Aneuploidy causes premature differentiation of neural and intestinal stem cells

Aneuploidy is associated with a variety of diseases such as cancer and microcephaly. Although many studies have addressed the consequences of a non-euploid genome in cells, little is known about their overall consequences in tissue and organism development. Here we use two different mutant conditions to address the consequences of aneuploidy during tissue development and homeostasis in Drosophila. We show that aneuploidy causes brain size reduction due to a decrease in the number of proliferative neural stem cells (NSCs), but not through apoptosis. Instead, aneuploid NSCs present an extended G1 phase, which leads to cell cycle exit and premature differentiation. Moreover, we show that this response to aneuploidy is also present in adult intestinal stem cells but not in the wing disc. Our work highlights a neural and intestine stem cell-specific response to aneuploidy, which prevents their proliferation and expansion.

D uring development the precise coordination between cell proliferation, death and differentiation ensures the assembly of functional tissues and organs of the correct size. Regulation of cell cycle length, number and outcome of stem cell divisions allows the generation of cell lineages in a spatial and temporal manner. Central to this control is the maintenance of a diploid genome as genome imbalance can result in decreased cell fitness, impaired viability and lead to developmental disorders or ultimately to death 1 . A current challenge in the developmental biology field is to understand how rapidly dividing cells maintain a diploid genome and coordinate different types of cell cycle parameters that sustains stem cell self-renewal, cell proliferation and differentiation within a developing organism 2 .
Aneuploidy, the gain or loss of whole chromosomes, resulting from errors occurring during mitosis has been recognized as a major feature of human cancers since the pioneer work of Boveri 3 . However, aneuploid conditions can lead to other disorders such as growth defects and mental retardation: Down and Turner syndromes 4,5 , mosaic variegated aneuploidy (MVA) 6 and microcephaly 7 . This raises the interesting question of whether aneuploidy and its consequences can predispose to different outcomes depending on cell type, developmental timing and age.
The Drosophila central nervous system is an excellent genetically tractable system to study the consequences of aneuploidy 8,9 . The central brain (CB) region contains neural stem cells (NSCs), also known as neuroblasts (Nbs) of embryonic origin that re-enter the cell cycle after a quiescence period during early larval stages 10 . Neuroblasts divide asymmetrically to selfrenew and to produce a ganglion mother cell (GMC) which will divide once more before differentiating into neurons or glia. Asymmetric cell division and the generation of two daughter cells with distinct cell fates rely on the differential segregation of polarity and cell fate determinants coupled to correct spindle position along the polarity axis during metaphase 9 .
In flies, defects in centrosome biogenesis cause spindle mispositioning and tumour formation in transplantation assays 11,12 , while aneuploid mutations do not 11 . Aneuploid mutants die at the end of larval stages, showing that accumulation of aneuploidy is not compatible with metamorphosis and adult life 8 . In contrast to the observations made in the brain, aneuploidy in other proliferative tissue, such as the wing disc was found to be a tumour-initiating event 13,14 . In mice, deregulation of the levels of checkpoint proteins caused tumours in a tissue-dependent manner 15,16 . Aneuploid mice displayed higher incidence of lymphomas and lung tumours but lower frequency of chemically induced tumours when compared with controls. It is therefore essential to understand the reasons why aneuploidy in certain tissues is permissive to tumour initiation, while in others inhibits tumour formation.
Here we use the Drosophila brain to investigate the consequences of aneuploidy in brain homeostasis and the outcome of combining aneuploidy with a tumour-permissive condition, centrosome amplification. We show that aneuploidy decreases the tumourigenic potential of NSCs. In addition, we found that aneuploid NSCs do not die by apoptosis. Instead, aneuploid NSCs display G1 lengthening and undergo premature differentiation. Further, we show that adult intestine stem cells (ISCs) present the same type of response. Our work identifies an outcome of aneuploid NSCs and adult ISCs, which inhibits the proliferation and accumulation of abnormal karyotypes in two proliferative tissues.

Results
Generating non-euploid cells in fly brains. To characterize the outcome of aneuploid NSCs during brain development in Drosophila, we used a previously described mutant SakOE,mad2 where the centriole kinase Sak, the Plk4 fly ortholog, is overexpressed (SakOE) leading to centrosome amplification 12 in the absence of the checkpoint protein Mad2 (ref. 17). SakOE thirdinstar larval (L3) NSCs (called neuroblasts-Nbs) of the CB possess extremely efficient clustering mechanisms of extra centrosomes and generate low aneuploidy levels 12 . mad2 mutants are viable and fertile since reduction of the mitotic timing does not affect cell division in flies 17 .
SakOE,mad2 mutants die at pupal stage and at larval stages present tissue size defects such as smaller imaginal discs, similarly to other aneuploid mutants 8 . Analysis of mid L3 SakOE,mad2 brains revealed a drastic reduction of both the neuroepithelium and CB regions, which appeared to be very disorganized (Fig. 1a).
