{"paper_id":"0f08a601-7766-4e3c-8bc4-ea2f919beb00","body_text":"Seizures, increased interhemispheric synchrony, altered brain \ntranscriptomics and a leaky blood-brain barrier result from loss \nof ap3b2 in a CRISPR tadpole model of DEE48. \nSulagna Banerjee1, Cabriana W. Earl1, Samuel C. Robson2,3, Paul Szyszka1, Caroline W. \nBeck*1,4 \n1Department of Zoology, University of Otago, Dunedin, New Zealand \n2Institute of Life Sciences and Healthcare, School of Medicine, Pharmacy and Biomedical Sciences, \nUniversity of Portsmouth, PO1 2DT, UK \n3Centre for Enzyme Innovation, School of Earth and the Environment, University of Portsmouth, \nPO1 2DT, UK \n4Genetics Otago Research Centre, University of Otago, Dunedin, New Zealand \n* Correspondence:  \nCaroline Beck \ncaroline.beck@otago.ac.nz  \n \nKeywords: Developmental and Epileptic Encephalopathy, Xenopus laevis, GCaMP6s, ap3b2, \nCRISPR, seizure, model organism, GABA pathway.  \nAbstract \nLoss-of-function variants in AP3B2, a neuronal adaptor protein required for synaptic vesicle \nformation, cause a severe early-onset neurodevelopmental epilepsy known as Developmental and \nEpileptic Encephalopathy 48 (DEE48). To investigate how AP3B2 loss alters brain development, \nleading to increased seizure susceptibility, we generated a Xenopus laevis model by targeting the \northologous gene using CRISPR/Cas9. Ap3b2.S-/- (mosaic) F0 tadpoles displayed increased \nlocomotor activity with frequent seizure-like episodes when compared to sibling controls. \nVisualisation of forebrain and midbrain activity using the genetically encoded Ca2+ sensor GCaMP6s \ndetected spontaneous, large amplitude, prolonged and widespread neural activity, alongside increased \ninterhemispheric synchrony of both regions. Comparison of whole-brain transcriptomes from ap3b2 \nCRISPants and unedited sibling controls detected mainly downregulation of brain expressed genes, \nwith significant over-representation of pathways involved in ion transport, axon formation and \nguidance, inhibitory (GABA) neurotransmission, and transport across the blood-brain barrier (BBB). \nIn a simple assay for BBB integrity, CRISPant tadpoles were confirmed to have faster leakage of \nsodium fluorescein. Acute exposure to the angiotensin receptor blocker losartan significantly reduced \nlocomotor hyperactivity, and CRISPant cohorts treated with losartan tended to have lower neural \nactivity, indicating incomplete rescue of the ap3b2.S CRISPant phenotype. These findings \ndemonstrate how AP3B2 loss of function alters brain development and the establishment of the BBB, \nwith the resulting alterations in neurotransmitter pathways predisposing the brain to spontaneous \nseizures. Our results suggest that traditional anti-seizure medications designed to alter ion transport \nand GABA metabolism could be augmented with drugs targeting neuroinflammation, as adjunct \nseizure control options in infants with DEE48.    \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n2 \n1 Introduction \nDevelopmental and epileptic encephalopathies (DEEs) are a heterogeneous group of severe, early-\nonset epilepsies that arise from genetic abnormalities and profoundly affect neurodevelopment. They \nare characterized by frequent, often refractory seizures, accompanied by global developmental delay, \ncognitive impairment, and behavioural regression (Scheffer and Liao, 2020, Scheffer et al., 2024). A \ndefining feature is the association between epileptiform EEG activity and developmental regression, \nunderscoring the dynamic interplay between seizure activity and brain maturation (Scheffer et al., \n2024). Advances in molecular genetics have revealed that DEEs reflect perturbations across diverse \nneuronal and developmental pathways. According to the Online Mendelian Inheritance in Man \ncatalogue (OMIM, https://www.omim.org), 119 genes are currently recognized as causative, but \nemerging studies suggest that variants in more than 800 genes may contribute to the broader DEE \nspectrum (Poke et al., 2023). These discoveries link DEEs to both neurodevelopmental and \nneurodegenerative mechanisms, offering new insight into disease pathogenesis (Riva et al., 2025). \nAlthough individual DEE syndromes are rare, their collective incidence approaches 1 in 590 children, \nmaking them a major cause of early-life neurological disability (Poke et al., 2023, Symonds et al., \n2021). \nAmong the expanding number of genes implicated in DEEs, AP3B2 exemplifies how disruption of \nsynaptic vesicle trafficking can result in severe and early neurodevelopmental failure. Biallelic \npathogenic variants in AP3B2 were first identified by Assoum et al. through whole-exome \nsequencing of individuals with early-onset epileptic encephalopathy, defining developmental and \nepileptic encephalopathy-48 (DEE48 OMIM#617276) (Assoum et al., 2016). Analysis of a cohort of \n12 individuals with DEE48 from 8 families revealed three nonsense mutations (p.Arg67*, p.Glu152* \nand p.958*), two frameshift mutations, each -4 bp (p.Leu841Glnfs*10 and p.Thr1060Ser*7), and \nthree splice variants deleting exons 10 and 14 (Assoum et al., 2016) (Figure1). Seizure onset \noccurred within the first year of life, ranging from birth to early infancy, and included infantile \nspasms, myoclonic seizures, and generalized or focal seizure types, frequently accompanied by \nmarkedly abnormal EEG patterns such as hypsarrhythmia. Neurodevelopmental impairment was \nprofound and often evident prior to or independent of seizure onset, with severe hypotonia, absent or \nminimal speech, poor visual engagement frequently associated with optic atrophy, and postnatal \nmicrocephaly, while early brain MRI findings were often unremarkable (Assoum et al., 2016). \nSubsequent case series have reinforced the consistency and severity of the DEE48 phenotype. Two \nfurther individuals with homozygous truncating frameshift variants in AP3B2 (p.Ala149Serfs*34 and \np.Pro993Argfs*5) presented with refractory seizures beginning at approximately 3–4 months of age, \nsevere global developmental delay, hypotonia, stereotypic movements, postnatal microcephaly, and \nprogressive intellectual disability (Dilber et al., 2022). Genomic analysis of a cohorts enriched for \nearly-onset epilepsy and intellectual disability, identified two cases of the frameshift \np.Glu613Ser*182 (Anazi et al., 2017). More recently, a novel homozygous AP3B2 missense variant \n(p.Val106Ile) was identified in a child with neonatal-onset seizures, microcephaly, developmental \nregression, and electroclinical features consistent with DEE48, extending the allelic spectrum while \npreserving the core clinical phenotype (Alizadeh et al., 2025). Together, these studies show that \nhomozygous loss of function of AP3B2 causes DEE. \nAP3B2 encodes the β3B subunit of the neuronal adaptor-protein-3 (AP-3) complex, which mediates \ncargo selection and vesicle budding from endosomes (Dell'Angelica et al., 1997, Faúndez et al., \n1998). The neuronal AP-3 complex, enriched in axons and presynaptic terminals, is essential for \nsynaptic vesicle biogenesis (Blumstein et al., 2001). Loss of AP3B2 disrupts the targeting of key \nvesicular proteins such as ZnT3 and ClC-3, causing altered vesicle composition and reduced \nvesicular zinc content, which compromises synaptic transmission (Seong et al., 2005) and leads to \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n3 \nhyperactivity, seizures, and synaptic dysfunction in Ap3b2⁻/⁻ mice (Nakatsu  et al., 2004). \nSubsequent studies have expanded the AP3B2 mutational spectrum, identifying novel homozygous \nand compound heterozygous loss-of-function (LOF) variants (Figure 1) associated with autosomal-\nrecessive DEE48 (Anazi et al., 2017, Dilber et al., 2022, Alizadeh et al., 2025). \nDespite rapid progress in genetic diagnosis, effective therapies for DEEs remain limited, highlighting \nthe need for experimental models that permit direct observation and manipulation of early \nneurodevelopmental processes in vivo. To address this, simple vertebrate systems such as the \nXenopus tadpole have emerged as powerful platforms for functional analysis of DEE-associated \ngenes. Early chemoconvulsant-based Xenopus laevis seizure models enabled electrophysiological and \nbehavioural quantification of seizure activity (Hewapathirane et al., 2008, Panthi et al., 2024, Bell et \nal., 2011). More recently, CRISPR/Cas9-mediated gene editing has allowed creation of tadpole \nmodels of NeuroD2 haploinsufficiency in both X. laevis and X. tropicalis that recapitulate the \nphenotype of DEE72, facilitating rapid assessment of pathogenic mechanisms and therapeutic \nresponses (Banerjee et al., 2024, Sega et al., 2019). \n \nFigure 1. Summary of reported AP3B2 mutations and inheritance patterns in patients. (a) \nSchematic representation of the human AP3B2 protein showing the locations and types of reported \npathogenic variants. The conserved Adaptin N terminal (Adaptin_N) and Clathrin-adaptor complex-3 \nbeta-1 subunit C-terminal (AP3B1_C) domains are indicated, with frameshift (green squares), \nmissense (blue circle), nonsense (pink triangles), and splice-site (orange diamonds) mutations \nannotated along the protein. (b) Summary of reported inheritance patterns across 24 documented \ncases, 75% were homozygous for AP3B2 variants, while 25% are compound heterozygous. The \nDEE48 pathogenic variants in AP3B2 shown here were described in Assoum et al. (2016), Anazi et \nal. (2016), Dilber et al. (2022) and Alizadeh et al. (2025). \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n4 \nTo investigate the aetiology of DEE48 and further understanding of how loss of function mutations \nin AP3B2 increase seizure susceptibility, we generated a Xenopus laevis tadpole model by targeting \nthe orthologous gene using CRISPR/Cas9. Xenopus are well suited for this purpose, due to efficient \nCRISPR introduction of disruptive frameshift indels, easy access to the developing tadpole brain and \na drug permeable skin (Li et al., 2022, Banerjee et al., 2024).We mined data from a previous study of \nbrain transcriptomics in this species (Ta et al., 2021) and found that ap3b2.S is expressed in the \ndeveloping tadpole brain. Ap3b2.S-/- (mosaic) CRISPants replicated the seizure phenotype previously \nshown in the mouse model (Nakatsu  et al., 2004). These tadpoles showed seizure-like behaviour, \ndefined as increased mean swimming velocity and runs of C-starts with abrupt directional changes \n(darting), compared to unedited sibling controls.GCaMP6s imaging of ap3b2 CRISPant tadpoles \nrevealed increased frequency and amplitude of Ca²⁺ events with increased cross-regional synchrony, \nconsistent with hypersynchronous epileptic activity. Despite a grossly normal brain morphology, \nsodium fluorescein tracking demonstrated early and pronounced blood-brain barrier (BBB) leakage. \nTranscriptomic analysis of the tadpole brain showed prominent dysregulation of both neuronal \nsignalling and development pathways as well as evidence of altered neuroinflammatory markers. \nLosartan, an angiotensin-II receptor blocker previously shown to reduce seizures in NeuroD2 \n(DEE72) CRISPants (Banerjee et al., 2024), significantly suppressed swimming velocity. \nCollectively, these findings establish the DEE48 X. laevis CRISPant tadpole as a rapid, scalable, and \nphysiologically relevant vertebrate model for dissecting AP3B2-associated epileptic encephalopathy \nand evaluating targeted interventions.  \n2 Materials and methods \n2.1 Production and maintenance of Xenopus laevis embryos. \nAdult Xenopus laevis frogs were maintained in temperature-controlled aquaria under standard \nhusbandry conditions and handled in accordance with institutional animal ethics requirements. All \nprocedures were approved by the University of Otago Animal Ethics Committee under protocols \nAUP22/12 and 22/24. To induce ovulation, adult females were primed by injection of human \nchorionic gonadotropin (hCG; 500 IU per 75 g body weight) into the dorsal lymph sac 16 hours \nbefore egg collection. Primed females were housed overnight in pairs in small holding tanks \ncontaining “frog water” (carbon-filtered tap water). Once egg laying commenced, each female was \ntransferred to individual tanks containing 1 L of 1× Marc’s Modified Ringers (MMR; 10 mM NaCl, \n0.2 mM KCl, 0.1 mM MgSO₄·6H₂O, 0.2 mM CaCl₂, 0.5 mM HEPES, 10 µM EDTA, pH 7.8), and \neggs were collected hourly. Eggs were fertilized in vitro using a sperm suspension prepared from \nfreshly isolated testes of a euthanized adult Xenopus laevis male. Fertilized eggs were left \nundisturbed until embryo rotation occurred (15–20 min), then dejellied in 2% L-cysteine (pH 7.9).  \nEmbryos were maintained at 14–18 °C, monitored regularly, and staged according to the Nieuwkoop \nand Faber (NF) normal table (Nieuwkoop and Faber, 1994). \n2.2 CRISPR/Cas9-mediated targeting of ap3b2.S. \nUsing published transcriptomic datasets (Ta et al., 2021), the ap3b2.S homeologue was confirmed to \nbe expressed in X. laevis brain tissue and identified as the ortholog of human AP3B2. Guide RNAs \ntargeting exonic regions of ap3b2.S were designed using ChopChop (Labun et al., 2016) \n(https://chopchop.cbu.uib.no/) and evaluated in InDelphi (Shen et al., 2018) \n(https://indelphi.giffordlab.mit.edu/) to confirm predicted on-target efficiency and minimize off-\ntarget editing (Supplementary figures S1–S2). Two sgRNAs, ChopChop rank 105 (sgRNA2) and \nrank 84 (sgRNA3), were selected to disrupt ap3b2.S (henceforth described as ap3b2). The sgRNA \nsequences and PCR primer sets used for genotyping are listed in Table 1. A scrambled control \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n5 \nsgRNA (CTTGTAGATCAGGTGCAAGCTGG) was designed using the method of (Hsu et al., \n2013). BLAST analysis confirmed that this scrambled sequence had no predicted targets in the X. \nlaevis genome. For sgRNA synthesis, long 54–55 bp oligonucleotides were designed using the \nEnGen sgRNA Template Oligo Designer (NEB), incorporating the 20-nt ChopChop target sequence \n(excluding the NGG PAM). A 5′ G was added when absent to optimize transcription efficiency. \nOligonucleotides were synthesized by IDT and used with the EnGen sgRNA Synthesis Kit, S. \npyogenes (NEB). Synthesized sgRNAs were dissolved in nuclease-free water, aliquoted, stored at -80 \n°C, and thawed on ice prior before injections. Immediately prior to use, 0.3 µL of EnGen S. pyogenes \nCas9-NLS protein was added to the sgRNA and incubated at 37 °C for 5 minutes to form Cas9–\nsgRNA ribonucleoprotein complexes. \nFertilized, de-jellied embryos (<1 hour post-fertilization) were examined for sperm entry points and \naligned in rows within a 2 mm × 40 mm trench cut into a 50 mm agar-coated dish filled with 5% \nFicoll PM400 in MMR. Cas9–sgRNA complexes were back-filled into pulled Drummond glass \nneedles and bilateral 5 nL injections (total 10 nL per embryo) were delivered adjacent tothe female \npronucleus using a Nanoject II injector mounted on a Narishige MM-3 micromanipulator. sgRNA \ndoses were 300 -350 pg per embryo. Embryos were injected in batches of 25 and transferred \nimmediately to 24 °C. For each sgRNA, 100 embryos from the same sibship were injected. Control \nembryos were injected with Cas9-NLS protein pre-incubated with the scrambled sgRNA in the same \namounts. At 2-3 hours post-injection, embryos were assessed for normal cleavage and transferred to \n2.5% Ficoll in 0.1× MMR. The following day, embryos were re-screened for normal development, \nand five were randomly selected for genotyping. Remaining embryos were maintained in 0.1× MMR \nuntil stages 46–47. For genotyping, whole embryos or individual tadpoles were homogenized in 200 \nµL of 5% Chelex in PBS with 1.5 µL Proteinase K (25 mg/mL) and incubated at 65 °C for 3 hours \n(embryos) or overnight (tadpoles). Reactions were terminated at 95 °C for 5-10 minutes and briefly \ncentrifuged to pellet the Chelex resin. One microlitre of supernatant was used directly as PCR \ntemplate (primer sequences in Table 1). Gene editing was confirmed using TIDE analysis (Brinkman \net al., 2014) (https://tide.nki.nl/), comparing CRISPant samples to scrambled sgRNA controls, and \nvalidated against InDelphi predictions. \n2.3 Behavioural recording and TopScan-based locomotor analysis.  \nBehavioural recordings were performed using a high-resolution locomotor tracking protocol \ndeveloped for Xenopus laevis tadpoles, adapted from the automated TopScan analysis workflow \ndescribed in Banerjee et al. (2024). Stage 47 tadpoles were placed individually into wells of a clear \n24-well plate containing 0.1× MMR and allowed to acclimate for 2–3 minutes before recording. \nPlates were positioned on a uniform full spectrum daylight LED back-illumination stage, and \noverhead recordings were acquired using two identical 12.3 Megapixel camera with 16 mm telephoto \nlens mounted at a fixed height and controlled by a Raspberry Pi5.. Tadpoles in each 24-well plate \nwas recorded for a continuous 1 hour period at 50 frames per second under constant illumination. \nRaw video files were converted to .mp4 format using HandBrake (https://handbrake.fr/) to \nstandardize encoding prior to analysis. Automated locomotor quantification was conducted in \nTopScan (CleverSys Inc.) following the arena-based workflow optimized for X. laevis tadpoles \n(Panthi et al., 2024). For each video, a static background image was generated, and circular arenas \ncorresponding to each well were manually defined. Identical detector thresholds, background \nparameters, and arena definitions were applied across all recordings to ensure analytical consistency. \nLocomotor metrics were extracted using the TopScan LocoMotion Super module, including total \ndistance travelled and mean swim velocity (mm/sec). Darting behaviour was quantified using the \nDrugAbuse detector, with darting defined as rapid burst movements exceeding the pre-set high-\nvelocity threshold. Event durations, velocities, and locomotor trajectories were exported as summary \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n6 \ntables for downstream statistical analysis. All recordings were processed using identical acquisition \nand analysis settings to ensure reproducibility across biological replicates. \n2.4. In vivo Ca²⁺ imaging and analysis. \nIn vivo Ca²⁺ imaging was performed using a widefield fluorescence protocol adapted from the \nneuroD2 DEE72 Xenopus laevis tadpole model and a previously published cranial-window workflow \n(Banerjee et al., 2024). Single-cell embryos were bilaterally injected with 500 pg GCaMP6s and 250 \npg mCherry mRNA together with the respective Cas9/sgRNA reagents. At NF stage 47, tadpoles \nwere anaesthetized in 1:4,000 MS-222 for 2 minutes, positioned dorsal-side-up in a Petri dish, and \nembedded in 2% low-melting-point agarose. Embedded animals were submerged in 0.1× MMR \ncontaining 200 µM pancuronium bromide to ensure complete neuromuscular blockade. A cranial \nwindow was created by gently removing the dorsal head skin with fine forceps to expose the dorsal \nbrain surface, which was stabilised under a thin cap of 1% low-melting-point agarose \n(Supplementary figure S3). Tadpoles were imaged using a Zeiss Axio Examiner D1 upright \nfluorescence microscope equipped with a 10×/0.3 NA water-dip objective (N-Achroplan, Zeiss). \nGCaMP6s fluorescence was excited at 480 nm using a Polychrome V light source (TillPhotonics) \nand detected through a 495 nm dichroic mirror and a 505 nm long-pass emission filter. Images were \nacquired with a PCO Sensicam CCD camera using 4×4 on-chip binning, as described in Banerjee et \nal (2024). Imaging was restricted to the forebrain and midbrain, as inclusion of the hindbrain reduced \nthe effective field of view and compromised spatial resolution and signal quality in these anterior \nregions. Spontaneous brain activity was recorded for 30 minutes at 2 frames/s (400 ms exposure per \nframe). Raw image sequences were processed in Fiji (Schindelin et al., 2012) to generate ΔF/F₀ (%) \nstacks. For each recording, a median Z-projection across the full 30-minute stack was used to \ncompute the baseline image (F₀); ΔF was calculated as F(t) − F₀, and ΔF/F₀ values were expressed as \npercentages (ΔF/F₀%). \nΔF/F₀% image stacks were analysed in MATLAB. A whole-brain mask was first manually delineated \non the maximum-fluorescence projection using interactive freehand drawing tools (imfreehand) and \napplied across the full image stack to restrict analysis to brain tissue. CalciSeg (Günzel et al., 2024) \nwas then applied within this masked region for automated, correlation-based detection and \nrefinement of Ca2+-active regions of interest (ROIs), as well as for initial preprocessing; all \nsubsequent signal extraction and quantitative analyses were performed in MATLAB (MathWorks). \nFor each ROI, a binary mask was applied to every frame of the ΔF/F₀% stack. At each time point, \nΔF/F₀% pixel values within the ROI were extracted and averaged to yield a single fluorescence \nintensity value. Repeating this operation across frames generated ΔF/F₀% traces for each ROI, which \nwas used for event detection and downstream analyses. Initial denoising occurred implicitly during \nCalciSeg processing, which suppresses pixel-level noise and background fluctuations by retaining \nspatially and temporally correlated Ca2+signals. No additional spatial or temporal denoising filters \nwere applied prior to event detection. \nTo remove slow baseline drift, ΔF/F₀% traces were high-pass filtered at 0.005 Hz. For event \ndetection, a single fixed threshold was derived from control animals by pooling high-pass–filtered \nwhole-brain mean traces and defining the cutoff as 3× the standard deviation (SD) of this pooled \ncontrol distribution. This control-derived threshold was applied uniformly to all replicates and \nexperimental groups; thresholds were not calculated separately for individual traces. For each \ntadpole, fluorescence values were averaged across all pixels within the manually defined whole-brain \nmask at each time point to generate a single whole-brain mean ΔF/F₀% trace. Peaks were identified \non the high-pass–filtered whole-brain trace using minimum width and distance constraints (5 s each), \nand peaks exceeding the fixed control-derived threshold were classified as Ca2+ events.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n7 \nFor frequency-domain analysis, whole-brain ΔF/F₀% traces were subjected to fast Fourier \ntransformation (FFT) on a per-ROI basis. Single-sided FFT amplitudes were squared to obtain power \nspectral density estimates ((ΔF/F₀)²/Hz), interpolated onto a common 0.01–1 Hz frequency grid, and \naveraged across Ca²⁺-active ROIs to generate a whole-brain spectrum per tadpole. Group spectra are \nshown as mean power ±95% confidence intervals across animals. Integrated low-frequency power \n(0.01–1 Hz) was calculated as the area under the power spectrum. \nTo quantify neural synchrony between brain regions, Pearson correlation coefficients were calculated \nbetween ΔF/F₀% traces of the left and right forebrain and left and right midbrain. To minimise light \nscatter from neighbouring regions, signals were extracted from fixed-area (500-pixel) central ROIs \npositioned at the geometric centroids of each region. Forebrain and midbrain boundaries were \nmanually delineated on maximum-fluorescence images split into left and right hemispheric masks, \nand centroids were computed to define circular ROIs of 500 pixels(regionprops, Centroid).  \n2.5 Losartan treatment for behavioural and Ca2+ imaging assays. \nBehavioural Phenotyping Following Losartan Treatment:  Losartan treatment assays were \nconducted using stage 47 Xenopus laevis tadpoles generated as described above. Individual tadpoles \nwere placed into wells of a clear 24-well plate containing 0.1× MMR and allowed to acclimate for 2–\n3 minutes before baseline recording. Baseline locomotor activity for each ap3b2.S CRISPant was \nrecorded for 1 hour using the Raspberry Pi high-resolution behavioural tracking setup described in \nSection 2.3. Following the baseline recording, 200 µL of 50 mM losartan (Sigma) dissolved in \nMilliQ water (MQW) was added directly to each well, resulting in a final concentration of 10 mM \nlosartan. Tadpoles were incubated in Losartan for 1 hour under identical environmental conditions. \nImmediately after incubation, a second 1-hour behavioural recording was performed using the same \nimaging setup and acquisition parameters. Raw video files were converted to .mp4 format and \nanalysed in TopScan (CleverSys Inc.) using the same detector thresholds, arena definitions, and \nworkflow described in Section 2.3. The LocoMotion Super and DrugAbuse (darting) detectors were \nused to extract locomotor and high-velocity event metrics. Behavioural parameters recorded before \nand after Losartan treatment were compared within the same CRISPant tadpoles to assess drug-\ninduced changes. \nCa2+ Imaging Following Losartan Treatment: To assess the effect of losartan on neuronal activity, \nin vivo Ca²⁺ imaging was performed on tadpoles treated with losartan or left untreated. CRISPants in \nthe control group received no drug exposure and were imaged using the standard cranial-window \nGCaMP6s protocol described in Section 2.4. For the treated group, ap3b2 CRISPants were incubated \nin 10 mM losartan for 1 hour immediately prior to cranial-window preparation. Following \nincubation, tadpoles were prepared for in vivo Ca²⁺ imaging as above.  \n2.6 Blood–brain barrier permeability assay. \nBBB permeability was assessed using a modified sodium fluorescein (NaF) leakage assay adapted \nfrom the neurod2 DEE72 X. laevis tadpole model (Banerjee et al., 2024). NF stage 47 tadpoles were \npositioned in a Petri dish and embedded in 2% low-melting-point agarose. Once the agarose had set, \nagarose-embedded animals were submerged in 1:4,000 MS-222 in 0.1× MMR for the duration of the \nexperiment. NaF dye (10 nL of a 0.1 mg/mL solution) was then injected into the fourth ventricle \nusing a glass capillary needle following the same injection approach used for embryos. Following \ninjection, tadpoles were kept protected from light. Dorsal images of the head were acquired at 2, 5, \n10, and 20 minutes post-injection using a Leica Fluo III stereomicroscope with a GFP2 filter set and \na DFC7000T camera under fixed exposure settings. Images were analyzed offline in Fiji. Mean \nfluorescence intensity (MFI) was extracted from the green channel within a defined region of interest \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n8 \n(ROI). To ensure that measurements reflected BBB permeability rather than injection artefacts, the \nROI was always drawn on the side opposite to the injection site, capturing NaF dispersion outside the \nbrain parenchyma. \n2.7 Transcriptomic sample preparation, RNA sequencing, and bioinformatic analysis \nNF stage 47 Xenopus laevis tadpoles were anesthetized in a 1:4,000 MS-222 solution until \nunresponsive to touch, then positioned dorsal-side-up on a custom agar dissection plate. Using fine \nforceps for stabilization, the brain-spinal cord junction was severed with Vanna iridectomy scissors, \nand the entire brain (forebrain, midbrain, and hindbrain) was excised as an intact unit by cutting \nalong the cranial margins. For each biological replicate, six brains were immediately pooled into pre-\nlabelled tubes on dry ice and stored at −80 °C. Total RNA was extracted using the RNeasy Mini Kit \n(Qiagen) with on-column DNase digestion, and purified RNA was stored at −80 °C until sequencing. \nRNA quality control and library preparation were performed by the Otago Genomics Facility, where \nhigh-quality samples were used to generate TruSeq stranded mRNA libraries. Indexed libraries were \nthen pooled and sequenced (Illumina Nextseq 2000 P3-200 XLEAP kit, 2 x 100bp paired end) to a \ndepth of 50-60 million reads per sample. \nRaw reads underwent standard quality control, adapter trimming, and alignment to the X. laevis v10.1 \nreference genome (XENLA_10.1 accessed May 2025, Xenbase.org) using STAR (Dobin et al., 2013) \nat the University of Portsmouth, UK. Sorted BAM files were used to generate a read count matrix in \nGalaxy.eu using FeatureCounts (Liao et al., 2014) (stranded_reverse, gene_id, paired count as one). \nDifferentially expressed genes were identified from normalised read counts, filtered to remove low or \nnon-expressed gene ids (CPM <=1 in 4 or more samples) using EdgeR (Robinson et al., 2010). Genes \nwith a false discovery rate (Benjamini-Hochberg corrected FDR) < 0.05 and a fold change of >2 \nwere considered differentially expressed between control and ap3b2 CRISPant brains. Sequencing \ndata and read counts are available at NCBI GEO under accession GSE312492  \nPrincipal component analysis (PCA) plot was generated in R v4.5.1 from normalised read counts \n(Log2 counts per million) using FactoMineR (Lê et al., 2008) and factoextra packages (Kassambara \nand Mundt, 2020) and plots rendered with ggplotR (Wickham, 2016). Volcano plots were generated \nin R from edgeR statistics. Functional enrichment of differentially expressed genes was performed \nusing Enrichr (https://maayanlab.cloud/Enrichr/) (Chen et al., 2013, Kuleshov et al., 2016, Xie et al., \n2021) against a custom background list of brain-expressed genes (Supplementary File 1), \nincorporating Gene Ontology (Biological Process and Molecular Function) and KEGG pathway \nlibraries. KEGG pathway visualisation was carried out using Pathview, and enrichment results were \ninterpreted alongside ClinVar 2025 annotations to identify overlap with known DEE- and seizure-\nassociated genes. Heatmaps were made from Z-scores of normalised read counts using the pheatmap \npackage (Kolde, 2025) in R v4.5.1 (RDevelopmentCoreTeam, 2024).  \n2.8 Graphs and statistical analysis \nMost graphs and statistical analyses were prepared in GraphPad Prism v10. Raw data, normality test \nresults and analysis for all figures can be found in Supplementary File 1. The Shapiro-Wilk test for \nnormality was used to confirm normal distributions, with nonparametric tests being used if \ndistributions were non-normal. Specific tests are included in figure legends. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n9 \n3 Results \n3.1 CRISPR/Cas9-mediated editing of ap3b2.S induces hyperactivity and seizure-\nassociated behaviour \nChemoconvulsant models have established that PTZ-evoked seizures in X. laevis tadpoles manifest \nas rapid, uncontrolled tail bends, excessive turning, and recurrent C-shaped body contractions, \nproviding robust behavioural markers of seizure onset and severity (Bell et al., 2011, Hewapathirane \net al., 2008, Panthi et al., 2024). Functional DEE models show similar behavioural patterns, as \nobserved in the neurod2 CRISPant tadpoles, where spontaneous seizure-like behaviours from early \nlarval stages, including abrupt C-shaped convulsions and sustained periods of high-intensity, \nuncoordinated swimming were reported in both X. laevis and X. tropicalis models (Banerjee et al., \n2024, Sega et al., 2019). We used the same CRISPR knockdown strategy to determine whether \ndisruption of ap3b2.S in X. laevis reproduces the phenotype of DEE48.  \nSince DEE48 results from homozygous loss of AP3B2, two sgRNAs were designed and tested to \ndetermine which would produce the greatest disruption of ap3b2.S. Preliminary behavioural \nscreening revealed that ap3b2 CRISPants generated with sgRNA2 exhibited markedly elevated swim \nvelocity, increased darting behaviour, and prolonged circling compared with both GFP and Cas9-\nonly controls, whereas sgRNA3 CRISPants showed a more variable phenotype (Supplementary \nfigure S4). Consistent with these behavioural differences, sgRNA2 also produced higher mean \nediting (85.6% +/-3.5) (Supplementary figure S4). For these reasons, sgRNA2 was selected for all \nsubsequent analyses. \nTo investigate whether loss of AP3B2 function contributes to DEE-associated behavioural \nabnormalities, we generated ap3b2 CRISPants using sgRNA2. The sgRNA2 cut site lies immediately \nupstream of the AP3B1_C domain, and Sanger sequencing of CRISPant tadpoles confirmed \nsubstantial sequence degradation at the targeted locus (Figure 2a). TIDE analysis revealed a \nheterogeneous mixture of frameshift and in-frame alleles across ap3b2 CRISPants, with three \npredominant deletion events: a 7 base pair (bp) frameshift deletion detected in 67% of sequenced \nsamples, an 11 bp frameshift deletion in 83%, and a 12 bp in-frame deletion in 93% of samples \n(Supplementary figure S5). The 7 bp and 11 bp deletions are predicted to introduce premature stop \ncodons shortly downstream of the cut site, resulting in truncated proteins lacking the conserved \nAP3B1_C domain (Figure 2b). In contrast, the 12 bp deletion removes four amino acids and would \nresult in an almost full length protein. However, this in-frame deletion lies within the conserved \nAP3B1_C domain. Since DEE48 patients have seizures from early infancy, we next examined \nwhether disruption of Ap3b2 in tadpoles leads to abnormal locomotor activity, indicative of \nspontaneous seizures. In the neurod2 DEE72 tadpole model, seizure-associated neural activity was \ncaptured by two characteristic behavioural signatures: darting, defined as rapid, high-velocity C-\nshaped contractions occurring in abrupt bursts, and elevated mean swimming velocity (Banerjee et \nal., 2024). High-speed video recordings captured clear darting episodes in ap3b2 CRISPants (Figure \n2c,; Supplementary video 1). Quantitative tracking using TopScan confirmed that the ap3b2 \nCRISPant group had significantly higher mean swimming velocities (Figure 2d) and increased time \nspent darting compared with sibling controls (Figure 2e). Mean editing in 7 randomly picked \nembryos from the cohort was 83.3% +/- 3.9 (Figure 2f). These levels of ap3b2 editing, while not a \ncomplete knockout, are therefore sufficient to elicit spontaneous seizure-like locomotor behaviour in \nmosaic F0 X. laevis tadpoles.   \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n10 \n \nFigure 2. CRISPR/Cas9 disruption of ap3b2 triggers premature protein truncation resulting in \nhyperactivity and seizure-like behaviour in CRISPants. (a) Schematic of the Xenopus laevis \nAp3b2.S protein showing the conserved Adaptin_N and AP3B1_C domains. sgRNA2 target site is \nindicated, along with representative Sanger sequencing traces show degradation at the sgRNA2 cut \nsite. (b) Predicted protein outcomes for the three most observed indels (-7, -11 and -12 bp deletions) \ngenerated by sgRNA2 (c) Representative frames from behavioural recordings at 50 fps of an ap3b2 \nCRISPant tadpole. Frames highlighted in purple boxes show characteristic seizure-related C-shaped \ndarting behaviour observed in CRISPants. (d-e) Comparison of (d) mean swim velocity (mm/s) and \n(e) time spent in darting behaviour (%) between ap3b2 CRISPants (N = 41) compared with sibling \ncontrols (N = 48), unpaired t-test with Mann–Whitney, *P < 0.05. Horizontal bars indicate group \nmeans and error bars denote SEM. (f) Summary of CRISPR/Cas9 editing outcomes in 7 arbitrarily \nselected ap3b2 CRISPant embryos from the same batch used in the behaviour experiments, \nconfirmed by Sanger sequencing and TIDE analysis. Raw count data and statistical analysis can be \nfound in Supplementary File 1. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n11 \n3.2 Ca2+ imaging of the fore and midbrain of ap3b2 CRISPant tadpoles reveals \nincreased neural activity and increased synchrony between brain hemispheres . \nThe marked behavioural hyperactivity and erratic swim episodes observed in ap3b2 CRISPants \nprompted us to investigate whether Ap3b2 loss disrupts normal neuronal signalling during early brain \ndevelopment in our model. Using the same Ca²⁺ imaging protocol established in our DEE72 neurod2 \nstudy (Banerjee et al., 2024), we recorded spontaneous in vivo activity from midbrain (MB) and \nforebrain (FB) regions in stage 47 tadpoles expressing GCaMP6s. CRISPants exhibited large, slow \nCa²⁺ events that frequently persisted for more than two minutes, in contrast to the infrequent, low-\namplitude fluctuations observed in controls (Figure 3a, Supplementary video 2). \nTo quantify this, ΔF/F₀  traces were first high-pass filtered at 0.005 Hz to remove baseline drift, and a \nglobal detection threshold, set at three times the SD of the filtered control cohort, was applied \nuniformly across all recordings (Figure 3b, Supplementary figure S6-S7). The brains of ap3b2 \nCRISPant tadpoles showed a marked increase in spontaneous neural activity. The mean number of \nCa²⁺ events was double that of controls (ap3b2 CRISPants 3.46 +/- 0.80, unedited group 1.67 +/- \n0.33, p=0.028, Figure 3c). Mean event amplitudes were more than twice as high (ap3b2 CRISPants \n8.20 +/- 0.84, unedited group 3.94 +/- 0.76, p=0.0004, Figure 3d), indicating significantly more \nfrequent and more intense events in CRISPants.  \nTo assess whether loss of Ap3b2 alters the temporal organization of spontaneous Ca²⁺ activity in the \ntadpole brains, ΔF/F₀% signals recorded across fore and midbrain were analysed in the frequency \ndomain using fast Fourier transform (FFT, Figure 3e). Consistent with prolonged and large-amplitude \nCa²⁺ events observed, ap3b2 CRISPant brains exhibited elevated low-frequency power across the \n0.01–1 Hz range. To enable statistical comparison at the level of individual animals, integrated low-\nfrequency spectral power (0.01–1 Hz) was calculated for each tadpole, (Figure 3f). Mean total \nspectral power was five times higher in ap3b2.S CRISPants compared with controls (CRISPants: \n0.58 +/- 0.14; unedited group: 0.11 +/- 0.02; P < 0.0001), representing an approximately five-fold \nincrease in low-frequency Ca²⁺ signal energy.  \nDuring qualitative inspection of the Ca2+ imaging data, spontaneous Ca²⁺ transients in ap3b2 \nCRISPants appeared highly synchronous between hemispheres. To quantify this interhemispheric \nsynchrony, we calculated Pearson correlation coefficients between left and right mid and forebrain \ncentroid-based regions (Figure 4a). Control tadpoles displayed moderate left–right correlation within \nboth mid- and forebrain regions (Figure 4b), consistent with normal developmental patterns of \ninterhemispheric coordination. In ap3b2 CRISPants, however, the correlation matrices showed \nuniformly elevated bilateral coupling across both brain regions (Figure 4c). \nQuantitative analysis showed that mean left–right midbrain synchrony was significantly higher in \nap3b2 CRISPants (0.96 +/- 0.01, N = 13 tadpoles) than in unedited controls (0.79 +/- 0.08, N = 12) (p \n= 0.04; Figure 4d). A similar increase was observed in the forebrain, where mean left–right forebrain \nsynchrony was also significantly elevated in CRISPants (0.88 +/- 0.03) compared with controls (0.72 \n+/- 0.05) (p = 0.01; Figure 4e). By contrast, heterotopic correlations did not differ significantly \nbetween ap3b2 CRISPants and controls (right midbrain-forebrain, Figure 4f, p=0.72, left midbrain-\nforebrain, Figure 4g, p=0.47). Together, the observed increased Ca2+ activity and enhanced \ninterhemispheric synchrony in ap3b2 CRISPant brains are consistent with a highly synchronised \nnetwork state, which could underlie the seizure susceptibility of both CRISPant tadpoles and DEE48 \npatients. \n \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n12 \n \nFigure 3. Ap3b2 CRISPant brains have elevated spontaneous Ca2+ activity compared to \nunedited controls in CRISPants. (a) Representative still frames acquired every 10 s from ap3b2 \nCRISPant tadpole brain during in vivo widefield Ca²⁺ imaging. GCaMP6s fluorescence intensity \n(lighter shades indicate higher ΔF/F₀%) is shown. Forebrain (FB) and midbrain (MB) hemispheres \nare outlined, the hindbrain lies outside the field of view. Time (s) is indicated in each frame. (b) \nRepresentative raw and high-pass-filtered ΔF/F₀% traces from a control tadpole (left) and an ap3b2.S \nCRISPant (right). Significant Ca²⁺ events were detected using a global threshold defined as 3× SD of \nall filtered control traces (black line); events exceeding this cutoff are marked by arrowheads. The \nred box indicates the time window in panel (a). (c,d) Comparison of Ca²⁺ event counts (c) and \nsignificant event amplitude (ΔF/F₀%) (d) in control (N = 12) and CRISPant tadpoles (N = 13). \nIndividual data points represent single tadpoles; horizontal bars denote group means and error bars \nindicate SEM. Groups were compared using unpaired t-test with Mann Whitney, *P < 0.05, ***P < \n0.001. (e) FFT-derived power spectral densities of whole-brain Ca²⁺ signals for control and ap3b2 \nCRISPant tadpoles. Solid lines represent group means and shaded envelopes indicate 95% \nconfidence intervals. (f) Integrated low-frequency spectral power (0.01–1 Hz; (ΔF/F₀)²/Hz, calculated \nas the area under the power spectrum for each tadpole. Data are from control (N = 12) and ap3b2.S \nCRISPant tadpoles (N = 13). Individual data points represent single tadpoles; horizontal bars denote \ngroup means and error bars indicate SEM. Groups were compared using unpaired t-test with Mann–\nWhitney correction, ***P < 0.001, ****P < 0.0001. (g) Summary of CRISPR/Cas9 editing outcomes \nin the CRISPant group, determined by Sanger sequencing and TIDE analysis. Raw count data and \nfull statistical details are provided in Supplementary File 1. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n13 \n \n \nFigure 4. Ap3b2 CRISPants exhibit increased interhemispheric synchrony in the midbrain and \nforebrain. (a) Schematic example illustrating the centroid-based regions of interest (ROIs) used for \ninterregional correlation analysis. Fixed-area central ROIs (500 pixels each) were positioned at the \ngeometric centroids of the left and right midbrain (MB), left and right forebrain (FB). (b, c) Group-\naveraged Pearson correlation coefficient matrices for unedited control tadpoles (b, N = 12) and  \nap3b2 CRISPants (c, N = 13). (d–e) Comparison of brain regional synchrony, quantified as Pearson \ncorrelation coefficients, in controls and ap3b2 CRISPants. Regions compared are (d) left-right MB \nand (e) left-right FB. Unpaired t-test with Mann Whitney, *P < 0.05, **P < 0.01. (f–g) Correlations \nbetween (f) right MB-FB, and (g) left MB-FB, unpaired t-tests with Mann Whitney, ns, not \nsignificant. Raw correlation values and full statistical details are provided in Supplementary File 1. \n3.3 Ap3b2 CRISPant brains have reduced expression of genes associated with \nmonovalent cation transport, BBB function, inhibitory GABA neurotransmission and \naxon guidance. \nThe pronounced changes in brain activity prompted us to examine whether Ap3b2 loss is \naccompanied by transcriptional changes in pathways governing neural circuit function and \nhomeostasis. Prior developmental transcriptome mapping in X. laevis tadpole brain development \nshowed both region and stage-specific shifts in neuronal, progenitor, and synaptic gene programmes, \nsupporting the suitability of this system for detecting mutation-induced network changes (Ta et al., \n2021). To determine what developmental changes predispose DEE48 brains to seizure activity, we \nperformed RNA-seq on pooled stage 47 unedited control and ap3b2.S CRISPant brains, the same \nstage at which behavioural and neural activity assays were conducted. \nPrincipal component analysis (PCA) showed that genotype was the strongest driver of gene \nexpression in our dataset, indicating an effect of ap3b2 loss on the developing brain transcriptome \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n14 \n(Figure 5a). EdgeR analysis was used to identify differentially expressed genes (DEG: FDR < 0.05, \nlog₂FC > 1) revealed a strongly asymmetric response: 1,078 genes were significantly downregulated \nin CRISPant brains compared to control unedited brains, whereas only 28 were upregulated, with \nmedian fold changes of ~2.3-fold decrease and ~2.1-fold increase, respectively. The volcano plot \nillustrates this bias towards downregulation, with top-ranked genes selected using a Euclidean \ndistance-based ranking that integrates both effect size (log₂ FC) and statistical significance (–log₁₀ \nadjusted P value). Top ranked genes included the endothelial junction gene jcad.S, the transporter \nslc16a2.L, and transcriptional regulators bach2.L and zbtb20.S, while the most upregulated genes \n(e.g., ikbke.S, cfi.L, cps1.S, mpc2.L, krt12.5.L, krt12.5.S, krt62.S) are associated with immune, \nmetabolic and cytoskeletal functions (Figure 5b). \nThe upregulated DEG group was too small to generate any significantly enriched pathway data. \nFunctional enrichment analysis of the downregulated set was conducted against a background list of \nstage matched brain-expressed genes, using Enrichr (Chen et al., 2013, Kuleshov et al., 2016, Xie et \nal., 2021). Many pathways and processes were found to be significantly overrepresented in this \nap3b2 CRISPant down-regulated gene set, including ion transport, inhibitory signalling, blood brain \nbarrier formation, and establishment of neural connectivity (axonogenesis/axon guidance). Both GO \nbiological process and KEGG pathway analyses highlighted monovalent and inorganic cation \ntransport, GABAergic synapse/signalling, axon guidance, endocytosis, and neuroactive ligand-\nreceptor interaction. Two ClinVar terms “Developmental and Epileptic Encephalopathy” and \n“Seizure” were significantly overrepresented (Figure 5c,d). Genes associated with transport across \nthe BBB (GO:0150104), included BBB-associated transporters and endothelial genes: slc16a2.L, \nslc11a2.L, slc11a2.S, abcg1.L, abcg1.S, cldn11.S, flt1.L, kdrl.L). Down regulated genes were also \nenriched for GABA signalling (GO:0007214), encompassing multiple GABA_A receptor subunits \nand synthetic machinery (gabra1.L, gabra2.L, gabrb1.L, gabrb1.S, gabrb2.L, gabrb2.S, gabrb3.L, \ngabrg2.L, gad2.L); and axon guidance (GO:0007411), including guidance receptors and ligands such \nas plxna4.L, sema3c.L, sema3c.S, sema4b.S, sema4g.L, sema4g.S, sema7a.L, unc5b.L, unc5b.S, \nunc5c.L, unc5c.S, unc5d.S (Figure 5d). These data indicate that ap3b2 knockdown leads to a broad, \npathway-level downregulation of BBB transport, inhibitory neurotransmission, and axon wiring \nprogrammes. High mean editing efficiency in the CRISPant cohort (~80%; Figure 5e) supports a \ndirect link between AP-3 disruption and these transcriptomic changes. \n3.4 Increased early BBB permeability in ap3b2 CRISPants \nIn our previous neurod2 CRISPant DEE72 model, we demonstrated that early-onset seizures \ncoincide with pronounced BBB leakage, with rapid sodium fluorescein (NaF) dye escape occurring \ndespite otherwise normal brain morphology. Notably, short-term losartan treatment reduced both dye \nleakage and seizure burden, supporting a mechanistic link between BBB permeability and \nepileptogenesis rather than a secondary consequence of seizures (Banerjee et al., 2024). Our analysis \nof down regulated genes in the brains of ap3b2 CRISPant tadpoles (Figure 5d) showed significant \nenrichment of pathways related to BBB transport suggesting impaired or developmentally delayed \nbarrier function. This raised the possibility that ap3b2 loss, similar to neurod2 haploinsufficiency, \ncompromises BBB integrity.  \nWe therefore assessed whether ap3b2 CRISPants have altered BBB integrity, by monitoring the \ndiffusion of intraventricularly injected NaF across four timepoints (2, 5, 10, and 20 minutes post-\ninjection). Visual inspection of CRISPant tadpoles revealed no gross abnormalities in overall brain \nmorphology compared with controls, consistent with observations previously reported in the DEE72 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n15 \n \nFigure 5. ap3b2 disruption downregulates BBB transport, GABA signalling, and axon guidance \npathways in CRISPant tadpole brains. a) PCA plot of normalized read counts of the five control \nsamples and four ap3b2 ⁻/⁻ (mosaic) samples, each sample is derived from six pooled brains. b) \nVolcano plot of EdgeR differentially expressed genes, (DEG) with FDR threshold of <0.05 and Log2 \nfold change of >1. The 10 most significant up and down regulated genes are labelled. c) Selected \noverrepresented ontologies calculated with EnrichR for down regulated DEG, the top 6 hits are \nshown for GO: biological process and KEGG_2021, for ClinVar the only two significant hits are \nshown (PAdj <0.05). d) Heatmap of z-scores for three down overrepresented ontologies, down \nregulated in CRISPants. PAdj ** <0.01, *** <0.001. e) Summary of tadpole editing in the ap3b2 \nCRISPant group, confirmed by Sanger sequencing and TIDE analysis. Supplementary data for (c) \nand (e), as well as the custom brain background gene list, can be found in Supplementary file 1. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n16 \ntadpole CRISPant model (Banerjee et al., 2024). As expected, control tadpoles showed a slow and \nprogressive spread of fluorescence beyond the ventricular space, consistent with low-level \nphysiological leakage from a normally developing BBB. In contrast, ap3b2 CRISPants displayed \nmarkedly accelerated dye escape, with significantly higher fluorescence outside the brain and in \nkidneys, apparent after 2 minutes (P= 0.0212, Figure 6c), indicating an immediate increase in BBB \npermeability. Although the difference between groups partially converged at 5–10 minutes, reflecting \nrapid equilibration, once the barrier is breached the later phase of the assay revealed a clear \ndivergence. By 20 minutes, most fluorescein had already cleared from both the brain and surrounding \ntissue in CRISPants, whereas controls continued to show a gradual outward leak and retained a \nvisible dye reservoir (Figure 6c, Supplementary figure S8). Together, these data demonstrate that loss \nof approx. 