Neuronal VPS13A depletion links diacylglycerol PKC signaling and synaptic spines Authors: Gisela Besa-Selva, Esther García-García, Alba Ramón-Lainez, Georgia Escaramis, Pol Garcia-Segura, Cristina Malagelada, Eulàlia Martí, Jordi Alberch, Mercè Masana, Manuel J Rodríguez Published: 2026 ABSTRACT AbstractBackgroundIntermembrane lipid transfer protein VPS13A (VPS13A) is a large protein whose cellular functions are still under investigation. VPS13A gene mutations leading to the absence of protein expression cause Chorea-acanthocytosis (ChAc), an ultra-rare inherited neurodegenerative movement disorder. Although the molecular mechanisms linking VPS13A loss to neuronal dysfunction remain unclear, its role as a bulk lipid transfer protein suggests that impaired lipid distribution may represent a primary pathogenic mechanism leading to neurodegeneration in ChAc. In this study, we investigated the effect of neuronal silencing of VPS13A in a murine model to phenocopy the human disease.ResultsLipidomics analysis revealed an increase in the concentration of several diacylglycerol species induced by VPS13A knockdown. We then explored the downstream molecular pathways related to the altered diacylglycerol levels and found that VPS13A knockdown induces a decrease in protein kinase C (PKC)βII concentration but an increase in PKCα/βII phosphorylation in cultured neurons. Finally, pharmacological inhibition of PKCβII reverted aberrant neuronal morphology and loss of spine density induced by VPS13A KD. These results underscore the importance of VPS13A in regulating neuronal lipid distribution, showing that its absence perturbs the diacylglycerol/PKC signaling pathway, with measurable effects on neuronal structure and synaptic density.ConclusionsOverall, our results underscore a previously underappreciated role for VPS13A in the structural organization of neurons through lipid-mediated signaling mechanisms, providing insight into the cellular dysfunction underlying ChAc.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12915-026-02649-w.Keywords: VPS13A disease, Basal ganglia disorders, Lipidomics, DAG, Glycerophospholipids, Synapse Background Intermembrane lipid transfer protein VPS13A (VPS13A) is a large, poorly understood protein whose cellular functions are still under investigation. Based on studies of its yeast homolog VPS13, cell models, and in silico studies, VPS13A has been proposed to mediate phospholipid transfer between membranes at organelle contact sites, functioning as a bulk lipid transporter [1–4]. In cellular models, VPS13A is necessary for forming or stabilizing endoplasmic reticulum-mitochondria contact sites and influences mitochondrial morphology and lipid droplet motility [5]. Additionally, VPS13A has been implicated in other membrane-associated processes, such as endocytic trafficking and lysosomal degradation [6], autophagy [7], and mitochondrial calcium homeostasis mechanisms [8]. Despite these findings, the exact role of VPS13A in maintaining cellular homeostasis is still emerging. Loss-of-function mutations in the gene encoding for VPS13A cause chorea-acanthocytosis (ChAc; VPS13A disease; MIM 200150), a rare and complex neurodegenerative disorder with autosomal recessive inheritance [9, 10]. ChAc is characterized by a range of basal ganglia-related symptoms, including chorea, dystonia, and psychiatric disturbances, which typically present in adulthood [11]. The hallmark neuropathological feature of ChAc is the progressive degeneration of the putamen and caudate nucleus due to massive neuronal loss [12]. Additionally, some patients exhibit morphologically altered erythrocytes known as acanthocytes [13]. However, the molecular mechanisms by which loss of VPS13A leads to neuronal dysfunction and degeneration remain poorly understood [14]. Evidence points to a potential role of VPS13A in synaptic activity and plasticity. Studies using human induced pluripotent stem cell (hiPSC)-derived striatal medium spiny neurons (MSNs) from ChAc patients have demonstrated increased synaptic activity linked to aberrant activation of the Lyn kinase pathway [15, 16]. Moreover, we have shown that a mouse model with cortico-striatal VPS13A knockdown (KD) exhibited altered locomotor behavior and impaired striatal long-term depression, highlighting the involvement of VPS13A in basal ganglia neuronal connectivity and synaptic plasticity [17]. Together, these findings suggest that disruption of VPS13A function may affect synaptic signaling pathways critical for basal ganglia function. Recent research has established a relationship between basal-ganglia neurodegeneration and aberrant tissue lipid composition [18, 19]. VPS13A functions as a bulk lipid transporter in this framework, suggesting that neuronal lipid dyshomeostasis may be a key pathological process in ChAc [20]. Indeed, disruptions in lipid metabolism in ChAc are increasingly recognized. Recent studies have identified elevated levels of certain sphingolipid and phospholipid species in the putamen [21] and red blood cells [22] of ChAc patients, pointing to lipid dysregulation as a key feature of the disease. However, the downstream consequences of these lipid alterations, particularly in neuronal function, remain unexplored. This study investigated whether VPS13A KD affects brain lipid composition. We conducted a hypothesis-free lipidomic approach, starting with an unbiased analysis to identify changes in the mouse brain profiles of glycerophospholipids, sphingophospholipids, and their precursors and derivatives. This revealed an increase in diacylglycerol (DAG) induced by VPS13A KD. Then, we performed a subsequent bioinformatics analysis of metabolic pathways and specific experimental validations focused on neuronal DAG metabolism. We finally explored whether the downstream DAG / protein kinase C (PKC, EC 2.7.11.13) signaling pathway was altered by VPS13A KD and its contribution to abnormal neuronal morphology and loss of synaptic spines in ChAc. Results Neuronal VPS13A knockdown increases diacylglycerol and sphingosine-1-phosphate levels We previously reported that the infection with adeno-associated virus (AAV)-VPS13A-KD in 8-week-old mice (Fig. 1A, B) reduces 28% of the VPS13A levels in the mouse motor cortex and striatum, when compared to AAV-Ctrl 30 days after the infection [17]. This VPS13A-KD disrupts the locomotor behavior pattern and impairs cortico-striatal synaptic plasticity, with a role of BDNF signaling [17]. Using this model of VPS13A neuronal silencing that partially mimics ChAc, we performed a lipidomics analysis (Fig. 1B). The brain-tissue-extracted lipid samples were analyzed blinded, using the LC-HRMS methodology to quantify glycerophospholipids, sphingolipids, and their precursors and derivatives. We identified 173 lipid species belonging to 17 different lipid subclasses (Fig. 1C). The heatmap of standardized lipid levels clustered by the Pearson’s distances revealed high variability of most lipid species and few differences between controls and VPS13A-KD samples. However, some lipids from the DAG subclass were