Neural stem cell temporal patterning and brain tumour growth rely on oxidative phosphorylation

Translating advances in cancer research to clinical applications requires better insight into the metabolism of normal cells and tumour cells in vivo. Much effort has focused on understanding how glycolysis and oxidative phosphorylation (OxPhos) support proliferation, while their impact on other aspects of development and tumourigenesis remain largely unexplored. We found that inhibition of OxPhos in neural stem cells (NSCs) or tumours in the Drosophila brain not only decreases proliferation, but also affects many different aspects of stem cell behaviour. In NSCs, OxPhos dysfunction leads to a protracted G1/S-phase and results in delayed temporal patterning and reduced neuronal diversity. As a consequence, NSCs fail to undergo terminal differentiation, leading to prolonged neurogenesis into adulthood. Similarly, in brain tumours inhibition of OxPhos slows proliferation and prevents differentiation, resulting in reduced tumour heterogeneity. Thus, in vivo, highly proliferative stem cells and tumour cells require OxPhos for efficient growth and generation of diversity.


Introduction
The observation that some cancer cells rely primarily on aerobic glycolysis for energy and biomass production (the Warburg effect) (Vander Heiden et al., 2009;Warburg, 1956) has often led to the assumption that the other main source of ATP, mitochondrial oxidative phosphorylation (OxPhos), is dispensable. However, it is becoming increasingly clear that many tumours do require mitochondrial activity for energy and biosynthesis and OxPhos is now frequently exploited as a therapeutic target in cancer (Gui et al., 2016;Molina et al., 2018;Shi et al., 2019;Weinberg and Chandel, 2015). OxPhos takes place at the inner mitochondrial membrane in five large protein complexes (Complex I-V), which together form the respiratory chain. Complexes I-IV transfer electrons from NADH to O 2 and use the released energy to translocate protons from the mitochondrial matrix into the intermembrane space. The resulting electrochemical gradient is then used by Complex V (ATP synthase) to generate ATP from ADP. Apart from the production of ATP, OxPhos is also directly involved in the generation of NAD + , orotate, fumarate and reactive oxygen species (ROS) and thus affects many cellular processes, such as nucleotide synthesis (Birsoy et al., 2015;Sullivan et al., 2018;Sullivan et al., 2015), signalling pathway activity (Chandel, 2014) and epigenetic modifications (Lu and Thompson, 2012). The Warburg effect has since been interpreted as a normal adaptation to the metabolic requirements of proliferation, both in cancer cells and proliferating stem cells (Vander Heiden et al., 2009). High glycolytic flux is thought to be required for a constant supply of biomass while OxPhos, apart from its role in production of ATP, primarily maintains the cellular redox balance (Birsoy et al., 2015;Sullivan et al., 2015;Titov et al., 2016).
However, metabolic flux in cancer cells can be influenced by extrinsic and intrinsic factors such as substrate availability, oncogenic mutations and the tumour's tissue and cell type of origin (Hu et al., 2013;Mayers et al., 2016;Vander Heiden and DeBerardinis, 2017). Brain tumours in particular recapitulate many features of their tissue of origin and grow along a hierarchy reminiscent of normal brain development (Azzarelli et al., 2018;Genovese et al., 2018;Lan et al., 2017;Lee et al., 2018;Tiberi et al., 2014). An integrated understanding of the interactions between metabolism and cell identity in vivo, during both tumourigenesis and normal development, is therefore crucial to translate advances in cancer research to clinical applications.
Development of the Drosophila central nervous system (CNS) has been used extensively as a powerful reductionist model of human brain development and tumourigenesis in vivo (Brand and Livesey, 2011;Hakes and Brand, 2019;Villegas, 2019). The CNS of Drosophila develops from rapidly cycling embryonic and larval neural stem cells (NSCs) that generate a wide variety of neurons and glia. Neuronal diversity is achieved primarily by spatial and temporal patterning, which confers specific identities on NSCs and their progeny according to their location and developmental time (Miyares and Lee, 2019;Technau et al., 2006). Neural stem cells (NSCs) in Drosophila and mammals are thought to generate ATP through aerobic glycolysis rather than OxPhos, whereas their neuronal progeny switch to mitochondrial respiration upon differentiation (Agathocleous et al., 2012;Beckervordersandforth et al., 2017;Hall et al., 2012;Homem et al., 2014;Lange et al., 2016;Tennessen et al., 2014;Tennessen et al., 2011;Zheng et al., 2016). Upregulation of aerobic glycolysis, reminiscent of the Warburg effect, has also been described in a number of Drosophila tumour paradigms (Eichenlaub et al., 2018;Wang et al., 2016;Wong et al., 2019). However, the interpretation that mitochondrial respiration is dispensable for normal Drosophila NSCs (Homem et al., 2014) contrasts with the clear requirement for OxPhos to support cell cycle progression in the Drosophila eye disc (Mandal et al., 2010;Mandal et al., 2005;Owusu-Ansah et al., 2008). Here, we investigate whether, and to what extent, Drosophila NSCs and brain tumours rely on oxidative phosphorylation.

