LDs have a neutral lipid core containing triacylglycerols (TAGs) and cholesterol esters (CEs), which is surrounded by a phospholipid monolayer and decorated by LD coat proteins (Fig. 1a). One type of coat protein is the Perilipin (PLIN) family, containing five members (PLIN1-5). While PLIN2 and PLIN3 are ubiquitously expressed, the other members have a more tissue-specific expression²⁴. We first assessed the expression of the Plins in proliferating adult SVZ and DG NSPCs in vitro by quantitative reverse transcription PCR (RT-qPCR). Plin2 was the highest expressed Plin mRNA in NSPCs from both SVZ (Fig. 1b) and DG (Supplementary Fig. 1a). As PLIN2 is a highly specific LD coat protein²⁴, antibodies against PLIN2 can be used to visualize LDs. To confirm that we visualized the majority of LDs by staining against PLIN2, we co-stained SVZ NSPCs with the fluorescent neutral lipid dye BODIPY493/503 (hereafter called BD493), which preferentially accumulates in LDs. Quantification of the co-localization with Plin2 showed that more than 90% of BD493 positive LDs were also positive for PLIN2 (Fig. 1c). Thus, these results validate PLIN2 as a robust marker of LDs in NSPCs.

We next characterized LD occurrence using high-resolution single-cell 3D-volume reconstructions (Supplementary Fig. 1b). Proliferating NSPCs from SVZ and DG showed a large variability in LD numbers, total LD volume and diameters of LDs, with neighbouring cells containing from a few LDs up to many hundreds of LDs (Fig. 1d, e; Supplementary Fig. 1c, d). Interestingly, DG NSPCs had lower numbers of LDs and smaller LD volume per cell (Supplementary Fig. 1d) than SVZ NSPCs, despite being grown under the same cell culture conditions as SVZ NSPCs, suggesting that LD accumulation in NSPCs is not only driven by nutrient availability. Theoretically, a cell can store the same number of lipids in a few large LDs or in many small LDs. Thus we tested whether NSPCs have a similar total stock of neutral lipids, despite the large differences in LD numbers. LD number per cell correlated with total LD volume in both NSPCs from SVZ and DG (Supplementary Fig. 1e), indicating that there is little compensation of lipid content in cells with low amounts of LDs. Cell size also could not explain the variability in LD accumulation, showing only a weak correlation with LD numbers (Supplementary Fig. 1f).

Proliferating NSPCs in vitro are not synchronized, thus cells fixed at a certain timepoint can be in any cell cycle stage. To address whether the observed variability in LD accumulation correlated with cell cycle stages, we used NSPCs isolated from the SVZ of adult FUCCI mice²⁵. The FUCCI system allows for direct visualization of cell cycle stages due to two fluorescent proteins that fluctuate with cell cycle, leading to differently coloured nuclei (G1 red, S-phase entry yellow, S/G2/M green, non-coloured just after division). The same LD analysis using high-resolution single-cell 3D-volume reconstructions was performed in FUCCI NSPCs. LD numbers and volumes were also variable within each cell cycle stage analysed separately, thus could not per se explain the large LD variability among NSPCs (Supplementary Fig. 1g).

Taken together, these results show that there is a high variability in the availability of stored lipids in NSPCs grown under the same in vitro conditions, which cannot be explained by cell size or cell cycle stage. This suggests that among a putatively homogenous population of NSPCs, differences in lipid metabolic pathway activity exist, which might influence stem cell behaviour.