To gain information on the mitotic defects generated in the presence of extra centrosomes and lack of spindle assembly checkpoint (SAC), we followed, by time-lapse spinning disc microscopy, mitosis of L3 Nbs of the CB expressing Tubulin-GFP and Histone 2B (H2B)-RFP. All wild-type (WT) Nbs contained two centrosomes and divide in a bipolar way, generating two cells of different size and fate 9,10 ( Fig. 1b and Supplementary Movie 1). SakOE Nbs that contained extra centrosomes also divided in a bipolar way, as observed previously 12   and SakOE,mad2 (right) brain lobes. In the WT lobe both the CB, highlighted by the yellow dashed line, and optic lobe (OL), highlighted by the red dashed line appear highly organized, while SakOE,mad2 lobes appear smaller and disorganized. Scale bar, 50 mm. (b-g) Stills of time-lapse movies of mitotic neuroblasts (Nbs) expressing Histone 2B-RFP and Tubulin-GFP, in red and in green, respectively. (b) Wild-type Nb with two centrosomes forms a bipolar spindle and divides asymmetrically to give rise to two cells. (c) SakOE Nb with at least five centrosomes that form a bipolar spindle due to centrosome clustering and inactivation generating two daughter cells. (d) SakOE,mad2 Nb with at least ten centrosomes and increased chromosome number. Not all centrosomes cluster and the cell undergoes a tripolar division. (e) SakOE,mad2 Nb with increased chromosome number and at least 15 centrosomes that cluster to form a bipolar spindle. Lagging chromosomes are noticed during anaphase. (f) SakOE,mad2 Nb divides in a bipolar way, but presents defects in cytokinesis. (g) SakOE, bubR1* (bubR1 DKEN in bubR1 mutant background) Nb divides in a bipolar way but shows extra lagging chromatids during anaphase. Scale bar, 3 mm. (h) Graph bars showing the quantification of mitotic defects in WT (n ¼ 110), SakOE,mad2 (n ¼ 166) and SakOE, bubR1* (n ¼ 64) Nbs, considering spindle morphology (left), chromosome segregation defects (middle) and cytokinesis defects (right). Statistical significance (SS) was determined using Fisher's exact test. ns, not significant, ***(Po0.0001), **(P ¼ 0.0061). (i) Fluorescent in situ hybridization with chromosomes II and III probes (green and red) in WT, SakOE, and SakOE,mad2 Nbs. Scale bar, 2 mm. (j) Graph bars showing the quantification of FISH in WT (n ¼ 227 cells), mad2 (n ¼ 166 cells), SakOE (n ¼ 153 cells) and SakOE,mad2 (n ¼ 109 cells) brains. (k) Dot plot chart showing the time spent in mitosis measured as the time elapsed between nuclear envelope breakdown and anaphase onset. Each dot represents a cell. Time is in minutes and the line represents the mean and the error bars the s.d. Statistical significance was assessed by unpaired t-test. *(0.01oPo0.05), **(0.001oPo0.01) and ***(Po0.001).
Movie 4), suggesting that chromosomes were mis-segregated even in clustered bipolar divisions, similarly to what has been described in tissue culture cells 18,19 . Failure in cytokinesis was observed in 17% (n ¼ 35 out of 196) of Nbs (Fig. 1f,h and Supplementary Movie 5). Cytokinesis defects were observed both in cells that divide bipolarly (10%) and multipolarly (7%) and frequently this correlated with the presence of lagging chromatids at the cytokinesis furrow. Importantly, in WT, SakOE or mad2 Nbs, these defects were never observed.
Using fluorescence in situ hybridization (FISH) we found that 19.1% (n ¼ 109) of SakOE,mad2 Nbs, were aneuploid (1.8% with loss of one chromosome, 11.8% with gain of one chromosome and 5.5% with gain or loss of two or more than two chromosomes), while 1.8% were polyploid (Fig. 1i,j Chromosome II Chromosome III * *** low levels of chromosome gain and losses were also noticed in WT, mad2 and SakOE brains (3.8%, 3.6% and 5.9%, respectively), these Nbs did not show polyploidy or gain/loss of more than one chromosome (Fig. 1i,j).
It has been shown recently that chromosome mis-segregation causes structural DNA aberrations 20,21 . We analysed DNA damage in SakOE,mad2 brains measured by damaged dependent phosphorylation of the Histone variant H2Av (g-H2Av). As control of g-H2Av antibody, we incubated the brains with the DNA synthesis inhibitor, aphidicolin to induce replication stress and consequently accumulate double-strand breaks. Incubation of WT brains with aphidicolin for 30 min was sufficient to detect an increase in g-H2Av-positive cells in the CB ( Supplementary Fig. 1A,B); however, untreated WT and SakOE,mad2 brains showed comparable levels of g-H2Avpositive cells. We conclude that chromosome mis-segregation in SakOE,mad2 brains does not result in DNA damage.
