Cortex glia encase lineage-related cells regardless of their identity

During niche formation, the first cell encased within a CG chamber is the NSC, suggesting that cell identity, specifically being a stem cell, may signal individual encasing. Entire NSC lineages, which can be tracked and controlled with the worniu-GAL4 (wor >) driver [32] from the NSC to immature neurons (S1B and S1C Fig), are ultimately encased later on.

To assess the importance of stem cell identity in chamber formation, we took advantage of genetic alterations known to dysregulate NSC division and differentiation and to lead to the formation of tumour-like, NSC-only, lineages [33]. In particular, pros knockdown in Type I lineages converts GMC into NSC-like, Dpn⁺ cells at the expense of neurons [34]. Surprisingly, in these conditions, we found that CG chambers contained not one, but several NSC-like, Dpn⁺ cells (Fig 2A–2A’). Similar results were obtained for other conditions that lead to Type I NSC-only lineages, including GMC dedifferentiation via Dpn overexpression or loss of asymmetric division via membrane-tethered aPKC overexpression (S2A Fig). We then asked how CG would adapt to the dysregulation of Type II NSCs, which generate bigger lineages than Type I NSCs. Since CG chamber formation was precisely described only for Type I NSCs [23], we first checked the dynamics of CG morphogenesis around Type II and found that they followed similar steps, albeit in a slower fashion (S2B Fig). We then knocked down the cell fate determinant brat [35,36], which is necessary for preventing immature INP (imINP) dedifferentiation into NSC-like cells. This led to the formation of large tumours (Fig 2B–2B’). CG were able to adapt to cell overproliferation, at least until this stage, and enwrapped many NSC-like cells within one chamber. These data show that both for Type I and Type II NSCs, stem cell identity is not sufficient to ensure their sorting into individual CG chambers.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. CG individually encase clonally related cells regardless of their identity.

(A, B) Adaptation of the CG network to NSC tumours. Type I pros (Nrv2::GFP, wor-GAL4; tub-GAL80^ts > pros RNAi) and Type II brat (Nrv2::GFP, wor-GAL4; tub-GAL80^ts > brat RNAi) tumours were induced, and the organisation of CG membrane was monitored by Nrv2::GFP (green). Control line for both (Nrv2::GFP, wor-GAL4; tub-GAL80^ts > w¹¹¹⁸). For pros, larvae were dissected after 48 h at 18°C, followed by 2 h heatshock at 37°C and 48–52 h at 29°C. For brat, larvae were dissected after 72 h at 29°C. See S1 Table for detailed genetics, timing, and conditions of larval rearing. (A) Left panel shows a representative whole thoracic VNC for control (n = 10 VNCs) and pros tumour (n = 10 VNCs), and (A’) middle and right panels show a close-up of the yellow boxes in (A). NSCs/tumour NSCs are labelled with anti-Dpn (grey), and neuron with anti-ElaV (magenta). (B) Left panel shows a whole representative CB for control (n = 10 CBs) and brat tumour (n = 10 CBs), and (B’) middle and right panels show a close-up of the yellow boxes in (B). NSCs/tumour NSCs are labelled with anti-Dpn (grey), Type I NSC and mINPs with anti-Asense (Ase, magenta), and GMCs plus newborn neurons with anti-Prospero (Pros, blue). For both, dashed white lines delineate one CG chamber. (C, D) Relationship between individual encasing and cell identity. Type I pros (wor > raeppli-NLS, pros RNAi) and Type II brat (wor > raeppli-NLS, brat RNAi) tumours were induced together with the multicolour lineage tracing Raeppli-NLS. One out of 4 colours (blue, white, orange, and red) is stochastically induced in the transformed NSCs using hs-Flp. For pros, larvae were dissected after 24 h at 18°C, followed by 2 h heatshock at 37°C and 48–52 h at 29°C. For brat, larvae were dissected after 48 h at 18°C, followed by 2 h heatshock at 37°C and 48–52 h at 29°C. See S1 Table for detailed genetics, timing, and conditions of larval rearing. (C) Top and orthogonal close-up views of representative control (n = 10 VNCs) and pros (n = 9 VNCs) Type I lineages marked with Raeppli-NLS. (D) Top and orthogonal close-up views of representative control (n = 8 CBs) and brat (n = 7 CBs) Type II lineages marked with Raeppli-NLS. For both, CG membrane is visualised with Nrv2::GFP (green). Dashed white lines delineate one CG chamber. White stars indicate Type II lineages. (E) brat RNAi tumours are individually encased by the CG. Representative confocal pictures of the CG network in control Type II NSC clone and brat RNAi Type II NSC clone. Top view and orthogonal views are shown. The CoinFLP system was used to generate wild-type and brat RNAi clones in the Type II NSCs, using grh^D4-FLP for induction. The grh^D4 enhancer drives in most NSCs from late embryonic stage. As such, clones in Type I NSCs can also be induced (marked by a cyan star) yet do not lead to any tumour. Larvae were dissected after 72–96 h at 25°C. See S1 Table for detailed genetics, timing, and conditions of larval rearing. The membrane of the clone is marked by mCD8::RFP (magenta), and CG membrane is visualised with Nrv2::GFP (green). Dashed white lines delineate one Type II NSC chamber. n(brat RNAi clones) = 20 clones in 15 CBs and n(control clones) = 15 clones in 11 CBs. All (100%) of the Type II NSC clones in both control and brat RNAi conditions were found individually encased by the CG. (F) Schematic of the different hypotheses explaining individual encasing of NSC lineages by CG. Panel I depicts the NSC-driven timing of NSC encapsulation, prior to lineage generation, as the instructive cue. Panel II describes the use of adhesion mechanisms: (1) specific CG to NSC lineage and (2) generic, based on difference in strength between intralineage adhesion (A(L^NSC)) and adhesion linking NSC lineages and CG (A(L^NSC-CG)); in this case, the prediction is that A(L^NSC-CG) < A(L^NSC). CB, central brain; CG, cortex glia; GMC, ganglion mother cell; NSC, neural stem cell; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g002

