Biallelic null RAB3GAP1 variants impair cortical development and autophagy in Warburg Micro syndrome: evidence from fetal brain tissue and patient fibroblasts Authors: Emma Noël, Fabien Guimiot, Yline Capri, Marianne Alison, Asha Baskaran, Clémence Delcour, David Germanaud, Sophie Lebon, Caroline Storey, Nicolas de Roux, Adeline Orts-Del’Immagine Published: 2025 ABSTRACT Warburg Micro syndrome (WARBM) is a rare autosomal recessive disorder characterized by ocular, neurodevelopmental, and neuroendocrine abnormalities, most commonly caused by RAB3GAP1 variants. RAB3GAP1 encodes the catalytic subunit of the RAB3 GTPase-activating protein (RAB3GAP), which regulates vesicular release and membrane trafficking through Rab GTPase modulation. Although in vitro studies suggest that RAB3GAP1 dysfunction impairs autophagy, the neuropathology of WARBM in the developing Human brain remains undocumented. Here, we present the first detailed neuropathological and molecular analysis of Human WARBM, based on two related cases: a 3-month-old infant and a 23-week gestation fetus with biallelic RAB3GAP1 pathogenic variants. Histological examination of the fetal cerebral cortex and cerebellum revealed selective vulnerability, with cortical plate thinning, while Purkinje cells appeared preserved. Consistently, immunohistochemistry demonstrated widespread RAB3GAP1 expression in the frontal cortex of controls, but no detectable staining in WARBM, coinciding with disrupted neurogenic niches, including reduced SOX2-positive progenitor cells, disorganized radial glia, and increased caspase-3 expression. These changes were associated with fewer DCX- and CTIP2-positive neurons, which may reflect a combination of altered neurogenesis, impaired neuronal migration and increased cell death. However, no gross brain malformations were observed in the fetus, whereas cortical malformations were evident in the 3-month-old infant, pointing to a progressive neurodevelopmental disruption potentially driven by fetal corticogenesis defects. These changes were accompanied by marked dysregulation of autophagy markers, linking impaired autophagy to disrupted cortical development. In contrast, autophagy disruption in the fetal lens was already associated with bilateral cataracts. Studies in patient-derived skin fibroblasts using immunocytochemistry, western blotting, and electron microscopy confirmed autophagy defects and tissue-specific consequences to RAB3GAP1 disruption in WARBM.Collectively, our findings provide the first in-depth neuropathological characterization of WARBM, highlighting the essential role of RAB3GAP1 in early Human and support autophagy as a key pathogenic feature in this severe disorder.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40478-025-02204-8. Introduction Warburg Micro syndrome (WARBM) is a rare autosomal recessive disorder characterized by ocular, neurodevelopmental, and neuroendocrine abnormalities [1]. Diagnosis is based on clinical presentation and characteristic brain MRI findings, including postnatal microcephaly, polymicrogyria, hypoplasia of the corpus callosum, cerebellar atrophy, and myelination defects, in association with biallelic variants in one of four genes: RAB3GAP1, RAB3GAP2, RAB18, or TBC1D20 [2–4]. Most cases are caused by disease-causing variants in RAB3GAP1, which encodes the catalytic subunit of the RAB3 GTPase-activating protein (RAB3GAP) complex. Together with its regulatory subunit, RAB3GAP2, it modulates the activity cycles of RAB family small GTPases, including RAB3 and RAB18 [5–7], which are essential for intracellular membrane trafficking and vesicular release [8–11]. Despite being described over three decades ago [1], the neuropathological features of WARBM remain poorly understood. Current knowledge is largely based on neuroimaging, genetic, and clinical studies in affected children [4, 12–14]. Only a single prenatal case has been reported to date, a 19-week fetus diagnosed via ultrasound and genetic testing due to bilateral cataracts, but this case provided no information on fetal brain development or pathology [15]. As a result, the pathophysiology of WARBM remains speculative, limiting efforts to identify therapeutic targets that might improve clinical phenotype. This lack of understanding stems not only from the extreme rarity of the disorder but also from the fact that Rab3gap1 knockout mice fail to faithfully recapitulate the human phenotype. For instance, Rab3gap1 knockout mice show synaptic transmission and plasticity defects linked to RAB3 dysregulation, but they do not develop the brain, ocular, or genital abnormalities observed in Humans [10]. In model organisms such as C. elegans, combined loss of function of Rab3gap1 (rbg-1) with Rab3gap2 (rbg-2) or Rab18 leads to impaired autophagy [16, 17], a fundamental pathway that maintains cellular homeostasis through the degradation and recycling of intracellular components [18, 19]. Consistently, depletion of RAB3GAP1 with RAB3GAP2, or RAB18 in Human fibroblasts similarly impairs autophagy [20, 21]. More recently, Ghate et al., showed that downregulation of RAB3GAP1 in Human iPSC-derived cortical neurons leads to reduced neurite outgrowth [22], suggesting a direct role in neuronal maturation and cortical circuit formation. However, the potential disruption of autophagy in these cells was not investigated, leaving this aspect unaddressed. In fact, autophagy defects have not yet been examined in Human WARBM brain tissue, and the effects of biallelic RAB3GAP1 pathogenic variants on neurodevelopment and corticogenesis remain unexplored. Here, we present the first in-depth neuropathological and molecular analysis of WARBM, based on two related cases harboring compound heterozygous RAB3GAP1 variants: a proband studied at 3 months of age and a fetus at 23 gestational weeks (GW). This unique pairing provides a rare opportunity to compare early developmental abnormalities with their postnatal consequences. Neuropathological analysis of the fetus revealed that RAB3GAP1 loss-of-function caused pronounced cellular alterations in the cerebral cortex without macroscopic brain defects, alongside major impairments in the autophagy pathway. Altogether, this study expands our understanding of the early pathophysiological events underlying cortical malformations in WARBM and lays the groundwork for future investigations into critical developmental windows and cellular mechanisms in this devastating neurodevelopmental disorder. Materials and methods Human subjects The proband is a boy born to healthy parents following an uneventful full-term pregnancy and spontaneous vaginal delivery. WARBM diagnosis was suspected at one month old because of the association of bilateral microphthalmia, cataract, axial hypotonia, 3rd percentile head circumference and micropenis with unilateral cryptorchidism (Additional File 1: Table S1). During a subsequent pregnancy, bilateral cataracts were detected prenatally, raising suspicion of WARBM; termination of pregnancy (TOP) was performed at 23 gestational weeks (GW), hereafter referred to as FWARBM. The parents later had a third child, a female, who is currently healthy. For neuropathological and western blot analyses, brains from 23 GW control fetuses (FControl) were used (Additional File 2: Table S2; ABM approval: PFS25-11599). Molecular genetic analysis DNA was extracted from blood lymphocytes. The coding sequence and donor/acceptor splicing sites of RAB3GAP1 were sequenced with a gene panel library prepared by the Sureselect method (Agilent). Sequencing reads were aligned to the Human reference genome (GRCh37/Hg19) and rare single nucleotide variants as well as structural variants were called with the CLC genomic workbench (QIAGEN) and annotated by Clinical Insight (QIAGEN). An exome analysis was also performed as described [23]. Total RNA was extracted from the proband’s blood lymphocytes (miRNeasy Mini Kit; 217004, Qiagen) and reverse-transcribed into cDNA using the SuperScript First-Strand Synthesis Kit (18091050, ThermoFisher Scientific), according to the manufacturer’s instructions. cDNA was used as a template to PCR-amplify exons 10 to 19 using the following primer pair: Forward: ACTCCATTGCCTCCAGTTAGT; Reverse: CGGGCACTCAGTTCTCCTTT. PCR products were gel-purified and subsequently sequenced. Cell culture and treatments Primary human skin fibroblasts were