The SAC guarantees mitotic fidelity by delaying anaphase onset until all chromosomes are correctly attached to microtubules 22 . Extra centrosomes are known to cause mitotic lengthening, which can serve as an advantage to cells that contain extra centrosomes 12,23 . To understand why some spindles manage to be bipolar while others maintain a multipolar status, we decided to separate mitotic timing from checkpoint function 24 . To this end, we used a mutant version of BubR1 (DBubR1 DKen in the bubR1 mutant background, that we will refer to as BubR1*), which has been shown to perturb BubR1 checkpoint activity while it did not affect mitotic timing 25 . SakOE, BubR1* mitotic Nbs analysis revealed that, even if lagging chromosomes (28%) and cytokinesis defects (8%) were identified in bipolar divisions, we never observed multipolar divisions (n ¼ 64) (Fig. 1g,h and Supplementary Movie 6). We quantified the time in mitosis as the time elapsed between nuclear envelope breakdown and anaphase onset. WT Nbs spent on average 7.9±0.2 min in mitosis, while mad2 and BubR1 Nbs spent 7.2 ± 0.2 and 7.5 ± 0.1 min in mitosis. SakOE Nbs underwent slower mitosis, consistent with previous findings (9.4 ± 0.3 min) 12 . Mitotic timing was also increased in SakOE, BubR1* (9.8 ± 0.5 min) and in bipolarly dividing SakOE,mad2 Nbs (9.2 ± 0.3 min). In contrast, SakOE,mad2 Nbs showing multipolar divisions spent on average less time in mitosis (7.4 ± 0.4 min) (Fig. 1k). These results show a correlation between mitotic timing and bipolar status in cells that contain extra centrosomes. As shown in other cell types 23,26 , mitotic timing but not the activity of the SAC per se contributes to bipolar spindle formation in the presence of extra centrosomes.
Aneuploid brains contain fewer neuroblasts. SakOE,mad2 brains appeared smaller than WT brains (Fig. 1a) and we investigated the CB cell population. At early L3, previously quiescent type I Nbs re-enter the cell cycle and divide asymmetrically several times to self-renew 9,10,27 (Fig. 2a) generating around 100 daughter cells 28,29 . SakOE,mad2 brains contained fewer Nbs positive for the Deadpan (Dpn) (Dpn þ ) marker when compared with WT brains (Fig. 2b,c). We then extended our analysis to the previously characterized aneuploid mutant with normal centrosome content, bub3 (refs 30,31,32). We used bub3 mutants because they die at the same developmental stage as SakOE,mad2 and bub3 Nbs display similar levels of aneuploidy although we did not detect polyploidy (Supplementary Fig. 2A,B). We characterized bub3 mitoses and found that all mitoses were bipolar (see below) and that frequently lagging chromosomes could be seen, as described before 31 . Similarly to SakOE,mad2 brains, we did not detect an increase in g-H2Av-positive nuclei ( Supplementary Fig. 2C,D). Importantly, we found that just like in SakOE,mad2 brains, the number of Nbs was reduced in bub3 mutant brains (Fig. 2b,c).
Since apoptosis is a common fate of aneuploidy cells 7,13,14,33 , we ascertained whether reduced number of Nbs was due to apoptosis. We investigated the presence of cleaved caspase 3-positive cells (CC3 þ ) in the CB. We found that the number of CC3 þ Nbs did not increase significantly in both SakOE,mad2 and bub3 CBs (Fig. 2d,e); however, it was detected in mutant optic lobes (OL) (Fig. 2d), similarly to what has been recently shown in asp mutant OL 34 . Interestingly, these results show that two different regions of the brain respond differently to aneuploidy. Importantly, suppression of apoptosis, using the pan-caspase p35 inhibitor did not rescue the decrease in Nb number of aneuploid brains (Fig. 2d-f). We controlled the efficiency of p35 in inhibiting apoptosis in the posterior compartment of bub3 mutant wing discs, using the engrailed-Gal4 (eng-Gal4) promoter. As described previously, suppression of apoptosis by expression of p35 in bub3 mutant wing discs resulted in highly abnormal discs ( Supplementary Fig. 3) due to the accumulation of aneuploid cells 13,14,32 . We conclude that the reduced Nb number found in aneuploid mutant brains cannot be explained by apoptosis. Certain brain Nb lineages do not seem to undergo apoptosis at least during larval stages 35 , but apoptosis is used to inhibit the proliferation of other lineages such as Mushroom body Nbs during pupal stages 36 . To test whether larval Nbs can undergo apoptosis in response to other insults, we treated WT brains with aphidicolin, to induce DNA damage ( Supplementary Fig. 1A,B). We observed CC3 þ Nbs, ( Supplementary Fig. 4), showing that even if Nbs can undergo apoptosis in response to DNA damage, aneuploidy does not trigger this cell death response.