Results
OxPhos is required for brain tumour growth and heterogeneity We first examined whether OxPhos is required in tumours generated by loss of the transcription factor, Prospero (Pros) (Caussinus and Gonzalez, 2005;Choksi et al., 2006), in which differentiating daughter-cells revert to a NSC-like fate (Choksi et al., 2006) (Figure 1h). pros tumours are invasive upon transplantation and exhibit genomic instability over time (Caussinus and Gonzalez, 2005). We used RNAi to knock down subunits of complex I (NDUFS1) or complex V (ATPsyna) in NSCs and tumour cells with a NSC-specific driver, Worniu-GAL4 (Albertson et al., 2004). The complex I RNAi line has been validated previously (Garcia et al., 2017;Hermle et al., 2017;Owusu-Ansah et al., 2013;Pletcher et al., 2019); expression of the complex V RNAi in NSCs strongly reduced the levels of ATPsyna (Figure 1-figure supplement 2a-c). We also assessed mitochondrial morphology by stimulated emission-depletion (STED) super-resolution microscopy of mitochondria-targeted GFP (Rizzuto et al., 1995). Both RNAi lines caused fragmentation of mitochondria (Figure 1-figure supplement 2d-f), a known consequence of OxPhos dysfunction in mouse and human cells (Duvezin-Caubet et al., 2006).
To our surprise, inhibition of OxPhos through knockdown of mitochondrial complex I or V in pros tumours caused a decrease in tumour growth and an overall reduction in brain size (Figure 1a-c,i). This result was comparable to the effect observed upon inhibition of glycolysis with an RNAi against aldolase ( Figure 1-figure supplement 1a-c). This suggests that neither glycolysis nor OxPhos are sufficient to support brain tumour growth in vivo.
Next, we tested the requirement for OxPhos in different types of brain tumours. Constitutive activation of aPKC (aPKC-CAAX) leads to symmetric division of NSCs in the Drosophila brain (Lee et al., 2006) (Figure 1h), whereas loss of brat results in dedifferentiation of the progeny of type II NSCs (Bowman et al., 2008) (Figure 1h). In both aPKC-CAAX and brat tumours we found that knockdown of the complex I subunit, NDUFS1, strongly inhibited tumour growth and decreased overall brain size i). This was accompanied by a significant decrease in the mitotic index of tumourigenic NSCs (Figure 1j), consistent with mitochondrial metabolism playing a key role in regulating the proliferation rate of brain tumour cells. There was no obvious increase in apoptosis upon OxPhos inhibition in pros tumours, as assessed by TUNEL-staining ( Figure 1-figure supplement 1d-f).
Growth of pros mutant tumours is sustained by a small proportion of highly proliferative stem cells that express Imp (IGF-II mRNA-binding protein) (Genovese et al., 2018;Narbonne-Reveau et al., 2016). These tumour stem cells self-renew and generate more differentiated Impnegative tumour cells with limited self-renewal capacity. We assessed whether OxPhos inhibition promotes the differentiation of these Imp-positive stem cells towards Imp-negative tumour cells, Figure 1. Brain tumours require OxPhos for growth. (a-g) phospho Histone H3 (pH3) staining in the CNS of third instar larvae (L3) with NSC-specific expression (Wor-GAL4;Tub-GAL80 ts ) of control RNAi (a), Pros-RNAi (b,c), aPKC-CAAX (d,e) or Brat-RNAi (f, g), either without (b,d,f) or with (c,e,g) RNAi against a complex I subunit (NDUFS1). Maximum intensity projections through the entire CNS; dashed lines outline the CNS. (h) NSC lineages before and after tumourigenic transformation. (i,j) Brain size (i) and mitotic index of Dpn+ tumour cells (j) from L3 larvae expressing the indicated transgenes in NSCs. Datapoints indicate individual brains from one to four biological replicates. (k,l) Dpn (red, k',l') and Imp (green, k'',l'') immunostaining in Pros-RNAi tumours, without (k) or with (l) a complex I RNAi. Scale bars are 100 mm (a-g) or 10 mm (k,l). DOI: https://doi.org/10.7554/eLife.47887.002 The following figure supplements are available for figure 1: which could result in inhibition of tumour growth (Genovese et al., 2018). However, after knockdown of complex I by targeted RNAi, most tumourigenic NSCs in pros and aPKC-CAAX tumours remained Imp-positive and differentiation into Imp-negative cells was reduced (Figure 1k,l-figure supplement 1g). Our results suggest that OxPhos inhibition does not lead to more aggressive tumours, but rather slows it down by decreasing the proliferation rate of the tumour cells.