obtained from the proband and age-matched children (ethics approval: Robert Debré Hospital, 2007*A0101251) and cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with Glutamax supplement (61965–026, GIBCO), 10% fetal bovine serum (FBS; A5256701, GIBCO), 1% Penicillin/Streptomycin (15140-122, GIBCO) and 1 mM Sodium Pyruvate (11360-070, GIBCO) at 37 °C in a 5% CO2 incubator. All treatments to modulate autophagy flux were applied for 4h at 37 °C in a 5% CO2 incubator. To block autophagy flux, we used 100 nM Bafilomycin A1 (19-148, Sigma-Aldrich) prepared in 10% FBS-DMEM. For starvation we used Hanks’ Balanced Salt Solution (HBSS; 14175095, GIBCO). Basal condition consisted in culture medium containing DMSO (1:1000; D8418, Sigma-Aldrich). All experiments were conducted in triplicate and cells were processed together. After treatment, cells were washed two times in cold 1X phosphate-buffered saline (PBS; 10010031, GIBCO) and either fixed with paraformaldehyde (PFA; 47340, Thermo Scientific) 4% for immunocytochemistry, with 2% PFA / 2% glutaraldehyde (G5882, Sigma-Aldrich,) for electronic microscopy, or lysed for protein extraction. Western blot Western blot analysis was performed on proteins extracted from Human fibroblasts and fetal brain extracts. Proteins were extracted in RIPA buffer (R-0278, Sigma-Aldrich) with a cocktail of protease and phosphatase inhibitors (78444, Thermo Scientific) and EDTA (1861283, Thermo Scientific). Lysates were incubated on ice for 45 min and centrifuged at 21,000 g for 30 min at 4 °C. The protein concentration in supernatant was determined using BCA assay. Equal amounts of protein were separated by SDS-PAGE and transferred to a TransBlot Turbo PVDF membrane (1704157, Biorad). The membranes were blocked with 5% milk in Tween 20 (P1379, Sigma-Aldrich) 0.1% in tris-buffered saline (TBST; 2h, room temperature: RT; T9141, Takara) and the membranes were incubated with primary antibodies listed in Additional File 3: Table S3, diluted in the same solution overnight at 4 °C. Membranes were washed twice with TBST and incubated with anti-Mouse or anti-Rabbit HRP-conjugated secondary antibodies (1h, RT, 1: 20 000 in TBST; 115-035-166 and 111-036-144 respectively, Jackson). The membrane was incubated for 5 min in Clarity Max ECL substrate (1705062, Biorad), and bands were imaged using a Syngene PXi without saturation and analyzed with Imagelab (Biorad). The relative density was calculated as the ratio of the intensity of the protein of interest to that of actin or alpha tubulin as the housekeeping protein, and data are presented as fold changes to the control DMSO group. Histology After autopsy, brains were fixed in formalin, embedded in paraffin, and stained with hematoxylin–eosin or analyzed by immunostaining. Neuropathological examination was performed on representative slides from the cerebellar hemispheres and vermis, eyes and frontal, parietal and temporal cortex as previously described [24]. Immuno-staining Immunostaining protocol was performed on fetal brain sections and fibroblasts. All antibodies used in this study are listed in Additional File 3: Table S3. Paraffin frontal brain sections 5-µm-thick coronal sections of the frontal lobe (corresponding to image 8 to 10 on the 21 post-conception weeks Human Allen Brain Atlas, Allen Reference Atlases: Atlas Viewer) were obtained using a Leica semi-automatic microtome (14051956472, HistoCore AUTOCUT). Before labeling, paraffin was removed with xylene and ethanol and antigen retrieval was performed by incubating in citrate buffer for 40 min at 94 °C (1.8mM citric acid and 8.2mM sodium citrate, pH6). Sections were then permeabilized with 0.3% Triton X-100 in 1X PBS, (15 min, RT), incubated with primary antibodies at the dilutions reported in Table S3 for 48h at 4 °C in Bond™ Primary Antibody Diluent (AR9352, Leica). Sections were then incubated with secondary antibodies in 0.1% Triton X-100, 10% Donkey Serum (DS; S30, Sigma-Aldrich) in PBS 1X solution (2h, RT). The secondary antibodies used were as follows: Alexa Fluor® 488 Anti-Mouse (1:1000; A21202, Invitrogen), Cy3- Anti-Rabbit (1:500; 711-165-152, Jackson), Cy3- Anti-Rat (1:1000; 112-165-143, Jackson). Nuclei were labeled with 4′,6-diamidino-2-phenylindole (DAPI) added to Fluoromount-G mounting medium (0100-01, SouthernBiotech) used to mount coverslips. Fibroblasts immunostaining Fibroblasts were cultured on poly-d-lysine (P6407, Sigma-Aldrich) coated cover slips and fixed with 4% PFA in PBS 1X (pH 7.4, 20 min, 4 °C). Following PBS wash, cells were permeabilized and blocked with 0.1% Triton X-100, 2% BSA, 3% DS in PBS 1X solution (30 min, RT). To observe global fibroblasts morphology, ActinRed-555 (R37112, Invitrogen) was incubated (30 min, RT) in PBS. For immunocytochemistry, cells were incubated with primary antibody (Additional File 3: Table S3) diluted in 0.1% Triton X-100, 2% DS in PBS 1X solution overnight at 4 °C, washed in PBS 1X, and incubated with Alexa Fluor® 488 Anti-Mouse (1:1000; A21202, Invitrogen) diluted in 0.1% Triton X-100, 2% DS in PBS 1X solution (1h, RT). For lipid droplet staining, PFA-fixed cells were permeabilized and blocked with 0.1% Saponin, 0.1% BSA in PBS 1X (10 min, RT), then incubated with BODIPY 493/503 (D3922, Invitrogen) in the same solution (1µg/mL, 30 min, 37 °C). Coverslips from each staining were mounted on glass slides using Fluoromount mounting medium containing DAPI (0100-20, SouthernBiotech). Imaging Confocal imaging Imaging was performed using a Leica TCS SP8 confocal scanning system (Leica Microsystems) equipped with 405-nm Diode, 488-nm Ar and 561-nm DPSS lasers. Twelve-bit (p62 quantification on fibroblasts) or eight-bit digital images were collected in a sequential mode using 40x (numerical aperture, NA 1) or a 20x (NA 0.75) HC PL APO CS2 oil-immersion Leica objectives with a 1X or 2X digital zoom. To ensure comparability, images used for quantification were acquired with identical laser intensities and settings. Pictures with composite colors and the corresponding Tiff files were generated and processed on Fiji [25] for sizing and eventually for improving color and contrast. All illustrated images of fetal brain are Z-max projections. Scan Whole fetal brain sections were scanned at 40 × magnification with the digital whole-slide NanoZoomer Digital Pathology 2.0 HT scanner (Hamamatsu Photonics), using the Fluorescence Unit option (L11600-05) and the NanoZoomer’s 3-CCD TDI camera. Images were then processed with NDP viewer software and exported as Tiff files for illustrations. Electron microscopy Fibroblasts in culture were fixed in 2% glutaraldehyde and 2% PFA in PBS overnight at 4 °C. Samples were washed 5 times and then post-fixed with 2% osmium tetroxide in distilled water (1h, RT). After an extensive wash (distilled water), cells were incubated for 1h in 2% uranyl acetate in water and then dehydrated in a graded series of ethanol solutions (10 min, from 50 to 100%). Samples were then progressively infiltrated with an epoxy resin, Epon 812® (EMbed 812, Electron Microscopy Sciences) overnight in 50% resin 50% ethanol at 4 °C in an airtight container, 2 × 2h in pure fresh resin (RT). After the resin polymerization (56 °C for 48h in a dry oven), embedded blocks were cut with an UC7 ultramicrotome (Leica Microsystemes). Ultra-thin sections (70nm thick) were cut and collected on copper grid. Sections were then contrasted with Reynolds lead citrate for 7min (Reynolds, ES (1963). Ultrathin sections were observed with a Hitachi HT7700 electron microscope operating at 100 kV. Pictures (2048 × 2048 pixels) were taken with a Nanosprint 12 camera. Analysis Quantifications on scanned fetal cortical sections Scanned Brain slices (FWARBM vs. three FControl) were analyzed by using QuPath-0.5.0. Cortical thickness measurement To determine mean cortical thickness, the cortical plate was identified manually, omitting the molecular layer of the plate and a line was drawn perpendicularly to the tissue surface and automatically measured (Additional File 4: Fig. S1). Cell density counting For each fetus, three to four regions of interest in two frontal, four parietal and four temporal Hematoxylin–eosin-stained cortex sections were analyzed. To determine cellular density, regions of interest were drawn using the polygon tracing tool to select either the whole depth of the cortical plate, still omitting