As NSPCs in vitro might show a different lipid metabolism than in their niche in the brain, we next addressed whether LDs are also present in NSPCs in vivo. Visualizing LDs in brain tissue is challenging, as the permeabilization required for antibody penetration affects LD stability²⁶, and milder permeabilization approaches that can be used for immunohistochemistry in vitro have a poor antibody penetration in vivo. Thus, we investigated the presence of Plin2 mRNA in vivo, using two different approaches: published scRNA-seq data and in situ hybridization. Plin2 mRNA was clearly expressed in NSPCs and astrocytes in two recent scRNA-seq datasets^27,28 of NSPCs and niche cells in the SVZ (Fig. 1f and g) and in the hippocampus (Supplementary Fig. 1h and i). To confirm the scRNA-seq data analysis, we used RNAscope technology, an advanced in situ hybridization assay, on brain tissue from NestinGFP reporter mice having GFP-labelled NSPCs²⁹. Plin2 mRNA was detectable in the SVZ (Fig. 1h and Supplementary Fig. 1j) and in the SGZ of the DG (Supplementary Fig. 1k and l), and co-localized with GFP-positive cells. Furthermore, PLIN2 immunostaining on brain sections from NestinGFP mice showed ring-like structures in NestinGFP-labeled cells in the ependymal layer/SVZ (Fig. 1i), and also within the DG (Supplementary Fig. 1m). Together, these data show that Plin2 is also expressed under normal physiological conditions in NSPCs in vivo.

NSPCs exist in either a quiescent or an activated state, which can be mimicked in vitro: addition of BMP4 in combination with a withdrawal of epidermal growth factor (EGF) leads to a quiescent state^30,31. As several studies have shown that proliferating and quiescent NSPCs are metabolically different and that lipid metabolism is altered^15,32,33, we next addressed whether LD accumulation differs under these two conditions. Similar to proliferating NSPCs, qRT-PCR revealed that Plin2 was the highest expressed among the Plin family members in quiescent NSPCs (Fig. 2a), with comparable levels of Plin2 mRNA between cell states (Supplementary Fig. 2a). In addition, Plin1 and Plin4 mRNA expression was significantly higher in quiescent NSPCs compared to proliferating NSPCs (Supplementary Fig. 2b). Co-staining of PLIN2 protein with BD493 showed that PLIN2 is a good LD marker in quiescent NSPCs (Supplementary Fig. 2c). However, the percentage of BD493 and PLIN2 double positive LDs was lower than in proliferating NSPCs (72.3% compared to 90.3%, Supplementary Fig. 2d and Fig. 1c), suggesting that other PLINs might play a role in LD regulation in quiescent NSPCs. While PLIN2-positive LD accumulation under quiescence did not significantly differ between proliferating and quiescent NSPCs (Fig. 2b and c), LDs appeared different, with some very large LDs in quiescent NSPCs (Fig. 2b). This observation was confirmed when measuring the diameter of LDs in proliferating and quiescent NSPCs, with a significant increase in the proportion of large LDs in quiescence (Fig. 2d). As LDs are sensitive to fixation and permeabilization, we next aimed to corroborate the observed differences in LDs using live cell imaging during quiescence induction of NSPCs. To do so, we generated a fluorescent LD-reporter system by stably overexpressing a Plin2-Gfp fusion construct in proliferating SVZ NSPCs, resulting in green fluorescent LDs. The LD-specific GFP signal was confirmed with co-staining against PLIN2 protein (Fig. 2e). As PLIN2 has been reported to stabilize LDs³⁴, we first tested whether overexpression of PLIN2-GFP in proliferating NSPCs changed the observed variability of LDs using the same single-cell 3D reconstruction quantification as done in wild-type NSPCs. LD numbers and total LD volume were still variable among cells (Supplementary Fig. 2e and 2f), suggesting that the overexpression of this reporter construct did not fundamentally alter LD accumulation in NSPCs, although PLIN2-GFP NSPCs had slightly less LDs than non-transfected NSPCs (Supplementary Fig. 2f).

