Neurodevelopmental disorders can have long-lasting effects, causing not only early pediatric symptoms but also a range of neurological issues throughout adulthood. West syndrome is a severe neurodevelopmental disorder marked by infantile spasms, an early symptom that typically subsides with age. However, many patients progress to other seizure forms, known as seizure evolution, which is closely linked to poor long-term outcomes. Despite its clinical significance, the neurobiological mechanisms behind seizure evolution in West syndrome remain poorly understood. Recent genetic studies have consistently identified the CYFIP2 p.Arg87Cys variant in West syndrome patients, and the Cyfip2^+/R87C mouse model carrying this mutation has been shown to recapitulate key symptoms of the disorder, including infantile spasms. In this study, we aimed to gain deeper insight into seizure evolution by conducting longitudinal deep phenotyping of the Cyfip2^+/R87C mouse model from the neonatal stage to seven months of age. We tracked seizure activity through behavioral and EEG recordings and employed multi-omic analyses, including tissue and single-cell level transcriptomics, ultrastructural analysis, proteomics, and lipidomics, to capture a comprehensive view of molecular and cellular changes. Our results showed that after an initial period of neonatal spasms, Cyfip2^+/R87C mice entered a seizure-free phase, followed by spontaneous recurrent seizures in adulthood, ultimately leading to premature death. This progression was associated with synaptic remodeling, sequential activation of different glial cell types, lipid droplet accumulation in astrocytes, and significant proteomic and lipidomic changes in the brain. These findings suggest that seizure evolution in West syndrome involves complex, time-dependent interactions between neurons and glial cells, along with alterations in lipid metabolism. Our study highlights the potential of longitudinal multi-omic approaches to uncover underlying mechanisms of seizure evolution and suggests that targeting these changes could offer novel therapeutic strategies. Additionally, the dataset generated here may provide valuable insights for other epilepsy and neurodevelopmental disorder models.

Spontaneous recurrent seizures and lethality in adult Cyfip2^+/R87C mice

In our previous study [22], we observed that Cyfip2^+/R87C mice exhibited spontaneous spasms during the neonatal stage (postnatal day 8, PD 8). However, long-term video-EEG recordings from the same study revealed no significant seizures in these mice during the juvenile stage (PD 30), indicating a seizure-free period following neonatal spasms. Unexpectedly, when maintaining Cyfip2^+/R87C mice in their home cages for longer periods, we found that lethality began around postnatal week 14 (PWK 14), with approximately 60% of the mice dying by PWK 30 (Fig 1A). In contrast, no significant lethality was observed in Cyfip2^+/− mice, a model for CYFIP2 haploinsufficiency [15], on either the C57BL/6N or C57BL/6J background. Additionally, Cyfip2^+/R87C mice, but not Cyfip2^+/−, displayed significantly smaller body size and lower body weight at PWK 28 compared to their respective wild-type (WT) littermates (S1 Fig).

To investigate whether the lethality in Cyfip2^+/R87C mice is associated with their seizures, we conducted long-term video recordings with different cohorts and found that these mice indeed experienced spontaneous recurrent seizures at both PWK 14 and 28 (Fig 1B and S1 and S2 Videos). Notably, once they occurred, behavioral seizure events were more clustered at PWK 28 compared to PWK 14, as evidenced by an increased number of seizure events per hour during the recording sessions at PWK 28. In contrast, no such behavioral seizures were observed in WT or Cyfip2^+/− mice at PWK 28. We also attempted simultaneous video-EEG recordings on Cyfip2^+/R87C mice at PWK 20 with different cohorts; however, due to a high mortality rate (approximately 80%) during or after surgery, likely linked to their poor physical condition, we were unable to analyze a significant number of mice. Nevertheless, in the few surviving Cyfip2^+/R87C mice (n = 3), we observed tonic-clonic seizures and myoclonic jerks that correlated with behavioral seizures, events that were never observed in WT mice (Fig 1C and S3 and S4 Videos). Unfortunately, it was not feasible to perform EEG recordings at both the earlier (PWK 14) and the later (PWK 28) time points due to the high perioperative mortality rate observed in Cyfip2^+/R87C mice, as described above. Given this limitation, we focused our video-EEG recordings on an intermediate time point (PWK 20), where a small number of animals survived the procedure. As a result, we were unable to directly assess age-dependent changes in seizure severity or frequency using electrophysiological data. Although we could not electrophysiologically assess age-dependent seizure progression, histopathological examination at PWK 28 revealed hippocampal pathology in Cyfip2^+/R87C mice, including shrunken CA1 regions and dispersion of dentate gyrus granule cells, hallmarks of epilepsy [23], while these changes were absent in Cyfip2^+/− mice (Fig 1D).

To further investigate the cause of lethality in Cyfip2^+/R87C mice, we conducted long-term video monitoring of animals aged beyond PWK 20 until spontaneous death (n = 5). In all cases, locomotor and general activity, encompassing both ambulatory movements and non-ambulatory behaviors such as head or body motions, declined markedly at least one day prior to death (S2 Fig). Even during this period of markedly reduced activity, behavioral seizure-like symptoms were occasionally observed, but these were less pronounced than those seen well before death, consisting only of mild manifestations such as forelimb extension (S5 Video). One possible explanation is that the animals were too debilitated to manifest more overt seizure activity. At the moment of death, some hindlimb rigidity was observed (S6 Video), although it was difficult to determine from behavioral observation alone whether this represented a seizure-related event.

To assess whether lethality might be attributable to dysfunction in peripheral organs, we performed gross and histological examinations of the heart, lung, liver, and kidney (S3 Fig). No significant abnormalities were detected; all organs exhibited normal architecture without evidence of inflammation, degeneration, or other pathological changes. Taken together, these findings suggest that recurrent seizures arising from brain dysfunction, rather than failure of other major organs, may contribute to the mortality observed in Cyfip2^+/R87C mice. Nevertheless, additional physiological recordings will be necessary to determine definitively whether this mortality reflects SUDEP. Moreover, our data indicate that Cyfip2^+/R87C mice undergo seizure evolution, from neonatal spasms to adult spontaneous recurrent seizures, with the latter stage closely associated with lethality. The absence of this phenotype in Cyfip2^+/− mice further supports the notion that the hotspot p.Arg87 variant likely confers a toxic gain-of-function, rather than simply destabilizing CYFIP2 proteins [22,24].

Brain transcriptomic signatures in Cyfip2^+/R87C mice during seizure evolution

Building upon the above findings, we explored longitudinal changes in the brains of Cyfip2^+/R87C mice during seizure evolution. We also analyzed Cyfip2^+/− mice in parallel to validate the toxic gain-of-function effect caused by the p.Arg87Cys variant at the gene expression level. Bulk-tissue transcriptomic analyses were conducted in the frontal cortex and hippocampus of Cyfip2^+/R87C mice, Cyfip2^+/− mice, and their respective WT littermates (all on the C57BL/6N background) at three postnatal stages (PWK 1, 7, and 14), covering the period from neonatal spasms to the onset of lethality and seizures in Cyfip2^+/R87C mice. The two brain regions were selected due to their high CYFIP2 expression, our previous findings which demonstrated morphological and functional changes in neurons within these regions in various Cyfip2 mouse models [15,16,19], and the above finding (Fig 1D) showing cytoarchitectural changes in the hippocampus of Cyfip2^+/R87C mice.

In Cyfip2^+/R87C mice, we observed a dramatic increase in differentially expressed genes (DEGs) in both brain regions as the mice aged, particularly in the hippocampus, with 20, 41, and 3,755 DEGs identified at the three stages, respectively (Fig 2A and S1 Table). Gene Ontology (GO) analysis of the DEGs at PWK 14, using the Database for Annotation, Visualization, and Integrated Discovery (DAVID), revealed significant terms, notably the downregulation of synapse-related terms (glutamatergic/cholinergic/dopaminergic synapses) and upregulation of extracellular matrix (ECM)-related terms, suggesting substantial synaptic and cytoarchitectural remodeling. In contrast, the number of DEGs in Cyfip2^+/− mice decreased with age, with only 19 DEGs detected in the hippocampus at PWK 14 (Fig 2B and S2 Table).

We also performed bulk-tissue transcriptomic analysis on the frontal cortex and hippocampus of Cyfip2^+/R87C mice at PWK 28, when their behavioral seizure events were more severe compared to PWK 14. The number of DEGs, particularly in the cortex, was even higher at PWK 28 than at PWK 14 (Fig 2C and S3 Table). DAVID GO analysis revealed some overlapping terms between PWK 28 and 14 in both brain regions. However, unlike at PWK 14, RNA regulation-related terms, such as RNA splicing and RNA binding, were significantly downregulated, while synaptic vesicle (i.e., presynapse)-related terms were more prominently upregulated at PWK 28. Terms related to general and postsynaptic functions, such as glutamatergic synapse and postsynaptic density, remained downregulated, which may suggest an imbalance between pre- and postsynaptic compartments, as further validated below.