the molecular layer, or the subplate (red rectangles). Nuclei were counted automatically in each rectangle using the Cell detection—Positive cell detection commands with a threshold of 0.1. Mean cellular density was obtained by normalizing the number of nuclei on the rectangle surface (Additional File 4: Fig. S1). DCX+ cells quantification For each fetus, three to four regions of interest in the cortical plate and in the subplate from two hemi-frontal lobe sections stained with anti-doublecortin antibody were analyzed. The cortical plate and subplate were identified manually, omitting the molecular layer. The cortical plate was defined based on the density of DAPI-stained nuclei in the dorsal part of the slice, located beneath the molecular layer. In this analysis, the subplate corresponds to the region between the cortical plate and the subventricular zone (Additional File 7: Fig. S3c). Regions of interest were drawn using the polygon tracing tool to select the whole depth of the area. To determine DCX+ cells proportion, positive cells were counted automatically in each region of interest using the Cell detection—Positive cell detection commands with a threshold of 15. Mean cellular percentage was obtained by normalizing the number of cells to the number of DAPI-stained nuclei, measured using the same method (Additional File 4: Fig. S1). Quantifications from confocal imaging SOX2+ cells quantification For each fetus, 3 to 6 regions of interest in the ventricular zone and in the subventricular zone from two hemi-frontal lobe sections stained with anti-SOX2 antibody were analyzed. SOX2-positive cells and DAPI-positive nuclei were manually counted on 2.5 µm depth max z-projection images (optical sections 0.5 µm) through the cell counter plug-in of Fiji. FWARBM and FControl images were processed together blindly by the same experimenter. CTIP2+ cells quantification A default filter (IsoData) was applied on 9 µm depth max z-projection images (optical section 0.35 µm) and CTIP2+ cells and DAPI+ nuclei were automatically counted through the analyze particle command of Fiji. CTIP2+ cells proportion was then normalized to the mean cortical plate thickness of each fetal section, measured on scanned images of the section on QuPath-0.5.0, as described above. p62+ particles quantification in fibroblasts A default threshold (1500) was applied on 6 µm depth max z-projection images (optical section 0.35 µm) and p62+ particles in single cell, manually delimitated, were automatically counted through the analyze particle command of Fiji. The number of particles detected for each cell, was then normalized to the surface area of the analyzed region in µm2. WARBM and Control images were processed together blindly by the same experimenter. Lipid droplet quantification Fibroblasts stained with Bodipy were manually delimitated, and the mean fluorescence intensity of each cell was determined using Fiji’s Measure command on 6 µm maximum z-projection images (optical section 0.35 µm), without any modification of brightness or contrast. WARBM and control images were processed together blindly by the same experimenter. Lipid droplets on electron microscopy images identified as round, electron-dense structures exhibiting uniform color and texture, without visible membranes and manually counted using the Cell Counter plugin in Fiji. The number of lipid droplets per cell was then normalized to the area of the drawn cytoplasmic region of interest, and data are presented as mean lipid droplet density per cell and per µm2. Statistical analysis All statistical analyses were performed with GraphPad Prism 10.3.1 software. Nonparametric Mann–Whitney test was used for image analysis, and One-way ANOVA followed by Bonferroni post-hoc test was used for western blot analysis. Protein levels were expressed as fold change relative to the control DMSO condition. A p value < 0.05 was always considered significant. All data are reported as mean ± SEM. Results Effects of complete RAB3GAP1 inactivation on brain development The first child of a healthy, unrelated couple was evaluated at birth for a constellation of anomalies, including bilateral microphthalmia, congenital cataract, axial hypotonia, 3rd percentile head circumference, micropenis and unilateral cryptorchidism. The proband was profoundly blind despite undergoing bilateral cataract extraction at 2 months of age. Targeted gene panel analysis identified a heterozygous nonsense variant in RAB3GAP1 (NM_012233.3:c.68G > A), inherited from the father, predicted to introduce a premature stop codon at position 23 (NP 036365.1:p(Trp23Ter)) (Fig. 1a). Subsequent exome sequencing identified a heterozygous deletion of exons 12–17 in RAB3GAP1 inherited from the mother (Fig. 1a). The 5′ and 3′ ends of the deletion were located in introns 11 and 17, respectively. Because the precise limits of the deletion were not established, the variant was designated NC_000002.12 (NM_012233.3):c.(79321 + 1_79322-1)_(103105 + 1_103106-1)del, according to HGVS nomenclature, and is hereafter referred to as g.ex12_ex17del (Fig. 1a). The canonical RAB3GAP1 transcript comprises 24 exons and encodes a 921–amino acid protein. The deletion is predicted to generate a truncated protein of 340 amino acids resulting from a frameshift-induced premature stop codon (NP_036365.1:p.(Asp327Thrfs*14). The truncation at amino acid 340 removes the C-terminal GTPase activating domain [6], and some residues involved in the interaction with RAB3GAP2 [5]. Together, these alterations are consistent with a loss-of-function mechanism, in agreement with previous reports linking dysfunctional RAB3GAP1 catalytic domain to WARBM [12]. Brain MRI performed at 3 months of age demonstrated bilateral frontoparietal polymicrogyria and a short, thin and dysmorphic corpus callosum consistent with callosal dysgenesis (Fig. 1b). During a subsequent pregnancy, the 18-week gestational ultrasound did not reveal any cerebral anomalies but detected bilateral increased echogenicity of the lenses, raising suspicion for cataracts and recurrence of the syndrome (Fig. 1c). Post-termination (TOP) genetic testing confirmed the fetus harbored the same biallelic RAB3GAP1 pathogenic variants as the proband (Fig. 1a). Western blot analysis of protein extracts from the cortex and cerebellum of the WARBM fetus (FWARBM) showed absence of both the expected 130 kDa RAB3GAP1 band and the smaller 85 kDa isoform observed in control fetal brain tissue (FControl) (Fig. 1d). Post-TOP examination at 23 weeks of gestation revealed no external malformations except a telecanthus, short palpebral fissures, micrognathism and slightly short hallux (Additional File 5: Table S4). Fetal growth parameters were within normal ranges (weight: 530 g, 50th percentile; crown-heel length: 30 cm, 50th percentile; crown-rump length: 21 cm, 50th percentile; head circumference: 20 cm, 25th–50th percentile; biparietal diameter: 53 mm, 25th percentile). The fetal brain appeared small (occipito-frontal diameter: 66 mm, 50th centile; weight:65 g, 5th–10th centiles) but structurally normal, with cortical gyration appropriate for gestational age (Additional File 6: Fig. S2). There were no anomalies observed after brain sectioning, the parenchyma was conserved, and the structures of the median line were visualized. Histological analysis of the eyes revealed rosettes of retinal tissue at the optic nerve insertion, consistent with bilateral focal retinal dysplasia, and small lenses composed of highly irregular, swollen, and extensively vacuolated fibers, confirming the presence of bilateral cataracts (Fig. 1e). The couple subsequently had a third healthy daughter, heterozygous for the paternal RAB3GAP1 nonsense variant (NM_012233.3:c.68G > A) (Fig. 1a).Fig. 1Brain and eye malformations in patients with biallelic RAB3GAP1 pathogenic variants. (a) Pedigree tree of the family with RAB3GAP1 genotype. (b) Coronal and axial T2-weighted brain MRI sequences and sagittal T1-weighted brain MRI sequence of the proband at 3 months of age. Yellow arrows point to fronto-parietal polymicrogyria and dotted arrows to callosal dysgenesis. The vermis as a normal appearance. (c) WARBM fetus ultrasound at 18 weeks of gestation demonstrating on the axial view of the orbits the bilateral increased echogenicity of the lens (yellow arrow) and on the midline sagittal brain view, a short