80% of ap3b2, resulting from editing (Figure 6d) causes early-onset and transiently \nheightened BBB permeability, which could represent an early pathological feature of AP3B2-\nassociated DEE. \n \nFigure 6 Rapid early sodium fluorescein dye leakage in ap3b2 CRISPants reveals a markedly \ncompromised BBB. (a) Schematic dorsal view of an NF stage 47 Xenopus laevis tadpole head \nshowing forebrain (FB), midbrain (MB), hindbrain (HB), and spinal cord (SC), and the site of \nsodium fluorescein (NaF) microinjection into the 4th ventricle (red arrow). The dashed rectangle \nindicates the ROI outside the brain from which fluorescence intensity was quantified. b) \nRepresentative images of NaF-injected controls (top row) and ap3b2.S CRISPants (bottom row) 2 \nminutes after microinjection (raw capture and green channel only). (c) Plot of mean fluorescence \nintensity (MFI) detected outside tadpole brain at 2, 5, 10 and 20 minutes post NaF injection in \nCRISPant tadpoles (N = 10) compared to controls (N = 7), Repeated measures 2-way ANOVA \nwith Tukey's multiple comparisons test, *P < 0.05. (d) Summary of CRISPR/Cas9 editing \noutcomes in the CRISPant group, determined by Sanger sequencing and TIDE analysis. Raw count \ndata and full statistical details are provided in Supplementary File 1. \n \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n17 \n3.5 Losartan treatment reduces mean swimming velocity, without significantly \naltering Ca2+ dynamics of ap3b2 CRISPant brains \nDrug-refractory seizures remain a major clinical challenge in developmental and epileptic \nencephalopathies (DEEs), with many patients showing limited or no response to conventional anti-\nseizure medications. This has prompted increasing interest in repurposed therapeutics that act on \nnon–ion-channel pathways, particularly agents with established clinical safety profiles and anti-\ninflammatory properties. One such candidate is losartan, a widely used angiotensin II type-1 receptor \nantagonist that modulates neuroinflammatory signalling by increasing thrombospondin-1 (TSP1) \nexpression and thereby regulating latent TGF-β activation (Figure 7a). Losartan has been shown to \nsuppress the development of chronic seizures in rodent models of traumatic brain injury and, more \nrecently, to acutely reduce seizure burden in the X. laevis neurod2 CRISPant model (Banerjee et al., \n2024). These prior findings implicate TGF-β–associated inflammatory and blood–brain barrier–\nlinked pathways as contributors to seizure susceptibility and raise the possibility that similar \nmechanisms contribute to the pathology of other DEEs, such as DEE48. \nTo see if there was evidence for this in our ap3b2 CRISPR brain transcriptomes, we looked at several \nkey anti-inflammatory regulators, including tgfbr2, tgfbr3, il6st, ahr, wdfy3, and cav1, each of which \nnormally constrains CNS inflammatory signalling ( \nTable 2). Loss of TGFBR2/3 weakens canonical TGF-β–SMAD signalling, a pathway required to \nmaintain microglia and astrocytes in a homeostatic, non-reactive state (Zöller et al., 2018, Luo, 2022, \nBlair et al., 2011, Duesman et al., 2023). Reduced IL6ST further diminishes STAT3-mediated \ncytokine negative feedback (Murakami et al., 2019, Rose-John, 2018, Cekanaviciute and Buckwalter, \n2016), while decreased AHR and WDFY3 remove important transcriptional and autophagic brakes \non glial activation (Wheeler et al., 2017, Wang et al., 2023, Filimonenko et al., 2010, Fox et al., \n2020). Concurrent downregulation of cav1 suggests impaired BBB stability, increasing susceptibility \nto peripheral inflammatory mediators (Huang et al., 2018, Trevino et al., 2024).  \nIn parallel, several potent pro-inflammatory genes were upregulated (Table 3), including card9, \nikbke, pstpip2, fosb, fosl1, gpr4 and cfi, which respectively activate NF-κB, interferon, AP-1, \ncomplement and endothelial inflammatory pathways (Hara et al., 2007, Zhong et al., 2018, Verhelst \net al., 2013, Clément et al., 2008, Cassel et al., 2014, He et al., 2022, Dong et al., 2013, Gomez-\nArboledas et al., 2021). Together, this pattern reflects a shift from protective, TGF-β–dependent \nimmune homeostasis toward a state of heightened innate immune activation. Thus, downregulation of \nTGF-β signalling components, combined with induction of cytokine- and danger-associated \ntranscripts, provides a mechanistic framework by which disrupted Ap3b2 function may predispose \nthe developing brain to persistent, TGF-β–associated neuroinflammation (Figure 7a).  \nTo test whether targeting TGF-β associated inflammatory signalling could mitigate the seizure \nphenotype observed in ap3b2 CRISPants, tadpoles were treated with 10 mM losartan and assessed \nusing behavioural and Ca²⁺ imaging assays (Error! Reference source not found.). The addition of i\nndividual paired measurements demonstrated decreased swim velocity of 9/11 ap3b2 CRISPant \ntadpoles following a 1 hour treatment with 10 mM losartan (Figure 7b, mean velocity 0.41+/-0.13 \nmm/sec before treatment and 0.11+/-0.03 mm/sec after treatment, p=0.02). Embryo sequencing from \nthis batch of CRISPants confirmed high editing efficiency (~76%). For assessment of Ca2+ signalling, \nit was not possible to use the same tadpoles, so ap3b2 CRISPants were arbitrarily assigned to \ntreatment (10 tadpoles) or control (9 tadpoles) groups.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n18 \n \nFigure 7. Losartan treatment reduces aberrant calcium activity and hyperactivity in ap3b2 \nCRISPants. (a) Conceptual schematic of ap3b2 CRISPant differentially expressed genes associated \nwith loss of anti-inflammatory regulatory control, BBB dysfunction, and neuronal hyperexcitability \nrelevant to TGF-β receptor–SMAD signalling, indicating a hypothetical mode of action for losartan. \n(Created in BioRender). (b) Paired comparison of mean swim velocity (mm/s, over 1 hour) in ap3b2 \nCRISPant tadpoles before and after 10 mM losartan treatment. Lines connect individual tadpoles, and \nthe effect of losartan was tested using a Wilcoxon paired t-test, *P < 0.05. (c) Summary of \nCRISPR/Cas9 editing outcomes for embryos used in the losartan phenotype experiments, determined \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n19 \nby Sanger sequencing and TIDE analysis. (d) Representative whole-brain Ca2+ signals (raw and \nfiltered ΔF/F₀ traces) showing activity from untreated ap3b2.S CRISPants (left) and losartan-treated \nCRISPants (1 hour, 10 mM, right). Black lines indicate the global event detection threshold (3× \ncontrol SD), with arrowheads marking significant Ca²⁺ events. (e,f) Comparison of the numbers of \nCa²⁺ events (e) and mean event amplitude (f) detected in untreated (N = 9) and losartan treated (10 \nmM; N = 10) ap3b2 CRISPant tadpoles. Groups were compared using unpaired Welch’s t-tests, ns = \nnot significant (P > 0.05). (g) FFT-derived power spectral densities of spontaneous whole-brain Ca²⁺ \nactivity in untreated and losartan-treated ap3b2 CRISPants, plotted on a logarithmic scale. Solid lines \nindicate group means and shaded envelopes represent 95% confidence intervals across animals. (h) \nComparison of integrated low-frequency spectral power density (0.01–1 Hz; (ΔF/F₀)²/Hz) between \nuntreated (N = 9) and losartan-treated (10 mM; N = 10) ap3b2 CRISPants. Individual data points on \nscatterplots represent single CRISPant tadpoles; horizontal bars denote group means and error bars \nindicate SEM. Statistical significance was assessed an unpaired Welch’s t-test, with the exact p value \nshown. (i,j) Distribution of CRISPR editing outcomes in untreated (i) and losartan-treated (j) \nCRISPant tadpoles, quantified by Sanger sequencing and TIDE analysis. Raw data and full statistical \nanalyses are provided in Supplementary File 1. \nGCaMP6S live Ca2+ imaging appeared to show a partial suppression of abnormal activity in the \nlosartan treated group (Figure 7d, Supplementary video 3, Supplementary figures S9 and S10). \nFurther analysis of the Ca2+ imaging data confirmed that untreated ap3b2 CRISPant brains \ndischarged frequent, high amplitude Ca²⁺ events, as previously shown. While the losartan treated \ngroup generated 31% fewer such events (control mean 3.33 ± 0.87; treated mean 2.30 +/- 0.73), this \ndid not represent a significant reduction (P = 0.256; Figure 7e). Similarly, although the amplitudes of \nthe losartan treated group Ca2+ events (ΔF/F₀% ) were on average 28% lower than in untreated ap3b2 \nCRISPants (control mean 10.17 +/- 2.56; treated mean 7.32 +/- 2.14), the effect was not significant \n(P=0.289, Figure 7f). Group-averaged power spectral density curves (Figure 7g, 0.01-1 Hz), \nindicated that losartan-treated ap3b2 CRISPants tend to show less slow calcium fluctuations. Total \nspectral power was 60% lower in the treated group (control mean 0.508 +/- 0.124; losartan treated \nmean 0.230 +/- 0.047; P = 0.061; Figure 7h). Analysis of CRISPR/Cas9 editing outcomes revealed \nno difference in editing efficiencies and indel distributions between untreated and losartan-treated \nap3b2 CRISPants (Mann-Whitney test, p=0.39, Figure 7i,j; Supplementary figure S11). The observed \nsignificant reduction in swimming velocity following losartan treatment of individual tadpoles, \ntogether with a trend of decreased seizure-like brain activity, measured through live Ca2+ monitoring \nof equivalent ap3b2 CRISPant groups, suggests that targeting neuroinflammatory pathways may \nhave beneficial effects in reducing seizure activity in DEE48.  \n \n4 Discussion \n4.1 Ap3b2 loss of function generates a robust DEE48-like phenotype in X. laevis \ntadpoles. \nWe set out to mimic the homozygous loss of function variants found in human patients with DEE48, \nusing F0 CRISPant tadpoles. While these tadpoles are mosaic, greater than 80% of ap3b2.S genes \nwere found to be edited on average. Two thirds of edits are predicted to result in a truncating \nframeshifts, seen in some patients, and the remainder correspond to a 12 bp in-frame deletion, of \nunknown consequence, but located in the conserved AP3B1-C terminal domain. Because AP3B2-\nassociated DEE48 is an autosomal-recessive disorder, the observed robust phenocopy in tadpoles \nstrongly suggests that loss of these four amino acids is also pathogenic. Disruption of the X. laevis \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n20 \nap3b2 homeologue produced a behavioural and neurophysiological phenotype that parallels the \nclinical presentation of DEE48. ap3b2 CRISPant tadpoles exhibited increased mean swimming \nvelocity, more time spent darting , behavioural signatures strongly reminiscent of the seizure-like \nmotor episodes reported in individuals with biallelic AP3B2 loss-of-function variants (Alizadeh et \nal., 2025, Dilber et al., 2022, Assoum et al., 2016, Anazi et al., 2017). Seizures and hyperactivity \nwere also reported in an ap3b2-/- mouse model (Nakatsu  et al., 2004), suggesting functional \nconservation at least across vertebrates. Ca²⁺ imaging of brain activity revealed increased occurrence \nof spontaneous, large-amplitude, prolonged Ca2+ transients and elevated interhemispheric synchrony \nconsistent with network hyperexcitability characteristic of epileptic encephalopathy. \nAP3B2 encodes the neuron-specific β-subunit of the adaptor protein-3 (AP-3) complex, required for \nsynaptic vesicle and endolysosomal cargo trafficking. In mice, selective loss of neuronal AP-3B \nalone is sufficient to cause spontaneous seizures and increased seizure susceptibility in the absence of \ngross brain malformations, driven by disturbed synaptic vesicle function (Nakatsu  et al., 2004). \nTogether with human genetic evidence, these findings have established defective synaptic vesicle \ntrafficking as a core pathogenic mechanism underlying AP3B2-associated DEE. Consistent with this \nmodel, increased spontaneous brain activity and interhemispheric synchrony observed in ap3b2 \nCRISPant tadpole brains likely arises from a primary synaptic trafficking defect that disrupts \ninhibitory–excitatory balance during critical periods of neural circuit assembly. \nOur results also align with a growing body of work demonstrating the translational power of rapid in \nvivo CRISPR-based modelling of rare genetic epilepsies in aquatic vertebrates. CRISPR targeted \nspout1 disruption in zebrafish resulted in epileptiform activity and neurodevelopmental abnormalities \nakin to those found in patients with compound heterozygous mutations in