We also performed Gene Set Enrichment Analysis (GSEA) to identify transcriptomic signatures based on coordinated transcriptional changes across many genes [25]. In testing for cell-type-specific gene sets, we found that the cortex transcriptome of Cyfip2^+/R87C mice shifted from positive to negative enrichment for neuron-specific gene sets between PWK 1 and 14, but became positive again at PWK 28 (Fig 2D). Meanwhile, enrichment for glia-specific gene sets in the cortex varied depending on the three glial cell types (i.e., astrocytes, oligodendrocytes, and microglia). In the hippocampus of Cyfip2^+/R87C mice, the age-progressive shift from positive to negative enrichment for neuron-specific gene sets followed a similar pattern to that in the cortex, but occurred earlier and persisted until PWK 28, particularly for the CA1 pyramidal neuron gene set. In contrast, overall enrichment for glia-specific gene sets in the hippocampus shifted from negative to positive with each glial cell type showing a distinct peak of positive enrichment: PWK 7 for oligodendrocytes, PWK 14 for microglia, and PWK 28 for astrocytes. Enrichment for other cell-type-specific gene sets was subtle in both brain regions, except for a notable positive enrichment for ependymal gene sets, particularly in the hippocampus at PWK 7.

GSEA of the Cyfip2^+/R87C transcriptome for biological processes supported the findings from the GO analysis of DEGs and the GSEA for cell-type analyses (Fig 2E). Neuron- and synapse-related terms were positively enriched at PWK 1 but became negatively enriched with age. Additionally, RNA regulation-related terms were negatively enriched in both the cortex and hippocampus, particularly at PWK 14 and 28. Meanwhile, terms related to tissue remodeling and immune responses, such as wound healing, cell migration, cell activation, response to bacterium, and immune response, became positively enriched, especially in the hippocampus at PWK 14 and 28.

Synapse remodeling in Cyfip2^+/R87C mice during seizure evolution

Given the significant age-dependent changes in synapse-related terms observed in the transcriptomic analyses of Cyfip2^+/R87C mice, we sought to validate these findings through additional approaches. qRT-PCR analysis confirmed that the expression levels of genes encoding synaptic proteins, selected from the DEG lists of PWK 14 and 28, notably increased during early postnatal stages but decreased at later stages in the cortex and hippocampus of Cyfip2^+/R87C mice compared to WT mice (Fig 3A). In contrast, none of these genes showed significant changes in the cortex and hippocampus of Cyfip2^+/− mice at PWK 28 (S4 Fig). Notably, some of these genes exhibited significant expression differences between PWK 14 and 28 within Cyfip2^+/R87C mice, which may be linked to the observed increase in seizure severity over time (Fig 3A). To further assess synaptic changes, we performed western blot analysis of representative excitatory postsynaptic markers, including SH3 and multiple ankyrin repeat domains 3 (Shank3), Homer protein homolog 1 (Homer1), and the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunit (GluA2). Our results showed a downregulation of these markers in the synaptosomes of both brain regions in Cyfip2^+/R87C mice at PWK 14 and 28 (Fig 3B). Interestingly, consistent with the DAVID GO analysis of the DEGs (Fig 2C), we observed increased levels of presynaptic markers, such as Synaptotagmin and Synapsin 1, suggesting that synapse remodeling in Cyfip2^+/R87C mice may involve an imbalance between post- and pre-synaptic compartments.

Fluorescence immunohistochemistry using representative excitatory presynaptic (vesicular glutamate transporter 1, vGluT1) and postsynaptic (Homer1) markers revealed a decrease in the number of total and contact puncta, as well as a reduction in average puncta intensity in the hippocampal CA1 region of Cyfip2^+/R87C mice compared to WT mice at PWK 14 and 28, indicating a reduced number of excitatory synapses (Figs 3C and S5). Interestingly, the average puncta volume for both vGluT1 and Homer1 increased in Cyfip2^+/R87C mice, suggesting that the remaining synapses may have altered structures. In contrast, vGluT1 and Homer1 parameters in the prelimbic medial prefrontal cortex (mPFC) of Cyfip2^+/R87C mice were largely normal, except for a decrease in Homer1 puncta number and average intensity at PWK 14 (S6 Fig).

To further investigate synaptic changes in the hippocampal CA1 region of Cyfip2^+/R87C mice, we conducted ultrastructural analysis using electron microscopy. Consistent with the immunohistochemistry findings, the number of excitatory (asymmetric) synapses decreased, while the number of inhibitory (symmetric) synapses remained normal in Cyfip2^+/R87C mice at PWK 28 (Fig 3D). Notably, we observed a significant reduction in dendritic spines, small protrusions on dendrites where excitatory synapses are typically formed, in Cyfip2^+/R87C neurons. Instead, excitatory synapses were predominantly located on dendritic shafts, with multiple presynaptic terminals clustering nearby (Fig 3D). This altered synaptic configuration was further highlighted by three-dimensional (3D) reconstruction of dendritic segments from CA1 pyramidal neurons in WT and Cyfip2^+/R87C mice (Fig 3E). In Cyfip2^+/R87C neurons, lacking dendritic spines, postsynaptic densities (PSDs) were shown on irregularly shaped dendritic shafts where axons made denser contacts compared to WT neurons. The distribution of PSD surface area differed significantly between WT and Cyfip2^+/R87C neurons, in line with immunohistochemical results showing an increase in synaptic puncta volume. Additionally, Cyfip2^+/R87C axons exhibited excessive myelination, consistent with increased myelin basic protein (MBP) signals in the hippocampus compared to WT mice (S7 Fig).

Gliosis in Cyfip2^+/R87C mice during seizure evolution

Beyond synapse remodeling, transcriptomic analysis of Cyfip2^+/R87C mice revealed a shift towards negative enrichment for neuron-specific gene sets and positive enrichment for glia-specific gene sets as they aged, particularly in the hippocampus (Fig 2). To further investigate this at the cellular level, we performed fluorescence immunohistochemistry using antibodies against markers for neurons (neuronal nuclei, NeuN) and various glial cell types (oligodendrocyte transcription factor 2 (Olig2) for oligodendrocytes, ionized calcium-binding adapter molecule 1 (IBA1) for microglia, and glial fibrillary acidic protein (GFAP) for astrocytes).

In the prelimbic mPFC, NeuN mean intensity in Cyfip2^+/R87C mice was elevated compared to WT mice at PWK 1 but became comparable at later stages (S8 Fig). Interestingly, glial markers in Cyfip2^+/R87C mice showed a sequential increase with age. Specifically, Olig2 mean intensity increased at PWK 7, followed by IBA1 at PWK 14, and GFAP at both PWK 14 and 28, compared to WT mice. In the mPFC of Cyfip2^+/− mice at PWK 28, Olig2 mean intensity increased on the C57BL/6N background, and GFAP mean intensity increased on the C57BL/6J background, relative to their respective WT mice (S8 Fig).

In the hippocampal CA1 region of Cyfip2^+/R87C mice, NeuN mean intensity decreased by 50% compared to WT mice at PWK 14 and 28, coinciding with the onset of seizures and lethality in these mice (Fig 4). A closer examination of the region revealed dispersion of CA1 pyramidal neurons in Cyfip2^+/R87C mice. To examine whether the observed changes in Cyfip2^+/R87C mice were associated with cell death of NeuN-positive neurons, we performed co-staining for cleaved Caspase3 and NeuN at PWK 1, 7, 14, and 20. Quantitative analysis of the fraction of cleaved Caspase3-positive area within the NeuN-positive region revealed no significant differences between WT and Cyfip2^+/R87C mice at any of these time points (S9 Fig). Interestingly, however, at PWK 20 in Cyfip2^+/R87C mice, but not in WT, we frequently observed small NeuN- and cleaved Caspase3-double-positive spots. If these structures indeed correspond to debris from dead NeuN-positive neurons, they may raise the possibility that a subset of CA1 pyramidal neurons in Cyfip2^+/R87C mice undergoes Caspase3-dependent cell death between PWK 14 and 20. Nonetheless, the potential involvement of alternative mechanisms, such as Caspase3-independent neuronal loss, cannot be excluded and warrants further investigation.

Similar to the mPFC, glial markers in the hippocampal CA1 region of Cyfip2^+/R87C mice exhibited an age-related increase compared to WT mice, with oligodendrocyte, microglia, and astrocytes in that order (Fig 4), consistent with the results of GSEA of the transcriptome for cell-type-specific gene sets (Fig 2D). However, unlike the mPFC, Olig2 mean intensity in the hippocampus increased at PWK 7 and persisted through PWK 14 and 28. In the hippocampus of Cyfip2^+/− mice at PWK 28, GFAP mean intensity, but not other markers, was mildly increased compared to WT mice on both the C57BL/6N and C57BL/6J backgrounds, although the increase was much less pronounced than in Cyfip2^+/R87C mice (S10 Fig).

These results suggest that neurons and multiple glial cell types undergo dynamic changes in Cyfip2^+/R87C mice during seizure evolution. Notably, the sequential nature of these phenomena indicates that changes in neurons and various glial cells may influence each other. Thus, in the next phase of our study, we focused on microglia and astrocytes to investigate these specific changes in greater detail.

Changes in microglial morphology and activity in Cyfip2^+/R87C mice

We analyzed the 3D morphology of microglia in the prelimbic mPFC and CA1 hippocampus of WT and Cyfip2^+/R87C mice at PWK 14 and 28. These stages were selected as they correspond to the onset of lethality and seizures in Cyfip2^+/R87C mice and because IBA1 intensity increased at PWK 14 and normalized by PWK 28 in Cyfip2^+/R87C brains (Fig 4). Understanding these changes may provide insights into the roles of microglia and their interactions with other cell types during seizure evolution.