corpus callosum (cc, yellow dashed line). (d) RAB3GAP1 expression in control and WARBM fetus brain tissue extracts (cropped gels). (e) Hematoxylin–Eosin-stained sagittal eye sections of the WARBM and control fetus at 23 GW. The black arrow indicates focal retinal dysplasia Brain and eye malformations in patients with biallelic RAB3GAP1 pathogenic variants. (a) Pedigree tree of the family with RAB3GAP1 genotype. (b) Coronal and axial T2-weighted brain MRI sequences and sagittal T1-weighted brain MRI sequence of the proband at 3 months of age. Yellow arrows point to fronto-parietal polymicrogyria and dotted arrows to callosal dysgenesis. The vermis as a normal appearance. (c) WARBM fetus ultrasound at 18 weeks of gestation demonstrating on the axial view of the orbits the bilateral increased echogenicity of the lens (yellow arrow) and on the midline sagittal brain view, a short corpus callosum (cc, yellow dashed line). (d) RAB3GAP1 expression in control and WARBM fetus brain tissue extracts (cropped gels). (e) Hematoxylin–Eosin-stained sagittal eye sections of the WARBM and control fetus at 23 GW. The black arrow indicates focal retinal dysplasia Neuropathological analysis of the fetal brain In the brain, cortical plate (CP) thickness was reduced by approximately 15 ± 2%, without a change in overall cell density, except in the temporal cortex (Fig. 2).Fig. 2Neurodevelopmental alterations in the cortex of a 23 GW fetus with WARBM. (a) Representative images of Hematoxylin–Eosin-stained coronal brain sections from a WARBM (FWARBM) and a control (FControl) fetus at 23 gestational weeks (GW), showing the different cortical regions of interest. CP: cortical plate; SP: subplate; SVZ/VZ: subventricular zone/ventricular zone; V: ventricle. (b) Higher magnification images of the frontal cortex in FWARBM and FControl, with an arrow indicating the boundary of the CP used for the quantification shown in (c). (c) Frontal, parietal and temporal cortical plate thickness measurements in FWARBM and FControl.. (d) Cell density quantification in frontal, parietal and temporal cortical plate and subplate of FControl and FWARBM. (**p < 0.01, *** p < 0.001, **** p < 0.0001; Mann–Whitney test) Neurodevelopmental alterations in the cortex of a 23 GW fetus with WARBM. (a) Representative images of Hematoxylin–Eosin-stained coronal brain sections from a WARBM (FWARBM) and a control (FControl) fetus at 23 gestational weeks (GW), showing the different cortical regions of interest. CP: cortical plate; SP: subplate; SVZ/VZ: subventricular zone/ventricular zone; V: ventricle. (b) Higher magnification images of the frontal cortex in FWARBM and FControl, with an arrow indicating the boundary of the CP used for the quantification shown in (c). (c) Frontal, parietal and temporal cortical plate thickness measurements in FWARBM and FControl.. (d) Cell density quantification in frontal, parietal and temporal cortical plate and subplate of FControl and FWARBM. (**p < 0.01, *** p < 0.001, **** p < 0.0001; Mann–Whitney test) Normal corticogenesis depends on the production of an adequate number of neural stem/progenitor cells and the elaboration of a large variety of intermediate progenitors that will give rise to cortical neurons (~ 7 GW to 27 GW) and subsequently glial cells (beginning at ~ 16 GW for astrocytes) [26–28]. These events take place in the germinative zones formed by the ventricular (VZ) and subventricular (SVZ) zones. After neurogenesis, newly born neurons migrate through the intermediate zone and subplate (SP) in an inside-out manner to form the layered CP where post-migration neuronal development and organization takes place [28, 29]. To explore further the consequences of RAB3GAP1 loss-of-function in corticogenesis, we examined RAB3GAP1 expression in the frontal lobe. We found that RAB3GAP1 was ubiquitously expressed in the VZ, SVZ, SP, and CP of control fetuses, whereas no staining was detected in FWARBM (Fig. 3). To assess early neurodevelopmental processes, we analyzed SOX2, a transcription factor expressed by neural stem cells and intermediate progenitors within the germinative zones of the developing neocortex [30, 31]. In FWARBM, the germinative zone area was reduced by approximatively 50% and remained significantly decreased (by approximatively 20%) after normalization to the total section surface area (8.9 ± 0.2% for FControl vs 7.1 ± 0.2% for FWARBM; p < 0.05) (Additional File 7: Fig. S3a and b).Fig. 3RAB3GAP1 is expressed in all regions of the Human developing frontal cortex. Representative images of RAB3GAP1 immunostaining in the coronal frontal cortex sections from a control (FControl) and a WARBM (FWARBM) fetus at 23 gestational weeks (GW), showing RAB3GAP1 expression in the cortical plate (a), subplate (b), subventricular zone (c) and ventricular zone (d). In each panel, a higher magnification of the area outlined by a yellow dashed square is shown for both FControl and FWARBM. In FControl, RAB3GAP1 is expressed in all compartments, while in FWARBM, the protein is absent from these regions. V: ventricle RAB3GAP1 is expressed in all regions of the Human developing frontal cortex. Representative images of RAB3GAP1 immunostaining in the coronal frontal cortex sections from a control (FControl) and a WARBM (FWARBM) fetus at 23 gestational weeks (GW), showing RAB3GAP1 expression in the cortical plate (a), subplate (b), subventricular zone (c) and ventricular zone (d). In each panel, a higher magnification of the area outlined by a yellow dashed square is shown for both FControl and FWARBM. In FControl, RAB3GAP1 is expressed in all compartments, while in FWARBM, the protein is absent from these regions. V: ventricle As expected, SOX2-positive (SOX2⁺) cells were localized in the VZ and SVZ in both FWARBM and FControl (Fig. 4a). However, the density of SOX2⁺ cells within the VZ of FWARBM was reduced compared to controls (Fig. 4b), despite no significant change in the overall cell density (Fig. 4c). In contrast, the SOX2⁺ cells density in the SVZ remained comparable between FWARBM and FControl (Fig. 4b), suggesting a preferential disruption of the proliferative VZ. At 23 GW, the prefrontal VZ is predominantly composed of radial glial (RG) progenitor population [32]. To further investigate their organization, we performed immunostaining for glial fibrillary acidic protein (GFAP), an astroglial marker prominently expressed in apical RG (aRG) and basal RG (bRG) neural progenitors at this stage of Human cortical development [26, 33]. In FControl, GFAP immunolabeling revealed a well-organized network of apical RG cells, with cell bodies in the VZ and long apical processes terminating in endfeet at the ventricular surface (Fig. 4d). In contrast, FWARBM exhibited fewer discernible GFAP⁺ cells in the VZ, and their apical processes appeared disorganized and poorly structured, lacking clearly identifiable endfeet at the ventricular wall (Fig. 4d). Morphological abnormalities of GFAP⁺ cells were also found in the SVZ, where basal processes, typically emitted by basal RG cells, were observed in FControl, but not in FWARBM (Fig. 4d). Disruption of basal RG cells morphology was further confirmed by co-immunostaining GFAP with HOPX, a marker of those cells [26] (Fig. 4e). Indeed, in the FWARBM VZ, early generated GFAP+/HOPX+ basal RG cells have no distinguishable apical process, contrary to FControl (Fig. 4e). Similarly, within the FWARBM SVZ the basal processes of basal RG cells observed in FControl were absent (Fig. 4f).Fig. 4Impaired radial glia in WARBM fetal brain. (a) Representative images of SOX2 immunostaining in the germinative zones of control (FControl) and WARBM (FWARBM) fetal brains at 23 GW. (b, c) Quantification of SOX2⁺ cell density (cells/mm2) and total cell density (cells/mm2) in the ventricular zone (VZ) and subventricular zone (SVZ). (d) GFAP immunostaining in the germinative zone; higher-magnification insets highlight altered apical processes (yellow arrows) in FWARBM with less clearly distinguishable endfeet ( white arrowheads). (e) Double immunostaining for GFAP and HOPX in the VZ, showing early basal radial glial cells (bRG, white arrows) in FWARBM lacking the typical apical process seen in controls. (f) In the SVZ, double immunostaining for GFAP and HOPX reveals altered bRG morphology in FWARBM, with a loss of basal processes, indicated by yellow arrowheads, normally