We then wondered whether tumour NSCs grouped within one chamber originated from the same mother NSC or had been encased randomly independently of their lineage of origin. To do so, we used a multicolour clonal analysis to label individual NSC lineages. The Raeppli system [37] results in the stochastic and irreversible labelling of a cell at the time of induction, allowing to mark and track a mother cell and its colour-sharing progeny. Heatshock-controlled induction before NSC reactivation of a nuclear tagged version of Raeppli (Raeppli-NLS) ensured most of lineages could be fully tracked, except a few in which the induction had failed. We first confirmed that cells found within each CG chamber belonged to the same lineage for wild-type Type I and Type II NSC lineages (Fig 2C and 2D), a property previously assumed [24,25] or reported for specific lineages only [23]. For Type I pros tumours, we found that clonal tumour-like growth coming from single dysregulated NSCs (marked by one colour) were contained within one CG chamber (Fig 2C). In brat tumours, most Type II chambers, found in a similar number to control (S2C and S2D Fig), showed a single, large tumour-like growth of one colour (Fig 2D). Yet, we sometimes observed multiple colours within a single CG chamber, mostly as scattered cells or small clones among unmarked cells (S2E and S2F Fig). We inferred this comes from later uncontrolled inductions of Raeppli-NLS in individual tumour NSCs from lineages in which recombination was not successfully induced during heatshock. This would result in differently marked NSC-like Dpn⁺ clones yet coming from the same Type II NSC mother cell. In support of this interpretation, we observed Raeppli-NLS stainings (isolated cells or small clones) in brat tumour without heatshock (S2E and S2F Fig), pinpointing a leakiness in the induction control for this background, which we were not able to bypass technically.

To strengthen our findings on brat RNAi, we then used a Coin-FLP clonal analysis [38] to generate brat RNAi clones in Type II NSC lineages (induced by grh^D4-FLP, expressed in most NSCs from late embryogenesis; see Methods and S1 Table). We found that 100% of Type II brat RNAi clones were individually encased (Fig 2E).

Altogether, these results demonstrate that cells from one NSC lineage, whether NSC, GMC, INP, or neurons, are kept together within one CG chamber regardless of their identity and suggest that NSC lineages contain intrinsic information allowing their individual encasing by the CG.

Individual encasing relies on intrinsic lineage cues

Chamber completion around NSCs occurs around the time of first division and is driven by NSC reactivation [23,26]. A simple explanation for keeping a lineage together and separated from others would thus be the sequential addition of newborn cells within a compartment already defined from the start by NSC-derived signals (Fig 2F, Panel I). Timing would thus be the instructive cue. Indeed, previous studies have shown that signals from reactivated NSCs are paramount to form CG niches [23], although the cues integrated by the glia have not been identified. Within this hypothesis, blocking CG morphogenesis until well after the first NSC division, followed by subsequent release, would result in aberrant chamber formation and random encasing of neurons from different lineages.

To test this hypothesis, we conditionally overexpressed the PI3K/Akt pathway inhibitor PTEN [23] to specifically block CG growth, using the binary QF system (whose repression can be chemically controlled), while Raeppli-NLS expression in NSC lineages was driven by the GAL4/UAS system (Fig 3A). Raeppli-NLS was induced at ALH0, and CG growth was impaired until NSCs cycled actively and neuronal progeny had been already produced (time T1). While CG membranes still infiltrated in between, they failed to form correct chambers to separate individual lineages from each other at this time (S3A and S3B Fig). We then allowed CG to resume their growth and later observed the establishment of a stereotyped chequerboard pattern, with most of the chambers containing a single NSC and its neuronal progeny (Fig 3B, time T2). This was confirmed by quantifying the chambers containing more than one NSC lineage (Methods; Fig 3C). These results thus show that the information directing individual encasing by CG is not exclusive to the NSC but is inherited by its progeny and can be sensed by CG later during larval development.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. CG use intrinsic and generic NSC lineage cues to individually encase NSC lineages.