We next investigated whether another type of cell death, necrosis, was responsible for Nb loss seen in aneuploidy mutants. Recently, the Fzy protein, the Cdc20 APC/C activator homologue, has been implicated in necrosis 37 . Interestingly, mutations in fzy, also lead to premature Nb loss and fzy Nbs accumulated several stress markers, showed loss of membrane integrity and upregulation of p53 (ref. 37). We first analysed Fzy levels by western blot in SakOE,mad2 and bub3 brains, and found that these were comparable to WT brains ( Supplementary Fig. 5A). We then analysed stress markers such as the presence of ubiquitin-conjugated protein aggregates, mitochondria aggregation and ROS accumulation. Aneuploid brains showed an increase in the percentage of cells positive for ubiquitinconjugated protein aggregates, when compared with WT brains (Supplementary Fig. 5B,C) (5.1% in SakOE,mad2, 22.5% in bub3 and 2.0% in WT), however not to the same extent as fzy mutants (40%) 37 . In contrast to fzy mutants, however, mitochondria aggregation or accumulation of ROS in aneuploid brains, were comparable to WT brains (Supplementary Fig. 5D-G). In addition, we never observed karyolysis or loss of membrane integrity ( Supplementary Fig. 5H) as reported in fzy mutants and in other cells undergoing necrosis 38 . We next analysed the contribution of p53 to Nb loss in aneuploid brains using a p53 GFP reporter (p53RGFP) 39 . We identified on average eight GFP-positive cells in SakOE,mad2 CB, while WT brains did not show p53 recruitment. However, in the large majority of the cases, the cells that appeared positive were not Nbs (Dpn À ), but rather GMCs, neurons (Pros þ or Elav þ , respectively) or lacked any of these markers (Supplementary Fig. 6A-C). Further, inhibition of p53 activity, either using a dominant negative version (R155H) or p53 RNAi40 exclusively in Nbs, using a specific Nb driver, did not rescue the Nb loss phenotype (Supplementary Fig. 6D-G). All together, these results show that reduction of the number of CB Nbs in aneuploid brains is not explained by the same mechanism as in fzy mutants. In agreement, fzy mutants did not show chromosome mis-segregation defects 37 . Aneuploid brains undergo premature differentiation. Since loss of Nbs in aneuploid brains is not caused by apoptosis or necrosis, we then tested whether these brains were undergoing premature differentiation. We stained L3 brains with Dpn, Prospero (Pros) and Elav antibodies to label Nbs, GMCs and neurons, respectively. Interestingly, the ratio between GMCs and Nbs was increased in SakOE,mad2 and bub3 brains when compared with WT ( Fig. 3a,b, and Supplementary Table 1) in staged mid L3 larvae. In addition, the density of neurons (Elav þ cells per CB area) was also increased in aneuploid mutants when compared with WT (Fig. 3c, Supplementary Fig. 7 and Supplementary Table 2).
We then performed a clonal analysis in mid L3 brains to ascertain self-renew capacity and Nb progeny of WT and bub3 mutant clones. Using the flip-out strategy 41 , clones were induced in staged L2 larvae by heat shock. After 48 h, brains were fixed and analysed. WT clones occupied a larger area (Fig. 3d,e)  bub3 aneuploid clones contained B25±1.5 cells (n ¼ 33 clones from eight brains) (Fig. 3f). Occasionally clones without any large Dpn þ cell (Fig. 3d) were also seen.
Aneuploid brains display a G1 lengthening. Accumulating evidence showed a correlation between the length of the cell cycle and proliferation/differentiation capacity. Indeed, some stem cells ARTICLE undergo shorter cell cycles than their differentiating progeny [42][43][44] . Moreover, cell fate decisions have been postulated to be dependent on cell cycle progression 45 and G1 lengthening is sufficient to promote neurogenesis of mouse neuronal progenitors [46][47][48] . Aneuploid yeast strains showed an extended G1 phase 49 and we decided to determine if this was also the case in aneuploid mutant brains. We performed 2 h 5-ethynyl-2'-deoxyuridine (EdU) incorporation assays and manually counted the number of cells in S-phase (EdU positive cells -EdU þ ), in mitosis (PH3 positive cells-Ph3 þ ) and all the nuclei. In addition, we also took into account the number of neurons (Elav þ ) to distinguish between cycling cells in G1 (EdU À , PH3 À , Elav À ) and terminally differentiated cells (EdU À , PH3 À , Elav þ ) (Fig. 4a). Cells positive for EdU and PH3 were considered in the G2/M category, as cells that had during the 2 h EdU incubation period exit S-phase and progressed to mitosis. In WT brains, several EdU þ and PH3 þ nuclei could be noticed (Fig. 4a,b) and our analysis shows that 53% of the cells were in G1 (n ¼ 4830 cells from eight brain lobes Fig. 4b). In SakOE,mad2 and bub3 brains, however, the number of EdU þ and PH3 þ cells was severely reduced (Fig. 4a,b). Importantly, more cells were arrested in G1, 84% (n ¼ 3220 cells from 10 lobes) in SakOE,mad2 brains and 88% (n ¼ 2973 cells from eight lobes) in bub3 (Fig. 4b).