SPOUT1, confirming \npathogenicity (Liu et al., 2024). In the neuroD2 (DEE72) haploinsufficient CRISPant model, \nspontaneous seizure-like behaviour and increased neural activity were observed despite preserved \ngross brain morphology (Banerjee et al., 2024), a pattern strikingly recapitulated in the ap3b2-/- \n(mosaic) CRISPants. The ap3b2.S CRISPant model therefore joins a growing class of vertebrate \nDEE systems in which targeted gene perturbation directly yields a reproducible encephalopathic \nphenotype, strengthening genotype–phenotype interpretation for rare variants. The early accessibility \nof both zebrafish and Xenopus CRISPant models highlights these model organisms as powerful \nplatforms for interrogating early disease mechanisms and evaluating candidate modifiers in DEE. \n4.2 Ap3b2 loss reveals early blood–brain barrier fragility and suggests altered \nneuroinflammation \nEarly BBB dysfunction is increasingly recognised as a key contributor to seizure susceptibility and \nepileptogenesis. Increased BBB permeability permits serum proteins, ions, and inflammatory \nmediators to enter the brain parenchyma, disrupting astrocytic regulation of potassium and glutamate \nhomeostasis and promoting network hyperexcitability (Swissa et al., 2019). In paediatric epilepsies, \nBBB instability correlates with higher seizure burden and drug resistance, supporting the view that \nbarrier dysfunction is a driver of disease severity rather than a secondary consequence (Kimizu et al., \n2018). Because the developing brain is particularly sensitive to neurovascular disruption, even \ntransient BBB opening may have lasting effects on circuit maturation and seizure risk (Moretti et al., \n2015). \nA striking phenotype in ap3b2 CRISPants was the rapid and pronounced leakage of sodium \nfluorescein from the ventricular system, compared to unedited siblings, indicating early and severe \nBBB compromise. This physiological defect was strongly supported by transcriptomic data showing \ncoordinated downregulation of endothelial solute carriers, transporters, adhesion molecules, and \ntight-junction–associated genes. The rapid leakage followed by accelerated washout is consistent \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n21 \nwith early, high-velocity barrier failure rather than gradual diffusion, suggesting that BBB integrity is \ncompromised almost immediately upon dye entry. Furthermore, transcriptome analysis of ap3b2 \nCRISPant brains found that genes associated with the BBB were overrepresented in the set of down-\nregulated differentially expressed genes. These genes included solute carriers, endothelial \ntransporters, adhesion molecules, and tight-junction components such as cldn11. Similar patterns are \ncharacteristic of epileptogenic tissue and experimental seizure models (Chen et al., 2020, Salman et \nal., 2017). Consistent with these molecular changes, ap3b2 CRISPants exhibited rapid sodium \nfluorescein leakage despite grossly normal brain morphology, indicating impaired barrier integrity. \nTogether, these data suggest that AP3B2 loss disrupts neurovascular maturation, creating a \npermissive environment for aberrant extracellular signalling and reduced homeostatic control. A \nsimilarly less robust BBB was demonstrated in our recent model of DEE72 (Banerjee et al., 2024), \nsuggesting a common mechanism leads to this phenotype. It is not yet known whether BBB leakiness \nis a cause or effect of seizure activity, but this commonality points towards the latter.  \nCurrently is it not known whether human DEE patients also have a compromised BBB, but the \nbroader epilepsy literature strongly supports this interpretation. Across neonatal hypoxic–ischaemic \ninjury, traumatic brain injury, and drug resistant epilepsy, early BBB opening permits albumin and \ncytokine entry into the brain, triggering astrocytic activation, impaired ion buffering, and \ndownstream network hyperexcitability (Dadas and Janigro, 2019, Goasdoue et al., 2019, Specchio et \nal., 2010). Albumin-driven activation of TGF-β signalling in astrocytes is a well-established \nictogenic mechanism, and blockade of this pathway prevents epileptogenesis in multiple \nexperimental models (Gorter et al., 2015, Librizzi et al., 2012).  \nIn human patients with DEE, infantile spasms are often among the first seizure types detected. It has \nlong been the custom to treat infantile spasms with ACTH or steroid therapy, which offers a drastic \nbut often effective solution to neuroinflammation, and the resulting damage to developing brains, that \ncomes with unrelenting seizure activity. More generally, the role of neuroinflammation in epilepsy is \nemerging as an important but often overlooked potential therapy target (Sanz et al., 2024). While we \ndid not find neuroinflammatory pathways to be overrepresented in our transcriptome analyses of \nap3b2 CRISPant brains, ikbke.S was one of the most significantly up-regulated genes, prompting us \nto look for other DEG associated with neuroinflammation (Tables 1 and 2). Additionally, ap3b2 \nCRISPants brains showed significant transcriptomic suppression of multiple components associated \nwith TGF-β responsiveness, including of tgfbr2, tgfbr3, il6st and skil. Although inflammatory \nactivation was not directly measured, these pathways are repeatedly implicated in seizure \nsusceptibility, BBB stability, and neurovascular regulation (Chen et al., 2020, Okamoto et al., 2010). \nThe convergence of reduced TGF-β–associated signalling, BBB dysfunction, and network \nhyperexcitability suggests a coordinated failure of neuronal, glial, and vascular regulatory systems \nrather than isolated defects. \nThe convergence of BBB leakage, suppression of neurovascular regulatory genes, and increased and \nhypersynchronous neural activity supports a model in which AP3B2 loss creates a permissive \nenvironment for persistent hyperexcitability during development. The shared neurovascular \nphenotype observed across DEE48 and DEE72 Xenopus models further highlights BBB fragility as a \nconserved mechanistic vulnerability in genetically driven epileptic encephalopathies, with important \nimplications for understanding disease severity and therapeutic responsiveness. \n4.3 Coordinated transcriptomic changes links molecular, behavioural, and Ca2+ \nimaging phenotypes \nTranscriptomic studies across human epilepsy and experimental seizure models consistently show \nthat epileptic encephalopathies arise from broad, coordinated disruption of neuronal, glial, immune, \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n22 \nand vascular gene networks rather than isolated pathway defects. Large-scale RNA sequencing of \nepileptic tissue reveals widespread suppression of neuronal and synaptic programmes alongside \nchanges in immune and vascular signalling (Guelfi et al., 2019, Iacobaş and Velíšek, 2018, Wen et \nal., 2024a) indicating that seizure phenotypes reflect systems-level transcriptional reprogramming.  \nWhole-brain RNA sequencing revealed a strongly directional transcriptional response dominated by \ndownregulation of genes governing inhibitory neurotransmission, ion transport, axon guidance, cell \nadhesion, endothelial transport, and neuroactive ligand–receptor signalling. This pattern closely \nmirrors transcriptomic signatures reported in human epilepsy, including reduced expression of \nGABA receptor subunits, glutamate transporters, and potassium channels (Guelfi et al., 2019, Kjær et \nal., 2019). In ap3b2 CRISPants, suppression of gabra1/2, gabrb1/2/3, gabrg2, slc1a2/6, and multiple \nkcn family members provides a molecular correlate for the hyperexcitable and hypersynchronous \nCa²⁺ dynamics observed in vivo. Beyond neurotransmission, extensive downregulation of axon-\nguidance and neurodevelopmental pathways indicates disruption of early circuit assembly and \nrefinement. Many DEE-associated genes converge on pathways regulating neuronal migration, axon \ngrowth, and synaptogenesis (Medyanik et al., 2025), and our findings suggest that AP3B2 loss \nperturbs developmental wiring processes, either resulting from, or independent of, synaptic vesicle \ntrafficking. This developmental instability likely contributes to the persistent network-level \nhyperexcitability detected by Ca²⁺ imaging. \nImportantly, since DEE is a genetically diverse but clinically identifiable umbrella disorder, we found \nmany down regulated genes in ap3b2 CRISPant brains that are associated with DEE. In total, 17 \nDEE-associated genes were identified by Enricher-GO analysis of the down regulated DEG list: \narhgef9 (DEE8), cdk19 (DEE87), gabra1 (DEE19), gabra2 (DEE78), gabrb1 (DEE45), gabrb2 \n(DEE92), gabrb3 (DEE43), gabrg2 (DEE74), grin2b (DEE27), hcn1 (DEE24), kcna2 (DEE32), \nkcnb1 (DEE26), kcnh5 (DEE112), kcnt1 (DEE14), slc1a2 (DEE41), slc25a22 (DEE3) and synj1 \n(DEE53). This illustrates the underlying common causes of DEE and may mean that therapies found \nto work in one DEE may well work in others, even when no known direct link has been found. \n4.4 Losartan provides partial rescue in two DEE models, suggesting potential for \nnew approaches to treatment. \nAcute losartan treatment produced a reproducible but incomplete improvement in the ap3b2 \nCRISPant phenotype. Behaviourally, losartan significantly reduced hyperlocomotion, and this was \nempowered by being able to use the same tadpoles before and after treatment. Due to the technical \nlimitations of immobilising tadpoles long term, effects on individual Ca²⁺ event amplitude, \nfrequency, and spectral power could only be compared between equivalent cohorts. While we \nobserved a trend towards less overt Ca2+ events, the differences between groups and were not \nsignificant.  \nThe relevance of losartan in this context lies in its capacity to modulate neurovascular and \nhomeostatic pathways rather than directly targeting synaptic excitability. Losartan enhances \nthrombospondin-1–dependent activation of latent TGF-β (Bar-Klein et al., 2014), a signalling axis \nrepeatedly implicated in seizure susceptibility in the setting of BBB dysfunction (Swissa et al., 2019). \nIn the present study, transcriptomic suppression of multiple components associated with TGF-β \nresponsiveness, including tgfbr2 and tgfbr3, coincided with rapid sodium fluorescein leakage, \nindicating impaired BBB integrity. Rather than demonstrating overt neuroinflammation, our data \nsupport a state in which developing neural circuits are rendered more vulnerable to dysregulated \nvascular and immune influences. Losartan’s partial efficacy therefore could be explained by \nstabilisation of this permissive pathological environment, rather than correction of the primary \nsynaptic trafficking defect caused by AP3B2 loss.