increased in the VPS13A-KD group (Fig. 1C). Also, we found no significant effect of sex in the general dataset (Fig. 1C). Experimental procedure and differential expression analysis of the lipidomic study. A Schematic representation of motor cortex and striatum brain regions (highlighted in orange in the drawing of a brain section), arrows indicate the sites of injection with control (Ctrl) and VPS13A knockdown (KD) AAV constructs for the lipidomic analysis. Scale bar 1 mm. B General procedure followed for brain VPS13A KD and sample processing for lipidomic analysis. The viral constructs contain the synapsin-1 promoter (SYN1) to drive neuronal expression of ZsGreen, along with either two control silencing RNAs (shCtrl 1 and shCtrl 2) or two VPS13A-targeting silencing RNAs (shVPS13A 1 and shVPS13A 2). C Heatmaps depicting standardized lipid levels in the samples of control (Ctrl; n = 6) and VPS13A KD (KD; n = 6). X-axes: samples from two sexes (three females and three males for each group). Y-axes: 173 lipid species of the 17 lipid groups. D Volcano plot of all lipid species (left) and magnified detail (right) showing lipid species with significantly increased concentration in VPS13A-KD samples (significance level, p < 0.00454; Bonferroni test corrected by the ENT value) To infer differences between VPS13A-KD and control samples in the lipid concentrations, a linear regression model was generated using the VPS13A expression group and sex as predictor variables over the level of every lipid species (Additional File 1: Suppl Info 1). In this model, of the 173 lipid species analyzed, we identified nine whose concentrations differed between VPS13A-KD and control samples (Fig. 1D). Eight of these lipid species belonged to the DAG subclass, while the ninth was sphingosine-1-phosphate (SP1). The list of these lipids with their detailed statistics is included in Table 1. Table 1Lipids enriched in VPS13A-KD according to the linear regression modelLipid classLipid specieVPS13A-KD effectR-squarep-valueDAGDAG (36:1)1.62950.73410.0008DAGDAG (36:2)1.57400.67600.0019DAGDAG (38:2)1.55060.65590.0025SphDerS1P1.52020.65820.0028DAGDAG (34:1)1.51380.62550.0038DAGDAG (38:1)1.51290.62430.0038DAGDAG (32:1)1.48540.62580.0042DAGDAG (40:2)1.49720.61770.0043DAGDAG (34:2)1.49560.61550.0044SphDer sphingosine derivatives, S1P sphingosine-1-phosphate Lipids enriched in VPS13A-KD according to the linear regression model SphDer sphingosine derivatives, S1P sphingosine-1-phosphate We then validated this result by performing a discriminant analysis. We first inferred differences between female and male samples in the lipid concentrations by fitting a second linear regression model to the standardized lipid levels with sex as a predictor variable (Additional File 1: Suppl Info 2). This model only presented a significant sex effect for Cer (d18:1/14:0) (r2 = 0.7206 p-value = 0.0005). We then performed a partial least squares differential analysis (PLS-DA) with the residuals of that second regression model to assess the effect of VPS13A-KD in lipid concentrations with no contaminant effects. The generated PLS-DA model split data into two different clusters, as evidenced in the sample plot that includes 95% confidence ellipses (Fig. 2A). Classification of the lipid profile expression of every sample in the prediction background plot showed a separation of VPS13A-KD and controls, with none of the samples misclassified (Fig. 2B). To estimate the risk of overfitting and evaluate the performance of the generated model, we calculated the optimal number of PLS-DA components and the area under the curve (AUC) scores, which estimates the specificity and sensitivity of the model, from training cross-validation with 150 repetitions. The analysis reached three as the optimal number of components in our model, with a low balanced error rate (BER = 7.77%) and high specificity and sensitivity (AUC = 97.54%). Finally, the accuracy of the whole model was validated through a permutation system with 1000 iterations, which provided a p = 0.006 as the probability that the model occurs by chance. Thus, the model generated denotes a significant difference in the lipid species concentration between VPS13A-KD and control samples. Partial-least square discriminant analysis (PLS-DA) of the detected lipid species. A PLS-DA sample plot of the lipid quantification data with the 95% confidence ellipses of control (Ctrl) and VPS13A-KD (KD) groups. B PLS-DA background prediction plot showing a separation between Ctrl and KD groups as an estimation of how novel samples would be classified according to the generated model. C Rate of the contribution of the 18 lipids with a major contribution (VIP values > = 1.4) to the generated PLS-DA model. D Sample distribution and density regions in the first principal components (PC) of the PCA, based on the 18 lipids selected from the PLS-DA analysis with VIP values ≥ 1.4. E Heatmaps and pairwise comparison of the 18 lipids with major contributions to the generated PLS-DA model. FCh, fold change. Ctrl, control; KD, VPS13A KD; n = 6 Ctrl (3 males + 3 females) and 6 KD (3 males + 3 females). F Bar chart of the VPS13A-KD effect in the total DAG levels. Ctrl, control; KD, VPS13A KD; n = 6 Ctrl (3 males + 3 females) and 6 KD (3 males + 3 females); **p < 0.01, Student’s t-test To estimate the contribution of every lipid species to the differences found in our PLS-DA model, we then calculated their variable importance in projection (VIP) scores. The 18 lipid species with VIP scores ≥ 1.4 were considered significant. Fourteen significant lipid species belonged to the DAG subclass (Fig. 2C) and included those DAG species identified as significant in the linear regression model. A final unsupervised principal component analysis (PCA) with the standardized values of these 18 lipid species split the samples into two groups, according to the experimental conditions (VPS13A-KD vs Control; F(1,10) = 20.656; R2 = 0.6738; p = 0.004) (Fig. 2D). Pairwise comparisons of the significant lipids showed a VPS13A-KD-induced increase in the expression of the fourteen DAG lipid species and S1P, while the expression PG (34:1), CDH (d18:1/22:0), and PE (38:0) was decreased in these samples (Fig. 2E). To determine whether the VPS13A-KD-induced changes in DAG species result in overall alterations in DAG levels, we calculated total DAG as the sum of all quantified DAG species. VPS13A-KD induced a significant 21% increase in the total DAG level in VPS13A-KD brain samples (p = 0.0039) (Fig. 2F). These results indicate that the imbalance of intracellular lipid transport in neurons by VPS13A KD specifically deregulates the cerebral levels of DAG. Diacylglycerol species enriched in VPS13A-KD samples are structurally related We analyzed which lipid properties were overrepresented in the dysregulated lipid subset in VPS13A-KD samples using the LION/web LION platform. Five LION terms were significantly enriched in VPS13A-KD samples compared with controls. These terms were “glycerolipids,” “diacylglycerols,” “headgroup with neutral charge,” “negative