We next live-imaged PLIN2-GFP reporter NSPCs during the 72 h of quiescence induction (Fig. 2f and Supplementary Fig. 2g, supplementary movie 1). We used two different plating densities, with more than 2-fold differences in the initial number of cells, to control for potential confluency effects and contact inhibition effects. The area covered by PLIN2-GFP normalized to cell numbers showed a significant increase over the 72 h when full quiescence is reached, independent of cell density (Fig. 2g and Supplementary Fig. 2h). LD diameters were significantly increased at 48 h and 72 h after quiescence induction in the PLIN2-GFP NSPCs, (Fig. 2h). The large GFP-positive rings were confirmed to be LDs using a lipid dye (Supplementary Fig. 2i). Cell numbers doubled within the first 24h–48h of quiescence induction and then remained stable (Supplementary Fig. 2j), indicating that contact inhibition is not required for quiescence. Taken together, these data show that with quiescence, LDs increase in size. This suggests that quiescence is accompanied by a change in lipid metabolic pathways, in line with previously published gene expression and proteomic data^15,32,33.

Besides being in an active or quiescent state, NSPCs can differentiate into astrocytes and neurons, and to a small extent also into oligodendrocytes. In vitro, differentiation occurs over five to seven days upon growth factor withdrawal and leads to a co-culture of astrocytes and neurons. To address expression of Plin family members in these NSPC-derived astrocytes and neurons separately, we established a separation protocol based on the different adherence properties of these two cell types (Supplementary Fig. 3a). We verified the enrichment of either neurons or astrocytes by analysing the expression of cell specific markers (Supplementary Fig. 3b) and analysed the expression of the Plins in these enriched fractions. For both astrocytes and neurons differentiated from NSPCs, Plin2 remained the highest expressed Plin family member (Fig. 3a) and showed a strong co-localization (87.3% in astrocytes, 78.2% in neurons) with BD493 positive LDs (Supplementary Fig. 3c and d). However, direct comparison of Plin2 mRNA showed that neurons had much lower Plin2 mRNA levels than NSPCs and astrocytes (Fig. 3b). 3D reconstructions of individual neurons and astrocytes confirmed this finding: neurons had significantly less LD numbers and total LD volume and reduced variability (Fig. 3c and d) as compared to astrocytes, which were full of LDs and displayed a large variability in number and total volume of LDs. As neurons and astrocytes differ in cell size (Supplementary Fig. 3e), we next tested whether these differences persist if cell size was considered. Even when normalizing to cell volume, neurons had significantly less LD numbers (Supplementary Fig. 3f), meaning that they have significantly less stored neutral lipids available per cell than astrocytes.

Given the variability of LDs in NSPCs (Fig. 1d and e; Supplementary Fig. 1c and d) and the significant difference in LDs upon differentiation, we next addressed whether there was a correlation between the natural range of LD accumulation and fate choice. We used the PLIN2-GFP reporter NSPCs and categorized them into low, medium, and high GFP signal. Then, we followed them with time-lapse imaging over 5 days of differentiation to determine their fate. Each category of NSPCs gave rise to a similar proportion of astrocytes, neurons and cells that underwent cell death without significant changes between the categories (Supplementary Fig. 3g). These results suggest that initial numbers of LDs available are not sufficient to predict cell fate. However, the difference in LDs between astrocytes and neurons at the end of differentiation was still evident (Supplementary Fig. 3h and i), suggesting that lipid metabolism might be altered during or following fate choice.