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n23 \nWe previously showed that losartan was effective in reducing seizures in the X. laevis neurod2 \n(DEE72) CRISPant tadpole model (Banerjee et al., 2024), suggesting that TGF-β–associated \npathways represent a shared downstream vulnerability across genetically distinct DEEs. Our findings \nare also concordant with evidence from other epilepsy models and from clinical studies. In rodent \nseizure models, losartan reduces BBB permeability and seizure burden (Hong et al., 2019, \nTchekalarova et al., 2016), while epidemiological studies report a reduced incidence of epilepsy \namong patients treated with angiotensin II receptor blockers, particularly losartan (Doege et al., \n2022). At the same time, the absence of anticonvulsant effects in ex vivo human cortical tissue \n(Reyes-Garcia et al., 2019) underscores that losartan does not act as a conventional anti-seizure \nmedication, and that its efficacy depends on engagement of intact neurovascular and immune-\nmodulatory mechanisms. \nTaken together, our data indicate that AP3B2-deficient networks retain sensitivity to pharmacological \nmodulation of downstream regulatory pathways. Although the rescue observed here is partial and \nacute, the responsiveness to losartan identifies TGF-β–linked neurovascular mechanisms as \nfunctionally relevant modifiers of network instability in this DEE48 model. In this light, \nepidemiological evidence linking angiotensin II receptor blocker use, particularly losartan, to a \nreduced incidence of new-onset epilepsy (Wen et al., 2024b) provides independent support for the \ntherapeutic relevance of targeting neurovascular and homeostatic pathways in epileptic \nencephalopathies. \n5 Conclusions and Limitations of the study \nThis work demonstrates that loss of 80 to 85% of Ap3b2 in X. laevis tadpoles is sufficient to generate \na robust DEE48-like phenotype, characterized by seizure-like behaviour, increased neural activity \nand interhemispheric synchrony and early blood–brain barrier leakage. Whole-brain transcriptomics \nrevealed coordinated downregulation of inhibitory synaptic components, ion channels, axon-\nguidance pathways, and neuroinflammatory genes, providing a mechanistic framework that links \nAP3B2 deficiency to circuit instability and BBB fragility. The behavioural normalisation achieved \nwith acute losartan treatment highlights the potential therapeutic relevance of targeting TGF-β–\nassociated neuroinflammatory mechanisms. While our results support the use of Xenopus CRISPants \nas a rapid, integrative model for dissecting genetic DEE, several limitations should be noted. F₀ \nCRISPant tadpoles are both mosaic and carry varying levels of overall editing, with the variability on \ngenotype likely contributing to variability in phenotype severity. Generating stable mutant lines  \nwould reduce such variability, but the severity of the human DEE phenotypes suggests raising and \nmaintaining these could be challenging. RNA-seq was performed on whole brains, limiting cell-type \nresolution, and pooling of brains to create large enough samples could also reduce power. Losartan \nwas assessed only acutely and at a single developmental stage and dose (albeit based on previous \ntesting in another model), so the practicality of using it therapeutically in DEE remains untested. \nFinally, while Xenopus provides rapid access to early neurodevelopmental mechanisms, \ncomplementary studies will be essential to confirm the translational relevance of these findings. \n6. Data Availability Statement \nAll data supporting the conclusions of this study are available from public repositories. Summary \ndatasets and supporting analyses are provided in Supplementary File 1.xlsx and \nSupplementary_Material.docx. Raw and processed RNA-sequencing data have been deposited in \nthe NCBI Gene Expression Omnibus (GEO) under accession GSE312492. The complete analysis \npipeline is available via GitHub (https://github.com/sulagna-banerjee/xenopus-calcium-imaging-\npipeline) and permanently archived on Zenodo (DOI: https://doi.org/10.5281/zenodo.17931981). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n24 \n7. Conflict of Interest \nThe author(s) declare no conflicts of interest. \n8. Author Contributions \nConceptualization: S.B., C.W.B., P.S.; Data curation: S.B., C.W.B.; Formal analysis and \nvisualization: S.B., C.W.B., C.W.E, S.C.R.; Investigation S.B., C.W.E., C.W.B.; Funding acquisition, \nProject administration, Supervision: C.W.B., P.S.; Methodology: S.B., C.W.B., P.S. S.C.R; Writing-\noriginal draft preparation: S.B.,C.W.B, P.S. Writing-reviewing and editing: S.B., C.W.B., P.S. \n9. Funding \nThis work was funded by the Neurological Foundation of New Zealand Project Grant 2346PRG \n10. Acknowledgments \nThe authors thank Nikita Woodhead for Xenopus care, Joanna Ward for general lab technical \nassistance, Jack O’Neill for assistance designing the scrambled control sgRNA, and Edward Ruthazer \nand Anne Schohl for the kind gift of GCaMP6s-CS2 and mCherry-CS2+ plasmids. \n11. Tables \nTable 1. ap3b2.S sgRNAs and primer sequences \nsgRNA Oligo sequence (PAM) Forward Primer Reverse Primer Amplicon size \n2 GTCTTTGATGGGACATAGGAGGG GGAATAACCCAGGTCCCGAA AAGGCTGGTAACAGGGGGTA 703bp \n3 TGGTCGGGATCCATAACATACGG AGCCAAACCCAGCTGCTATC CCCGTATCAGGAAAACCCCA 558bp \n \nTable 2. Neuroinflammation-associated DEG downregulated in ap3b2 CRISPant tadpole brains \nX. laevis gene Human ortholog Log2FC PValue FDR \nil6st.L IL6ST  -1.0822 0.0003 0.0098 \ntgfbr2.L \nTGFBR2 \n-1.6293 0.0018 0.0133 \ntgfbr2.S -1.4140 0.0002 0.0098 \ntgfbr3.L \nTGFBR3 \n-1.0863 0.0056 0.0224 \ntgfbr3.S -1.7020 0.0046 0.0203 \nahr.L \nAHR  \n-1.6638 0.0062 0.0236 \nahr.S -1.4367 0.0011 0.0115 \nwdfy3.S WDFY3  -1.1333 0.0007 0.0105 \ncav1.L CAV1  -1.2217 0.0014 0.0123 \n \nTable 3. Neuroinflammation-associated genes upregulated in ap3b2 CRISPant tadpole brains \nX. laevis gene Human ortholog Log2FC PValue FDR \ncard9.L CARD9 1.1067 0.0064 0.0239 \nikbke.S IKBKE  1.0658 0.0000 0.0098 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n25 \npstpip2.S PSTPIP2 1.1003 0.0144 0.0401 \nfosb.L FOSB  1.2522 0.0127 0.0369 \nfosl1.S FOSL1 1.1257 0.0031 0.0168 \ngpr4.L GPR4 1.2102 0.0097 0.0307 \ncfi.L CFI  1.0237 0.0014 0.0124 \n \n12. References \nAlizadeh, P., Babadi, A. J., Ghadiri, N., Neissi, M. & Zeinali, M. 2025. Gene Variant Analysis In \nPediatrics With Early-Onset Epilepsy: Identification Of Novel Variants. Practical Laboratory \nMedicine, 45, E00462. \nAnazi, S., Maddirevula, S., Faqeih, E., Alsedairy, H., Alzahrani, F., Shamseldin, H., Patel, N., \nHashem, M., Ibrahim, N. & Abdulwahab, F. 2017. Clinical Genomics Expands The Morbid Genome \nOf Intellectual Disability And Offers A High Diagnostic Yield. Molecular Psychiatry, 22, 615-624. \nAssoum, M., Philippe, C., Isidor, B., Perrin, L., Makrythanasis, P., Sondheimer, N., Paris, C., \nDouglas, J., Lesca, G., Antonarakis, S., Hamamy, H., Jouan, T., Duffourd, Y ., Auvin, S., Saunier, A., \nBegtrup, A., Nowak, C., Chatron, N., Ville, D., Mireskandari, K., Milani, P., Jonveaux, P., Lemeur, \nG., Milh, M., Amamoto, M., Kato, M., Nakashima, M., Miyake, N., Matsumoto, N., Masri, A., \nThauvin-Robinet, C., Rivière, J.-B., Faivre, L. & Thevenon, J. 2016. Autosomal-Recessive Mutations \nIn Ap3b2, Adaptor-Related Protein Complex 3 Beta 2 Subunit, Cause An Early-Onset Epileptic \nEncephalopathy With Optic Atrophy. The American Journal Of Human Genetics, 99, 1368-1376. \nBanerjee, S., Szyszka, P. & Beck, C. W. 2024. Knockdown Of Neurod2 Leads To Seizure-Like \nBehavior, Brain Neuronal Hyperactivity And A Leaky Blood-Brain Barrier In A Xenopus Laevis \nTadpole Model Of Dee72. Genetics, 227, Iyae085. \nBar-Klein, G., Cacheaux, L. P., Kamintsky, L., Prager, O., Weissberg, I., Schoknecht, K., Cheng, P., \nKim, S. Y ., Wood, L., Heinemann, U., Kaufer, D. & Friedman, A. 2014. Losartan Prevents Acquired \nEpilepsy Via Tgf-Β Signaling Suppression. Annals Of Neurology, 75, 864-875. \nBell, M. R., Belarde, J. A., Johnson, H. F. & Aizenman, C. D. 2011. A Neuroprotective Role For \nPolyamines In A Xenopus Tadpole Model Of Epilepsy. Nature Neuroscience, 14, 505-512. \nBlair, C. R., Stone, J. B. & Wells, R. G. 2011. The Type Iii Tgf-Β Receptor Betaglycan \nTransmembrane-Cytoplasmic Domain Fragment Is Stable After Ectodomain Cleavage And Is A \nSubstrate Of The Intramembrane Protease Γ-Secretase. Biochim Biophys Acta, 1813, 332-9. \nBlumstein, J., Faundez, V ., Nakatsu, F., Saito, T., Ohno, H. & Kelly, R. B. 2001. The Neuronal Form \nOf Adaptor Protein-3 Is Required For Synaptic Vesicle Formation From Endosomes. The Journal Of \nNeuroscience, 21, 8034-8042. \nBrinkman, E. K., Chen, T., Amendola, M. & Van Steensel, B. 2014. Easy Quantitative Assessment Of \nGenome Editing By Sequence Trace Decomposition. Nucleic Acids Research, 42, E168-E168. \nCassel, S. L., Janczy, J. R., Bing, X., Wilson, S. P., Olivier, A. K., Otero, J. E., Iwakura, Y ., \nShayakhmetov, D. M., Bassuk, A. G., Abu-Amer, Y ., Brogden, K. A., Burns, T. L., Sutterwala, F. S. \n& Ferguson, P. J. 2014. Inflammasome-Independent Il-1β Mediates Autoinflammatory Disease In \nPstpip2-Deficient Mice. Proceedings Of The National Academy Of Sciences, 111, 1072-1077. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n26 \nCekanaviciute, E. & Buckwalter, M. S. 2016. Astrocytes: Integrative Regulators Of \nNeuroinflammation In Stroke And Other Neurological Diseases. Neurotherapeutics, 13, 685-701. \nChen, E. Y ., Tan, C. M., Kou, Y ., Duan, Q., Wang, Z., Meirelles, G. V ., Clark, N. R. & Ma'ayan, A. \n2013. Enrichr: Interactive And Collaborative Html5 Gene List Enrichment Analysis Tool. Bmc \nBioinformatics, 14, 128. \nChen, Q.-L., Xia, L., Zhong, S.-P., Wang, Q., Ding, J. & Wang, X. 2020. Bioinformatic Analysis \nIdentifies Key Transcriptome Signatures In Temporal Lobe Epilepsy. Cns Neuroscience & \nTherapeutics, 26, 1266-1277. \nClément, J.-F., Meloche, S. & Servant, M. J. 2008. The Ikk-Related Kinases: From Innate Immunity \nTo Oncogenesis. Cell Research, 18, 889-899. \nDadas, A. & Janigro, D. 2019. Breakdown Of Blood Brain Barrier As A Mechanism Of Post-\nTraumatic Epilepsy. Neurobiology Of Disease, 123, 20-26. \nDell'angelica, E. C., Ohno, H., Ooi, C. E., Rabinovich, E., Roche, K. W. & Bonifacino, J. S. 1997. \nAp‐3: An Adaptor‐Like Protein Complex With Ubiquitous Expression. The Embo Journal, 16, 917-\n928. \nDilber, C., Yücel, G. & Şahin, Y . 2022. Novel Homozygous Ap3b2 Mutations In Four Individuals \nWith Developmental And Epileptic Encephalopathy: A Rare Clinical Entity. Clinical Neurology And \nNeurosurgery, 223, 107509. \nDobin, A., Davis, C. A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., Batut, P., Chaisson, M. & \nGingeras, T. R. 2013. Star: Ultrafast Universal Rna-Seq Aligner. Bioinformatics, 29, 15-21. \nDoege, C., Luedde, M. & Kostev, K. 2022. Association Between Angiotensin Receptor Blocker \nTherapy And Incidence Of Epilepsy In Patients With Hypertension. Jama Neurology, 79, 1296-1302. \nDong, L., Li, Z., Leffler, N. R., Asch, A. S., Chi, J.-T. & Yang, L. V . 2013. Acidosis Activation Of \nThe Proton-Sensing Gpr4 Receptor Stimulates Vascular Endothelial Cell Inflammatory Responses \nRevealed By Transcriptome Analysis. Plos One, 8, E61991. \nDuesman, S. J., Ortega-Francisco, S., Olguin-Alor, R., Acevedo-Dominguez, N. A., Sestero, C. M., \nChellappan, R., De Sarno, P., Yusuf, N., Salgado-Lopez, A., Segundo-Liberato, M., De Oca-Lagunas, \nS. M., Raman, C. & Soldevila, G. 2023. Transforming Growth Factor Receptor Iii (Betaglycan) \nRegulates The Generation Of Pathogenic Th17 Cells In Eae. Frontiers In Immunology, V olume 14 - \n2023. \nFaúndez, V ., Horng, J.-T. & Kelly, R. B. 1998. A Function For The Ap3 Coat Complex In Synaptic \nVesicle Formation From Endosomes. Cell, 93, 423-432. \nFilimonenko, M., Isakson, P., Finley, K. D., Anderson, M., Jeong, H., Melia, T. J., Bartlett, B. J., \nMyers, K. M., Birkeland, H. C., Lamark, T., Krainc, D., Brech, A., Stenmark, H., Simonsen, A. & \nYamamoto, A. 2010. The Selective Macroautophagic Degradation Of Aggregated Proteins Requires \nThe Pi3p-Binding Protein Alfy. Mol Cell, 38, 265-79. \nFox, L. M., Kim, K., Johnson, C. W., Chen, S., Croce, K. R., Victor, M. B., Eenjes, E., Bosco, J. R., \nRandolph, L. K., Dragatsis, I., Dragich, J. M., Yoo, A. S. & Yamamoto, A. 2020. Huntington's \nDisease Pathogenesis Is Modified In Vivo By Alfy/Wdfy3 And Selective Macroautophagy. Neuron, \n105, 813-821.E6. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n27 \nGoasdoue, K., Chand, Kirat K., Miller, Stephanie M., Lee, Kah M., Colditz, Paul B., Wixey, Julie A. \n& Bjorkman, Stella T. 2019. Seizures Are Associated With Blood-Brain Barrier Disruption In A \nPiglet Model Of Neonatal Hypoxic-Ischaemic Encephalopathy. Developmental Neuroscience, 40, \n560-575. \nGomez-Arboledas, A., Acharya, M. M. & Tenner, A. J. 2021. The Role Of Complement In Synaptic \nPruning And Neurodegeneration. Immunotargets Ther, 10, 373-386. \nGorter, J. A., Van Vliet, E. A. & Aronica, E. 2015. Status Epilepticus, Blood–Brain Barrier \nDisruption, Inflammation, And Epileptogenesis. Epilepsy & Behavior, 49, 13-16. \nGuelfi, S., Botia, J. A., Thom, M., Ramasamy, A., Perona, M., Stanyer, L., Martinian, L., Trabzuni, \nD., Smith, C., Walker, R., Ryten, M., Reimers, M., Weale, M. E., Hardy, J. & Matarin, M. 2019. \nTranscriptomic And Genetic Analyses Reveal Potential Causal Drivers For Intractable Partial \nEpilepsy. Brain, 142, 1616-1630. \nGünzel, Y ., Couzin-Fuchs, E. & Paoli, M. 2024. Calciseg: A Versatile Approach For Unsupervised \nSegmentation Of Calcium Imaging Data. Neuroimage, 298, 120758. \nHara, H., Ishihara, C., Takeuchi, A., Imanishi, T., Xue, L., Morris, S. W., Inui, M., Takai, T., Shibuya, \nA., Saijo, S., Iwakura, Y ., Ohno, N., Koseki, H., Yoshida, H., Penninger, J. M. & Saito, T. 2007. The \nAdaptor Protein Card9 Is Essential For The Activation Of Myeloid Cells Through Itam-Associated \nAnd Toll-Like Receptors. Nat Immunol, 8, 619-29. \nHe, Y .-Y ., Zhou, H.-F., Chen, L., Wang, Y .-T., Xie, W.-L., Xu, Z.-Z., Xiong, Y ., Feng, Y .-Q., Liu, G.-\nY ., Li, X., Liu, J. & Wu, Q.