intrinsic curvature,” and “lipid-mediated signaling” (Fig. 3A, B). Therefore, to further investigate the meaning of this result, we analyzed the structural characteristics of the enriched lipids. One hundred seventy-three lipid species detected in our LC-HRMS analysis provided a dataset of 1181 molecular structures collected from the LIPID MAPS® Structure Database. t-Distributed Stochastic Neighbor Embedding (t-SNE) representation of the Morgan fingerprints of the molecular structures followed by the DBSCAN clustering analysis of these structures split them into 22 different clusters. Within the cluster where DAG subclass species aggregated, we found a subcluster with most DAG species enriched in VPS13A-KD samples (Fig. 3C, D, Additional File 1: Suppl Info 3). This analysis revealed that the DAG species increased in VPS13A-KD tissue share highly similar structural features. Lipid enrichment and structure analysis of significant lipid species. A LION network plot with the significant pathways that enriched considering the 18 lipids selected in the PLS-DA analysis. B Histogram with the estimated statistical significance of the enriched terms in the LION network. FDR, false discovery rate. C t-SNE plot with the DBSCAN clustering analysis of the 1181 molecular structures from all 173 lipid species detected in the lipidomic analysis. Dots highlighted in green correspond to the structures of the 18 lipids selected in the PLS-DA analysis. The subcluster inside the red square is amplified in D Enzymes involved in the diacylglycerol and sphingosine-1-phosphate metabolism are not deregulated We then performed a lipid pathway analysis using BioPAN, a bioinformatics tool that integrates quantitative lipidomics data to infer dysregulated enzymatic reactions and lipid metabolic pathways based on changes in substrate–product relationships. Using our dataset comprising the quantification of 173 lipid species, we aimed to predict the enzymatic reactions and lipid metabolism pathways that may be altered by neuronal VPS13A KD and underlie the observed changes in DAG and S1P. Although the analysis included 173 quantified lipid species, the BioPAN lipid network revealed only two main lipid pathways altered by VPS13A KD (Fig. 4A): the balance between DAG and phosphatidylcholine synthesis, and between S1P and sphingosine synthesis, suggesting a dysregulation of interconversion reactions within these lipid axes. No specific enzymes have been described yet mediating the direct synthesis of DAG from PC. But BioPAN also identified the synthesis of different species of PC and PE from DAG as the most suppressed reaction chains in VPS13A-KD samples when compared to controls (Z-scores = 2.036 and 1.948, respectively). The main enzymes in these reactions are choline/ethanolamine phosphotransferase 1 (Cept1, EC: 2.7.8.1) and choline phosphotransferase 1 (Chpt1, EC: 2.7.8.2) (Fig. 4B, Additional File 1: Suppl Info 4). The enzymes involved in the turnover pathway between S1P and sphingosine that VPS13A KD may alter were sphingosine kinase (Sphk, EC 2.7.1.91) 1 and 2, which mediate the synthesis of S1P from sphingosine; and sphingomyelin synthase 2 (Sgms2, EC 2.7.8.27), which synthesizes SM by transferring phosphocholine from PC onto ceramide, rendering DAG as a side product (Fig. 4B). Lipid metabolism pathway analysis. A Lipid networks generated from BioPAN with all the lipidomic data and comparing Ctrl (n = 3 females + 3 males) and VPS13A-KD (n = 3 females + 3 males) groups. Values are presented as the Z-score of the given reaction, whereby a Z-score > 0 represents an active reaction (green), and a Z-score < 0 represents a suppressed reaction (red). Reactions with a Z-score > 1.645 or < -1.645 are considered significant. B Table with the main enzymes catalyzing the significant reactions from the BioPAN network analysis VPS13A-KD does not modify the levels of DAG-related enzymes in neuronal cultures As VPS13A in the mouse brain is mainly expressed in neurons [23], our AAV construct is designed to KD the VPS13A levels specifically in this cell type. The infection with AAV-VPS13A-KD of cultured cortical neurons at 3 days in vitro (DIV) reduced by 82% the levels of VPS13A at DIV 15 when compared with AAV-Ctrl infection detected by Western blot (Fig. 5A), as previously published [17] and by immunofluorescence (Fig. 5B). Using this model, we first assessed the effect of VPS13A-KD on neuronal DAG concentration by incubation with a Red Downward DAG fluorescent sensor [24]. VPS13A-KD induced a significant 37% increase in the total DAG level in cultured neurons when compared with controls (Fig. 5C). We then experimentally analyzed whether the DAG increase affects the expression of any of the enzymes predicted by BioPAN and assessed the levels of Chpt1, Cept1, Sgms2, and Sphk2 (the main Sphk isoform in neurons) in primary cultured neurons by Western blot. We found similar Chpt1, Cept1, Sgms2, and Sphk2 protein levels between control and VPS13A-KD neuronal cultures (Fig. 5D). Then, because the activation of the gamma isoform of phosphatidylinositol phospholipase (PLCγ1, EC 3.1.4.11) increases neuronal DAG levels in response to BDNF [25], and because VPS13A KD alters neuronal BDNF levels [17], we also examined PLCγ1 and phosphorylated PLCγ1 (pPLCγ1) levels in primary cultured neurons by Western blot. We found no differences in PLCγ1 or pPLCγ1 protein levels between control and VPS13A-KD neuronal cultures (Fig. 5D). This result suggests that VPS13A KD does not deregulate the concentration of those enzymes. Effect of VPS13A KD on DAG levels and lipid metabolism enzymes in cultured neurons. A Representative images and quantification of VPS13A levels in Ctrl and VPS13A KD neurons by Western blot (n = 3 embryos/condition). B Representative confocal images of Ctrl and VPS13A KD neurons showing expression of the infection vector (ZsGreen, green) and VPS13A labeling (red); the bar chart shows the quantification of VPS13A-labeling intensity with respect to Ctrl. n = 6 embryos from two independent cultures. **p < 0.01, analyzed by Wilcoxon rank test. C Drawing of the DAG-sensor structure. DAG binding to the C1 domain induces a conformational change that is transmitted to the circularly permuted RFP (cpRFP), perturbing its structure and decreasing fluorescence. Representative confocal images of Ctrl and VPS13A KD neurons showing DAG sensor labeling. Red fluorescence intensity reflects DAG levels inversely, with higher intensity indicating lower DAG. The bar chart shows the quantification of inverted DAG sensor fluorescence intensity with respect to Ctrl, from 6 embryos from two independent cultures; **p < 0.01, analyzed by Wilcoxon rank test. D Concentrations of Chpt1, Cept1, Sgms2, Sphk2, PLCγ1, and pPLCγ1 in Ctrl and KD neuronal cultures, quantified by Western blot. Data represent mean ± SEM. Differences were analyzed by Student’s t-test. n = 6–8 embryos from two independent cultures. Some data points were excluded after outlier analysis VPS13A-KD disrupts protein kinase C signaling and is linked to abnormal neuronal morphology