In the mPFC, Sholl analysis showed decreased intersections at certain distances for microglia in Cyfip2^+/R87C mice compared to WT mice at PWK 14, and slightly increased intersections at some distances at PWK 28 (S11 Fig). However, average values for total branch length, area, branch points, and maximal diameter of microglia were comparable between genotypes at both ages. Morphological changes in microglia were more pronounced in the hippocampus of Cyfip2^+/R87C mice (Fig 5A). Sholl analysis revealed significantly fewer intersections at most distances in Cyfip2^+/R87C hippocampus compared to WT at both PWK 14 and 28. Additionally, total branch length and area were reduced in Cyfip2^+/R87C microglia at both ages. At PWK 28, but not PWK 14, the number of total branch points decreased, while the maximal branch diameter increased in Cyfip2^+/R87C microglia. These morphological properties suggest the presence of “ameboid microglia” in the hippocampus of Cyfip2^+/R87C mice, which are highly active (i.e., phagocytic) and often associated with pathological brain conditions [26].

To explore whether the normalization of IBA1 intensity at PWK 28 was related to microglial death, we investigated apoptotic markers. Cleaved Caspase3 signals in microglia were significantly higher in the Cyfip2^+/R87C hippocampus at PWK 28, but not PWK 14, compared to WT, suggesting increased apoptotic cell death at this stage (Fig 5B). Consistent with the observed morphological changes, CD68 signals, a marker for activated microglia, were significantly higher in the hippocampus of Cyfip2^+/R87C mice compared to WT mice (Fig 5C). Additionally, expression levels of multiple genes associated with microglial activation, such as proinflammatory cytokines, significantly increased in Cyfip2^+/R87C mice, particularly in the hippocampus at PWK 14 (Fig 5D).

Lastly, we examined whether activated microglia contributed to the reduction of excitatory synapses in the hippocampal CA1 region of Cyfip2^+/R87C mice at PWK 14 through synapse engulfment [27]. We observed an increase in the total volume of lysosomes (labeled by Lamp1) within microglia in the Cyfip2^+/R87C hippocampus compared to WT (Fig 5E). Although the amount of Homer1 within microglial lysosomes, representing engulfed excitatory synapses, showed a trend towards increase, it did not reach statistical significance. This may be due to other mechanisms contributing to synapse reduction or because microglial phagocytosis of synapses occurred earlier than PWK 14, by which time the number of excitatory synapses had already significantly decreased (Fig 3).

Lipid droplets and crystals in astrocytes of Cyfip2^+/R87C mice

We next investigated astrocytes, marked by GFAP, which showed increased intensity in the mPFC and hippocampus of Cyfip2^+/R87C mice at both PWK 14 and 28 (Fig 4). In the prelimbic mPFC, astrocytes from Cyfip2^+/R87C mice exhibited increased intersections in Sholl analysis, although no significant differences in other branch parameters were observed compared to WT mice (S12 Fig). In the hippocampus, however, Sholl analysis revealed decreased intersections at most distance points for Cyfip2^+/R87C astrocytes relative to WT astrocytes at both PWK 14 and 28 (Fig 6A). Additionally, average branch length, area, and points were reduced, while branch diameter increased in Cyfip2^+/R87C astrocytes at both ages.

Given the persistent increase in GFAP intensity and morphological changes in Cyfip2^+/R87C mice, coupled with recurrent seizures, we hypothesized that additional features in Cyfip2^+/R87C astrocytes might be associated with these seizures. Using electron microscopy, we discovered unexpected alterations in the lysosomes of Cyfip2^+/R87C astrocytes in the hippocampus at PWK 28 (Fig 6B). Specifically, lysosomes in Cyfip2^+/R87C astrocytes contained larger lipid droplets (LDs) compared to those in WT astrocytes. Correlative light and electron microscopy (CLEM) with BODIPY, a marker for neutral lipids, confirmed the presence of LDs in the lysosomes (Fig 6C). Quantitative analysis through fluorescence immunohistochemistry revealed that both the number and volume of LDs within astrocytic lysosomes were significantly increased in the hippocampus of Cyfip2^+/R87C mice compared to WT mice at PWK 28, but not at PWK 14 (Figs 6D and S13).

While examining electron microscopy images, we frequently observed crystals on or within the LDs of astrocytes in the hippocampus of Cyfip2^+/R87C mice, but never in WT mice (Fig 6E). Although the exact identity of these crystals remains unknown, cholesterol crystallization is well-documented in hepatocyte LDs in nonalcoholic steatohepatitis (NASH) and on LD surfaces in atherosclerotic lesions [28,29]. Using CLEM, we confirmed that these crystals could be visualized through reflection imaging [30], and quantification showed a higher number of crystal spots in the hippocampus of Cyfip2^+/R87C mice compared to WT mice at PWK 28 (Fig 6F). Since no significant differences in LDs were observed at PWK 14 (S13 Fig), crystal analysis was not performed at that time point. Notably, we identified various types of LDs, ranging from those without crystallization to those with moderate and extensive crystallization, which may represent different stages of lipid accumulation or processing (S14 Fig).

Single-nucleus transcriptomic analysis in the Cyfip2^+/R87C hippocampus

Although we observed changes in LDs mainly in astrocytes, not in neurons or other glial cell types of Cyfip2^+/R87C mice, we hypothesized that lipid-related pathways might also be affected in other cell types, as lipids are transported in various forms among neurons and glial cells to maintain proper lipid metabolism and storage in the brain [31]. Furthermore, we aimed to gain a deeper understanding of the molecular characteristics of astrocytes in Cyfip2^+/R87C mice, particularly focusing on whether and how they influence neuronal function and contribute to seizures. To test this hypothesis in a cell-type-specific manner, we conducted single-nucleus transcriptomic analysis on the hippocampus of WT and Cyfip2^+/R87C mice at P28 WK.

After applying stringent quality controls, including doublet removal, we retained 54,998 high-quality nuclei for downstream analyses (S15A Fig). Uniform Manifold Approximation and Projection (UMAP) visualization identified 12 distinct cell populations (Fig 7A), including two major neuronal groups (excitatory and inhibitory), five glial lineages (astrocytes, committed oligodendrocyte precursor cells (OPCs), OPCs, oligodendrocytes, and microglia), a rare neuronal subtype (Cajal-Retzius), a fibroblast population, endothelial and ependymal cells, and a “mixed” cluster co-expressing markers characteristic of both neurons and oligodendrocytes (Fig 7B and S4 Table). While some subsets of neurons and oligodendrocytes exhibited partial segregation by genotype, the substantial intermixing of replicates within each genotype indicated minimal batch effects, making batch correction unnecessary (S15B Fig).

Notably, a global differential expression (DE) analysis revealed significant upregulation of genes involved in lipid metabolism and transport in the Cyfip2^+/R87C condition, with apolipoprotein D (Apod) and prostaglandin D2 synthase (Ptgds) being particularly prominent (Figs 7C and S15C and S5 Table). Focusing on the neuronal compartment, initial sub-clustering separated excitatory neurons from inhibitory neurons (S15D Fig). The excitatory neurons were further divided into subgroups, including pyramidal, granule, and mossy cells (S15E Fig). Among the excitatory neurons, pyramidal neurons were classified into five distinct subtypes, such as canonical CA1 and CA2 populations, defined by specific marker genes and functional annotations (S15F Fig). Further sub-clustering of CA1 neurons identified six distinct subsets, with notable differences in abundance between the CA1-0 (WT) and CA1-1 (Cyfip2^+/R87C) subsets (Fig 7D). Regulatory network analysis using Single-Cell rEgulatory Network Inference and Clustering (SCENIC) [32] revealed transcription factors with significantly increased regulon activity in CA1-1 compared to CA1-0. These transcription factors are associated with processes such as cellular metabolism and transport (Fig 7E). Additionally, GSEA of DEGs in CA1-1 neurons demonstrated enrichment for catabolic and metabolic processes, including phosphatidylinositol dephosphorylation (S15G Fig).

Among the glial populations, oligodendrocytes were divided into two distinct subtypes, one of which was notably characterized by elevated expression of lipid-associated genes, including Apod and Ptgds (S16A Fig). Microglia were categorized into six condition-dependent subtypes (S16B Fig); however, neither oligodendrocytes nor microglia showed significant genotype-dependent differences in their relative abundance. In contrast, astrocytes, which were divided into five subtypes (Fig 7F), exhibited more pronounced genotype-specific changes. Notably, Astrocyte-4, which was more abundant in Cyfip2^+/R87C samples, showed a significant increase in the expression of apolipoprotein E (Apoe), a lipid-associated gene, and Gfap compared to the other astrocyte subtypes (Fig 7G). SCENIC analysis identified transcription factors with significantly elevated regulon activity in Astrocyte-4 compared to Astrocyte-1, regulating gene networks linked to cellular metabolism, stimulus response, and transport processes (Fig 7H). Complementing these findings, GSEA of Astrocyte-4 highlighted metabolic pathways, including oxidative phosphorylation, fatty acid metabolism, and related processes (Fig 7I).