-P. 2022. The Fra-1: Novel Role In Regulating Extensive Immune Cell \nStates And Affecting Inflammatory Diseases. Frontiers In Immunology, V olume 13 - 2022. \nHewapathirane, D. S., Dunfield, D., Yen, W., Chen, S. & Haas, K. 2008. In Vivo Imaging Of Seizure \nActivity In A Novel Developmental Seizure Model. Experimental Neurology, 211, 480-488. \nHong, S., Jiancheng, H., Jiawen, W., Shuqin, Z., Guilian, Z., Haiqin, W., Ru, Z., Zhen, G. & \nHongwei, R. 2019. Losartan Inhibits Development Of Spontaneous Recurrent Seizures By \nPreventing Astrocyte Activation And Attenuating Blood-Brain Barrier Permeability Following \nPilocarpine-Induced Status Epilepticus. Brain Research Bulletin, 149, 251-259. \nHsu, P. D., Scott, D. A., Weinstein, J. A., Ran, F. A., Konermann, S., Agarwala, V ., Li, Y ., Fine, E. J., \nWu, X., Shalem, O., Cradick, T. J., Marraffini, L. A., Bao, G. & Zhang, F. 2013. Dna Targeting \nSpecificity Of Rna-Guided Cas9 Nucleases. Nature Biotechnology, 31, 827-832. \nHuang, Q., Zhong, W., Hu, Z. & Tang, X. 2018. A Review Of The Role Of Cav-1 In Neuropathology \nAnd Neural Recovery After Ischemic Stroke. Journal Of Neuroinflammation, 15, 348. \nIacobaş, D. A. & Velíšek, L. 2018. Regeneration Of Neurotransmission Transcriptome In A Model Of \nEpileptic Encephalopathy After Antiinflammatory Treatment. Neural Regeneration Research, 13, \n1715-1718. \nKassambara, A. & Mundt, F. 2020. Factoextra: Extract And Visualize The Results Of Multivariate \nData Analyses. R Package Version 1.0.7. \nKimizu, T., Takahashi, Y ., Oboshi, T., Horino, A., Omatsu, H., Koike, T., Yoshitomi, S., Yamaguchi, \nT., Otani, H., Ikeda, H., Imai, K., Shigematsu, H. & Inoue, Y . 2018. Chronic Dysfunction Of Blood-\nBrain Barrier In Patients With Post-Encephalitic/Encephalopathic Epilepsy. Seizure, 63, 85-90. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n28 \nKjær, C., Barzaghi, G., Bak, L. K., Goetze, J. P., Yde, C. W., Woldbye, D., Pinborg, L. H. & Jensen, \nL. J. 2019. Transcriptome Analysis In Patients With Temporal Lobe Epilepsy. Brain, 142, E55-E55. \nKolde, R. 2025. Pheatmap: Pretty Heatmaps. R Package Version 1.0.13. \nKuleshov, M. V ., Jones, M. R., Rouillard, A. D., Fernandez, N. F., Duan, Q., Wang, Z., Koplev, S., \nJenkins, S. L., Jagodnik, K. M., Lachmann, A., Mcdermott, M. G., Monteiro, C. D., Gundersen, G. \nW. & Ma'ayan, A. 2016. Enrichr: A Comprehensive Gene Set Enrichment Analysis Web Server 2016 \nUpdate. Nucleic Acids Res, 44, W90-7. \nLabun, K., Montague, T. G., Gagnon, J. A., Thyme, S. B. & Valen, E. 2016. Chopchop V2: A Web \nTool For The Next Generation Of Crispr Genome Engineering. Nucleic Acids Research, 44, W272-\nW276. \nLê, S., Josse, J. & Husson, F. 2008. Factominer: An R Package For Multivariate Analysis. Journal Of \nStatistical Software, 25, 1 - 18. \nLi, V . J., Schohl, A. & Ruthazer, E. S. 2022. Topographic Map Formation And The Effects Of Nmda \nReceptor Blockade In The Developing Visual System. Proceedings Of The National Academy Of \nSciences, 119, E2107899119. \nLiao, Y ., Smyth, G. K. & Shi, W. 2014. Featurecounts: An Efficient General Purpose Program For \nAssigning Sequence Reads To Genomic Features. Bioinformatics, 30, 923-30. \nLibrizzi, L., Noè, F., Vezzani, A., De Curtis, M. & Ravizza, T. 2012. Seizure-Induced Brain-Borne \nInflammation Sustains Seizure Recurrence And Blood–Brain Barrier Damage. Annals Of Neurology, \n72, 82-90. \nLiu, W., Gao, K., Du, X., Wen, S., Yan, H., Wang, J., Wang, Y ., Song, C., Lin, L., Ji, T., Gu, W. & \nJiang, Y . 2024. Spout1 Variants Associated With Autosomal-Recessive Developmental And Epileptic \nEncephalopathy. Acta Epileptologica, 6, 42. \nLuo, J. 2022. Tgf-Β As A Key Modulator Of Astrocyte Reactivity: Disease Relevance And \nTherapeutic Implications. Biomedicines, 10, 1206. \nMedyanik, A. D., Anisimova, P. E., Kustova, A. O., Tarabykin, V . S. & Kondakova, E. V . 2025. \nDevelopmental And Epileptic Encephalopathy: Pathogenesis Of Intellectual Disability Beyond \nChannelopathies. Biomolecules, 15, 133. \nMoretti, R., Pansiot, J., Bettati, D., Strazielle, N., Ghersi-Egea, J.-F., Damante, G., Fleiss, B., \nTitomanlio, L. & Gressens, P. 2015. Blood-Brain Barrier Dysfunction In Disorders Of The \nDeveloping Brain. Frontiers In Neuroscience, V olume 9 - 2015. \nMurakami, M., Kamimura, D. & Hirano, T. 2019. Pleiotropy And Specificity: Insights From The \nInterleukin 6 Family Of Cytokines. Immunity, 50, 812-831. \nNakatsu , F., Okada , M., Mori , F., Kumazawa , N., Iwasa , H., Zhu , G., Kasagi , Y ., Kamiya , H., \nHarada , A., Nishimura , K., Takeuchi , A., Miyazaki , T., Watanabe , M., Yuasa , S., Manabe , T., \nWakabayashi , K., Kaneko , S., Saito , T. & Ohno , H. 2004. Defective Function Of Gaba-Containing \nSynaptic Vesicles In Mice Lacking The Ap-3b Clathrin Adaptor. Journal Of Cell Biology, 167, 293-\n302. \nNieuwkoop, P. D. & Faber, J. 1994. Normal Table Of Xenopus Laevis (Daudin): A Systematical And \nChronological Survey Of The Development From The Fertilized Egg Till The End Of Metamorphosis. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n29 \nOkamoto, O. K., Janjoppi, L., Bonone, F. M., Pansani, A. P., Da Silva, A. V ., Scorza, F. A. & \nCavalheiro, E. A. 2010. Whole Transcriptome Analysis Of The Hippocampus: Toward A Molecular \nPortrait Of Epileptogenesis. Bmc Genomics, 11, 230. \nPanthi, S., Chapman, P. A., Szyszka, P. & Beck, C. W. 2024. Characterisation And Automated \nQuantification Of Induced Seizure-Related Behaviours In Xenopus Laevis Tadpoles. Journal Of \nNeurochemistry, 168, 4014-4024. \nPoke, G., Stanley, J., Scheffer, I. E. & Sadleir, L. G. 2023. Epidemiology Of Developmental And \nEpileptic Encephalopathy And Of Intellectual Disability And Epilepsy In Children. Neurology, 100, \nE1363-E1375. \nRdevelopmentcoreteam 2024. R: A Language And Environment For Statistical Computing (V 4.5.1). \nVienna, Austria.: R Foundation For Statistical Computing. \nReyes-Garcia, S. Z., Scorza, C. A., Ortiz-Villatoro, N. N. & Cavalheiro, E. A. 2019. Losartan Fails To \nSuppress Epileptiform Activity In Brain Slices From Resected Tissues Of Patients With Drug \nResistant Epilepsy. Journal Of The Neurological Sciences, 397, 169-171. \nRiva, A., V olpedo, G., Zara, F., Fassio, A., Striano, P. & Falace, A. 2025. Pathophysiological \nMechanisms Fostering Developmental And Epileptic Encephalopathies (Dee): A Complex Interplay \nBetween Genetics, Inflammation And Neurodegeneration. Current Neurology And Neuroscience \nReports, 25, 65. \nRobinson, M. D., Mccarthy, D. J. & Smyth, G. K. 2010. Edger: A Bioconductor Package For \nDifferential Expression Analysis Of Digital Gene Expression Data. Bioinformatics, 26, 139-40. \nRose-John, S. 2018. Interleukin-6 Family Cytokines. Cold Spring Harb Perspect Biol, 10. \nSalman, M. M., Sheilabi, M. A., Bhattacharyya, D., Kitchen, P., Conner, A. C., Bill, R. M., \nWoodroofe, M. N., Conner, M. T. & Princivalle, A. P. 2017. Transcriptome Analysis Suggests A Role \nFor The Differential Expression Of Cerebral Aquaporins And The Mapk Signalling Pathway In \nHuman Temporal Lobe Epilepsy. European Journal Of Neuroscience, 46, 2121-2132. \nSanz, P., Rubio, T. & Garcia-Gimeno, M. A. 2024. Neuroinflammation And Epilepsy: From \nPathophysiology To Therapies Based On Repurposing Drugs. Int J Mol Sci, 25. \nScheffer, I. E. & Liao, J. 2020. Deciphering The Concepts Behind “Epileptic Encephalopathy” And \n“Developmental And Epileptic Encephalopathy”. European Journal Of Paediatric Neurology, 24, 11-\n14. \nScheffer, I. E., Zuberi, S., Mefford, H. C., Guerrini, R. & Mctague, A. 2024. Developmental And \nEpileptic Encephalopathies. Nature Reviews Disease Primers, 10, 61. \nSchindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V ., Longair, M., Pietzsch, T., Preibisch, S., \nRueden, C., Saalfeld, S., Schmid, B., Tinevez, J. Y ., White, D. J., Hartenstein, V ., Eliceiri, K., \nTomancak, P. & Cardona, A. 2012. Fiji: An Open-Source Platform For Biological-Image Analysis. \nNat Methods, 9, 676-82. \nSega, A. G., Mis, E. K., Lindstrom, K., Mercimek-Andrews, S., Ji, W., Cho, M. T., Juusola, J., \nKonstantino, M., Jeffries, L., Khokha, M. K. & Lakhani, S. A. 2019. De Novo Pathogenic Variants In \nNeuronal Differentiation Factor 2 (Neurod2) Cause A Form Of Early Infantile Epileptic \nEncephalopathy. J Med Genet, 56, 113-122. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n30 \nSeong, E., Wainer, B., Hughes, E., Saunders, T., Burmeister, M. & Faundez, V . 2005. Genetic \nAnalysis Of The Neuronal And Ubiquitous Ap-3 Adaptor Complexes Reveals Divergent Functions In \nBrain. Molecular Biology Of The Cell, 16, 128-140. \nShen, M. W., Arbab, M., Hsu, J. Y ., Worstell, D., Culbertson, S. J., Krabbe, O., Cassa, C. A., Liu, D. \nR., Gifford, D. K. & Sherwood, R. I. 2018. Predictable And Precise Template-Free Crispr Editing Of \nPathogenic Variants. Nature, 563, 646-651. \nSpecchio, N., Fusco, L., Claps, D. & Vigevano, F. 2010. Epileptic Encephalopathy In Children \nPossibly Related To Immune-Mediated Pathogenesis. Brain And Development, 32, 51-56. \nSwissa, E., Serlin, Y ., Vazana, U., Prager, O. & Friedman, A. 2019. Blood–Brain Barrier Dysfunction \nIn Status Epileptics: Mechanisms And Role In Epileptogenesis. Epilepsy & Behavior, 101, 106285. \nSymonds, J. D., Elliott, K. S., Shetty, J., Armstrong, M., Brunklaus, A., Cutcutache, I., Diver, L. A., \nDorris, L., Gardiner, S., Jollands, A., Joss, S., Kirkpatrick, M., Mclellan, A., Macleod, S., O’regan, \nM., Page, M., Pilley, E., Pilz, D. T., Stephen, E., Stewart, K., Ashrafian, H., Knight, J. C. & Zuberi, S. \nM. 2021. Early Childhood Epilepsies: Epidemiology, Classification, Aetiology, And Socio-Economic \nDeterminants. Brain, 144, 2879-2891. \nTa, A. C., Huang, L.-C., Mckeown, C. R., Bestman, J. E., Van Keuren-Jensen, K. & Cline, H. T. \n2021. Temporal And Spatial Transcriptomic Dynamics Across Brain Development In Xenopus \nLaevis Tadpoles. G3 Genes|Genomes|Genetics, 12. \nTchekalarova, J. D., Ivanova, N., Atanasova, D., Pechlivanova, D. M., Lazarov, N., Kortenska, L., \nMitreva, R., Lozanov, V . & Stoynev, A. 2016. Long-Term Treatment With Losartan Attenuates \nSeizure Activity And Neuronal Damage Without Affecting Behavioral Changes In A Model Of Co-\nMorbid Hypertension And Epilepsy. Cellular And Molecular Neurobiology, 36, 927-941. \nTrevino, T. N., Almousawi, A. A., Robinson, K. F., Fogel, A. B., Class, J., Minshall, R. D., Tai, L. M., \nRichner, J. M. & Lutz, S. E. 2024. Caveolin-1 Mediates Blood-Brain Barrier Permeability, \nNeuroinflammation, And Cognitive Impairment In Sars-Cov-2 Infection. J Neuroimmunol, 388, \n578309. \nVerhelst, K., Verstrepen, L., Carpentier, I. & Beyaert, R. 2013. Iκb Kinase Ε (Ikkε): A Therapeutic \nTarget In Inflammation And Cancer. Biochem Pharmacol, 85, 873-80. \nWang, Y ., Sun, J., Zhu, K., Wang, D., Zhao, X., Zhang, H., Wu, S., Wang, Y . & Wang, J. 2023. \nMicroglial Aryl Hydrocarbon Receptor Enhances Phagocytic Function Via Syk And Promotes \nRemyelination In The Cuprizone Mouse Model Of Demyelination. Journal Of Neuroinflammation, \n20, 83. \nWen, F., Tan, Z., Huang, D. & Xiang, J. 2024a. Molecular Mechanism Analyses Of Post-Traumatic \nEpilepsy And Hereditary Epilepsy Based On 10× Single-Cell Transcriptome Sequencing Technology. \nCns Neuroscience & Therapeutics, 30, E14702. \nWen, X., Otoo, M. N., Tang, J., Brothers, T., Ward, K. E., Asal, N. & Meador, K. J. 2024b. \nAngiotensin Receptor Blockers For Hypertension And Risk Of Epilepsy. Jama Neurology, 81, 866-\n874. \nWheeler, M. A., Rothhammer, V . & Quintana, F. J. 2017. Control Of Immune-Mediated Pathology \nVia The Aryl Hydrocarbon Receptor. Journal Of Biological Chemistry, 292, 12383-12389. \nWickham, H. 2016. Ggplot2: Elegant Graphics For Data Analysis. New York: Springer-Verlag. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint \n\n \n31 \nXie, Z., Bailey, A., Kuleshov, M. V ., Clarke, D. J. B., Evangelista, J. E., Jenkins, S. L., Lachmann, A., \nWojciechowicz, M. L., Kropiwnicki, E., Jagodnik, K. M., Jeon, M. & Ma'ayan, A. 2021. Gene Set \nKnowledge Discovery With Enrichr. Current Protocols, 1, E90. \nZhong, X., Chen, B., Yang, L. & Yang, Z. 2018. Molecular And Physiological Roles Of The Adaptor \nProtein Card9 In Immunity. Cell Death & Disease, 9, 52. \nZöller, T., Schneider, A., Kleimeyer, C., Masuda, T., Potru, P. S., Pfeifer, D., Blank, T., Prinz, M. & \nSpittau, B. 2018. Silencing Of Tgfβ Signalling In Microglia Results In Impaired Homeostasis. Nature \nCommunications, 9, 4011. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted December 30, 2025. ; https://doi.org/10.64898/2025.12.29.696940doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}