NSC proliferation depends on OxPhos
We found that, as for tumour cells, inhibition of OxPhos in NSCs throughout development resulted in smaller brains (Figure 1i; Figure 2a-d). This could not be explained by an overall developmental delay, as larval and pupal body length was similar to controls ( Figure 2-figure supplement 1). In contrast, inhibition of glycolysis by NSC-specific knockdown of phosphofructokinase (PFK), aldolase or phosphoglycerate kinase (PGK) had no effect on brain size and knockdown of pyruvate kinase (PyK) only caused a slight reduction (   To investigate whether RNAi-mediated OxPhos inhibition affects ATP production in NSCs, we measured ATP concentration in vivo using a genetically encoded ATP FRET sensor (Tsuyama et al., 2013). ATP concentration in NSCs in the L3 VNC was similar between controls and complex V knockdown (Figure 1-figure supplement 2g-i). Acute pharmacological inhibition of glycolysis through application of 2-deoxyglucose to ex vivo cultured brains caused a drop in ATP levels in both conditions. However, this drop was significantly more rapid and severe in NSCs with prior complex V inhibition (Figure 1-figure supplement 2g-i). This suggests that mitochondrial dysfunction results in rewiring of NSC metabolism to rely more on glycolysis for ATP production.

OxPhos is required for temporal patterning of NSCs and their progeny
In order to generate the diversity of neurons and glia within the CNS, NSCs undergo temporal patterning. This allows them to generate progeny with different identities according to their developmental time (Miyares and Lee, 2019). Drosophila NSCs in the larval VNC progress from an early identity marked by cytoplasmic Imp and nuclear Chinmo, to a late identity marked by cytoplasmic Syncrip (Syp) and nuclear Broad (Liu et al., 2015;Maurange et al., 2008) (Figure 3a). We found that inhibition of OxPhos caused a defect in temporal patterning of larval NSCs. After knockdown of complex I, one third of NSCs in the VNC (32.0 ± 4.0%, mean ± s.e.m., n = 11 VNCs) failed to downregulate Imp expression (Figure 3b-e-figure supplement 1a-c) and some (9.1 ± 1.4%, n = 14 VNCs) even failed to differentiate into Syp-positive NSCs at the end of larval life (Figure 3-figure supplement 1d-g). This is reminiscent of the failure to downregulate Imp and reduced differentiation in NSC-derived tumours. Immunostaining for other temporal markers (Maurange et al., 2008;Miyares and Lee, 2019) revealed a delay in the downregulation of the early temporal factors Castor and Chinmo, a decreased peak of expression of the switching factor Sevenup, and delayed upregulation of the late temporal factor Broad (Figure 3e-figure supplement 1h). Similar results were observed after knockdown of other subunits of complex I (NDUFA10, NDUFV1) or V (ATPsyna, ATPsyng) in the VNC (Figure 3d-figure supplement 1i-m), and after OxPhos inhibition in the central brain (CB) (Figure 3-figure supplement 1n-p). Importantly, this was accompanied by a significant reduction in the number of NSC progeny expressing Broad and lacking Chinmo, indicators of late neuronal identity (Figure 3f-h-figure supplement 2). We conclude that OxPhos is required for NSCs to progress from an early to a late temporal fate.
