, no significant differential expression of NRXNs and NLGNs was observed in cortical tubers (TSC) in GABAergic neurons, glutamatergic neurons, or astrocytes, except for a small upregulation of NLGN4X in cortical tubers (TSC) glutamatergic neurons. Additionally, CNTN6 was upregulated in astrocytes and GABAergic neurons in cortical tubers (TSC), while CHL1 showed upregulation in GABAergic neurons. In contrast, FCD_mTOR exhibited no differential expression in NRXNs, NLGNs, CNTN6, or CHL1 across any cell type. In FCD_DEPDC5, NLGN1 was upregulated in astrocytes, and SLC1A3 was upregulated in both GABAergic and glutamatergic neurons. These molecular findings align with the observed cell-type-specific communication patterns, further emphasizing the differential mechanisms underlying mTORopathies. To corroborate these findings at the protein level, we performed immunohistochemistry for NRXN1 and NLGN2 on brain section from control, FCD IIa, and FCD IIb samples. These two molecules were selected because they showed the highest contribution to the NRXN-NLGN signaling pathway in our ligand-receptor communication analysis. Immunostaining confirmed the presence of NRXN1 (supplementary Fig. [139]5) and NLGN2 protein (supplementary Fig. [140]6) in neurons across all conditions, with additional prominent staining observed in dysmorphic neurons and balloon cells in cortical tubers (TSC) and FCD samples. In summary, the NRXN-NLGN signaling network in mTORopathies is characterized by cell-type- and disease-specific alterations. cortical tubers (TSC) showed increased endothelial and GABAergic involvement, including unique NRXN-NLGN signaling between glutamatergic neurons and endothelial cells and NCAM1-NCAM1 signaling between glutamatergic and GABAergic neurons. FCD_mTOR displayed diverse glutamatergic interactions, whereas FCD_DEPDC5 retained relatively preserved signaling pathways, with enhanced glutamatergic signaling and astrocytic involvement. These findings suggest differential mechanisms of dysregulated cell–cell communication underlying the pathophysiology of these conditions, with distinct contributions from neuronal and glial populations in each pathology. Discussion This study provides a comprehensive analysis of cellular heterogeneity, pathology-specific changes, and cell–cell communication networks in mTORopathies, with a focus on cortical tubers (TSC), FCD_mTOR, and FCD_DEPDC5. Using single-cell RNA sequencing (scRNA-seq), we identified transcriptionally distinct clusters across major neural and non-neural cell types, revealing pathology-specific differences in cellular composition and signaling networks that contribute to the unique pathophysiology of each condition. Cellular heterogeneity and disease-specific changes Across control and pathological samples, we identified 33 transcriptionally distinct clusters encompassing neuronal, glial, vascular, and immune cell types. Glutamatergic neurons showed the greatest diversity with nine clusters, reflecting their complexity and their central role in cortical network organization. Pathology-specific differences included the absence of a glutamatergic neuron cluster (Cluster 21) and a microglial cluster (Cluster 29) in TSC and FCD subtypes, alongside the emergence of a pathology-specific endothelial cluster (Cluster 31). The absence of Cluster 21, characterized by genes related to synapse organization and calcium signaling, suggests a selective loss of excitatory subtypes crucial for maintaining excitatory–inhibitory balance. Similarly, the absence of Cluster 29 microglia, enriched for synapse remodeling and neuroimmune regulation, indicates impaired microglia-mediated synaptic support. In contrast, Cluster 31 endothelial cells appeared exclusively in disease samples and were enriched for axonogenesis and synaptic signaling pathways, suggesting disrupted neurovascular interactions as a shared hallmark of mTORopathies. Importantly, quantitative analyses of cell numbers and proportions per cluster (Supplementary Tables [141]3–[142]4) provide context for these findings. Nearly all cell classes were present across all conditions, arguing against sampling artifacts. ANOVA testing across clusters revealed that only astrocytic Cluster 15 differed significantly between conditions, with increased representation in TSC, consistent with gliosis observed histologically. In other cases, such as oligodendrocytes, no major abundance differences were detected, despite the well-established hypomyelination phenotype in TSC [[143]15]. This discrepancy underscores that transcriptional state changes can occur independently of overall cell number and highlights the need to interpret shifts in both abundance and gene expression. Dysregulated cell–cell communication – NRXN-NLGN signaling Cell–cell communication plays a crucial role in maintaining neural network integrity and regulating brain function, and its dysregulation is often implicated in various neurological disorders, including epilepsy [[144]13, [145]30]. In mTORopathies, including cortical tubers (TSC) and other genetics FCDs, the disruption of signaling pathways and altered interactions between neurons and glial cells contribute to the pathological features of these diseases, including cortical malformations and hyperexcitability [[146]5, [147]19]. Previous research has highlighted the importance of cell–cell communication in epilepsy, particularly through synaptic transmission and neuroimmune signaling [[148]16], where impaired communication between excitatory and inhibitory neurons, as well as between neurons and glial cells, can lead to network dysfunction and seizure