Using gene expression profiles, we analyzed ligand-receptor (LR) interactions between neurons and astrocytes to investigate their potential influence on each other in Cyfip2^+/R87C mice. Notably, neuron-astrocyte pairs specific to Cyfip2^+/R87C mice (i.e., CA1-1 to Astrocyte-4 and Astrocyte-4 to CA1-1) exhibited 27 and 18 unique LR interactions, respectively, compared to control neuron-astrocyte pairs (e.g., CA1-0 to Astrocyte-1 and Astrocyte-1 to CA1-0 in WT mice) (Fig 7J). Among the 27 CA1-1 to Astrocyte-4 LR interactions, a disintegrin and metalloproteinase domain 10 (Adam10), amyloid-beta precursor protein (App), brain derived neurotrophic factor (Bdnf), and inhibin subunit beta A (Inhba) acting as ligands on CA1-1, were predicted to have significant regulatory potential, influencing 12 target genes in Astrocyte-4, including Gfap, prion protein (Prnp), phospholipid phosphatase 3 (Plpp3), and protocadherin 7 (Pcdh7) (S17A Fig) [33]. Notably, Gfap, Prnp, and Plpp3 were significantly upregulated, whereas Pcdh7 was downregulated, in Astrocyte-4 compared to Astrocyte-1 (Figs 7K and S17B). In the opposite LR direction, among the 18 Astrocyte-4 to CA1-1 LR interactions, four ligands (angiopoietin 1 (Angpt1), App, macrophage migration inhibitory factor (Mif), and prosaposin (Psap)) on Astrocyte-4 were predicted to have significant regulatory potential, influencing a total of 28 target genes in CA1-1 (S17C Fig). These 28 genes were most significantly associated with glutamatergic synaptic transmission (S17D Fig). Of these, 19 target genes, including adenosine kinase (Adk, downregulated) and glutamate ionotropic receptor kainate type subunit 5 (Grik5, upregulated), were significantly altered in CA1-1 compared to CA1-0 (Figs 7K and S17E).

Brain proteomic and lipidomic changes in Cyfip2^+/R87C mice

The single-nucleus transcriptomic analysis suggested alterations in lipid and metabolic homeostasis, as well as changes in intercellular crosstalk in the Cyfip2^+/R87C brain, prompting further investigation into the proteomic and lipidomic profiles of these mice. We performed proteomic analysis on the hippocampus of WT and Cyfip2^+/R87C mice at both PWK 14 and 28. Principal component analysis (PCA) revealed a distinct separation in proteomic profiles between WT and Cyfip2^+/R87C mice at both ages (Fig 8A). A total of 3,102 and 3,748 differentially expressed proteins (DEPs) were identified in the hippocampus of Cyfip2^+/R87C mice compared to WT mice at PWK 14 and 28, respectively (Fig 8B and S6 Table). Notably, an increased abundance of Apod and Ptgds was confirmed in Cyfip2^+/R87C mice relative to WT mice, particularly at PWK 28 (Fig 8C). Clustering analysis of the entire proteome separated them into four (A to D) groups based on age-dependent relative expression levels in Cyfip2^+/R87C mice compared to WT mice (Fig 8D and S7 Table). We focused on groups B and D, which represent proteins that exhibited further decreases and increases, respectively, in Cyfip2^+/R87C mice compared to WT mice as they aged. We hypothesized that these proteins might be associated with the age-dependent seizure severity observed in Cyfip2^+/R87C mice.

GO analysis of group B revealed significant enrichment in mRNA splicing and processing (Figs 8E and S18A), which is consistent with the downregulation of genes involved in these pathways observed in bulk-tissue transcriptomic analyses from the cortex and hippocampus of Cyfip2^+/R87C mice at PWK 28 (Fig 2). Additionally, protein–protein interaction (PPI) networks of group B proteins highlighted processes related to mRNA splicing, processing, and transport (S19 Fig). To further explore these findings, we examined alternative splicing events (ASEs) in the cortex and hippocampus of Cyfip2^+/R87C mice compared to WT mice using bulk-tissue transcriptome data. In line with the progressive downregulation of genes involved in mRNA splicing and processing in Cyfip2^+/R87C mice, only three ASE changes were observed at PWK 1 (two in the cortex and one in the hippocampus), and none at PWK 7 (S8 Table). However, at PWK 14, we identified three significant ASE changes in the cortex and 15 in the hippocampus, predominantly involving exon skipping, in Cyfip2^+/R87C mice compared to WT mice. Notably, altered exon skipping in the Neurexin 2 (Nrxn2) gene, which encodes a transmembrane protein regulating excitatory synapse formation [34], was consistently detected in both the cortex and hippocampus of Cyfip2^+/R87C mice. At PWK 28, we observed 26 significant ASE changes in the cortex and 21 in the hippocampus of Cyfip2^+/R87C mice compared to WT mice. Altered exon skipping in Nrxn3, another member of Nrxn gene family [35], was identified in the cortex but not in the hippocampus. Additionally, changes in U2 small nuclear RNA auxiliary factor 1 (U2af1), spermine oxidase (Smox), and phosphoseryl-tRNA kinase (Pstk) genes were detected in both brain regions of Cyfip2^+/R87C mice.

GO analysis of group D proteins revealed enrichment in mitochondrial functions, including oxidative phosphorylation and electron transport pathways (Figs 8E and S18B). Additionally, PPI networks of group D proteins highlighted several mitochondria-associated metabolic processes (Fig 8E). To validate these proteomic changes, we revisited the electron microscopic images of astrocytes in Cyfip2^+/R87C mice at PWK 28. Intriguingly, astrocytic mitochondria in Cyfip2^+/R87C mice were smaller and lacked distinct cristae structures compared to those in WT mice (Fig 8F), a mitochondrial characteristic often linked to aging and disease states [36].

We conducted lipidomic analysis on the hippocampus of WT and Cyfip2^+/R87C mice at both PWK 14 and 28. PCA revealed distinct separations in lipidomic profiles based on genotype and age (Fig 8G). A total of 745 lipid species were identified, with phosphatidylethanolamine (PE) and phosphatidylcholine (PC) being the most abundant lipid classes (Fig 8H), consistent with recent reports on the mouse brain lipidome [37]. Similar to the proteomic analysis, clustering analysis of the entire lipidome categorized them into six groups based on age-dependent relative expression levels in Cyfip2^+/R87C mice compared to WT mice (S20 Fig and S9 Table). Groups 1–3 represented lipid classes upregulated in Cyfip2^+/R87C mice compared to WT mice (Fig 8I). Specifically, group 1 lipids were upregulated at PWK 14 in Cyfip2^+/R87C mice compared to WT mice at both PWK 14 and 28, as well as compared to Cyfip2^+/R87C mice at PWK 28. Group 2 lipids were upregulated at PWK 28 in Cyfip2^+/R87C mice compared to WT mice at both PWK 14 and 28, as well as compared to Cyfip2^+/R87C mice at PWK 14. Group 3 lipids were upregulated in Cyfip2^+/R87C mice at both PWK 14 and 28 compared to WT mice at any age. PE, PC, and triacylglycerol (TG) together constituted the major lipid classes in both groups 1 (60%) and 3 (47%). However, diacylglycerol (DG) and cholesteryl ester (CE) were uniquely enriched, accounting for 14% in group 1 and 11% in group 3, respectively. Notably, TG was predominant in group 2, representing 52% of the total lipids. Group 4 included lipid classes that were downregulated in Cyfip2^+/R87C mice at both PWK 14 and 28 compared to WT mice at any age. PE and PC together accounted for 67% of the lipids in group 4, while TG was notably low, comprising only 0.51%. Group 5 represented lipid classes that were upregulated in Cyfip2^+/R87C mice compared to age-matched WT mice. Group 6 consisted of lipid classes upregulated at PWK 28 compared to PWK 14 in WT mice but remained unchanged in Cyfip2^+/R87C mice. This group was notably enriched in hexosylceramide (HexCer). Taken together, these results highlight significant alterations in the proteome, particularly involving mRNA splicing and processing, mitochondrial metabolic pathways, and in the lipidome during seizure progression in Cyfip2^+/R87C mice.

Fig 9. Summary of the seizure evolution in Cyfip2^+/R87C mice.

Seizure phenotypes in Cyfip2^+/R87C mice begin with infantile spasms during the neonatal stage (PWK 1), followed by a seizure-free latent period during the juvenile and young adult stages. Around PWK 14, Cyfip2^+/R87C mice start to exhibit spontaneous recurrent seizures, which worsen with age and are associated with premature death. This seizure evolution is accompanied by various temporal cellular and molecular changes in the brain, including alterations in neurons, glial cells, and their lipid and metabolic profiles, which may interact to contribute to the seizure phenotype.

https://doi.org/10.1371/journal.pbio.3003192.g009

Ethics statement

All mouse maintenance and experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Korea University College of Medicine (protocol no. KOREA-2024-0073) and by the IACUC of the Korea Advanced Institute of Science and Technology (protocol no. KA2022-004). All procedures were conducted in accordance with the Laboratory Animal Act of the Republic of Korea, the Animal Protection Act of the Republic of Korea, and the United States Public Health Service Policy on Humane Care and Use of Laboratory Animals.

Animals

Cyfip2^+/R87C (C57BL/6N background) and Cyfip2 heterozygous (Cyfip2^+/−, C57BL/6N or C57BL/6J background) mice used in this study have been described previously [15,22,78]. Both male and female mice were used for the experiments.