One of the LION terms significantly enriched in VPS13A-KD samples compared with controls was “Lipid-mediated signaling” (Fig. 3A, B). Following this, DAG acts as a second messenger and activates different isoforms of calcium-dependent protein kinase C (PKC, EC 2.7.11.13) [26, 27]. These kinases are involved in synaptic plasticity [28–30], dendritic tree and spine complexity [31], and have also been implicated in neurodegeneration-related synaptic dysfunction and cognitive decline (see Singh et al., [32] for a review). Because of that, we next assessed the effect of neuronal VPS13A-KD on PKC levels and activation in cultured cortical neurons. We first analyzed the DAG-dependent PKCα and PKCβII levels in primary cortical neurons by Western blot. We found a tendency to decrease PKCα protein levels and a significant 25.8% decrease in the levels of PKCβII (t = 2.214, p = 0.0256; Student’s-t test) induced by VPS13A KD (Fig. 6A, B). We then analyzed pPKCα/βII levels by Western blot as a measure of protein activation (Fig. 6C). We observed a 30.8% increase in pPKCα/βII concentration in cultured VPS13A-KD neurons compared with controls (t = 1.791, p = 0.0492; Student’s-t test). By contrast, when we analyzed the PKCδ isoform, we found similar protein levels between the control and VPS13A-KD groups (Fig. 6D). This result suggests that VPS13A KD leads to a deregulation of conventional PKC activation in cultured neurons. Effect of VPS13A-KD in PKC signaling, neuronal morphology and spine density in primary cortical neurons. A–D Quantification and illustrative immunoblots of PKCα (A), PKCβII (B), pPKCα/βII (C), and PKCδ (D). *p < 0.05, Student’s t-test. E Representative images of neuron morphology and spine density in control (Ctrl) and VPS13A-KD (KD) cortical culture neurons in the presence of 6 nM enzaustaurin (Enz) or vehicle (Veh). Scale bar = 20 μm in neuron pictures and 10 μm in spine pictures. F Histograms represent the quantification of the morphological assessment of neurons; n = 7–9 embryos from three independent cultures. Data represent mean ± SEM and differences were analyzed by Two-way ANOVA followed by the LSD-post hoc test. **p < 0.01 vs Ctrl–Veh; #p < 0.05; ##p < 0.01; ###p < 0.001 vs KD–Veh; $p < 0.05 vs Ctrl–Enz. Some data points were discarded after outlier analysis To assess a possible relationship between VPS13A-KD and neuronal morphology in our model, we first analyzed the effect of VPS13A-KD on neuronal morphology independently of treatment, as an initial VPS13A-KD-only analysis. We found that VPS13A-KD significantly reduced dendritic area (45.6%), perimeter (42.5%), skeleton length (42.9%), and branching points (47.6%) compared to controls (Additional File 1: Suppl Info 5). Then, to explore the potential relationship between increased PKCβII activity and neuronal morphology and synaptic spine density, we incubated Ctrl and VPS13A-KD neurons with either 6 nM enzastaurin (Enz), a PKCβII inhibitor, or vehicle (Veh), and analyzed neuronal morphometry and spine density (Fig. 6E–F). Two-way ANOVA revealed a significant Enz treatment effect and a VPS13A-KD × Enz interaction, indicating that the effect of VPS13A-KD on dendritic morphology was treatment-dependent. Precisely, significant Enz treatment effects and VPS13A-KD × Enz interactions were observed for dendritic area (F(1,8) = 6.26, p = 0.0368 for Enz treatment; F(1,8) = 12.26, p = 0.0081 for the interaction), dendritic perimeter (F(1,8) = 7.73, p = 0.0239 for Enz treatment; F(1,8) = 12.26, p = 0.0057 for the interaction), and skeleton length (F(1,8) = 5.63, p = 0.0451 for Enz treatment; F(1,8) = 12.73, p = 0.0073 for the interaction). Post hoc pairwise comparisons further showed that Enz treatment counteracted the VPS13A-KD-associated reductions in dendritic area, dendritic perimeter, and skeleton length. Additionally, a significant VPS13A-KD × Enz interaction was observed for the branching parameter (F(1,8) = 5.70, p = 0.0440) (Fig. 6E–F). We also analyzed the effects of VPS13A-KD and Enz treatment on neurite width and neuronal spine density. Two-way ANOVA revealed significant effects of VPS13A-KD and Enz treatment, as well as a significant VPS13A-KD × Enz interaction for neurite width (F(1,13) = 4.695, p = 0.0494 for VPS13A-KD; F(1,9) = 10.66, p = 0.0098 for Enz treatment; F(1,9) = 11.13, p = 0.0087 for the interaction), indicating that the effect of VPS13A-KD on neurite width was treatment-dependent. Post hoc pairwise comparisons showed that VPS13A-KD neurons displayed reduced neurite width under vehicle conditions, an effect that was counteracted by Enz treatment (Fig. 6E–F). Similarly, a significant VPS13A-KD × Enz interaction was observed for synaptic spine density (F(1,7) = 17.33, p = 0.0042), and post hoc analyses revealed that Enz treatment restored spine density in VPS13A-KD neurons (Fig. 6E–F) (see Additional File 1: Suppl Info 6 for pairwise-comparison results). Together, these results indicate that PKCβII activity is associated with cytoskeletal and synaptic alterations induced by VPS13A deficiency. Discussion Aberrant lipid composition in the brain is emerging as a novel neuropathological mechanism in neurodegenerative diseases [33]. As VPS13A functions as a bulk lipid transporter, disruption of neuronal lipid homeostasis has been proposed as a key pathological process in ChAc [20, 22]. In our murine model, neuronal VPS13A KD led to a selective increase in several DAG species. These changes occurred independently of the levels of enzymes involved in DAG metabolism, suggesting a disturbance in lipid trafficking rather than in metabolic fluxes. We also found that PKCβII, a protein directly responsive to DAG levels, may contribute to cytoskeletal alterations associated with VPS13A deficiency, potentially linking disrupted lipid metabolism to structural neuronal dysfunction. Recent research has established a relationship between basal-ganglia neurodegeneration and aberrant tissue lipid composition [18]. In Huntington’s disease (HD), lipidomic analyses revealed an altered sphingolipid profile in the caudate nucleus of patients, correlating with disease severity [19]. Similarly, increased DAG species have been detected in the frontal cortex of Parkinson’s disease patients and experimental models, where they co-occur with phosphoglyceride imbalance and α-synuclein dyshomeostasis [34, 35]. Our findings contribute to this body of evidence, showing that ChAc, a neurodegenerative disorder affecting basal ganglia circuits [11], is also characterized by altered lipid homeostasis. Although previous lipidomic profiling of post-mortem ChAc brain tissue reported sphingolipid disbalance without DAG changes [21], this discrepancy may reflect differences between the mouse model and the human disease, and/or the time frame of the neurodegenerative process. In our model, DAG increase observed at 30 days may be transient, suggesting it may represent an early-stage event, while the chronic absence of VPS13A in patients could trigger compensatory mechanisms that normalize DAG levels but sustain phospholipid imbalance. This possibility