Temporal patterning of NSCs is regulated at the G 1 /S transition To test directly whether increasing cell cycle length inhibits NSC temporal progression, we slowed the cell cycle by expression of Myt1, Wee1 (Price et al., 2002) or both, which delay the G 2 /M transition ( Figure 4g) and strongly decrease final brain size ( Figure 4-figure supplement 1a). However, this did not affect NSC temporal progression and no Imp-positive NSCs could be detected at the end of neurogenesis (Figure 4d-f,h). Next, we tested whether inhibition of the G 1 /S transition affects temporal progression by expression of Dacapo (Dap; the p21/p27/p57 homologue), or an activated form of Rb (Rbf280). Strikingly, many NSCs in the VNC expressed Imp continuously (Dap: 7.8 ± 1.2%, n = 15 VNCs; Rb: 24.2 ± 3.4%, n = 10 VNCs) (Figure 4a-c,h). When Dap and Rbf280 were co-expressed, a majority of NSCs remained positive for the early NSC marker Chinmo (80.4 ± 1.1%, n = 4 VNCs) (Figure 4-figure supplement 1b,c). The block in temporal patterning correlated with the decrease in mitotic index (Figure 4h-figure supplement 1d,e). Our data suggest that temporal patterning and generation of neuronal diversity are linked to cell cycle progression and that regulation occurs at the G 1 /S rather than the G 2 /M transition.
Activation of the G 1 /S checkpoint upon downregulation of OxPhos activity has been observed in various tissues in Drosophila (DiGregorio et al., 2001;Mandal et al., 2005). In the eye disc, G 1 /S delay upon complex I dysfunction was caused by increased production of ROS and JNK-pathway activity, while complex IV dysfunction decreased the ATP/AMP ratio and activated the G 1 /S checkpoint through AMPK and p53 (Mandal et al., 2010;Mandal et al., 2005;Owusu-Ansah et al., 2008). Our preliminary data suggest that decreasing ROS does not rescue the proliferation or temporal patterning defects of complex I or V inhibition (data not shown) and nor does knock down of AMPK or p53 (data not shown). Moreover, clones mutant for ampk in a background where all NSCs continue to express complex I or V RNAi enhanced rather than suppressed the temporal patterning defect (Figure 4-figure supplement 2a-d). Therefore, it remains to be seen which pathway activates the G 1 /S checkpoint in NSCs with mitochondrial dysfunction.

OxPhos dysfunction and prolonged G 1 /S interfere with termination of proliferation
The adult CNS in Drosophila does not normally contain NSCs Siegrist et al., 2010;von Trotha et al., 2009). NSCs stop dividing in the first 20-30 hr after pupariation at which van den Ameele and Brand. eLife 2019;8:e47887. DOI: https://doi.org/10.7554/eLife.47887 time they differentiate or undergo apoptosis (Figure 3a) (Homem et al., 2014;Ito and Hotta, 1992;Maurange et al., 2008;Siegrist et al., 2010;Truman and Bate, 1988). It was previously shown that knocking down complex III or IV subunits in NSCs prevents termination of proliferation at the onset of pupal life (Homem et al., 2014). The authors suggested that pupariation is accompanied by a metabolic switch from glycolysis to OxPhos that results in NSC shrinkage and cell cycle  exit. Similarly, we found that when complex I or V subunits were knocked down, NSCs, identified by Dpn-expression and continued expression of GFP from a NSC-specific GAL4-driver (Worniu-GAL4), were maintained into the adult VNC and CB (Figure 5a-c-figure supplement 1a-d). Of the 133 NSCs in the larval VNC (Birkholz et al., 2015;Lacin and Truman, 2016), an average of 30.8 ± 3.1  and 20.2 ± 3.3 persisted into adulthood when complex I or V were inhibited respectively (Figure 5c). These NSCs continued to proliferate and generate neuronal progeny (Figure 5a-c-figure supplement 1a-j). NSCs also persisted in the adult CB and VNC when the G 1 /S, but not G 2 /M, transition was delayed, independent of OxPhos dysfunction (Figure 5d-g-figure supplement 1k-n).
Timely cell cycle exit of Drosophila NSCs at the end of neurogenesis was shown to depend on normal progression through the larval temporal cascade (Maurange et al., 2008;Yang et al., 2017). We therefore asked whether the defect in termination of proliferation caused by OxPhos inhibition could be due to delayed temporal patterning during larval life, as opposed to a metabolic switch at pupariation. To test this, we restored the temporal identity in NSCs in which complex I was downregulated by removing Imp at 48 hr or 72 hr ALH. Deletion of Imp significantly decreased adult neurogenesis (Figure 5h-k), consistent with a direct relationship between temporal patterning defects and the adult persistence of NSCs upon OxPhos dysfunction. Together, our data indicate that the previously observed defect in termination of NSC proliferation is a consequence of the earlier temporal patterning defects caused by OxPhos dysfunction.