Bulk-tissue RNA sequencing and analysis

Bulk-tissue RNA sequencing and analysis were performed as described previously [15,79,80]. RNA extraction, library preparation, cluster generation, and sequencing were performed by Macrogen (Seoul, Korea). RNA samples for sequencing were prepared using a TruSeq Stranded Total RNA LT Sample Prep Kit (Illumina) according to the manufacturer’s instructions. An Illumina’s platform was used for sequencing to generate 101-bp paired-end reads. Transcript abundance was estimated in pseudo-mapping-based mode for the Mus musculus genome (GRCm38) using Salmon (v1.1.0) [81]. Differential gene expression analysis was performed using R/Bioconductor DEseq2 (v1.30.1) [82] by importing the estimated abundance data into R (v.4.1.3) using the tximport package [83]. The P values were adjusted for multiple testing with the Benjamini-Hochberg correction. Genes with an adjusted P value of less than 0.05 were considered as differentially expressed. Volcano plots were generated using the R ggplot2 (v.3.3.3) package. The Gene Ontology (GO) enrichment analyses were performed using DAVID (https://davidbioinformatics.nih.gov/) software (v2021) [84]. Mouse gene names were converted to human homologs using the Mouse Genome Informatics (MGI) database (http://www.informatics.jax.org/homology.shtml). Gene Set Enrichment Analysis (GSEA) (https://www.gsea-msigdb.org/gsea) [25] was used to determine whether gene expression of a specific set of genes changes in a consistent direction using GSEAPreranked (gsea-4.2.3.jar) module on gene set collections downloaded from Molecular Signature Database (MSigDB) (v7.5.1). GSEAPreranked was applied using the list of all genes expressed, ranked by the fold change and multiplied by the inverse of the P value with recommended default settings (1,000 permutations and a classic scoring scheme). The gene sets with a false discovery rate (FDR) of less than 0.05 were considered as significantly enriched. Integration and visualization of the GSEA results were performed using the EnrichmentMap Cytoscape App (version 3.8.1) [85,86]. The raw data have been submitted to the Gene Expression Omnibus (GEO) repository under the accession numbers GSE292102 (Cyfip2^+/− mice) and GSE292221 (Cyfip2^+/R87C mice).

Alternative splicing events analysis

We used rMATS v4.1.2 [87] to identify differential alternative splicing events (ASEs) in the cortex and hippocampus of Cyfip2^+/R87C mice compared to WT mice across different postnatal weeks (PWK 1, 7, 14, and 28). To ensure high-confidence results, we filtered out cases where the sum of inclusion junction counts (IJC) and skipping junction counts (SJC) was below 25 in either sample set. Differentially spliced exons were defined using a splicing ratio threshold of |∆PSI (Percent-Spliced-In)| > 0.2 and a FDR < 0.05.

RNA purification and real-time quantitative reverse transcription polymerase chain reaction

RNA purification and real-time quantitative reverse transcription polymerase chain reaction (qRT-PCR) were performed as described previously [16,88]. RNA was purified using the miRNeasy Minikit (QIAGEN, 217,004), and cDNA libraries were synthesized using the iScript cDNA Synthesis Kit (Bio-Rad, BR170-8,891). The target mRNAs were detected and quantified using a real-time PCR instrument (CFX96 Touch, Bio-Rad) with SYBR Green Master Mix (Bio-Rad, BT170–8884AP). The results were analyzed using comparative Ct method and normalized against the housekeeping gene Gapdh. The primer sequences are listed in S10 Table.

Western blot analysis

Brain synaptosomal lysate preparation and western blot analyses were conducted as described previously [89,90]. Briefly, mouse brains were homogenized in buffered sucrose (0.32 M sucrose, 4 mM HEPES, 1 mM MgCl₂, 0.5 mM CaCl₂, pH 7.3) with freshly added protease and phosphatase inhibitors (Sigma–Aldrich, 05892970001 and 04906837001, respectively). The homogenate was centrifuged at 900 × g for 10 min, and the resulting supernatant was centrifuged again at 12,000 × g for 15 min. The pellet was resuspended in buffered sucrose and centrifuged at 13,000 × g for 15 min, resulting in the synaptosome pellet. From each sample, 15–20 μg of protein was loaded onto 4%–15% Mini-PROTEAN TGX Precast Protein Gels (Bio-Rad, 4561084). The proteins were then transferred to a nitrocellulose membrane (GE Healthcare, 10600001). Tris-buffered saline with Tween 20 (TBST, 0.05%) solution was used for washing, and 5% skim milk in TBST was used for blocking. The western blot images were captured using the ChemiDoc Touch Imaging System (Bio-Rad) and analyzed for quantification using ImageJ software. The primary antibodies used are listed in S11 Table.

Necropsy and histology

Four WT and four Cyfip2^+/R87C mice (two males and two females per group) at PWK 20 were used for histology. Mice were euthanized and perfused with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). The gross anatomy of four major organs including heart, lungs, liver and kidney were visually examined, and the organs were subsequently harvested for fixation in 4% PFA and embedding in paraffin. Tissue blocks were sectioned at 4 μm thickness. Standard Hematoxylin and Eosin (H&E) staining was performed as previously described [91,92]. Slides were scanned using a Zeiss Axio Scan.Z1 digital slide scanner equipped with a Plan Apochromat 20×/0.8 and a high-end 3CCD progressive scan color camera (1,600 × 1,200 resolution). Representative regions of interests were imaged at 20× and 40× magnification using a Thermo Fisher EVOS FL Auto 2 imaging system. Two experts, a biologist and an anatomic pathologist, independently reviewed the slides and discussed their findings to reach a consensus.

Fluorescence immunohistochemistry

Mice were anesthetized with isoflurane and transcardially perfused with heparinized (20 units/ml) PBS, followed by 4% PFA in PBS. Brains were then extracted and post-fixed overnight in 4% PFA. After post-fixation, brain tissue was washed with PBS and cryoprotected in 30% sucrose in PBS for 48 h. The brain tissues were frozen in O.C.T compound (SAKURA Tissue-Tek, 4583) and coronally sectioned at 60 μm using a cryostat microtome (Leica, CM3050S). Sections were collected from the following approximate Bregma coordinates: mPFC: +2.245 mm to +1.645 mm; hippocampus: −1.555 mm to −2.355 mm. Whole-brain slice imaging was conducted using a Slide Scanner (AxioscanZ1 Fluorescence), and quantification analysis was referenced to the Allen Brain Atlas. For neuronal (NeuN) and glial cell (Olig2, IBA1, and GFAP) markers, we measured intensities instead of performing cell counting, as it was difficult to separately count cells due to their dense clustering and crowding. The mean fluorescence intensity in the mPFC was measured across total layers 1–5 using a 500 µm × 500 µm region of interest (ROI). In the hippocampal CA1 region, the same ROI width was applied to both WT and Cyfip2^+/R87C mice, but the ROI height was adjusted to account for structural differences and to capture dispersed NeuN, particularly in Cyfip2^+/R87C mice. All confocal images were acquired using a ×60 water objective (1.2 numerical aperture). A Z-stack interval of 0.5 µm was used for synapse and BODIPY assessments, while a 0.24 µm Z-stack interval was used for microglia and astrocyte morphology analysis. The primary and secondary antibodies used are listed in S11 Table.

Imaris quantification

EEG recordings

Electroencephalographic (EEG) recordings and quantifications were performed as described previously [22,93]. After anesthetizing the mice with a mixture of isoflurane and O₂ (4%, v/v; 1 NL per min for O₂ flow and 1.5 L per min for mixed flow) for induction, the mice were positioned on a stereotactic plate (Kopf Instruments, Model 940). The head was fixed, while respiratory anesthesia (2%, v/v) was continuously maintained. Once the skull was exposed, the bregma-lambda axis was aligned with a length of 4.2 mm. Small stainless-steel screws (1 mm × 3 mm) were then implanted on the skull. Two screws were placed in the bilateral prefrontal region (+1.8 mm AP, ±1.0 mm ML from the bregma), and two in the bilateral parietal region (−2.0 mm AP, ±1.8 mm ML from the bregma). Ground and reference screws were positioned on the cerebellum (−1.5 mm AP, ±1.5 mm ML from the lambda). The screws were connected with a connector-housing socket (Hirose Electric, H2021-ND) using soldered-coated stainless-steel wires. The complex was then fixed on the skull using dental acryl. After a week of recovery period, the mice were placed in a white opaque acrylic box (25 × 25 × 35 cm), allowing them to explore freely and to feed ad libitum under a 12-h light–dark cycle (13:00–01:00). The EEG recording was performed for 6 days using a Cheetah Data Acquisition System (Neuralynx) synchronized with video recording, 4 days of which were analyzed. The seizure properties were analyzed using Neuraview (Neuralynx), and the representative traces were exported using NeuroExplorer 5 (NeuroExplorer).

Reflection imaging for visualization of crystals

For the visualization of crystals in Cyfip2^+/R87C mice, tissue sections were fixed and stained with 4′,6-diamidino-2-phenylindole (DAPI), then imaged using Nikon A1R confocal microscope, where a combination of laser reflection and fluorescence confocal microscopy was employed to identify the crystalline materials. Reflectance images were acquired using the confocal system by adjusting the dichroic mirror to the transmitted/reflectance setting (BS20/80), with all light paths set to ‘through’.

Electron microscopy

Using brain samples from 3 WT and 3 Cyfip2^+/R87C mice at PWK 28, we employed various electron microscopy techniques, generating distinct results for each approach. For sample preparation, the mice were perfused with 2% paraformaldehyde and 2.5% glutaraldehyde in 0.15 M cacodylate buffer with 2 mM Ca²⁺. The brains were post-fixed in the same solution overnight. Fixed brains were sliced into 150 μm-thick sections using a vibratome (Leica, VT1000S) and the hippocampal CA1 region was dissected out.