opens the door to exploring the time-dependent progression of molecular and cellular changes caused by the neuronal loss of VPS13A, ultimately leading to neurodegeneration. Augmented DAG levels have been reported in Alzheimer’s, Parkinson’s, and Lewy body diseases [34–36] and may contribute to neuronal dysfunction. Although the mechanisms linking DAG accumulation to neurodegeneration remain unclear, DAG plays central roles in both membrane phospholipid metabolism and signal transduction, among other molecular functions [37]. Given that VPS13A acts as a bulk lipid transporter and is implicated in membrane homeostasis [14], its depletion may impair lipid trafficking and metabolism, leading to DAG accumulation. When we analyzed the putative causes of DAG accumulation, we found no changes in the concentration of Cept1, Chpt1, and Sgms2, the main enzymes involved in DAG metabolism [38–40]. However, the accumulation of DAG species strongly points to a change in those enzymatic activities or substrate availability. Indeed, VPS13A depletion affects phospholipid transport [14] and membrane organization [41], which may indirectly shift the lipid fluxes without necessarily altering the expression levels of key metabolic enzymes. In this line, the absence of changes in PLCγ1 suggests that the increase in DAG is not a consequence of overactivation of this pathway, but rather results from the activation of phospholipid metabolic pathways. Structural analysis further revealed that the DAG species increased in VPS13A-KD tissue share structural similarities, suggesting possible specificity in VPS13A lipid transport, as seen in its yeast homolog VPS13 [42, 43]. Alternatively, this pattern may reflect a broader membrane imbalance rather than substrate-specific accumulation. Clarifying whether VPS13A actively shapes or passively reflects membrane lipid composition is essential to understanding its role in neuronal lipid homeostasis. Our analysis of PKC signaling suggests dysregulated activation of conventional PKCs, likely driven by elevated DAG. This effect appears isoform-specific, as PKCδ remained unchanged. Dysregulation of PKC isoforms has been observed in multiple neurodegenerative contexts. For example, PKCβII and PKCδ are reduced in HD models [44–46], and PKCα/βII alterations have been reported in Alzheimer’s disease [47, 48]. In our model, changes in dendritic complexity and synaptic spine loss are consistent with cytoskeletal disorganization in which PKC signaling may contribute. Stanslowsky et al. [15] also reported altered neuronal morphology and synaptic spine loss as consistent consequences of VPS13A depletion, underscoring the critical role of this lipid transporter in maintaining neuronal architecture and synaptic integrity. PKCα is well established as a regulator of cytoskeletal plasticity, particularly through direct interaction with actin filaments [49]. It integrates NMDA-Ca2⁺ and BDNF-TrkB signaling [50] and modulates Rac1 activation during synaptic plasticity [30]. Notably, these pathways are disrupted in ChAc [17, 51], and VPS13A has been shown to influence cytoskeletal remodeling via Rac1 [52]. Reduced Rac1 activity and cytoskeletal abnormalities have also been observed in neurons derived from iPSCs of ChAc patients [14, 15]. This mirrors the cytoskeletal dysfunction seen in red blood cells, where the impaired actin dynamics are central to the formation of acanthocytes [52, 53], a hallmark feature of the disease [13]. Compared to PKCα, the role of PKCβII in neuronal morphology is less characterized. PKCβII has been associated with microtubules and influences their dynamics [54], and its activation affects actin organization and contractility in growth cones [55]. While PKCβII has been implicated in neurodegeneration [56–58], no prior studies have directly linked it to neuronal morphology or synaptic structure. In our study, PKCβII inhibition led to the recovery of neuronal morphology and synaptic spine density, suggesting that it contributes to cytoskeletal alterations. Further studies are needed to elucidate the PKCβII-specific role in ChAc pathophysiology and other neurodegenerative diseases. Conclusions In summary, impaired lipid transport in neurons due to VPS13A deficiency is associated with DAG accumulation and altered PKC signaling, potentially contributing to aberrant neuronal morphology and synaptic spine loss. These findings support a role for lipid homeostasis in maintaining the structural organization of neurons and suggest that disruptions in DAG/PKC signaling may be involved in the pathophysiology of ChAc. These results offer new perspectives on the molecular mechanisms potentially underlying synaptic dysfunction in ChAc and underscore the need for further investigation into lipid signaling in neurodegeneration. Methods Animals Female and male B6CBA wild-type mice were housed under a standard 12:12 h light/dark cycle with access to food and water ad libitum in a colony room kept at 40–60% humidity and 19–22 °C. All animal procedures were performed in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the local animal care committee of the Universitat de Barcelona (225/21, 245/22, OB_146.23), in accordance with the Spanish RD 53/2013 and the 2010/63/EU Directive of the European Commission. Experiments were reported following the ARRIVE guidelines, when applicable. AAV vector plasmids Vector plasmids were designed to KD VPS13A expression in vitro and in vivo using a pCAG-miR30 shRNA-KD system as previously described [17]. In brief, an AAV containing two VPS13A shRNA and the fluorescent marker ZsGreen was developed under synapsin 1 promoter (AAV8[2miR30]-SYN1 > ZsGreen1:shVPS13A-1:shVPS13A-2:WPRE; KD). Another AAV containing two scramble shRNA and ZsGreen (AAV8[2miR30]-SYN1 > ZsGreen1:shCtrl-1:shCtrl-2:WPRE; Ctrl) was used as control. Virus production, amplification, and purification were performed by VectorBuilder (titers, ~ 5 × 1013 genomic particles/ml). Stereotaxic surgery and tissue sampling Stereotaxic surgery was performed in 8-week-old mice under isofluorane anesthesia (5% induction, and 1.5% maintenance), as previously described [17]. A volume of 0.5 µL of corresponding viral constructs was injected bilaterally in two regions of the striatum at the following coordinates: + 0.8 mm AP, ± 1.8 mm L, − 2.6 mm DV and + 0.6 mm AP, ± 2 mm L and − 2.6 mm DV from bregma and dura, and in the frontal cortex at the following coordinates: + 2.46 mm anteroposterior (AP), ± 1 mm mediolateral (L), and − 0.8 mm dorsoventral (DV) from bregma and dura mater, using a 5 µL Hamilton syringe with a 33-gauge needle at 0.1 µL/min (Fig. 1A, B). The injection needle was left for an additional 5 min period to allow diffusion of virus particles and avoid reflux. For biochemical analysis, 30 days after surgery mice were decapitated, and the brains were removed. Immediately, the frontal cortex and striatum were dissected (Fig. 1A), frozen in liquid N2 and kept at − 80 °C until use. Lipid extraction and analysis from brain tissue samples Lipidomic analyses were performed at the