We next tested whether we could influence fate by artificially increasing or decreasing LDs beyond the natural variability, prior to NSPC differentiation. Extracellular neutral lipids can be taken up by the cells and stored in LDs if not immediately used. Thus, we exposed proliferating NSPCs to two different doses of oleic acid coupled to fatty acid (FA) free BSA (0.1 mM and 0.5 mM) or FA-free BSA alone (0.1% and 0.5%) for 15 h (Fig. 3e). LD accumulation massively increased with oleic acid loading and led to a more than 9-fold increase in the area covered by PLIN2 compared to non-treated NSPCs (Fig. 3f and g). LDs decreased more than half with lipid efflux (Fig. 3f and g), demonstrating that LD content in NSPCs can indeed be artificially altered. Such treated NSPCs were then exposed to differentiation conditions. Analysis of the proportion of neurons after 7 days showed a significant increase of MAP2AB positive neurons in the loading condition (Fig. 3h and i). Remarkably, a significant increase in LD accumulation was observed in the loading condition even after 7 days of differentiation (Supplementary Fig. 3j and k), indicating that the initial lipid load in NSPCs was carried on with differentiation. To rule out that this effect was due to remaining oleic acid in the medium or attached to the plastic of the cell culture plate, proliferating NSPCs were treated for 15 h with oleic acid or FA-free BSA, followed by cell resuspension and two washing steps before the cells were plated into fresh cell culture dishes for differentiation. Under these conditions, the percentage of neurons was also significantly increased in oleic acid-loaded NSPCs (Supplementary Fig. 3l), pointing towards a cell intrinsic effect rather than remaining oleic acid in the medium. The increase in neurons was at least partially at the cost of astrocyte production, as the ratio of neurons (MAP2AB) to astrocytes (GFAP) was significantly changed in the oleic acid loading condition (Supplementary Fig. 3m). Taken together, these data suggest that artificially increasing LDs prior to NSPC differentiation leads to more neurons. Whether this is due to increased neuronal production, better survival of the newly generated neurons, a fate-influencing effect of oleic acid, or a yet unknown mechanism remains to be determined.

While lipid loading or efflux allows to manipulate LD accumulation in NSPCs, it leads to massive LD changes beyond the natural variability, thus might not reveal more subtle influences of LDs on NSPC behaviour. We therefore next used the natural variability of LD content to probe if LDs correlated with the proliferative ability of NSPCs. We performed cell cycle analysis of proliferating PLIN2-GFP NSPCs using live Hoechst, and split the population retrospectively into the top 25% of PLIN2-GFP signal versus low 25% of PLIN2-GFP signal for analysis. This revealed that NSPCs with the top 25% of PLIN2-GFP signal had a significantly larger population of cells in S/G2/M phase than those with lower PLIN2-GFP signal (Supplementary Fig. 4a and b), suggesting that a higher number of LDs gives a proliferative advantage. Using fluorescent activated cell sorting (FACS) we next collected two populations of GFP-positive NSPCs, namely the ones containing the 25% highest and the 25% lowest PLIN2-GFP signal (Fig. 4a). Viability of these cells was confirmed using the viability marker Calcein red (Supplementary Fig. 4c). Following sorting, cells were cultured under proliferative conditions for 48 h. Interestingly, the collective PLIN2-GFP signal remained higher in the high GFP cells compared to the low-GFP cells, indicating that these two distinct populations remained stable over this time (Supplementary Fig. 4d and e). The high PLIN2-GFP NSPCs had an increased proportion of cells positive for the proliferation marker phospho-histone 3 (pH3), compared to the low PLIN2-GFP NSPCs (Supplementary Fig. 4d and f). Moderate proliferation effects, as seen at 48 h after sorting, might become more pronounced with prolonged time in culture. We thus sorted equal numbers of low and high GFP cells into 96-well plates at low density and let them grow as neurospheres for 7 days, after which the area covered by cells was assessed using the live nuclear dye Hoechst (Fig. 4b). This analysis revealed a significant increase in the area covered by cells in the high GFP NSPCs compared to the low-GFP NSPCs (Fig. 4c), suggesting that over a prolonged time in culture, proliferating NSPCs with high LD numbers generate more progeny. Taken together, this set of experiments suggest that within a population of NSPCs displaying a variable number of LDs, a higher number of LDs correlates with increased proliferative ability.