To assess synaptic density, tissue blocks were prepared following transmission electron microscopy (TEM) protocols as previously described [94]. Briefly, tissue was post-fixed in 2% osmium tetroxide for 1 h, stained en bloc with 1% uranyl acetate, dehydrated, and embedded in Epon 812 resin (EMS). Seventy-nanometer-thick sections were mounted on 200-mesh grids (Gilder Grids, G200N) and post-stained with uranyl acetate followed by lead citrate. For each animal, 20 images were recorded on a Tecnai F20 TEM (Thermo Fisher Scientific) at 3,500× magnification and 120 kV. A counting frame (4 × 4 μm²) was placed on each image, and synapses with a clear postsynaptic density (PSD) and presynaptic vesicles were counted using stereological methods. Synapses were classified as asymmetric (excitatory) or symmetric (inhibitory) based on their distinct PSD shapes where asymmetric synapses have a thicker PSD and symmetric synapses have a thinner, more uniform PSD. Additionally, the dissected CA1 tissue blocks were prepared for ATUM-SEM imaging using a modified ROTO (Reduced Osmium-Thiocarbohydrazide-Osmium) protocol [15]. This process enabled the three-dimensional visualization of axonal and dendritic structures, as well as the reconstruction of their synaptic connections. The blocks were post-fixed in 2% osmium tetroxide/1.5% potassium ferrocyanide for 1 h, transferred in 1% thiocarbohydrazide for 20 min, and further fixed in 2% osmium tetroxide for 30 min. Samples were incubated in 1% uranyl acetate at 4°C overnight, followed by incubation in lead aspartate solution at 60°C for 30 min. After serial dehydration in ethanol, samples were infiltrated with acetone and embedded in 7% (w/v) conductive Epon 812 resin mixed with Ketjen black powder to enhance electrical conductivity. The specimens, mounted on the aluminum stub, were cured at 60°C for 48 h. After trimming each side of the sample at a 90-degree angle, 50 nm-thick sections were cut with an ultra-Maxi knife (DiATOME, Biel, Switzerland) and collected on plasma-hydrophilized carbon nanotube (CNT)-coated PET tapes (Boeckeler Instruments) using an ATUMtome (Boeckeler Instruments, Tucson, USA). The sections on CNT tapes were cut into strips, mounted on wafers, and imaged using a Gemini 300 SEM (Carl Zeiss Microscopy GmbH, Oberkochen, Germany). Large-area imaging was conducted using Atlas 5 software (Fibics Incorporated, Ottawa, Canada). Serial SEM imaging was performed at 5 kV with BSD detection, a dwell time of 7 µs, and a resolution of 5 nm × 5 nm per pixel in the X and Y directions, with the images stitched into a single plane. Serial images were aligned using TrackEM2, and one EM stack per brain sample was obtained for both the WT and Cyfip2^+/R87C groups, covering approximately 100 μm × 150 μm in the XY plane with 200 sections at 50 nm each. We reconstructed the dendrites of CA1 pyramidal neurons in the stratum radiatum, followed by reconstruction of the axons forming synapses with these dendrites using an open-source reconstruction software (https://synapseweb.clm.utexas.edu/software-0). Additionally, the surface area of the PSD was measured.

Correlative light and electron microscopy (CLEM)

To simultaneously detect fluorescence signals of lipid droplets and crystals in EM images, we combined confocal imaging with ATUM-SEM, utilizing the CLEM method. Hippocampal slices, 150 μm thick, were washed three times for 10 min each in 0.15 M sodium cacodylate buffer with 2 mM Ca²⁺, and then incubated with BD493/503 (Thermo Fisher Scientific, D3922) in cacodylate buffer for 30 min at room temperature. The crystal fluorescence signal was detected directly using reflection microscopy, without any additional staining. After staining, DAPI was added to visualize nuclei and blood vessels. In the CA1 region of a coronal brain slice, ROIs containing fluorescence signals were imaged using the Nikon A1R confocal microscope, specifically detecting BODIPY signal or reflected fluorescence. The slices containing the imaged ROIs were dissected and laser-cut into small blocks (PALM MicroBeam, Carl Zeiss). The dissected tissues were processed for heavy metal staining, dehydration, and resin embedding using the modified ROTO protocol [15]. The tissues were then serially sectioned at a thickness of 50 nm using an ATUMtome (Boeckeler Instruments, Tucson, USA). Low-resolution images of the serial sections were initially acquired to confirm their position relative to marked structures, such as blood vessels and nuclei. Once the target structure was identified, high-resolution (5 nm × 5 nm) serial SEM imaging was performed. Finally, the EM images were matched with the fluorescence images, and the fluorescence-positive structures were three-dimensionally reconstructed.

Single-nucleus RNA sequencing data analysis

The 10× Genomics Chromium platform was employed to prepare snRNA-seq libraries, enabling high-throughput capture and barcoding of single-nucleus transcripts. Initial data processing and quality control were performed using the Seurat R package (version 5.1.0) [95] to exclude low-quality cells and potential contaminants. Normalization was performed using LogNormalize, and 2,000 highly variable genes were selected. Dimensionality reduction was then conducted via PCA, selecting the number of PCs that explained 90% of the variance. Subsequently, neighbors were identified, and cell clustering was performed using Seurat’s FindClusters function. The results were visualized with UMAP.

Marker genes for each cell cluster were identified through pairwise differential expression analysis using MAST [96], and cell types were annotated by comparing cluster-specific markers to known expression profiles from PanglaoDB and the Human Protein Atlas [97,98]. Downstream analyses included differential expression analysis (Seurat’s FindMarkers function with MAST method, Log₂ fold change > 1, adjusted P value < 0.05), gene set enrichment analysis (fgsea R package (version 1.27.0)) [99], Over representation analysis (enrichR (version 0.0.31)) [100], differential abundance analysis of cell neighborhoods (Milo R package (version 3.2) [101], DAseq R package (1.0.0)) [102], Transcription factor analysis (SCENIC R package (version 1.3.1)) [32]. Cell–cell communication analysis was performed using NATMI [103], Connectome [104], logFC Mean [105], and SingleCellSignalR [106] methods. The results from these methods were aggregated and ranked [107] using the liana R package (version 0.1.14) [108], enabling comprehensive prioritization of ligand-receptor interactions. The predicted ligand-receptor pairs were ranked by aggregate rank, and their distribution was plotted to identify potential ligands. Target genes were then identified using the nichenetr R package (version 2.2.0) [33]. The raw data has been submitted to the Gene Expression Omnibus (GEO) repository under the accession number GSE293769.

Sample preparation for proteomic analysis

Hippocampal tissues were isolated from WT and Cyfip2^+/R87C brains at PWK 14 and 28 and homogenized and lysed with the 1× sodium dodecyl sulphate (SDS) buffer (5% SDS, 50 mM triethylamonium bicarbonate (TEAB), pH 8.5). A 500 mg of proteins were reduced and alkylated with final 10 mM Tris (2-carboxyethyl) phosphine (TCEP) and 20 mM indole 3-acetic acid (IAA). Tryptic digestion was performed using an S-trap mini digestion kit (ProtiFi, Huntington, NY) following the provided protocol. In this case, 16.7 mg amounts of mass-spectrometry-grade Trypsin Gold (Promega, Madison, WI) were used for each sample (1:30). The final eluted samples were dried in a speed vacuum. The dried peptides were quantified with a Pierce quantitative colorimetric peptide assay kit (Thermo Fisher Scientific, Rockford, IL). A 300 mg of trypsin-digested peptides from each sample were labeled using 18-plex TMT reagent (Thermo Fisher Scientific, Rockford, IL) according to the manufacturer’s instructions. Peptides from WT mice at PWK 14 were labeled with the 126, 127N, 127C, 128N tags and peptides from Cyfip2^+/R87C mice at PWK 14 were labeled with the 128C, 129N, 129C, 130N tags. Peptides from WT mice at PWK 28 were labeled with the 130C, 131N, 131C, 132N tags and peptides from Cyfip2^+/R87C mice at PWK 28 were labeled with the 132C, 133N, 133C, 134N tags. For quality control in TMT experiments, reference samples were prepared by pooling of same amount of all samples and labeled with 134C and 135N TMT tags. Each TMT channel was freshly dissolved in anhydrous acetonitrile (ACN) with a ratio of 0.5:20 (w:v, mg:mL). After incubation for 1 h at room temperature, the reaction was quenched by adding 5 mL of 5% hydroxylamine and incubated for 15 min. Then, samples were combined, dried, and desalted with Pierce peptide desalting spin columns (Thermo Fisher Scientific, Rockford, IL). Total peptides were fractionated into 20 fractions by basic reverse phase liquid chromatography and each eluted peptide sample was vacuum dried. For the liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis, fractionated peptides were diluted by mobile phase A (99.9% water with 0.1% FA).

LC-MS/MS and protein identification for total proteomic analysis

Dissolved samples were analyzed using an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific) coupled to an UltiMate 3,000 RSLCnano system (Thermo Fisher Scientific) equipped with a nano electrospray source. Samples were trapped on 75 mm × 2 cm C18 precolumn (nanoViper, Acclaim PepMap100, Thermo Fisher Scientific) before being separated on an analytical C18 column (75 mm × 50 cm PepMap RSLC, Thermo Fisher Scientific) for 140 min at a flow rate of 250 nL/min. The mobile phases A and B were composed of 0 and 99.9% acetonitrile containing 0.1% formic acid, respectively. The LC gradient started with 5% B for 20 min. It was ramped to 13% of B buffer for 50 min followed by a gradient from 13% to 25% B buffer for 65 min, then 25% to 95% for 5 min, and 95% for 5 min, after which they were equilibrated for 15 min with 5% B buffer for next run. The voltage applied to produce an electrospray was 2,000 V. During the chromatographic separation, the Orbitrap Exploris 480 was operated in data-dependent mode, automatically switching between MS1 and MS2. The full scan resolution was 120,000 at m/z 400. The MS2 scans were performed with HCD fragmentation (37.5% collision energy). Previously fragmented ions were excluded for 30 s within 10 ppm. The electrospray voltage was maintained at 2.0 kV, and the capillary temperature was set to 275°C.