Institut de Química Avanzada de Catalunya-Consejo Superior de investigaciones científicas (iQAC-CSIC, Barcelona, Spain). For each mouse, frontal cortex and striatum samples were mixed and homogenated in phosphate-buffered saline (PBS) at a final concentration of 5 mg/ml protein and spiked with multiple class-based internal standards of similar ionization efficiencies. Glycerophospholipids, sphingolipids, and their derivatives and precursors were separated by liquid chromatography-high resolution mass spectrometry (LC-HRMS) using an Acquity ultra-high-performance liquid chromatography (UHPLC) system (Waters) connected to a time-of-flight detector (LCT Premier XE). Lipid extracts were injected onto an Acquity UHPLC BEH C8 column (1.7 μm particle size, 100 mm × 2.1 mm, Waters, Ireland) at a flow rate of 0.3 mL/minute and a column temperature of 30 °C. The mobile phases were methanol with 2 mM ammonium formate and 0.2% formic acid (A)/water with 2 mM ammonium formate and 0.2% formic acid (B). Full-scan spectra from 50 to 1800 Da were acquired, and individual spectra were summed to produce data points of 0.2 s each. Mass accuracy at a resolving power of 10,000, and reproducibility were maintained by using an independent reference spray via LockSpray interference. Internal standards representative of each lipid class were used to monitor the extraction recovery and validate the data treatment process. The table in Fig. 1B defines the abbreviations of the lipid classes determined in this study. Identification of compounds was based on accurate mass measurements with an error of B5 ppm and an LC retention time of 92% compared with that of a standard. Quantification was carried out using the extracted ion chromatogram of each compound using 50 mDa windows. The linear dynamic range was determined by injecting mixtures of internal and natural standards. Since standards for all identified lipids were unavailable, the amounts of lipids are given as pmol/mg protein. Lipidomic statistical analysis The lipid profiles of brain tissue samples were compared between the VPS13A-KD and control groups using multivariate analyses, including sex as a concomitant variable. Variable distribution was scaled by calculating Z-scores for all pre-processed data. Then two different complementary analyses were approached. Firstly, a linear ordinary least square regression model was generated using the VPS13A expression group (Ctrl and VPS13A-KD) and sex as predictor variables over the level of every lipid species. The lipids whose concentration was significantly modified by VPS13A-KD were determined by calculating the p-value of the model corrected by the effective number of tests (ENT). The ENT was calculated by using the method of eigenvalues derived from the correlation matrix [59]. The number of tests with eigenvalues > 1 [60] was applied to the Bonferroni correction, and p < 0.004545 was considered significant. Secondly, a partial discriminant analysis was performed to validate the results of the linear regression model. In this analysis, to subtract the effect of sex in lipid levels, we generated a model of linear ordinary least square regression only with sex as the predictor variable. Then, we used the residuals of this model to perform a PLS-DA with repeated cross-validation. To evaluate the performance of the PLS-DA classification, we calculated the optimal number of components, the BER, and the AUC scores, which estimate the specificity and sensitivity of the model from training cross-validation with 150 repetitions. The accuracy of the whole PLS-DA model was validated through a permutation system with 1000 iterations. Finally, we calculated the VIP scores to estimate the importance of each variable in the projection used in the PLS-DA model and to select those lipids that contribute most to the underlying differences. VIP scores ≥ 1.4 were considered significant when selecting lipids for the final model. Unsupervised PCA, validated with a PERMANOVA with 999 permutations, was also performed. All analyses were performed with R. The PLS-DA model was generated and validated with the MixOmics R package [61]. R packages were executed in RStudio 2022.12.0–353 for Windows (Posit.co) and R 4.2.2 for Windows (The Comprehensive R Archive Network -CRAN-). Lipid enrichment and pathway analysis Lipid enrichment analysis was performed using the Lipid Ontology (LION)/web application (http://www.lipidontology.com/) [62]. Based on a library of lipid species, LION/web reports whether specific physicochemical properties, biological functions, or cellular localization are enriched in the studied conditions. The lipid species dysregulated in VPS13A-KD samples were uploaded as target list, with all lipids detected in the LC-HRMS analysis as background list to identify LION enrichment. In these analyses, a given term is enriched if the lipids belonging to the term are overrepresented in a target list than expected by chance. Lipid terms were sorted by enrichment using -log(FDR q-values) and a threshold was set at 1.4 to determine those significantly enriched in the VPS13A-KD group. The characteristics of the lipid three-dimensional structures were explored to identify regions of interest within the enriched lipids. To do that, the structures of the lipid species detected in our LC-HRMS analysis were collected from the LIPID MAPS® Structure Database, which encompasses structures and annotations of biologically relevant lipids, using the lipidmapsR interface to LipidMaps REST API. Morgan fingerprints (nBits = 2048) for each lipid structure were generated using the RDKit package (https://www.rdkit.org) in Python and, then, t-SNE dimensionality reduction was used to visualize the structural relationships between molecules based on their fingerprints and to identify clusters of similar structures. After that, a clustering analysis was performed by applying the density-based spatial clustering of applications with noise (DBSCAN) algorithm with the parameters MinPts = 5 and eps = 3, based on the k-NN distance plot, using the Dbscan package for R [63]. Lipid pathway analysis was performed using Bioinformatics Methodology For Pathway Analysis (BioPAN; https://lipidmaps.org/biopan/) [64] on the LIPID MAPS® Lipidomics Gateway. Our dataset with the lipid quantification was loaded into the BioPAN webpage and analyzed following the tool’s standard workflow. BioPAN provides a Z-score and calculates the p-value for a given reaction. Then the reaction is classified as activated (Z-score > 1.645) or suppressed (Z-score < − 1.645) depending on the direction of the change. To implement the analysis the following options were selected: condition of interest, VPS13A-KD, control: Ctrl; type, lipid; status, active; level, lipid subclass; subset of lipid data, reactions; p-value, 0.05; paired data, no. Generation of VPS13A-knockdown in primary neuronal cell cultures Cortical primary cell cultures were performed as previously described [17]. Briefly, the cerebral cortex of E18 mouse embryos was dissected, placed in Neurobasal medium (Cat: 