observed in FControl. White arrows indicate bRG. (****p < 0.0001; Mann–Whitney test) Impaired radial glia in WARBM fetal brain. (a) Representative images of SOX2 immunostaining in the germinative zones of control (FControl) and WARBM (FWARBM) fetal brains at 23 GW. (b, c) Quantification of SOX2⁺ cell density (cells/mm2) and total cell density (cells/mm2) in the ventricular zone (VZ) and subventricular zone (SVZ). (d) GFAP immunostaining in the germinative zone; higher-magnification insets highlight altered apical processes (yellow arrows) in FWARBM with less clearly distinguishable endfeet ( white arrowheads). (e) Double immunostaining for GFAP and HOPX in the VZ, showing early basal radial glial cells (bRG, white arrows) in FWARBM lacking the typical apical process seen in controls. (f) In the SVZ, double immunostaining for GFAP and HOPX reveals altered bRG morphology in FWARBM, with a loss of basal processes, indicated by yellow arrowheads, normally observed in FControl. White arrows indicate bRG. (****p < 0.0001; Mann–Whitney test) Given the critical role of RG cells scaffolds in supporting neuronal migration, we next assessed the organization of migrating neurons by examining doublecortin (DCX), a marker of immature migrating neurons [3]. In FWARBM, the leading processes of DCX⁺ neurons migrating within the subplate were either markedly shorter compared to those in FControl or absent (Fig. 5a). In the CP, FWARBM DCX⁺ neurons displayed a rounded morphology, contrasting with the elongated shape observed in controls and suggesting an alteration of neuronal polarity (Fig. 5b). Quantification revealed a significant reduction in the number of DCX⁺ neurons in FWARBM compared to FControl in both the CP and SP (Fig. 5c). While in the SVZ of FControl, DCX immunolabeling highlighted organized columns of migrating neurons exiting the SVZ, this pattern was absent in FWARBM, where only rare, scattered DCX⁺ cells were detectable (Additional File 7: Fig. S3c and d).Fig. 5Impaired neuronal migration and reduced cortical CTIP2 neuronal populations in the frontal cortex of FWARBM. (a, b) Doublecortin (DCX) immunostaining in the subplate and cortical plate of FControl and FWARBM. Yellow arrows indicate DCX⁺ cells, and white arrows indicate their leading processes (c) Quantification of the percentage of DCX⁺ cells. (d) CTIP2 immunostaining in the cortical plate of FControl and FWARBM. (e) Quantification of the percentage of CTIP2⁺ cells, shown with and without normalization to cortical plate thickness. (**p < 0.01, *** p < 0.001, **** p < 0.0001; Mann–Whitney test) Impaired neuronal migration and reduced cortical CTIP2 neuronal populations in the frontal cortex of FWARBM. (a, b) Doublecortin (DCX) immunostaining in the subplate and cortical plate of FControl and FWARBM. Yellow arrows indicate DCX⁺ cells, and white arrows indicate their leading processes (c) Quantification of the percentage of DCX⁺ cells. (d) CTIP2 immunostaining in the cortical plate of FControl and FWARBM. (e) Quantification of the percentage of CTIP2⁺ cells, shown with and without normalization to cortical plate thickness. (**p < 0.01, *** p < 0.001, **** p < 0.0001; Mann–Whitney test) We next examined the generation and positioning of deep-layer excitatory neurons by assessing CTIP2, a marker of layer V cortical neurons [34]. CTIP2⁺ cells displayed a similar distribution within the CP of both FWARBM and FControl, with mean distances to the marginal zone of 588.6 ± 48.5 µm in FControl and 513.6 ± 23.8 µm in FWARBM, indicating substantial early cortical lamination is preserved (Fig. 5d, Additional File 7: Fig. S3e and f). These distances were not affected after normalization to CP thickness (Additional File 7: Fig. S3e and f). However, the percentage of CTIP2+ cells within the CP was significantly reduced in FWARBM even after normalization to CP thickness (Fig. 5e). The associated reduction in immature DCX⁺ migrating neurons and CTIP2⁺ cortical neurons together with defects in RG cell density and morphology, suggest that multiple neurodevelopmental processes, including cell proliferation, neurogenesis and migratory scaffolding, may contribute to the corticogenesis defect. Given that cortical plate thinning may involve both proliferative zone disruption and increased apoptosis, we assessed cleaved caspase-3, a reliable marker of cell death, in paraffin-embedded sections [35]. In FControl and FWARBM, cleaved caspase-3 positive cells were rarely detected among postmitotic neurons within the cortical plate and subplate (Additional File 8: Fig. S4a and b). In contrast, apoptotic cells were highly observed within the proliferative zones of FWARBM, particularly in the subventricular zone, with a 6.7 fold increase in caspase-3⁺ cells compared to the control (Additional File 8: Fig. S4c-e). This revealed an increased cell death, that may preferentially affect progenitor populations or immature neurons. Alongside the marked cortical abnormalities, we also investigated the cerebellum, a region frequently impacted in this syndrome [36]. At 23 GW, no gross malformations were observed. However, Purkinje cells appeared less well delineated and discernible in FWARBM compared to controls, although calbindin expression remained similar between groups (Additional File 9: Fig. S5), indicating that these cells were not markedly affected at 23 GW. Altogether these observations reveal that the early development of the cerebral cortex is particularly vulnerable at 23 GW to the disruption of functional RAB3GAP1. Autophagy impairments in fetal tissues affected by WARBM Autophagy is a dynamic and tightly regulated cellular process, comprising four main stages following its induction: membrane nucleation (phagophore formation), membrane elongation with cargo recognition, autophagosome closure, and subsequent fusion with the lysosome to form the autolysosome, where degradation occurs [37]. The autophagosome is a double-membrane vesicle that sequesters cytoplasmic material, including proteins and organelles, targeted for degradation [37]. To investigate potential autophagy dysregulation in FWARBM, we analyzed key molecular markers involved in autophagosome biogenesis. Membrane expansion involves a tightly regulated sequence of events driven by core autophagy-related (ATG) proteins. In particular, two ubiquitin-like conjugation systems are central to this process: ATG3, an E2-like conjugating enzyme, and the ATG5-ATG12-ATG16L1 complex, which acts as an E3-like ligase. These systems mediate the lipidation of LC3-I with phosphatidylethanolamine (PE), generating LC3-II, its membrane-associated form essential for autophagosome maturation [38]. We also assessed p62/SQSTM1, a selective autophagy receptor that binds ubiquitinated cargo and interacts with LC3-II and that is widely used as an indicator of impaired autophagic flux [18, 39, 40]. Strikingly, we observed large p62- and LC3-positive aggregates in the lenses of FWARBM, while these proteins were almost undetectable in FControl (Fig. 6a and b). In the frontal cortex and cerebellum of FWARBM, p62 was undetectable, ATG3 was markedly reduced, and in the cerebral cortex both ATG16L1 and LC3-I accumulated abnormally (Fig. 6c).Fig. 6Autophagy impairments in a WARBM fetus lens and brain. (a). p62 immunohistochemistry in the lens of a WARBM (FWARBM) and a control (FControl) fetus at 23 GW. (b) p62 and LC3 immunostaining in the lens of FWARBM and FControl at 23 GW. (c) Expression of autophagy markers in control and WARBM fetal frontal cortex and cerebellum extracts (cropped gels) Autophagy impairments in a WARBM fetus lens and brain. (a). p62 immunohistochemistry in the lens of a WARBM (FWARBM) and a control (FControl) fetus at 23 GW. (b) p62 and LC3 immunostaining in the lens of FWARBM and FControl at 23 GW. (c) Expression of autophagy markers in control and WARBM fetal frontal cortex and cerebellum extracts (cropped gels) These findings suggest a disruption of the LC3 lipidation step, thereby impairing its insertion to the autophagosome membrane and disrupting autophagosome maturation [38]. Altogether, our data indicate that autophagy is disrupted in both the brain and lens of a 23 GW fetus with WARBM, supporting a role for defective autophagy in the pathophysiology of the disease. Basal and modulated autophagy flux is altered in patient fibroblasts To complement these findings of