MS/MS spectra were analyzed using the following software analysis protocol with the Uniprot mouse database. The reversed sequences of all proteins were added to the database for calculation of FDR. ProLuCID [109] in Integrated Proteomics Pipeline software; IP2 (Integrated Proteomics Applications, San Diego, CA, USA) was used to identify the peptides, with a precursor mass tolerance of 20 ppm, and a fragment ion mass tolerance of 200 ppm. The output data files were filtered and sorted to compose the protein list with two and more peptides assignments for protein identification at a false positive rate less than 0.01. TMT reporter ion analysis was established with Census [110] in Integrated Proteomics Pipeline software within mass error of 20 ppm. Statistical analysis was performed with Perseus [111] software (version 1.6.15) The expressions of proteins between samples were compared using Welch’s t test with P value set at <0.05. All raw MS data files from this study have been deposited to the repository MassIVE with identifier PXD059911.

Proteomic data enrichment analysis

Proteomic changes in the hippocampus of Cyfip2^+/R87C mice compared to WT mice were classified into four groups based on relative expression levels with mouse age using clustering analysis. For each protein group, Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed using DAVID (https://davidbioinformatics.nih.gov/) software (v2021) [84]. To further capture the relationships between the enriched terms, Metascape (https://metascape.org/gp/) v3.5.20240901) [112] was applied and a subset of enriched terms has been selected and rendered as a network plot, where terms with a similarity >0.3 are connected by edges. The top 20 clusters consisting of terms with the highest P values were displayed, constrained to no more than 15 terms per cluster and no more than 250 terms in total. Protein–protein Interaction enrichment analysis was also performed using Metascape. Only physical interactions in STRING (physical score > 0.132) and BioGrid are used, and the resultant network contains the subset of proteins that form physical interactions with at least one other member in the list. If the network contains between 3 and 500 proteins, the Molecular Complex Detection (MCODE) algorithm has been applied to identify densely connected network components and the three best-scoring terms by P value have been retained as the functional description of the corresponding components.

Lipidomic analysis

Global lipidomic analysis of hippocampus was performed using ultra-performance liquid chromatography (UPLC) coupled with trapped ion mobility (TIMS) time-of-flight (TOF). Weighed hippocampus tissue samples were placed into homogenizer tubes, each containing three 2.8 mm zirconium oxide beads. For solvent extraction, 1.2 mL of 50% methanol/chloroform (1:1, v/v) was added to each tube, based on a tissue weight of 25 mg, ensuring a consistent weight-to-solvent ratio across all samples, after which the tissues were homogenized. After homogenization, the samples were left to stand at 4°C for 10 min, followed by centrifugation at 12,700 rpm for 20 min. The lower lipid phase was aliquoted into 100 μL, and the solvents were removed using nitrogen gas. The dried extracts were reconstituted in 200 μL of 80% isopropanol, vortexed for 1 min. They were then centrifuged at 12,700 rpm for 10 min at 4°C, and 100 μL of the supernatants transferred to vials for lipidomic analysis.

Untargeted lipid profiling was performed using high-performance liquid chromatography (HPLC; 1290 Infinity; Agilent) coupled with trapped ion mobility spectrometry time-of-flight (TIMS-TOF; Bruker). The column oven and auto-sampler temperatures were maintained at 55°C and 4°C, respectively. The samples were eluted and separated using an Acquity UPLC CSH C18 column (2.1 × 100 mm with 1.7-μm particles; Waters). The LC mobile phase comprised, in positive ion mode, 10 mM ammonium formate and 0.1% formic acid in an acetonitrile/water mixture (6:4, v/v) (solvent A) and 0.1% formic acid in an acetonitrile/isopropanol mixture (1:9, v/v) (solvent B). In negative ion mode, the LC mobile phase comprised 10 mM ammonium acetate in an acetonitrile/water mixture (4:6, v/v) (solvent A) and an acetonitrile/isopropanol mixture (1:9, v/v) (solvent B). Gradient elution began with 40% B, increased to 43% B by 2 min, and then gradually increased to 50% B by 2.1 min. The composition was further increased to 54% B by 12 min, followed by a continuous increase to 99% B by 18 min. The composition was held at 99% B for 0.1 min, it was returned to initial conditions after 18.1 min, and was finally maintained for 2 min until equilibration. The flow rates were set to 400 μL/min. For lipid analysis, 2 μL of lipid extract was injected for positive ion mode, and 4 μL was injected for negative ion mode into the HPLC/TIMS-TOF system. All samples were pooled in equal amounts to generated a quality control (QC) sample. For MS analysis, the mass range was set to m/z 50–1,500. The operating parameters were as follows: capillary voltage of 4,500 V, end plate offset of 500 V, nebulizer pressure of 2.2 bar for positive ion mode and 2.0 bar for negative ion mode, drying gas flow rate of 10.0 L/min, and source temperature of 220°C.

The acquired data were processed using MetaboScape 5.0 (Bruker) to alignment and peak picking, and identified utilizing the MSDIAL-TandemMassSpectralAtlas-VS69 library. Principal component analysis (PCA) was performed to visualize scatter score plots using SIMCA-P+ software, version 16.0 (Umetrics, Umea, Sweden). Statistical analyses were conducted using Statistical Package for the Social Sciences software, version 28.0 (SPSS, Armonk, NY), and data were plotted using GraphPad Prism, version 8.0 (GraphPad Software, San Diego, CA). The z-scores were calculated based on the entire sample set. The Kruskal–Wallis U test was conducted to evaluate overall differences of z-scores among the groups. For pairwise comparisons, the Mann–Whitney test was performed, and Bonferroni correction was applied to adjust for multiple comparisons. The results of the lipidomic analysis were deposited in the Korea BioData Station (K-BDS) (BioProject: KAP241486).

S1 Fig. Representative whole-body images of mice at postnatal week 28 (PWK 28), along with comparisons of body weights.

Scale bar, 2 cm (n = 9–32 mice per genotype, unpaired two-tailed Student t test). N = C57BL/6N background, J = C57BL/6J background. ***P < 0.001. Data are represented as mean ± standard error of the mean. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s001

(TIF)

S2 Fig. Locomotor and general activity of Cyfip2^+/R87C mice during the final 3–4 days before death.

(A) Percentage of time spent in locomotor activity for five Cyfip2^+/R87C mice (thin lines) and their average (thick red line) over the final 3–4 days before death. On the day of death, the exact time of death differed among individual mice. (B) Percentage of time spent in general activity of Cyfip2^+/R87C mice during the final 3–4 days before death. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s002

(TIF)

S3 Fig. Normal histology of four major organs in Cyfip2^+/R87C mice at postnatal week 20.

(A) Representative histological images of WT (upper panels) and Cyfip2^+/R87C (lower panels) mouse hearts show intact myocardial architecture with well-organized cardiac muscle fibers and normal myocyte morphology in both genotypes. The left ventricular (LV) and atrial (LA) walls, mitral valve (MV), and intramyocardial vessels appear unremarkable. No pathological changes, including but not limited to, inflammatory cell infiltration, cardiomyocyte hypertrophy or degeneration, fibrosis, necrosis, calcification, or valve thickening, are observed in either WT or Cyfip2^+/R87C mice. Scale bars, 500 µm (10× magnification, left panels) and 100 µm (40× magnification, right panels). (B) Representative histological images of lungs show normal pulmonary architecture in both genotypes. Alveolar spaces are well-expanded with thin alveolar septa (AS), and bronchioles (Br) and alveolar ducts (AD) appear unremarkable. No pathological changes such as infection, inflammation, hemorrhage, septal thickening, edema, or fibrosis are observed in either genotype. (C) Representative liver histology shows well-preserved hepatic architecture in both genotypes. Hepatocytes are arranged in uniform plates radiating from central veins, with clear cytoplasm and round, centrally located nuclei. No pathological changes such as inflammatory cell infiltration, steatosis, hepatocellular degeneration, necrosis, fibrosis, or cholestasis are observed in either genotype. (D) Representative kidney histology shows normal renal architecture in both genotypes. The cortex exhibits well-defined glomeruli (G) with intact capillary loops and surrounding Bowman’s capsules, along with normal proximal and distal tubules. The medulla shows preserved tubular structures and collecting ducts without signs of congestion or degeneration. No pathological changes such as glomerular hypercellularity, tubular atrophy, interstitial inflammation, fibrosis, or vascular abnormalities are observed in either genotype.

https://doi.org/10.1371/journal.pbio.3003192.s003

(TIF)

S4 Fig. qRT-PCR analysis of genes encoding synaptic proteins in the cortex and hippocampus of Cyfip2^+/− mice compared to WT mice at postnatal week 28 (PWK 28).

No significant differences were observed between genotypes (n = 6 mice per genotype, unpaired two-tailed Student t test). N = C57BL/6N background. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s004

(TIF)

S5 Fig. Imaging and quantification of excitatory synapses in the hippocampus.

A schematic diagram illustrating the hippocampal CA1 region, where the confocal images were acquired. The images on the bottom show examples of converting the hippocampal confocal fluorescence images into spot images using Imaris rendering for WT and Cyfip2^+/R87C mice. Scale bar, 10 μm.

https://doi.org/10.1371/journal.pbio.3003192.s005

(TIF)

S6 Fig. Changes in excitatory synapses in the mPFC of Cyfip2^+/R87C mice.