These results are in agreement with our live imaging observations of mid L3 brains. While in WT brains we can frequently follow consecutive Nb cell divisions due to the fast cell cycle of these cells (19 out of 93 Nbs re-enter mitosis), this event was very rare in SakOE,mad2 brains (9 out of 127), suggesting that interphase was prolonged in aneuploid brains.
The Cdk inhibitor p27/dacapo is associated with terminal divisions and differentiation 51 . We investigated the role of Dacapo in premature differentiation, but found no consequence when dacapo levels were decreased ( Supplementary  Fig. 8B). We also tested the involvement of the p38 stress kinase, a known inhibitor of proliferation of aneuploid cells 33 , but found no effect (Supplementary Fig. 8A).
Aneuploid neuroblasts differentiate prematurely. Larval Nb size is progressively reduced at the end of larval stages throughout pupation 9,28,52 . We reasoned that if aneuploid Nbs were undergoing premature differentiation, the committed progenitors and resulting neurons might show increased nuclear size. In WT Nbs, the transcription factor Pros is sequestered by the adaptor Miranda (Mira), which is normally cortically localized. During anaphase, Mira is segregated to the daughter GMC, carrying Pros. Mira is subsequently degraded and Pros translocates to the nucleus to initiate its transcriptional dependent response 53 . In WT Nbs, Pros was never seen associated with the Nb nucleus (Fig. 5a,b) and was localized to the small GMC nucleus. In SakOE,mad2 and bub3 mutant brains; however, 5.6 ± 2.2% (n ¼ 360 Nbs from 16 lobes) and 15.3 ± 4.1% (103 Nbs from 11 lobes) of Nbs showed co-localization of Dpn and Pros (Fig. 5a,b), respectively. In addition large Pros þ nuclei could also be noticed, which were never seen in the WT CB (Fig. 5a,b). In WT GMCs Mira degradation allows Pros release into the nucleus 53 . We then investigated whether abnormal Pros nuclear localization, seen in Dpn þ Nbs from aneuploid brains, correlated with alteration in Mira levels. Even if 30.5% (n ¼ 42 Nbs from 10 lobes) of Dpn þ ,Pros þ Nbs showed decreased Mira levels in bub3 mutant brains, we also observed 69.5% of Dpn þ ,Pros þ Nbs where Mira levels were comparable to WT brains ( Supplementary  Fig. 9A,B). We concluded that if Mira degradation might contribute to Pros nuclear accumulation in the nucleus, other Mira independent mechanisms also participate in this process.
We then tested the role of Pros in premature differentiation as this transcription factor represses genes required for self-renewal while it activates genes necessary for differentiation 29,54 . We knocked down Pros exclusively in Nbs, using the Worniu-Gal4 driver. Since GMC maintenance requires Pros 54 to distinguish between the consequences of Pros depletion in Nbs and GMCs, we took into account only Nbs with nuclear diameter superior to 8 mm as described previously 55 . Analysis of mid L3 brains, showed an increase in the number of CB Nbs (Fig. 5f,g). In addition, we also tested whether decreasing Pros levels would also result in similar rescue. For this purpose we analysed bub3 mutant or bub3 RNAi brains in the pros heterozygous background. As before, a significant (using a Fischer's exact test) increase in the number of Dpn þ cells displaying 48 mm diameter, were detected ( Supplementary Fig. 9C-F). Our results suggest that the premature differentiation seen in aneuploid brains recapitulates cell cycle exit and the normal differentiation of late L3 WT brains in a Pros-dependent manner.
Aneuploidy in other tissues. To determine the fate of aneuploid cells in other tissues than the brain, we analysed aneuploid wing discs, which are highly proliferative non-stem epithelial cells and aneuploid intestinal stem cells of the adult midgut (ISCs). In contrast to CB Nbs, SakOE,mad2 wing disc cells showed similar cell cycle parameters when compared with WT wing disc cells ( Supplementary Fig. 10A,B). Furthermore, these discs showed high levels of apoptosis, in agreement with published results 13,14 .
We then analyse adult ISCs using the Mosaic analysis with a repressible cell marker (MARCM) technique to induce expression of bub3 RNAi . ISCs divide asymmetrically to self-renew and to   generate one enteroblast (EB). EBs differentiate either into enteroendocrine cells (EEs) or enterocytes (ECs) that undergo polyploidization (Fig. 6a). First, to ascertain that bub3 RNAi triggers aneuploidy in ISCs, we used the temporal and regional gene expression targeting (TARGET) method (EsgGal44GFP, tubGal80 ts ) to express bub3 RNAi in adult ISCs and progenitor cells by shifting the temperature during 12 consecutive days to degrade the Gal4 inhibitor, tubGAL80. FISH revealed that 14.2% of bub3 RNAi Esg þ cells were aneuploid with only 4.3% in WT (Fig. 6b,c).