autophagy disruption in post-TOP extracts analyses, we monitored basal, activated (starvation-induced), and inhibited (bafilomycin) autophagy flux in proband fibroblasts. Proband fibroblasts exhibited no major morphological changes, and RAB3GAP1 was undetectable, as in fetal brain (Additional File 10: Fig. S6). In proband cells, p62+ particles were increased compared with controls under basal conditions. Moreover, the large increase induced by bafilomycin in control cells (⁓622%) was markedly attenuated in proband fibroblasts, reaching only 35% (Fig. 7a and b). This reduced effect of bafilomycin on autophagy flux in proband fibroblasts was confirmed by western blot (Fig. 7c and d). In contrast to control cells, starvation failed to further increase p62+ particles (Fig. 7a and b). This mild response to starvation was confirmed by the lack of an increase in the LC3-II/LC3-I ratio in proband fibroblasts (Fig. 7e and f). ATG3 expression did not differ between control and proband fibroblasts, either under basal conditions or starvation (Fig. 7c-f).Fig. 7Basal and modulated autophagy is altered in WARBM patient fibroblasts. (a, b) p62 immunostaining in proband and control fibroblasts in basal (DMSO), blocked (Bafilomycin, 100 nM) and activated (Starvation) autophagy flux conditions and quantification of the number of p62+ particles (b). (c, d) Expression of autophagy markers in fibroblasts protein extracts in basal (DMSO) condition and after Bafilomycin treatment (cropped gels) and its quantification (d). (e, f) Expression of autophagy markers in fibroblasts protein extracts in basal (DMSO) and starved conditions (cropped gels) and its quantification (f). (*p < 0.05,, ** p < 0.01, **** p < 0.0001; a,b: Mann–Whitney test; d,f: One-Way ANOVA + Bonferroni) Basal and modulated autophagy is altered in WARBM patient fibroblasts. (a, b) p62 immunostaining in proband and control fibroblasts in basal (DMSO), blocked (Bafilomycin, 100 nM) and activated (Starvation) autophagy flux conditions and quantification of the number of p62+ particles (b). (c, d) Expression of autophagy markers in fibroblasts protein extracts in basal (DMSO) condition and after Bafilomycin treatment (cropped gels) and its quantification (d). (e, f) Expression of autophagy markers in fibroblasts protein extracts in basal (DMSO) and starved conditions (cropped gels) and its quantification (f). (*p < 0.05,, ** p < 0.01, **** p < 0.0001; a,b: Mann–Whitney test; d,f: One-Way ANOVA + Bonferroni) Autophagy disruption was confirmed by electron microscopy, showing an accumulation of non-degraded materials free in the cytosol or engulfed in autophagic vacuoles in proband fibroblasts (Fig. 8a). These cells also displayed a high number of lysosomes, and an accumulation of lipid droplets and multilamellar bodies (MLB), whose clearance depends on functional autophagic flux [41, 42] (Fig. 8). Lipid droplet accumulation was confirmed by electron microscopy ultrastructural analysis and BODIPY-based fluorescence intensity measurements (Fig. 8b-d). All these characteristics were never observed in control cells and demonstrated that RAB3GAP1 loss-of-function was associated with an inability to complete cargo degradation in Human fibroblasts.Fig. 8Ultrastructural evidence of impaired clearance and lipid droplet accumulation in WARBM patient fibroblasts. (a) Transmission Electron microscopy images of control and proband fibroblasts. LD, lipid droplets; m, mitochondria; N, nucleus; RER, rough endoplasmic reticulum; black arrows indicate autophagic vacuoles; black arrowheads indicate multilamellar bodies, white stars indicate lysosomes, orange arrows indicate undegraded materials and white arrowheads indicate undegraded materials delimited by membrane portions. (b) Quantification of lipid droplet number per cell and per µm2 based on electron microscopy images. (c) Representative BODIPY-stained images of control and proband fibroblasts. (d) Quantification of BODIPY mean fluorescence intensity in control and proband fibroblasts.(* p < 0.05, ** < 0.01; Mann–Whitney test) Ultrastructural evidence of impaired clearance and lipid droplet accumulation in WARBM patient fibroblasts. (a) Transmission Electron microscopy images of control and proband fibroblasts. LD, lipid droplets; m, mitochondria; N, nucleus; RER, rough endoplasmic reticulum; black arrows indicate autophagic vacuoles; black arrowheads indicate multilamellar bodies, white stars indicate lysosomes, orange arrows indicate undegraded materials and white arrowheads indicate undegraded materials delimited by membrane portions. (b) Quantification of lipid droplet number per cell and per µm2 based on electron microscopy images. (c) Representative BODIPY-stained images of control and proband fibroblasts. (d) Quantification of BODIPY mean fluorescence intensity in control and proband fibroblasts.(* p < 0.05, ** < 0.01; Mann–Whitney test) Discussion Although WARBM was first described in 1993, the mechanisms by which RAB3GAP1 loss-of-function disrupts Human brain development remain poorly defined. Here, we provide the first in-depth neuropathological characterization of a Human fetal case of RAB3GAP1-related WARBM at 23 GW, offering critical insights into the prenatal onset of cortical pathology. Our analysis reveals early alterations in cortical development, preceding macroscopic abnormalities, and highlights the disruption of key cellular processes, including radial glial density and organization, cell polarity, and autophagy along with increased cell death near the ventricle. All of which may individually or in combination contribute to defective cortical development, such as cortical plate thinning and reduced neuronal populations. The availability of both fetal and postnatal material from siblings carrying the same disease-causing variant enabled a unique comparison across developmental timepoints, providing a rare opportunity to trace the temporal emergence of pathological features. Our results suggest that the brain MRI cortical anomalies observed in WARBM children may result from cumulative neurodevelopmental defects, with the third trimester of gestation and early postnatal period further exacerbating cortical abnormalities. Altogether, our findings identify RAB3GAP1 as a critical regulator of Human corticogenesis and cellular homeostasis, advancing our understanding of the early pathophysiological events that contribute to brain malformations and dysfunctions in WARBM and related neurodevelopmental disorders. High sensitivity of the lens to autophagy defects In the lens of the fetus with WARBM, the loss of functional RAB3GAP1 has major early consequences, leading to in utero bilateral cataracts observable at 19 GW [15]. Induction of the Human lens occurs at around 28 days of gestation by the formation of the lens vesicle from the ectoderm that will be filled by fiber cells derived from epithelial cells in the lens germinative zone [43]. During their migration and differentiation, fiber cells lose their organelles and are compacted to create the organelle-free zone of the lens [44]. These processes continue throughout life, and their disruption during development or later in life leads to cataract formation [44]. Similarly to mice lacking Tbc1d20, encoding another WARBM-associated RAB GTPase-activating protein [45], we observed significant alterations in fetal lens fiber cells, which was associated with high accumulation of p62 and LC3, suggesting that RABGAP-associated cataracts result from an autophagy process alteration. This is supported by the role of autophagy in lens organelle removal [44] and its link to cataract observed in autophagy-related gene disorders [46]. Corticogenesis alterations in a fetus with WARBM at 23 GW WARBM patients present severe neurodevelopmental abnormalities affecting both brain morphology and function [3]. However, the precise timing and cellular origins of these defects remain poorly understood. In this study, we show that polymicrogyria, a cortical malformation characteristic of WARBM, is already evident by 3 months of age. In contrast, the 23 GW fetal brain appears structurally intact, its size is reduced but still within the expected range for this developmental stage. The most notable change is a marked reduction in CP thickness. At this developmental stage, CP thickness predominantly reflects the number of newly generated