21,103–049, Thermo Fisher Scientific, Waltham, MA, USA), and dissociated through a glass Pasteur pipette. Afterward, cells were seeded at a density of 60.000 cells/well onto 12-mm glass coverslips, previously coated with 0.1 mg/mL Poly-D-lysine (Cat: 16,021,412, Thermo Fisher Scientific, Waltham, MA, USA), for immunocytochemistry samples, and at 300.000 cells/well density onto 12-well dish pre-coated with 0.1 mg/mL Poly-D-lysine, for Western Blot samples. We supplemented Neurobasal medium with B27 (Cat: 17,504–044, Thermo Fisher Scientific, Waltham, MA, USA) and Glutamax (Cat: 13,462,629, Thermo Fisher Scientific, Waltham, MA, USA) in serum-free conditions. Cultures were maintained at 37 °C in a humified atmosphere, containing 5% CO2 until 15 DIV. For the different experiments, neuronal cultures were performed on embryos from two or three independent mothers. VPS13A-KD was generated through infection with AAV at 2 DIV. Shortly, half of the medium from each well was collected and stored at 4 °C. Neurons were infected with AAV-Ctrl or AAV-KD at a multiplicity of infection of 1:50.000 in Neurobasal medium supplemented with B27 and Glutamax. After 24 h, the virus-containing medium was discharged and replaced with a conditioned medium, made with that kept at 4 °C and new Neurobasal supplemented with B27 and Glutamax (50%–50%). Thirteen days after infection (15 DIV), cells were harvested or fixed. VPS13A and DAG labeling and fluorescence quantification in neuronal cultures For VPS13A immunolabeling in primary cortical neurons, cell cultures were seeded on 12-mm glass coverslips. At 15 DIV, cultures were fixed with 4% paraformaldehyde (PFA) for 15 min and washed with 0.1 M PBS. Afterwards, coverslips were permeabilized with 0.1 M PBS containing 0.1% saponin and 1% BSA, for 10 min at RT. After permeabilization, coverslips were incubated with a blocking solution of 0.1 M PBS containing 15% Normal Goat Serum (NGS) for 30 min at RT. Then, coverslips were incubated overnight at 4 °C in an antibody primary solution containing anti-VPS13A (1:200, Invitrogen) and 5% NGS. Coverslips were incubated with secondary antibody solution containing Alexa Fluor 555 Goat anti-mouse (1:100, Thermo Fisher Scientific) for 1 h at RT and, finally, incubated with Hoechst (1:10,000), mounted on microscope slides with ProLong (Thermo Fisher Scientific) and kept in the dark. For DAG labeling in primary cortical neurons, the Red Downward DAG sensor (#D0300R, Montana Molecular) was used following the manufacturer’s protocol. Briefly, 15 DIV cells were transduced with 40 µl of BacMam DAG sensor, supplemented with 2 mM of sodium butyrate in complete culture medium to a final volume of 250 µl per well. Cells were then kept for 24 h in the incubator under normal growth conditions (5% CO2 and 37 °C), protected from light. Afterwards, culture medium was replaced with 0.1 M PBS, containing Ca2+ and Mg2+, and allowed to rest for 25 min at RT, protected from light, before acquiring the images. Imaging was performed using a ZEISS LSM880 confocal microscope (Zeiss, Jena, Germany) at × 20 magnification. The fluorescence intensity of every single neuron was calculated using the measurement plugin of the image processing software Fiji ImageJ (Madison, WI, USA). Mean fluorescence intensity values were extracted from manually defined regions of interest corresponding to single neurons. Background-corrected fluorescence value was obtained by subtracting the median pixel value from a background region devoid of cells within the same field of view. For the computation of the VPS13A labeling, fluorescence intensity values were normalized to the mean fluorescence of the Ctrl condition, which was set to 1. Because the BacMam DAG sensor exhibits inverse fluorescence response (i.e., lower fluorescence corresponds to higher DAG levels), for the computation of the DAG labeling, intensity values were inverted before normalization (I′ = Imax/I), where Imax corresponds to the maximum background-corrected intensity measured within the same experiment. This transformation yields values where higher inverted intensity values reflect higher DAG levels. Drug treatment in primary cultures To pharmacologically inhibit the activation of PKCβII, cortical neuron cultures were treated using Enz as a selective inhibitor of PKCβII (Selleckchem, Houston, TX, USA) at a concentration of 6 nM [65, 66]. Enz was diluted in 0.5% DMSO. This same DMSO concentration was also used as a Veh in the control condition. Cell cultures were treated at 3 DIV. Western blotting Primary cultures were homogenized in lysis buffer (50 mM Tris–HCl pH 7.5, 10% glycerol, 1% Triton X-100, 150 mM NaCl, 100 mM NaF, 5 μM ZnCl2, and 10 mM EDTA, supplemented with proteases inhibitors (1 μg/ml aprotinin, 1 μg/ml leupeptin, 1 mM Na3VO4, and 2 mM PMSF) immediately before use and protein samples were resolved in NuPage gels. Proteins were transferred to nitrocellulose membranes through the iBlot2 Gel Transfer Device (Thermo Fisher Scientific, Waltham, MA, USA). Immunoblots were probed with either anti-VPS13A (1:1500, HPA021662, Sigma-Aldrich, St. Louis, MI, USA); anti-Chpt1 (1:1000, Bioss, USA); anti-Cept1 (1:1000, Proteintech, Chicago, IL, USA); anti-Sphk2 (1:3000, Proteintech); anti-Sgms2 (1:1000, Thermo Fisher Scientific); anti-PLCγ1 (1:1000, Cell Signaling Technologies); anti-phosphorylated PLCγ1 (pPLCγ1) (1:1000, Cell Signaling Technologies); anti-PKCα (1:200, Santa Cruz Biotechnologies, Dallas, TX, USA); anti-PKCβII (1:200, Santa Cruz Biotechnologies); anti-PKCδ (1:1000, Cell Signaling Technologies, Danvers, MA, USA); anti-phosphorylated PKCα/βII (pPKCα/βII) (1:1000, Cell Signaling Technologies); anti-VPS13A (1:1500, Sigma-Aldrich, St. Louis, MI, USA); or anti-Vinculin (1:1000, Cell Signaling Technologies). Membranes were incubated with anti-rabbit HRP conjugated IgG (H + L) (1:2000, Promega, Madison, WI, USA) or with anti-mouse HRP conjugated IgG (1:2000, Promega) Immunoreactive bands were visualized using ECL Prime Western Blotting Detection Reagent (Cytiva, Marlborough, MA, USA) or using SuperSignal West Femto (Thermo Fisher Scientific). Images were acquired using ChemidocTM (Bio-Rad, Hercules, CA, USA) and quantified by a computer-assisted densitometer (ImageLabTM, Bio-Rad). Uncropped immunoblots for all Western blot experiments are provided in Additional File 2. Actin labeling and morphometric analysis in neuronal cultures Cortical primary cultures grown on glass coverslips were fixed at 15 DIV with 4% PFA for 15 min at 4 °C and washed with 0.1 M PBS. After fixation, cells were permeabilized with 0.1 M PBS containing 0.1% Triton X-100 for 15 min at room temperature. Afterward, coverslips were incubated with AF647 Phalloidin (1:400, Cat: A22287, Invitrogen, Waltham, MA, USA) for 60 min at RT. Nuclei staining was performed using Hoechst (1:10.000, Thermo Fisher Scientific) for 15 min at RT, coverslips were mounted on microscope slides with ProLong™ (Invitrogen) and kept in the dark. For the morphometric analysis, digital images were acquired using the LSM880 confocal microscope (Zeiss, Jena, Germany) at × 40 magnification. Images were analyzed using the measurement plugin of the image processing software Fiji ImageJ (Madison, WI, USA). Single neurons were selected from the confocal images, and the soma was identified as an independent object. Then, a masked image was created, and the neuronal skeleton was identified, marking the branching points and the endpoints (termini branches). The dendrite area, the dendritic perimeter, the skeleton length, and the number of primary branches, branching points, and end points were obtained from the skeleton image. For the spine density analysis, images were obtained at × 63 magnification, and the Dendritic Spine Counter plugin from Fiji ImageJ was used. The image contrast was maximized by creating a grayscale working image. Dendritic spines were individually counted, and the density was obtained by dividing the number of dendritic spines by the length of the neurite in micrometers. In the same pictures, the neurite width was also measured. In all analyses, neurons from 5 random fields per coverslip were analyzed, and the mean of all neurons per field was acquired. For each experimental condition, six embryos (three coverslip/embryo) from three independent mothers were used. Statistical analysis of protein concentration and cellular morphology analyses The effect of VPS13A-KD on the VPS13A labeling, the neuronal DAG levels, the proteins quantified by Western blot, and neuronal morphology was assessed by mono-variant or bi-variant statistical analyses. Normal data distribution was assumed when at least one of the following tests was positive for normality: Shapiro–Wilk, Kolmogorov–Smirnov, and D’Agostino-Pearson. Statistical analyses were performed using unpaired Student’s t-test or two-way ANOVA as appropriate. All experiments were blinded and randomized, and results are reported as mean ± SEM. All statistical tests were performed on GraphPad Prism 10.0 (GraphPad Software, San Diego), and values of p < 0.05 were considered statistically significant. Supplementary Information Additional file 1: Supporting methods and figures. Suppl Info 1: regression model. Suppl Info 2: analysis of the sex effect on lipid levels. Suppl Info 3: structure of the DAG species enriched in VPS13A-KD samples. Suppl Info 4: significant BioPAN reactions. Suppl Info 5: Effect of VPS13A knockdown on neuronal morphology. Suppl Info 6: pairwise comparisons of the neuronal morphology analysis. Additional file 2: Uncropped immunoblots. Pictures of uncropped immunoblots for Western blot experiments in Figs. 5 and 6. Acknowledgements We are very grateful to Dr. Josefina Casas (Institut de Química Avanzada de Catalunya; iQAC-CSIC) for LC-MS lipid determinations. We are grateful to Carmen Andrade (Dept. Biomedical Sciences, UB) for excellent technical support, to the staff of the Confocal Microscopy Service and the Animal Experimental Unit of the Scientific and Technological Centers of the University of Barcelona (CCiTUB), and to Ana Maria Lopez and Silvia Artigas for excellent mouse colony management. Abbreviations Adeno-associated virus Area under the curve Balanced error rate Bovine serum albumin Choline/ethanolamine phosphotransferase 1 Chorea-acanthocytosis Diacylglycerol choline phosphotransferase 1 Diacylglycerol Days in vitro Effective number of tests Enzaustaurin Huntington’s disease Human induced pluripotent stem cell Knockdown Liquid chromatography-high resolution mass spectrometry Medium spiny neurons Normal goat serum N-methyl-D-aspartate receptor subunit 1 Phosphate-buffered saline Principal component analysis Paraformaldehyde Protein kinase C Gamma isoform of phosphatidylinositol phospholipase C Phospholipid phosphatase Partial least squares differential analysis Sphingomyelin synthase 2 Sphingosine kinase T-Distributed Stochastic Neighbor Embedding Ultra-high-performance liquid chromatography Vehicle Variable importance in projection Intermembrane lipid transfer protein VPS13A Authors’ contributions All authors read and approved the final manuscript. GB-S: methodology, investigation, formal analysis, visualization, writing—review & editing. EG-G: Conceptualization, methodology, investigation, writing—review & editing. AR-L: Investigation, visualization, writing—review & editing. GE: Formal analysis, writing—review & editing. PG-S: formal analysis, visualization, writing—review & editing. CM: Investigation, funding acquisition, writing—review & editing. EM: Investigation, funding acquisition, writing—review & editing. JA: Conceptualization, supervision, resources, funding acquisition, writing—review & editing. MM: Conceptualization, funding acquisition, resources, methodology, writing—review & editing. MJR: Conceptualization, resources, formal analysis, data curation, funding acquisition, writing—original draft. Funding This study was supported by grants from the Ministerio de Ciencia e Innovación (Spain), under projects no. PID2023-150728OB-I00 (J.A. and M.J.R.), PID2023-150592OB-I00 (C.M.), no. PID2023-146279OB-I00 (E.M.), and no. PID2021-124896OA-I00 (M.M); Instituto de Salud Carlos III, Ministerio de Ciencia, Innovación y Universidades and European Regional Development Fund (ERDF) (CIBERNED, to J.A.), Spain; and ChAc Foundation (J.A., M.M. and M.J.R.) Spain. PG-S was supported by an FPU grant from the Spanish Ministry of Science and Innovation (grant #FPU21/02928). Also, the project has been supported by María de Maeztu Unit of Excellence (CEX2021-001159), Institute of Neurosciences of the University of Barcelona, Ministry of Science, Innovation, and Universities. Data availability The datasets generated during the current study are available from the corresponding author upon reasonable request. The lipidomics data are publicly available in the Figshare repository and can be accessed via 10.6084/m9.figshare.32083563. [67]. Declarations Ethics approval and consent to participate This study was performed in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the local animal care committee of the Universitat de Barcelona (225/21, 245/22, OB_146.23), in accordance with the Spanish RD 53/2013 and the 2010/63/EU Directive of the European Commission. Experiments were reported following the ARRIVE guidelines, when applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Additional file 1: Supporting methods and figures. Suppl Info 1: regression model. Suppl Info 2: analysis of the sex effect on lipid levels. Suppl Info 3: structure of the DAG species enriched in VPS13A-KD samples. Suppl Info 4: significant BioPAN reactions. Suppl Info 5: Effect of VPS13A knockdown on neuronal morphology. Suppl Info 6: pairwise comparisons of the neuronal morphology analysis. Additional file 2: Uncropped immunoblots. Pictures of uncropped immunoblots for Western blot experiments in Figs. 5 and 6. Data Availability Statement The datasets generated during the current study are available from the corresponding author upon reasonable request. The lipidomics data are publicly available in the Figshare repository and can be accessed via 10.6084/m9.figshare.32083563. [67].