Discussion
Significant progress has been made in identifying the signalling pathways and transcription factors that regulate stem cell transitions during brain development and homeostasis (Taverna et al., 2014;Tiberi et al., 2012). In contrast, our understanding of the metabolic changes that accompany, or drive, these transitions is still limited (Knobloch and Jessberger, 2017). Here we show that the metabolic requirements of highly proliferative NSCs in the Drosophila brain, as well as the tumour cells they generate upon transformation, cannot be met by aerobic glycolysis alone. Instead, Drosophila NSCs require OxPhos for key aspects of their behaviour: proliferation, generation of diversity through temporal patterning, and termination of proliferation ( Figure 6). Respiratory activity may provide an explanation for the strong increase in ROS production that has been observed in NSCs upon hypoxia (Bailey et al., 2015) and for the developmental lethality caused by CNS-specific mutation of the mitochondrial genome . While OxPhos dysfunction affects both normal NSCs and tumour cells in the brain, inhibition of glycolysis only affects tumour growth (Figure 1figure supplement 1) but not normal brain development (Figure 2). This is reminiscent of the upregulation of aerobic glycolysis in Hipk, EGFR or PDGF/VEGF-induced tumours in the Drosophila wing disc (Eichenlaub et al., 2018;Wang et al., 2016;Wong et al., 2019). Future experiments will determine the origin and consequences of this tumour-specific reliance on glycolysis in the brain.
Our results contrast with previous findings suggesting that OxPhos is dispensable during normal NSC development and in brain tumours, and is only activated at the end of neurogenesis as part of a metabolic switch to induce termination of NSC proliferation (Homem et al., 2014). While our experiments do not directly address whether this metabolic switch takes place, the results provide an alternate interpretation. We find that sustained OxPhos activity throughout NSC development is required for normal temporal patterning. Prolonged expression of early temporal markers makes NSCs unresponsive to the developmental cues that govern cell cycle exit (Maurange et al., 2008;Yang et al., 2017) and we show that restoring temporal progression by timely depletion of the early temporal factor Imp enhances termination of proliferation in spite of continued OxPhos inhibition. Our findings thus integrate key aspects of NSC and tumour cell biology ( Figure 6) : OxPhos-dependent proliferation is required for temporal patterning and differentiation at the G 1 /S transition of the cell cycle. This enables NSCs to undergo normal aging and to respond to the developmental cues that instruct termination of proliferation. Interestingly, adult neurogenesis in the subventricular zone of the mammalian brain depends on p57-induced slowing of the cell cycle during embryonic development (Furutachi et al., 2015). It is not known whether p57 expression or mitochondrial dysfunction also affects the temporal identity of mammalian NSCs. Importantly, the effects we observed are specific to the G 1 /S transition: activation of the G 2 /M checkpoint did not affect temporal patterning or termination of proliferation. Our results therefore demonstrate that the size and composition of Drosophila NSC lineages are not strictly predetermined (Birkholz et al., 2015) but rather controlled by both intrinsic and extrinsic factors. Single-cell sequencing data indicate that metabolic differences exist between NSCs in different regions of the brain or at different developmental stages (Davie et al., 2018;Genovese et al., 2018) and it will be interesting to assess whether all NSCs are similarly affected by OxPhos dysfunction and G 1 /S delay or whether specific lineages show van  stereotypical responses, as has been shown for entry into quiescence, where arrest in G 2 or G 0 is predetermined (Otsuki and Brand, 2018).
Our study indicates that OxPhos might constitute a targetable metabolic vulnerability of cancer. Small molecule inhibitors of OxPhos are currently being developed and tested in clinical trials to treat various forms of cancer (Gui et al., 2016;Molina et al., 2018;Shi et al., 2019;Weinberg and Chandel, 2015). However, we find that the in vivo impact of OxPhos dysfunction is much more Figure 6. Model of the role of OxPhos in Drosophila NSCs and tumour cells. We propose a model, whereby highly proliferative Drosophila NSCs also rely on OxPhos for most aspects of their behaviour. In particular, the G 1 /S transition depends on OxPhos activity and perturbation of this transition, either directly, or indirectly through OxPhos inhibition, results in delayed temporal patterning. This in turn prevents NSCs from terminating proliferation at the appropriate time, causing neurogenesis to persist into the adult. A similar dependence on OxPhos can be seen in brain tumours, where both proliferation and differentiation require mitochondrial activity, presumably through a similar mechanism to that found in normal NSCs. DOI: https://doi.org/10.7554/eLife.47887.018 van den Ameele and Brand. eLife 2019;8:e47887. DOI: https://doi.org/10.7554/eLife.47887 complex than mere inhibition of proliferation. A better understanding of the interactions between metabolism, differentiation and tumour heterogeneity in vivo has the potential to uncover novel therapeutic approaches.