Specified organelle distribution is critical during cell division and can drive downstream daughter cell behaviours³⁵. However, little is known about the mitotic segregation and inheritance of LDs in general and specifically in NSPCs. To address this, we used high-resolution single-cell 3D-volume reconstructions of paired-daughter NSPCs (anaphase to telophase). As observed previously, LD variability was evident in paired-daughter cells, with pairs having from a few to many LDs, confirming that variability was not per se linked to cell cycle (Figs. 4d and S4h). Surprisingly, we observe many daughter pairs in proliferative conditions that displayed a clear asymmetric distribution of LDs among the two daughter cells (Figs. 4d and S4h). When defining asymmetry as at least a 1.5-fold difference between daughters, we found that 31.4% of dividing SVZ NSPCs distributed their LDs asymmetrically (Fig. 4d and e). The percentage of asymmetry increased to 51.4% when analysing total LD volume (Fig. 4e), suggesting that asymmetrical inheritance of LDs leads to unequal neutral lipid distribution among daughter cells. A more stringent analysis of asymmetric distribution for discrete values such as LD numbers, using the square root of n to determine the threshold of asymmetry, confirmed these results (Supplementary Fig. 4g). Similarly, dividing DG NSPCs showed an asymmetric distribution of LDs, with 40.1% of the cells having asymmetry in LD numbers and 56.1% asymmetry when considering total LD volume (Supplementary Fig. 4h–j). This asymmetric inheritance occurred in proliferative conditions, where both daughters will go on to remain as NSPCs, suggesting this asymmetric inheritance can occur in the absence of cell fate determinants.

To test whether such an asymmetric inheritance of neutral lipids is directly impacting NSPC behaviour, we next used the PLIN2-GFP reporter system to address whether asymmetric inheritance of LDs influences the time to next division. Asymmetrically dividing cells were followed individually until their next division using time-lapse microscopy (Fig. 4f). Intriguingly, asymmetric distribution of LDs led to a significant difference in the time to next division in the paired daughters, with cells that inherited more LDs dividing earlier than their sisters with less LDs (Fig. 4g and h, supplementary movie 2). Thus, these data show that asymmetric inheritance of LDs influences the proliferative behaviour of NSPCs and further confirms our finding that inheriting more LDs provides a proliferative advantage.

To better understand the differences between NSPCs with high and low amounts of LDs, we next performed scRNA-seq on sorted PLIN2-GFP NSPCs from the two categories used before, namely the 25% highest versus the 25% lowest PLIN2-GFP NSPCs. Cells were directly sorted into 384-well plates and a total of 304 cells passed all the subsequent quality controls. Low and high PLIN2-GFP NSPCs were dispersed among the 3 clusters determined by cell cycle phases, confirming that the number of LDs does not predict cell cycle phase (Supplementary Fig. 5a and b). As cell cycle was not a determining factor to cluster the two populations, it was regressed out for subsequent analyses (Supplementary Fig. 5c). Both low and high PLIN2-GFP NSPCs expressed comparable levels of NSPC markers such as fatty acid binding protein 7 (Fabp7), Vimentin (Vim), Nestin (Nes) and Hairy/Enhancer of Split 1 (Hes1), suggesting that both populations are indeed NSPCs (Supplementary Fig. 5d). Plin2 gene expression levels were increased in high PLIN2-GFP NSPCs (Supplementary Fig. 5e), which also was confirmed by qRT-PCR on bulk sorted cells (Supplementary Fig. 5f). This indicates that increased PLIN2-GFP protein levels are also reflected by a change in Plin2 mRNA on the gene expression level.

A small number of genes was significantly upregulated (9 genes) in the high vs low LD-containing NSPCs (Fig. 5a), such as insulin growth factor binding protein 2 (IGFBP2). Interestingly, IGFBP2 has been recently identified as a protein which is critically involved in promoting neural stem cell maintenance and proliferation³⁶, suggesting that the high PLIN2-GFP NSPCs might have an advantage over the low PLIN2-GFP in keeping their stem cell potential. In line with this, the transcription factor 4 (TCF4, also called E2-2) was among the few significantly downregulated genes (6 genes) in the high PLIN2-GFP NSPCs (Fig. 5a). TCF4 has been shown to be important in neurogenesis, and neuronal maturation during development³⁷. It also plays an important role in postnatal and adult neural stem cell differentiation³⁸. While overexpression of TCF4 leads to increased neuronal differentiation, downregulation keeps the NSPCs in a more radial glial-like state³⁸. Thus, the mild but significant transcriptional changes, with upregulation of a pro-stem cell gene (IGFBP2) and downregulation of a pro-differentiation gene (TCF4) in the PLIN2-GFP high NSPCs (Fig. 5a and b) suggest that the high PLIN2-GFP NSPCs are slightly more stem-like than the low PLIN2-GFP NSPCs. Both candidate genes were also confirmed to be altered by qRT-PCR on bulk-sorted low and high PLIN2-GFP NSPCs (Fig. 5c).