excitatory neurons that have successfully migrated to their appropriate laminar positions, as neuronal maturation, neurite extension, and synaptogenesis are only beginning around 22–24 GW [27]. Thus, the early reduction in CP thickness likely reflects fewer neurons reaching the cortical plate, as no increase in cell death was observed within the CP itself. Our findings reveal a marked reduction in SOX2⁺ progenitor cells within the VZ of the FWARBM brain, consistent with impaired generation and/or maintenance of the apical RG progenitor pool. Additionally, GFAP immunostaining further showed that the radial glia scaffold in the germinative zone was severely disorganized and morphologically altered, reflecting disrupted progenitor architecture and polarity. Elevated cell death within the germinative zone, along with these defects, likely reduces neuronal numbers and disrupts corticogenesis, ultimately leading to cortical plate thinning and the later emergence of cortical malformations [47]. In line with this, we observed a significant decrease in the number of immature DCX⁺ migrating neurons in the SP and CP, many of which displayed abnormal rounded morphologies and loss of leading processes, consistent with impaired neuronal polarity and migration. These alterations were accompanied by a substantial reduction in CTIP2⁺ deep-layer cortical neurons, which may result from a combination of reduced neuronal production, migration failure, and/or increased vulnerability to apoptosis in the FWARBM cortex. Whether these apoptotic cells correspond to newly generated neurons that fail to reach their proper cortical destinations, defective neuronal progenitors, or both remains to be determined. Given the critical role of CTIP2 in corticospinal motor neuron development [48], its reduced expression may underlie, at least in part, the motor deficits observed in WARBM patients [2, 3]. Interestingly, these profound cortical abnormalities contrast with the cerebellum, where Purkinje cells appear morphologically intact at this developmental stage. This suggests a region- and/or time-specific vulnerability to RAB3GAP1 dysfunction during brain development. Since most of the WARBM patients suffer from cerebellar hypoplasia [36], it is plausible that cerebellar pathology emerges later in development or may primarily affect granule cells population. This is supported by the protracted timeline of cerebellar neurogenesis, with the majority of granule cells generated after birth [28]. Whether this temporal specificity reflects differential expression dynamics of RAB3GAP1, distinct molecular interactors in cerebellar versus cortical cells, or both, remains an important area for future investigation. Taken together, our findings indicate that RAB3GAP1 loss-of-function disrupt corticogenesis in the Human fetal brain, with progenitor depletion, RG disorganization, reduced number of neurons, and increased cell death. While based on a single developmental time point (23 GW), these early defects likely contribute to postnatal cortical atrophy and malformations in WARBM. Autophagy & cortex developmental alterations in WARBM Despite the presence of defective autophagy in FWARBM, no macroscopic abnormalities are observed at 23 GW, aligning with prior evidence that pathogenic variants in autophagy-related genes exert limited effects on early brain morphology [49]. However, autophagy plays a central and pleiotropic role during neurodevelopment, controlling multiple steps of central nervous system development, including stem and progenitor cell proliferation, neuronal migration, differentiation and maturation [38]. This explains why variants in autophagy-related genes are associated with neurodevelopmental disorders [38, 50]. Defective autophagy may compromise cellular quality control and the renewal of cell content, triggering apoptosis and likely contributing to the increased cell death observed in the germinative zones of FWARBM, a mechanism also described in aging-related brain diseases [51, 52]. Such early disruptions in cellular homeostasis can be particularly critical around birth, when autophagy becomes increasingly important to cope with environmental and nutritional changes [3]. Consequently, constitutive defects in autophagy may exacerbate neurodevelopmental abnormalities postnatally and contribute to the dramatic differences in cortical development observed between 23 GW and 3 months. In this context, RAB3GAP1 deficiency may further impair post-migratory cortical organization, such as neuronal maturation, synaptogenesis, synaptic and neuronal activity [53]. This is further supported by recent findings showing that RAB3GAP1 downregulation disrupts neuronal maturation and neurite outgrowth in Human iPSC-derived cortical neurons [22]. Given its established role in synaptic homeostasis [10] and autophagy, RAB3GAP1 pathogenic variants could significantly impact the formation and refinement of neural circuits during early postnatal life. We observed that RAB3GAP1 deficiency also led to morphological abnormalities in polarized cells of the fetal frontal cortex. Prior in vitro studies have shown that RAB3GAP1 regulates intracellular trafficking [22, 54], a process essential for oriented transport in polarized cells, which underlies proper brain development [55, 56] and supports efficient autophagic flux [57, 58]. Structural analyses suggest that RAB3GAP1 shares similarities with Zwilch, a component of the outer kinetochore that recruits dynein, a motor protein important for retrograde transport and which contributes to neuronal migration and cell division [5, 59]. These findings in the human fetal brain support previous evidence, confirming the tight interconnection between intracellular trafficking and autophagy and their joint importance for neurodevelopment and brain function [38, 50, 60]. Clarifying the respective contributions of these processes to the clinical manifestations of WARBM will be essential for a better understanding of the disease’s pathophysiology. Importantly, WARBM appears to represent a valuable model for investigating the interplay between intracellular trafficking and autophagy, and their roles in developmental brain abnormalities. Molecular clues linking RAB3GAP1 to autophagy In the WARBM fetal brain, our results suggest that LC3-I lipidation is altered and support a role for RAB3GAP1 in autophagosome biogenesis [21]. In contrast, in the fetal lens and proband fibroblasts, no evidence supported a defect in autophagy initiation, but a defect in autophagosome degradation. The reduced activation of autophagy by starvation suggests a complex alteration in autophagic flux during stress-induced autophagy, as reported in RAB3GAP1 KO HeLa cells [61]. Interestingly, p62 accumulated in the fetal lens and proband fibroblasts but was almost undetectable in the brain, supporting previous thoughts that autophagy defects are tissue-dependent [3, 16]. Subcellular ultrastructural analysis in RAB3GAP1-mutated Human neurons will be necessary to assess the exact consequences of RAB3GAP1 pathogenic variants on autophagic flux in post-mitotic cells, as well as the possible tissue specificity, which may depend on tissue-specific molecular partners. RAB3GAP1 was initially described as a modulator of synaptic release through RAB3 activation [10, 62], which in turn can act on autophagy [63]. In Human fibroblasts, the functions of RAB3GAP1 in autophagy are RAB3-independent [21] but appear to rely on the upstream regulation of RAB18 [20]. By contrast, in HeLa cells, RAB3GAP1 contributes to autophagosome formation through RAB3A modulation [64]. These observations underscore the context-dependent nature of autophagy regulation, which can differ both by cell type and by the method used to modulate autophagic flux. Interestingly, RAB3GAP1 and RAB18 are involved in lipid droplet catabolism adjusting the availability of lipid-derived metabolites [61], a suggested lipid source for autophagosome formation [65], more important under nutrient-deprived conditions [61]. In this context, the accumulation of lipid droplets observed in patient-derived fibroblasts suggests an impairment in lipid mobilization, potentially explaining their inability to upregulate autophagy in response to starvation, when cellular lipid demand is heightened. In the C. elegans model of Vici syndrome (EPG5 variants), loss-of-function