Clonal analysis revealed a significant reduction of the size of bub3 RNAi clones, which contained on average 5.4 cells per clone (n ¼ 45 clones from six guts), as compared with WT clones having 27 cells (n ¼ 31 clones from eight guts Fig. 6d,e). ARTICLE Furthermore, bub3 RNAi clones presented less ISCs (Delta-Dl þ ) per clone and even clones without ISCs (Fig. 6f,g). In addition, analysis of single cell clones resulting from MARCM labelling of differentiating daughter cells demonstrated that these were more abundant in bub3 RNAi conditions than WT (bub3 ¼ 12.7% compared with WT ¼ 3.1%; Fig. 6h). Thus, bub3 RNAi limits the ability of ISCs to produce progeny, promotes ISC loss and results in more single differentiated cells.
To further assess the cause of reduced clone growth, we inactivated bub3 in all Esg þ cells (ISCs and progenitor cells). In WT, EsgGAL44GFP marks diploid ISCs and progenitor cells. Upon bub3 RNAi expression, large polyploid nuclei could be seen marked by GFP, which was particularly evident in the anteriormost part of the posterior midgut (Fig. 6i). Although only 7.5% of WT guts presented these cells, 57.4% of esgGAL44bub3 RNAi guts showed this phenotype (Fig. 6l). Many of the large polyploid nuclei were positive for the differentiated enterocyte marker Pdm1 suggestive of ISC terminal differentiation (Fig. 6k). In addition, some of the large polyploid nuclei were positive for the ISC marker, Dl, suggestive of polyploidization of aneuploid ISCs (Fig. 6j).
Together these data suggest that bub3 aneuploidy in ISCs leads to a reduction in the ability of ISCs to produce progeny, which is at least in part due to ISC differentiation and polyploidization.
SakOE,mad2 brains display decreased tumourigenic potential. Aneuploidy can be seen as tumour suppressor or oncogenic 16,56 . SakOE brains induced tumour formation, due to spindle positioning defects, which resulted in the expansion of the Nbs pool 12 . We wondered whether the addition of aneuploidy to a tumour-permissive condition would influence the tumourigenic capacity of SakOE Nbs by transplanting SakOE,mad2 brain pieces into the abdomen of WT hosts. Interestingly, we found a clear decrease in the tumourigenic potential of SakOE,mad2 brains (Fig. 7a,b). Nevertheless, these brains were still tumourigenic, while other aneuploid mutants such as bub3 (ref. 11) were not.
Since brains that contain Nbs that display asymmetric cell division defects were found to be tumourigenic in transplantation assays 11,57 , we reasoned that SakOE,mad2 Nbs might maintain the tumourigenic potential because of defects in spindle positioning and asymmetric cell division. We analysed and compared spindle positioning relative to the apical aPKC crescent in SakOE,mad2 and bub3 brains. As expected, SakOE,mad2 Nbs presented defects in spindle positioning in a significant (using a student T-test) proportion of cells, while in bub3 Nbs, which contained always two centrosomes, spindles were correctly positioned (Fig. 7c,d).
Together, we conclude that in the fly brain, aneuploidy decreases NSCs tumourigenic potential.

Discussion
Maintenance of an euploid genome in the stem cell compartment is essential to control stem cell self-renewal and to produce genetically stable daughter cells. Cells and organisms have therefore developed surveillance mechanisms that detect the presence of abnormal karyotypes that are normally eliminated by apoptosis 13,14,33 . In this study we have found a novel response of Drosophila NSCs or ISCs to aneuploidy, which prevent cells with abnormal genomes from cycling. We show that these cells undergo premature differentiation. Importantly, this response is not exclusive to Drosophila. Premature differentiation of NSCs was observed in the mouse brain, but only when apoptosis was inhibited 7 . Thus, while in Drosophila premature differentiation represents a primary response, in the mouse brain this might represent a backup mechanism when the mechanisms that regulate cell death are not efficient.
One challenge in the future will be to identify the sensor(s) and effector(s) present in NSCs and ISCs that influences the shift towards differentiation. Several studies have recently shown that cell cycle timing dictates cell fate decisions. A functional organ of the correct size requires a tight temporal control of cell cycle progression and differentiation. Reduced proliferation and cell death in the CNS leads to the formation of microcephalic brains 7,58 (and our results), whereas over proliferation and/or lack of differentiation might lead to tumour formation. Neurogenic divisions of neural progenitors are accompanied by G1 lengthening prior to differentiation in the developing mouse 47 and Xenopus 59 nervous systems. This is also the case in both SakOE,mad2 and bub3 mutant NSCs that presented a significant delay in the G1 phase, due to aneuploidy.