As LD accumulation likely reflects a certain metabolic state, we next performed functional metabolic measures using Seahorse technology to assess potential metabolic differences between low and high LD-containing NSPCs. Oxygen consumption rate (OCR), a measure for oxidative phosphorylation, and extracellular acidification rate (ECAR), a measure for glycolysis, were assessed in sorted low and high PLIN2-GFP NSPCs. Indeed, high PLIN2-GFP NSPCs had significantly higher baseline OCR (Fig. 5d) and ECAR (Fig. 5e) than low PLIN2-GFP NSPCs, indicating a general higher metabolic activity in high LD-containing NSPCs. Adding oligomycin, an ATPase inhibitor, allows the measurement of the capacity of cells to increase glycolysis. Adding FCCP, a proton gradient uncoupler, allows to assesses the maximal oxygen consumption. Both OCR and ECAR were significantly increased upon addition of oligomycin and FCCP in high PLIN2-GFP NSPCs compared to low PLIN2-GFP, showing that they have a higher metabolic capacity (Fig. 5f).

The differences in metabolic activity might come from differential FAO activity. We thus assessed the effect of the FAO inhibitor etomoxir on both low and high PLIN2-GFP NSPCs. Blocking FAO for 1 h resulted only in a mild decrease in OCR and a mild increase in ECAR in both NSPC populations (Supplementary Fig. 5g), indicating that FAO activity is not sufficient to explain the differences in the metabolic activity and capacity of low versus high PLIN2-GFP NSPCs.

To test whether manipulation of the LD content alters metabolic activity and capacity, we next sorted low and high PLIN2-GFP NSPCs and assessed their metabolic potential after overnight loading with a low dose of oleic acid. Indeed, such a lipid loading slightly increased the baseline OCR and the metabolic capacity of the low PLIN2-GFP NSPCs (Supplementary Fig. 5h and i). However, the effects of oleic acid loading were mild and did not reach statistical significance. Whether loading with a more complex lipid mixture would be able to increase metabolic activity and capacity in the low LD-containing NSPCs remains to be determined.

Taken together, while there is a clear metabolic difference between low and high LD-containing NSPCs, lipid availability and potential differences in FAO are not sufficient to explain these differences, suggesting that the underlying mechanisms are more complex.

Increased metabolic activity can be accompanied by increased generation of reactive oxygen species (ROS). To test whether the increased metabolic activity of high PLIN2-GFP NSPCs did generate more ROS, we used a fluorescent ROS sensor and analysed bulk populations of PLIN2-GFP NSPCs by flow cytometry (Fig. 6a, Supplementary Fig. 6a and b). When back-gating to the 25% highest and 25% lowest GFP NSPCs, a clear difference in ROS signal intensity was apparent, with the higher GFP NSPCs having significantly increased ROS levels compared to the lower GFP NSPCs (Fig. 6b and c). ROS can induce lipid peroxidation, which is toxic to cells. LDs in a Drosophila neural stem cell niche have recently been shown to have an antioxidant role, protecting lipids from lipid peroxidation²³. To assess whether high LD-containing NSPCs were protected from lipid peroxidation, we used the lipid peroxidation product 4-hydroxynonenal (HNE) as readout. HNE is stably added to proteins and can be revealed with an antibody. Sorted low and high PLIN2-GFP NSPCs had comparable HNE staining intensity (Fig. 6d and e), despite the significant difference in ROS levels between the two populations (Fig. 6b and c). This suggests that the higher numbers of LDs in high PLIN2-GFP NSPCs (Fig. 6e and f) might at least partially protect them from ROS induced lipid peroxidation.