Fly husbandry
Drosophila melanogaster were reared in cages at 25˚C. For most experiments, embryos were collected on food plates for 3 hr and transferred to 29˚C until analysis. Unless indicated otherwise, larvae were matched for developmental timing at wandering third instar (L3). For time-course experiments, embryos were collected on yeasted apple juice plates and larvae were transferred to a fresh yeasted food plate within 2 hr of hatching (designated 0 hr ALH) and grown at 25˚C until the desired stage. For clonal analysis, embryos and larvae were grown at 25˚C and heat-shocked when indicated for 20 min in a 37˚C water bath.

Imaging and image processing
Fluorescent images were acquired using a Leica SP8 confocal microscope and analysed using ImageJ. For the larval CNS, we imaged the thoracic segments of the VNC from the ventral side until the neuropil, or the ventral regions of the CB; for the adult CNS, the entire VNC or CB was imaged. All images are single sections, unless indicated otherwise. For live imaging, third instar larval brains were dissected at room temperature in Schneider's insect medium (Sigma S0146), mounted in Schneider's medium with 10% FBS on low 35 mm Ibitreat dishes (Ibidi 80136) and imaged on an inverted Leica SP8 confocal microscope at room temperature. Z-stacks of the ventral side of the thoracic VNC were made at the indicated intervals for 3 hr. For live in vivo ATP measurements with an ATP FRET sensor for Drosophila (Imamura et al., 2009;Tsuyama et al., 2013), confocal settings were as follows: 405 nm excitation and simultaneous detection at 445-490 nm (CFP) and 530-760 nm (FRET); 2-DG (Sigma D8375) was added to the medium to a final concentration of 200 mM. Ratios were calculated for mean FRET/CFP intensity per NSC. Stimulated emission depletion (STED) super-resolution imaging was performed on a custom STED microscope as described in Trovisco et al. (2016) with a 100x UPlanSApo 1.35 NA silicone oil immersion objective lens (Olympus, Japan) over a region of 20 mm 2 (1024 Â 1024 pixels). Images were processed using ImageJ. Timestamps were generated with a custom-built OverTime ImageJ plugin (Richard Butler). Figures were compiled in Adobe Illustrator.

Quantifications and statistical analysis
For quantification of NSCs, Dpn-or Mira-positive NSC on the ventral side of the thoracic VNC at the indicated stage were counted. For TUNEL-staining all TUNEL-positive cells were quantified throughout the entire thickness of the thoracic VNC. To quantify adult NSCs, all GFP-positive lineages were counted throughout the entire VNC or CB; in the control CB, GFP perdures until pharate adult stage in eight mushroom body lineages, which terminate proliferation only at the end of pupal life. To quantify adult NSCs upon Imp-mutation (Figure 5j), all Dpn-positive cells were counted in the VNC or CB. Mitotic index is the number of pH3-positive cells among Dpn-positive cells. For quantification of tumour mitotic index, over 200 Dpn-positive cells were quantified in each thoracic VNC. For brain size, the area of CNS maximum projections was measured. Graphs were generated in R or Excel. Box-and-whisker plots depict median, interquartile range (box) and 1.5IQR below and above the first and third quartiles respectively (whiskers). Bar graphs, line graphs and values in the text indicate mean ± s.e.m. Datapoints indicate the value of individual VNCs or CBs, apart from Figure 5-figure supplement 1j where datapoints depict individual clones. One biological replicate is defined as the result of one parental cross.
Statistical tests were performed in R. All datasets were first checked for normal distribution with a Shapiro-Wilk test, and then ANOVA was performed with a post-hoc Tukey test. When data were not normal distributed, Kruskal-Wallis test was performed with post-hoc Dunn test and Bonferroni adjustment for multiple comparisons. For time course experiments (Figure 3e-figure supplement  1h), the two conditions at individual time-points were compared with a two-sided Mann-Whitney U test. ATP measurements (AT1.03-NL) were normalised for each NSC to t = 0 when 2-DG was added to the medium, and to the mean values from VNCs that expressed an ATP-insensitive sensor (AT1.03-RK) and were imaged in the same experiment. Modelling of the dynamics of ATP levels was done in R, based on the assumption of exponential decay. Significance is shown compared to control samples, unless indicated otherwise, with the following symbols: *p<0.05; **p<0.01; ***p<0.001; n.s. p!0.05.