activity [[149]17, [150]54, [151]59]. Additionally, cell–cell communication networks involving astrocytes, microglia, and endothelial cells are increasingly recognized for their role in modulating neuroinflammation, neurovascular integrity, and synaptic plasticity in epilepsy and related disorders [[152]58, [153]68–[154]70]. Our analysis of intercellular communication networks revealed a general increase in interaction numbers and strength in all three mTORopathies compared to controls, with disease-specific patterns. cortical tubers (TSC) vexhibited reduced interactions involving microglia, neurons, and oligodendrocyte precursor cells (OPCs), indicating impaired crosstalk between these cell types. FCD_mTOR displayed reduced astrocyte-mediated interactions, essential for neural homeostasis, while FCD_DEPDC5 demonstrated weakened astrocyte-OPC communication and disrupted neuroimmune signaling. Astrocytes emerged as key regulators of cortical network dysregulation. In cortical tubers (TSC) and FCD_mTOR, astrocytic signaling through laminin and collagen pathways was downregulated, reflecting impaired extracellular matrix (ECM) maintenance. Conversely, FCD_DEPDC5 astrocytes exhibited upregulated NECTIN signaling, potentially compensating for weakened structural support. These molecules are implicated in ECM remodeling, neuron-glia communication, and immune signaling, potentially reflecting compensatory but maladaptive processes [[155]21, [156]63]. Laminin and collagen pathways exhibited opposing trends: downregulated in astrocytes, reflecting weakened ECM maintenance, but upregulated in glutamatergic neurons, likely representing compensatory neuronal ECM remodeling to stabilize hyperactive circuits. These shifts highlight possible neuron-astrocyte imbalances driving cortical instability in FCD_mTOR. This downregulation may exacerbate network dysregulation by failing to buffer neuronal hyperactivity [[157]3, [158]48]. Notably, strong differential interaction strengths were observed in astrocytic collagen and laminin pathways, indicating active ECM remodeling in response to neuronal hyperactivity or tissue disruption, a process absent in neurons. These findings show the critical role of astrocyte-neuron interactions in maintaining cortical stability and their disruption in disease states. Astrocytes are essential for maintaining the homeostasis of the brain’s microenvironment, playing a pivotal role in neuronal development, synaptic communication, and cognitive functions [[159]27, [160]35, [161]65]. Neuron-astrocyte interactions are particularly crucial for synapse formation and their dynamic modulation in response to neuronal activity [[162]2, [163]18, [164]25]. In this study, we explored the disruption of neuron-astrocyte crosstalk in mTORopathies, with a specific focus on the neuroligin (NLGN)-neurexin (NRXN) signaling network, in line with previous research [[165]51]. Both NRXN and NLGN are well-established autism-risk genes involved in maintaining the structural stability of neurons by regulating neurite outgrowth, synapse formation, and synaptic plasticity [[166]29, [167]34, [168]40, [169]57]. Interestingly, in the context of mTORopathies such as cortical tubers (TSC), alterations in NRXN-NLGN signaling were observed. In cortical tubers (TSC) and FCD_mTOR, increased NRXN-NLGN signaling was observed in both glutamatergic and GABAergic neurons, reflecting, heightened synaptic activity and plasticity, characteristics of mTOR dysregulation. However, astrocytic signaling in these pathways was reduced, pointing to a disruption in neuron-glia communication that may contribute to the neurodevelopmental deficits seen in these disorders. FCD_DEPDC5 displayed distinct features, including preserved NRXN-NLGN signaling in astrocytes and reduced collagen signaling, which may point to unique mechanisms of extracellular matrix dysregulation. Together, these findings suggest that NRXN and NLGN not only function as autism-risk genes but also play a critical role in maintaining synaptic and structural stability within neural networks. The dysregulation of NRXN-NLGN signaling in mTORopathies underscores the potential contribution of these pathways to the pathogenesis of these disorders, highlighting the importance of neuron-glia interactions in maintaining cortical network stability. Phenotypic correlates: dyslamination, neuronal misplacement, and ECM changes A defining feature of both TSC and FCD is cortical dyslamination and neuronal misplacement [[170]4, [171]19, [172]66]. These anatomical hallmarks likely shape the molecular landscape we observed. For instance, neurons residing outside their expected laminar environment may experience distinct extracellular matrix (ECM) conditions and signaling cues. In line with this, we found astrocytic laminin and collagen signaling reduced in TSC and FCD_mTOR, whereas glutamatergic neurons upregulated ECM-related pathways, likely reflecting compensatory remodeling. Such imbalances between neuronal and astrocytic ECM contributions may destabilize cortical circuits, further exacerbating hyperexcitability. These findings align with prior evidence that ECM dysregulation contributes to cortical malformations and epilepsy [[173]8, [174]26, [175]37]. Aberrant ECM remodeling not only disrupts synaptic stability but also alters migration cues, potentially amplifying the effects of neuronal misplacement [[176]44, [177]63]. Thus, our data suggest a convergence of transcriptional, spatial, and extracellular factors that together drive network instability in mTORopathies. Post-mortem gene expression and seizure-related influences A potential