To further understand the role of LDs for NSPC behaviour, we next manipulated their breakdown or build-up, by inhibiting or knocking out several key players (Fig. 7a), and assessed the effects on LD numbers and proliferation.

We first assessed if the breakdown of LDs is required for FAO. To do so, we used two doses of the Cpt1a inhibitor etomoxir. Surprisingly, blocking FAO led to a significant reduction in LDs (Fig. 7b and c). This might be due to an increased lysis of LDs, as no fatty acids are reaching the mitochondria due to the Cpt1a blockage. While the ratio of NSPCs in S/G2/M phase assessed by pH3 did not significantly change (Supplementary Fig. 7a), we observed a significant reduction in the total number of NSPCs (Fig. 7c). This finding suggests that blocking FAO and concomitantly reducing LDs might affect survival and/or proliferation of NSPCs, as has been previously shown^14,15. Further supporting a reduced survival or proliferation, the number of differentiated progeny after etomoxir treatment was strongly reduced (Supplementary Fig. 7b), while the number of neurons among the remaining cells was not affected (Supplementary Fig. 7c).

TAGs stored in LDs have to be broken down before they are available for use, for instance as an energy source or as membrane building blocks. Adipose triglyceride lipase (ATGL, also known as patatin-like phospholipase domain containing protein 2, PNPLA2) is a key lipase initiating the first step of TAG breakdown. To reduce the expression of this key lipase in proliferating NSPCs, we used a combination of shRNAs against Atgl. Indeed, in NSPCs expressing shRNAs against Atgl, LDs accumulated compared to NSPCs expressing a non-targeting control shRNA (Supplementary Fig. 7d and e). Proliferation was significantly lower after Atgl knockdown compared to control NSPCs (Fig. 7e and f). The same effect was seen when using the ATGL inhibitor Atglistatin: LDs accumulated in a dose-dependent manner (Fig. 7g and h), while proliferation was already significantly reduced with the lowest dose and was consistently lower in all doses tested (Fig. 7g and i). These data suggest that NSPCs need to access the TAGs in LDs through ATGL for proliferation. Interestingly, Atglistatin treatment during the first 2 days of differentiation initiation did not affect NSPC differentiation into neurons (Supplementary Fig. 7f, g and h), indicating that for differentiation initiation, access to LDs via ATGL is not critical. Whether they can access the LDs via lipophagy instead remains to be determined.

Lipids in LDs either originate from nutritional sources or from de novo lipogenesis. To assess the importance of de novo lipogenesis for LDs in NSPCs, we genetically ablated FASN in Fasn floxed NSPCs (Fasn KO) by virus mediated Cre-recombinase (Cre-GFP). Fasn KO NSPCs had significantly reduced LD numbers and total LD volume compared to control NSPCs (Fasn Ctrl) (Fig. 7j and k), suggesting that de novo lipogenesis is a major driver of LD formation in NSPCs. As previously reported¹², Fasn KO NSPCs were significantly less proliferative than Fasn Ctrl NSPCs, as assessed by the proliferation marker pH3 (Supplementary Fig. 7i and j). These data are in line with the proliferation differences seen between low and high LD-containing NSPCs (Fig. 4b and c). Thus, reduced build-up of LDs by inhibiting de novo lipogenesis is associated with the reduced NSPC proliferation.

Taken together, these manipulations show that FAO and LDs are interconnected, that LDs can be broken down by ATGL in proliferating NSPCs, and that LDs in NSPCs store de novo produced lipids. Inhibiting these pathways leads to changes in LD accumulation in NSPCs, accompanied by a decrease in proliferation and/or survival, emphasizing the important role of LDs for NSPCs.