Aneuploidy has been proposed to be at the origin of a general stress response in cells, mediated by p53-dependent mechanisms. Moreover, extension of G1 due to prolonged prometaphase arrest p38 dependent has also been implicated in eliminating proliferative cells containing spindle abnormalities 60 . Nonetheless, we ruled out the possibility that aneuploid fly NSCs undergo a p38-mediated arrest, as lowering p38 levels did not restore proliferation of aneuploid NSCs. Moreover, even if p53 was upregulated in aneuploid brains, Nbs did not show frequently p53 increased levels. Instead other brain cell types, such as GMCs, neurons and other cell types, probably glia, showed an increase in p53. It is thus possible that p53 upregulation in the brain reflects a stress response that unlike in other cell types, does not culminate with cell death by apoptosis.
Therefore other mechanisms induced during or in response to G1 lengthening seem to be responsible for premature differentiation in fly Nbs. A likely possibility is that during this abnormal G1 lengthening the transcription factor Pros, which promotes differentiation and inhibits proliferation, translocates to the nucleus to promote differentiation. In agreement, Pros co-localizes with Dpn in Nbs from aneuploid brains and reduction of Pros levels results in increased Nb number. Through binding to its adapter Mira, Pros is targeted to the basal cortex and inherited by the GMC and upon separation from Mira, Pros translocates to the GMC nucleus 9,53 . Importantly, Mira degradation is not required for Pros release 53 . Therefore, it is possible that an extended G1 phase allows for Pros release, which would be sufficient to trigger differentiation of Nbs.
Our work extends the list of adverse situations to aneuploidy and might reflect a common response of stem cells to hazardous situations to inhibit the proliferation of damaged genomes.
Clonal expression of bub3 RNAi was achieved using a flip-out strategy 41 , using the Tub-FRT-GAL80-FRT-GAL4 transgene. Clones were induced on second-instar larvae by heat shock (40 min 37°C) and mid third-instar larval brains were analysed. In the analysis of intestines, only female flies were analysed. MARCM clones were generated as previously described 63 . To measure cell cycle parameters in p38 RNAi and dacapo (p27) RNAi conditions, recombinants lines mad2,p38 RNAi and mad2,dacapo RNAi were crossed with worGAL4; SakOE,mad2 lines. Cyo; SakOE,mad2/mad2 larvae were used as controls. Adult flies were kept at 25°C (unless otherwise noted) in freshly yeasted tubes, changed every 2-3 days. Crosses were raised at 25°C and 3-5-days-old adults were heat shocked for 35 min at 36.5°C. For temperature shift experiments crosses were set up and maintained at 18°C until adulthood. Adult flies (3-5 days old) were shifted to 29°C for 12 days.
Staging fly cultures. Six hours egg collection were performed at for each genotype. Vials were kept at 25°C until larvae were reaching the required developmental stage (early, mid or late third instar at 72, 96 or 120 h after egg deposition).
Immunohistochemistry and fixed tissue imaging. For immunohistochemistry, brains from third instar larvae were dissected in PBS, fixed for 30 min in 4% paraformaldehyde in PBS or in PBS with 0.1% Triton X-100; They were washed three times in PBS-T (PBS, 0.3% Triton X-100) and incubated O/N at 4°C with primary antibodies diluted in PBS-T. After washing in PBS-T three times, brains were incubated O/N at 4°C with secondary antibodies and Hoechst (0.5 mg ml À 1 in PBS-T), washed once more in PBS and mounted in mounting media (1.25% n-Propyl Gallate, 75% glycerol, 25% H 2 0).
Midgut fixation and immunofluorescence staining were performed as described previously 61 . Briefly adult female intestines were dissected in PBS and fixed for 2 h in 4% paraformaldehyde. Intestines were rinsed in PBT (PBS containing 0.1% Triton X-100), trimmed and incubated for at least 30 min in PBS containing 50% glycerol, then equilibrated in PBT to osmotically clean the lumen before antibody incubations.
EdU (5-ethynyl-2'-deoxyuridine) incorporation. Mid third-instar larval brains were dissected in Schneider's Drosophila Medium used for live imaging and incubated for 2 h at 25°C in the same medium implemented with 100 mM EdU (C10338, Invitrogen). Brains were then washed in PBS, fixed and immunostained. EdU detection was realized after the secondary antibody detection, according to the manufacturer instructions.
DNA damage assay. Mid third instar larval brains were dissected in Schneider's Drosophila Medium used for live imaging and incubated for 30 min at 25°C in medium implemented with 100 mg ml À 1 of aphidicolin (A0781, Sigma-Aldrich). Brains were washed three times in PBS and incubated 1 h in medium without drug. Brains were fixed and stained as described above.
Reactive oxygen species detection. Mid third instar larval brains were dissected in PBS supplemented with dihydroethidium (final concentration 30 mM, D-1168, Life Technologies) for 10 min, washed two times in PBS (5 min per wash) and immediately mounted in a glass slide for live imaging.