limitation of this work is the use of post-mortem control tissue. While post-mortem changes may affect transcript stability, all control samples had low post-mortem intervals and high RNA integrity numbers, minimizing degradation [[178]11]. Comparative studies, including our own work in temporal lobe epilepsy, have demonstrated that post-mortem and resection controls are highly comparable when matched for quality [[179]45]. Nevertheless, we acknowledge that differential expression patterns may reflect both primary disease mechanisms and secondary seizure-related effects. Seizures induce gene expression changes in inflammation, synaptic remodeling, and stress response pathways, some of which are compensatory or protective [[180]24, [181]53, [182]62]. While these influences cannot be fully disentangled in the current dataset, they provide an authentic representation of the diseased cortical environment. We therefore interpret our findings as reflecting both pathogenic drivers and seizure-induced adaptations. Methodological considerations Our study shows the identification of pathology-specific transcriptional clusters, a feature that has not been consistently reported in previous single-nuclei studies on mTORopathies and epilepsy [[183]9, [184]12]. Several methodological factors likely contributed to our ability to discern these unique clusters, including the samples used and differences in sequencing platforms. Compared to our prior study [[185]61] using 10 × Genomics 3’ chemistry (median ~ 3,017 genes and ~ 8,445 UMIs per cell), the present dataset using Chromium Single Cell Flex yielded substantially higher coverage (median ~ 5,607 genes and ~ 15,073 UMIs per cell). This increase in transcript capture improved the resolution of rare clusters and enhanced characterization of glial and vascular populations. While improved chemistry contributed, we acknowledge that other factors, including sample size, cortical region, and cellular mosaicism, also influence cluster detection. Therapeutic implications and future direction These findings reveal shared and disease-specific mechanisms underlying mTORopathies, emphasizing the importance of targeting cell-type-specific pathways for therapeutic interventions. Strategies to restore synaptic and ECM integrity, such as modulating astrocytic laminin and collagen signaling, could offer new possibilities for treatment, and warrant further investigations to delineate more detailed mechanisms of the proposed involvement of the ECM [[186]22, [187]38]. While our current analysis focused on broader cell types to establish an overarching view of communication dysregulation, we acknowledge that distinct GABAergic subtypes and glial subpopulations may differ in their contribution to pathological signaling. Future studies using larger samples sizes and complementary datasets may help resolve these subtype-specific communication dynamics, particularly in SST + and PV + interneurons known to play divergent roles in cortical circuitry and epileptogenesis. Moreover, future studies should focus on validating these findings in larger cohorts and exploring their relevance to developmental and functional outcomes. Integrating multi-omics approaches, such as spatial transcriptomics and proteomics, could provide deeper insights into the spatial and temporal dynamics of these interactions. Furthermore, preclinical models targeting NRXN-NLGN and ECM pathways may elucidate their potential for therapeutic modulation. In conclusion, this study highlights the complex interplay between cellular and molecular mechanisms in mTORopathies, providing a foundation for future research into targeted therapies for these debilitating conditions. Supplementary Information [188]Additional file 1.^ (10.2MB, docx) Author contributions E.A, J.D.M, and M.S conceptualized the project. F.E.J, W.V.H, and A.M, helped with collection, selection and/or revision of human brain tissues and clinical data. M.S selected all scRNAseq samples and J.J.A. prepared and cut all samples for sequencing. All processing and bioinformatical analysis was performed by M.S, under supervision of E.A and J.D.M. Immunohistochemical stainings were performed by J.J.A. The original draft of the manuscript was prepared by M.S with review and editing by J.D.M and E.A. All authors read, refined and approved the final version of the manuscript. Funding This research has received funding from the ZonMw, Programme Translational Research no. 95105004. Data availability The sequencing data underlying this study are currently available from the corresponding author upon reasonable request. These data will be deposited in a public repository and made publicly accessible prior to publication. The accession number will be provided in the final version of the manuscript. Declarations Ethics approval and consent to participate All procedures with human tissue were obtained with informed consent for the use in research and access to medical records in accordance with the Declaration of Helsinki and the Amsterdam UMC Research Code provided by the Medical Ethics Committee. Consent for publication Not applicable. Competing interests EA has served on scientific advisory boards for UCB and Nutricia. FEJ has served on scientific advisory boards for UCB, Jazz pharma, Novartis, and Nutricia. The remaining authors declare no conflict of interest. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. James D. Mills and Eleonora Aronica have contributed equally to this work. Contributor Information Mirte Scheper, Email: m.scheper@amsterdamumc.nl. Eleonora Aronica, Email: e.aronica@amsterdamumc.nl. References