(A) Schematic diagram illustrating the prelimbic medial prefrontal cortex (mPFC), where the confocal images were acquired. The images on the right show examples of converting the mPFC confocal fluorescence images into spot images using Imaris rendering for WT and Cyfip2^+/R87C mice. Scale bar, 10 μm. (B) Representative fluorescence immunohistochemistry images and quantification show changes in excitatory presynaptic (vGluT1) and postsynaptic (Homer1) markers in the prelimbic mPFC of Cyfip2^+/R87C mice compared to WT mice at PWK 14 and 28 (n = 7–9 mice per genotype, two-way ANOVA with Šídák’s multiple comparisons test). Images for automated foci counting, obtained using Imaris software, are also shown. A.U. = arbitrary units. *P < 0.05. Data are represented as mean ± standard error of the mean. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s006

(TIF)

S7 Fig. Changes in axonal myelination in the hippocampus of Cyfip2^+/R87C mice.

(A) Representative 3D electron microscopy images of axonal segments in the hippocampal CA1 region from WT and Cyfip2^+/R87C mice at PWK 28. The graph displays the g-ratio (axon diameter/total myelinated fiber diameter) for WT and Cyfip2^+/R87C mice (n = 36–41 axons per genotype, unpaired two-tailed Student t test). (B) Representative fluorescence immunohistochemistry images and corresponding quantification show increased myelin basic protein (MBP) mean intensity in the hippocampus of Cyfip2^+/R87C mice compared to WT mice at PWK 28 (n = 8–9 mice per genotype, unpaired two-tailed Student t test). DG = dentate gyrus, SO = stratum oriens, SP = stratum pyramidale, SR = stratum radiatum. **P < 0.01; ***P < 0.001. Data are represented as mean ± standard error of the mean. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s007

(TIF)

S8 Fig. Neuronal and glial changes in the mPFC of Cyfip2^+/R87C mice during seizure evolution.

Representative fluorescence immunohistochemistry images and quantifications show age-dependent changes in the intensities of neuronal and glial cell markers in the prelimbic medial prefrontal cortex (mPFC) of Cyfip2^+/R87C mice compared to WT mice (n = 7–9 mice per genotype, two-way ANOVA with Šídák’s multiple comparisons test). The results of the same analysis for Cyfip2^+/− mice (on either the C57BL/6N (N) or C57BL/6J (J) backgrounds) at PWK 28 are also presented (n = 6–9 mice per genotype, two-way ANOVA with Šídák’s multiple comparisons test). *P < 0.05; ***P < 0.001. Data are represented as mean ± standard error of the mean. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s008

(TIF)

S9 Fig. NeuN and cleaved Caspase3 co-staining in the CA1 region of WT and Cyfip2^+/− mice.

Co-staining for NeuN and cleaved Caspase3 was performed at PWK 1, 7, 14, and 20. Quantitative analysis of the fraction of cleaved Caspase3-positive area within the NeuN-positive region showed no significant differences between WT and Cyfip2^+/R87C mice. At PWK 20, however, small NeuN- and cleaved Caspase3-double-positive spots were frequently observed in Cyfip2^+/R87C mice but not in WT mice, potentially representing debris from dead NeuN-positive neurons. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s009

(TIF)

S10 Fig. Neuronal and glial changes in the hippocampus of Cyfip2^+/− mice.

Changes in the intensities of neuronal and glial cell markers in the hippocampal CA1 region of Cyfip2^+/− mice at PWK 28 (n = 7–9 mice per genotype, two-way ANOVA with Šídák’s multiple comparisons test). SO = stratum oriens, SP = stratum pyramidale, SR = stratum radiatum. **P < 0.01. Data are represented as mean ± standard error of the mean. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s010

(TIF)

S11 Fig. Changes in morphology of microglia in the mPFC of Cyfip2^+/R87C mice.

Representative 3D images of microglia in the prelimbic mPFC of WT and Cyfip2^+/R87C mice at PWK 14 and 28, processed using Imaris software. Graphs show quantifications from Sholl analysis as well as measurements of branch length, area, points, and diameter (n = 7–8 mice per genotype, two-way ANOVA with Šídák’s multiple comparisons test). *P < 0.05; **P < 0.01; ***P < 0.001. Data are represented as mean ± standard error of the mean. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s011

(TIF)

S12 Fig. Changes in morphology of astrocytes in the mPFC of Cyfip2^+/R87C mice.

Representative 3D images of astrocytes in the mPFC prelimbic region of WT and Cyfip2^+/R87C mice at PWK 14 and 28, processed using Imaris software. Graphs show quantifications from Sholl analysis as well as measurements of branch length, area, points, and diameter (n = 7–8 mice per genotype, two-way ANOVA with Šídák’s multiple comparisons test). *P < 0.05; ***P < 0.001. Data are represented as mean ± standard error of the mean. The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s012

(TIF)

S13 Fig. Lysosomal lipid droplets in astrocytes of WT and Cyfip2^+/R87C hippocampus at PWK 14.

Representative 3D images show astrocytes (GFAP), lysosome (Lamp1) within the astrocytes, and lipid droplets (LDs, BODIPY) within astrocytic lysosome (indicated by white arrows in merged images) in the hippocampal CA1 region of WT and Cyfip2^+/R87C mice at PWK 14, analyzed using Imaris software. The Graph shows quantification of the number of LDs within astrocytic lysosomes in WT and Cyfip2^+/R87C mice (n = 7–8 mice per genotype, unpaired two-tailed Student t test). The data underlying this Figure can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003192.s013

(TIF)

S14 Fig. Various types of lipid droplets observed in the hippocampal astrocytes of Cyfip2^+/R87C mice.

Serial electron microscopic images of various types of lipid droplets (within lysosome, with or without crystals) in the hippocampal astrocytes of Cyfip2^+/R87C mice at PWK 28. Schematic diagrams for each type of lipid droplet are provided on the left side.

https://doi.org/10.1371/journal.pbio.3003192.s014

(TIF)

S15 Fig. Single-nucleus transcriptomic analysis on the hippocampus of Cyfip2^+/R87C mice.

(A) Violin plots showing quality control metrics for each sample: the total number of Unique Molecular Identifier (UMI) counts (left), the number of observed genes (middle), and percent of mitochondrial genes (right). KI, knock-in. (B) UMAP visualization of all hippocampal cells, grouped by sample. Each sample is represented by a distinct color. (C) Volcano plot showing global differential expression (DE) analysis results for all hippocampal cells based on genotype. (D) Subclustering results of neurons. The left panel presents a UMAP visualization of neuronal subtypes, while the right panel shows a dot plot of cell-specific marker gene expression, with dot size representing the percentage of cells expressing the gene and color intensity indicating average expression levels. (E) Subclustering results of excitatory neurons. The left panel displays a UMAP visualization of excitatory neuronal subtypes, while the right panel features a dot plot highlighting cell-specific marker genes. (F) Subclustering results of pyramidal neurons. The left panel presents a UMAP visualization of neuronal subtypes, while the right panel shows a dot plot of cell-specific marker gene expression. (G) Bar plot showing significant GSEA results for CA1-1 compared to CA1-0. The Y-axis represents Gene Ontology Biological Process (GOBP) terms, while the X-axis represents Normalized Enrichment Scores (NES). Color intensity reflects the significance level, represented by −Log₁₀ (adjusted P value).

https://doi.org/10.1371/journal.pbio.3003192.s015

(TIF)

S16 Fig. Single-nucleus transcriptomic analysis on the glial cells in the hippocampus of Cyfip2^+/R87C mice.

(A) Subclustering results of oligodendrocytes. The left panel presents a UMAP visualization of oligodendrocyte subtypes, while the right panel features a dot plot highlighting the top DEGs for each cluster. (B) Subclustering results of microglia. The left panel presents a UMAP visualization of microglia subtypes, while the right panel features a dot plot highlighting the top DEGs for each cluster.

https://doi.org/10.1371/journal.pbio.3003192.s016

(TIF)

S17 Fig. Ligand-target gene predictions between CA1-1 and Astrocyte-4.

(A) Regulatory potential of ligands (Adam10, App, Bdnf, and Inhba) from CA1-1 on 12 predicted target genes expressed in Astrocyte-4. (B) Expression level comparisons of the 12 predicted target genes between Astrocyte-1 and Astrocyte-4. (C) Regulatory potential of Angpt1, App, Mif, and Psap from Astrocyte-4 on their 28 target genes expressed in CA1-1. (D) Over-Representation Analysis (ORA) result for the 28 target genes. (E) Expression level comparisons of the 28 target genes between CA1-0 and CA1-1.

https://doi.org/10.1371/journal.pbio.3003192.s017

(TIF)

S20 Fig. Heatmap representing lipid species clustered into six groups based on age-dependent relative expression levels in Cyfip2^+/R87C mice compared to WT mice.

Z-scores are significantly different among different groups (n = 9–14 mice per genotype, Kruskal–Wallis test).

https://doi.org/10.1371/journal.pbio.3003192.s020

(TIF)

S5 Video. Mild seizure-like behavior in Cyfip2^+/R87C mice during the terminal stage.

Representative video showing a Cyfip2^+/R87C mouse during the final days of life. Despite markedly reduced activity, occasional seizure-like behaviors were observed, limited to mild manifestations such as forelimb extension.

https://doi.org/10.1371/journal.pbio.3003192.s036

(MP4)