of rgb-1 (RAB3GAP1 in Humans) activates lysosomal biogenesis and function through RAB7 modulation, which restores autophagy [17]. Such a mechanism seems unlikely in Human brain where RAB3GAP1 dysfunction leads to a defect in autophagosome biogenesis. In Human cells, proteomic analysis suggested RAB3GAP1 as a potential interactor of the ATG8 family [66]. Identifying the specific molecular partners of RAB3GAP1 in autophagy will be a central point of investigation for understanding the molecular pathophysiology of WARBM, and may provide insights for future therapeutic exploration. Supplementary Information Below is the link to the electronic supplementary material. Additional file 1: Table S1. Clinical, neurological, ophthalmic and brain imaging findings of WARBM patients and fetus. Additional file 1: Table S1. Clinical, neurological, ophthalmic and brain imaging findings of WARBM patients and fetus. Additional file 2: Table S2. Gestational age and cause of death of selected control cases. IHC, Immunohistochemistry; IUFD, in utero fetal death; TOP, medical termination of pregnancy; GW, Gestational weeks; F, female; M, male; WB, Western blot. Additional file 2: Table S2. Gestational age and cause of death of selected control cases. IHC, Immunohistochemistry; IUFD, in utero fetal death; TOP, medical termination of pregnancy; GW, Gestational weeks; F, female; M, male; WB, Western blot. Additional file 3: Table S3. List of primary antibodies used for western blot (WB), immunocytochemistry (ICC), and immunohistochemistry (IHC). Additional file 3: Table S3. List of primary antibodies used for western blot (WB), immunocytochemistry (ICC), and immunohistochemistry (IHC). Additional file 4: Fig. S1. Cell counting method using the QuPath-0.5.0 software. (a) Images of a hemi-frontal cortical section showing the regions of interest (red rectangles) in the cortical plate and subplate for cell density quantification. Red lines correspond to the region of cortical plate that has been used to measure its thickness. (b) Representative images of a hemi-frontal cortical section showing the regions of interest in the cortical plate and subplate (yellow rectangles) used to quantify the percentage of DCX+ cells. CP: cortical plate; SP: subplate. Additional file 4: Fig. S1. Cell counting method using the QuPath-0.5.0 software. (a) Images of a hemi-frontal cortical section showing the regions of interest (red rectangles) in the cortical plate and subplate for cell density quantification. Red lines correspond to the region of cortical plate that has been used to measure its thickness. (b) Representative images of a hemi-frontal cortical section showing the regions of interest in the cortical plate and subplate (yellow rectangles) used to quantify the percentage of DCX+ cells. CP: cortical plate; SP: subplate. Additional file 5: Table S4. Autopsy analysis of 23 GW fetus with WARBM. Additional file 5: Table S4. Autopsy analysis of 23 GW fetus with WARBM. Additional file 6: Fig. S2. Macroscopic description of the WARBM fetal brain. (a, b) Left and right lateral view of the brain after removal of the meninges. (c) Ventral view of the brain. Additional file 6: Fig. S2. Macroscopic description of the WARBM fetal brain. (a, b) Left and right lateral view of the brain after removal of the meninges. (c) Ventral view of the brain. Additional file 7: Fig. S3. Neuropathological analysis of WARBM fetal brain at 23 GW. (a) Representative DAPI-stained hemi-frontal cortical sections from control and WARBM fetuses. The dashed line indicates the delineated subventricular zone (SVZ) used for quantification. (b) Quantification of the SVZ area normalized to the total section area. (c) Schematic representation of a hemi-frontal cortical section including the cortical plate (CP), subplate (SP), subventricular/ventricular zone (SVZ/VZ), and ventricle (V). The green dashed square indicates the region selected for staining analysis in the germinative zone. (d) DCX immunostaining in the VZ and SVZ of control and WARBM fetuses. Yellow arrows show DCX+ cells joining the radial migration pathway. (e) CTIP2 immunostaining in the frontal cortex of control and WARBM fetuses. Dash line delimitated the CTIP2-positive cells layer. (f) Quantification of the mean distance from the CTIP2-positive cell layer to the molecular layer, shown both as raw values and normalized to cortical plate thickness. Additional file 7: Fig. S3. Neuropathological analysis of WARBM fetal brain at 23 GW. (a) Representative DAPI-stained hemi-frontal cortical sections from control and WARBM fetuses. The dashed line indicates the delineated subventricular zone (SVZ) used for quantification. (b) Quantification of the SVZ area normalized to the total section area. (c) Schematic representation of a hemi-frontal cortical section including the cortical plate (CP), subplate (SP), subventricular/ventricular zone (SVZ/VZ), and ventricle (V). The green dashed square indicates the region selected for staining analysis in the germinative zone. (d) DCX immunostaining in the VZ and SVZ of control and WARBM fetuses. Yellow arrows show DCX+ cells joining the radial migration pathway. (e) CTIP2 immunostaining in the frontal cortex of control and WARBM fetuses. Dash line delimitated the CTIP2-positive cells layer. (f) Quantification of the mean distance from the CTIP2-positive cell layer to the molecular layer, shown both as raw values and normalized to cortical plate thickness. Additional file 8: Fig. S4. Increase in cell death vulnerability in the germinative zone of FWARBM. Representative images of cleaved caspase-3 immunostaining in the coronal frontal cortex sections from a control (FControl) and a WARBM (FWARBM) fetus at 23 gestational weeks (GW), showing cleaved caspase-3 expression in the cortical plate (a), subplate (b), subventricular zone (c) and ventricular zone (d). In each panel, a higher magnification of the area outlined by a yellow dashed square is shown below for both FControl and FWARBM. (e) Quantification of total cell density (cells/mm²), caspase-3⁺ cell density (cells/mm²), and caspase-3⁺ cells as a percentage of total DAPI⁺ cells in the subventricular zone (SVZ). Additional file 8: Fig. S4. Increase in cell death vulnerability in the germinative zone of FWARBM. Representative images of cleaved caspase-3 immunostaining in the coronal frontal cortex sections from a control (FControl) and a WARBM (FWARBM) fetus at 23 gestational weeks (GW), showing cleaved caspase-3 expression in the cortical plate (a), subplate (b), subventricular zone (c) and ventricular zone (d). In each panel, a higher magnification of the area outlined by a yellow dashed square is shown below for both FControl and FWARBM. (e) Quantification of total cell density (cells/mm²), caspase-3⁺ cell density (cells/mm²), and caspase-3⁺ cells as a percentage of total DAPI⁺ cells in the subventricular zone (SVZ). Additional file 9: Fig. S5. Neuropathological analysis of WARBM fetal cerebellum. (a) Hematoxylin-Eosin-stained sagittal cerebellum sections of WARBM and control fetuses at 23 GW showing a disorganization of the Purkinje cell layer. EGL: External granular layer; IGL: Internal granular layer; ML: Molecular layer; PCL: Purkinje cell layer. (b) Calbindin expression in 23 GW control and WARBM fetus cerebellum extracts. Additional file 9: Fig. S5. Neuropathological analysis of WARBM fetal cerebellum. (a) Hematoxylin-Eosin-stained sagittal cerebellum sections of WARBM and control fetuses at 23 GW showing a disorganization of the Purkinje cell layer. EGL: External granular layer; IGL: Internal granular layer; ML: Molecular layer; PCL: Purkinje cell layer. (b) Calbindin expression in 23 GW control and WARBM fetus cerebellum extracts. Additional file 10: Fig. S6. Loss of functional RAB3GAP1 does not affect global cell morphology despite autophagy alteration. (a) RAB3GAP1 immunostaining in proband and control fibroblasts, actin was stained with ActinRed-555 to show the overall cell morphology. (b) RAB3GAP1 expression in proband and control fibroblasts. Additional file 10: Fig. S6. Loss of functional RAB3GAP1 does not affect global cell morphology despite autophagy alteration. (a) RAB3GAP1 immunostaining in proband and control fibroblasts, actin was stained with ActinRed-555 to show the overall cell morphology. (b) RAB3GAP1 expression in proband and control fibroblasts.