Fluorescent in situ hybridization. Oligonucleotides probes for AACAC repeats (chromosome II) and dodeca sequence (Chromosome III) 64,65 were synthesized with a 5 0 CY3 and FAM488 fluorescent dye, respectively. We used the following sequences: 5' CY3-AACACAACACAACACAACACAACACAACACAACAC, FAM488-CCCGTACTGGTCCCGTACTGGTCCCG. Our FISH protocol was adapted from previously described methods 64,65 (with the exception that all the steps were done on whole brains in the PCR machine instead of glass slides). Fluorescent in situ hybridization in the brains was performed by dissecting third-instar larval brains in PBS and fixed 30 min in 4% formaldehyde in PBS with 0.1% tween 20 (162312, Panreac Sintesis). For the intestine, the identification of esgGFP þ cells is required. Intestines were dissected and fixed for 40 mn in PFA 4%, washed 3 times in PBS-T (PBS, 0.3% triton) and incubated with an anti-GFP antibody (1:250) overnight (O/N) washed three times in PBS-T and incubated with secondary antibody O/N at 4°C. Intestines were then washed three times in PBS-T and fixed another time for 20 mn in PFA 4% before proceeding with the probe hybridization. Brains or intestines were then washed three times in PBS, once in 2xSSCT (2XSSC (EU0300, Euromedex)/0.1% tween-20), once in 2XSSCT/50% formamide (47671, Sigma). For the pre-hybridization step, brains were transferred to a PCR tube containing 92°C pre-warmed 2XSSCT/50% formamide and denaturated 3 min at 92°C. Brains were then hybridized 5 min at 92°C with previously denaturated DNA probe (40-80 ng) in hybridization buffer (20% dextran sulfate (D8906, sigma)/2XSSCT/50% deionized formamide (F9037, Sigma), 0.5 mg ml À 1 salmon sperm DNA (D1626, Sigma)). 3 min after denaturation at 92°C, tubes were left O/N at 37°C. Samples were then washed with 60°C pre-warmed 2XSSCT for 10 min, and one time 5 min in 2XSSCT at room temperature. Samples were then stained with DAPI (in 2XSSC) for at least 30 min and mounted using our standard mounting medium (1.25% n-propyl gallate (Sigma, P3130), 75% glycerol (bidistilled, 99.5%, VWR, 24388-295), 25% H 2 O). FISH signals were quantified by manual scoring of the individual slices for each z-stack.
Image analysis quantification and statistical analysis. All analysis or measurements were performed on Fiji or ImageJ software. Quantifications were performed manually using the cell counter tool. Brain areas were measured by ImageJ. In the quantifications of Dpn þ , Pros þ and Elav þ cells (Figs 3 and 5), only nuclear Dpn, Pros and Elav signals were considered as scoring positive. To quantify type-I Nbs, only Dpn þ cells displaying nuclear sizes 48 mm were considered. Quantifications were always done on the same area of brains (ventral side) using 3 Z (steps of 1 mm) around the plane were the maximum of Nbs are present. For CC3 quantification, the whole CB region was analysed.
In Fig. 6, only posterior midgut tissue was analysed and only clones of minimum two GFP-positive cells were scored and at least six different midguts of each genotype were analysed. The presence of Esg þ cells with nuclear size above 7 mm was only observed and scored in the distal midgut loop previously reported as highly proliferative and being a dynamic part of the tissue 66 . Statistical analysis was performed with Graph Pad Prism using the tests mentioned in the figure legends.
Analysis of spindle positioning. Analysis of spindle position relative to the apical aPKC crescent position was performed by staining fixed brains with anti-aPKC, anti-a-tubulin and anti-Cnn antibodies. For SakOE,mad2 Nbs analysis, we only scored mitotic cells with more than two centrosomes and where an aPKC crescent could be clearly distinguished. The angle between the spindle axis and the aPKC crescents was determined at early anaphase to avoid any bias in possible rotation during early mitotic stages, using the measurement tool in ImageJ. The data were analysed for statistical significance using the two-tailed t-test function in Prism.
Transplantation assays. The injection protocol was adapted from previously described method 57 . Mid third-instar larval brains expressing tubulin-GFP in either WT or mutant background were dissected in PBS, rinsed three times, sliced in several pieces and transplanted into adult host females abdomen using a elongated Pasteur pipette (drawn out in a flame to a tip diameter of approximately 100 mm) connected to a hand-driven microinjector. After injection, flies were transferred to 18°C for 24 h and then to 25°C in tubes with males. Five days after transplantation, flies were scored on a daily basis. Hosts were scored positive for tumour formation when a high GFP intensity signal could be easily detected in a homogeneous way within the abdomen. This normally occurs, independently of the genotypes tested in the study, between days 10 and 15. Injected flies were monitored until day 25 and we have never observed late onset tumours.