(A) Schematic of the timing and genetic conditions used to probe the importance of NSC-driven timing for the encapsulation of individual NSC lineages. To delay chamber formation around NSC lineages (“Delayed chamber formation” condition), CG growth is initially blocked using PTEN expression in the CG through the use of the Q system (cyp-QF2 driving QUAS-PTEN and combined with tub-QS under quinic acid–supplemented food). At T1 (72 h at 29°C from ALH0), CG growth is allowed by preventing PTEN expression (no quinic acid in the food), and CG structure is assessed at T2 (24–30 h at 29°C from T1, so after 100 h at 29°C from ALH0). In the control condition, QUAS-PTEN is not present, while the larvae are subjected to the same temperature, time, and quinic acid regimens. See S1 Table for detailed genetics, timing, and conditions of larval rearing. (B) Representative confocal picture at T2 of the extent of individual encasing of NSC lineages by CG after the regimen described in (A) both for Control and Delayed chamber formation conditions. Top panel shows the whole thoracic VNC, and bottom panel a close-up of the yellow box. NSC lineages were marked with the multicolour lineage tracing Raeppli-NLS (blue, white, orange, and red), induced under the control of the GAL4/UAS system (wor-GAL4 driver) at ALH0 using hs-Flp (wor > raeppli-NLS). Larvae are dissected after 100 h at 29°C. See S1 Table for detailed genetics, timing, and conditions of larval rearing. CG membrane is visualised with Nrv2::GFP (green). Two representative CG chambers displaying individual encasing of NSC lineages are delineated with a white dashed line. (C) Quantification of the percentage of NSC lineages non-individually encased from (B). Control T2 (n = 8 VNCs) and Delayed chamber formation T2 (n = 5 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution). Results are presented as box and whisker plots, where whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. (D) Schematic of the experiment designed to probe whether specific adhesions exist between individual CG cells and individual NSC lineages. Mosaic clonal analysis is used to impair the growth of a few CG cells in a random fashion (magenta), preventing the encasing of corresponding NSC lineages. If specific adhesions exist between individual CG cells and individual NSC lineages, neighbouring CG cells would not be able to bind to and encase these lineages. (E) Representative confocal picture of the CG network in control CG clone and in clone in which CG growth was blocked (PTEN OE). The CoinFLP system was used to generate rare wild-type and PTEN OE clones in the CG. Clones were induced at late embryogenesis/early larval stage through the expression of cyp4g15-FLP. Larvae were dissected after 48–72 h at 25°C. See S1 Table for detailed genetics, timing, and conditions of larval rearing. The membrane of the clone is marked by mCD8::RFP (magenta) and CG membrane is visualised with Nrv2::GFP (green). (F) Quantification of the clone volume (μm3), the number of NSCs per clone, and the number of non-encased NSCs around the clone between wild-type and PTEN OE clones from (E). Clone volume and number of NSCs per clone decrease in PTEN OE clones compared to control, indicating that PTEN-overexpressing CG cells were not able to reach their normal size and average NSC encasing. Yet, all NSCs neighbouring the clones were individually encased. Control (n = 14 clones) and PTEN OE (n = 18 clones). Data statistics: unpaired Student t test for clone volumes and Mann–Whitney U test for number of NSCs per clone. For the number of non-encased NSCs around the clone, there is no variance and, thus, statistics cannot be applied (NA, non-applicable). Results are presented as box and whisker plots, where whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. ALH, after larval hatching; CG, cortex glia; NSC, neural stem cell; OE, overexpression; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g003

One elegant way cells could be kept together would be their physical binding through adhesion mechanisms. We hypothesised 2 possibilities (Fig 2F, Panel II). The most complex mechanism would rely on the existence of lineage-specific adhesions, with a code of unique molecular interactions between specific CG cells and all cells of specific NSC lineages (Fig 2F, Panel II.1). In this scenario, CG cells would not be interchangeable regarding their adhesive properties. The simplest solution would see all NSC lineages relying on the same adhesion mechanisms between cells (Fig 2F, Panel II.2). The differential adhesion hypothesis proposes that cells with similar adhesive strength cluster together, ultimately sorting cell populations with different adhesions and creating cellular compartments [39]. The existence of intralineage adhesions (A(L^NSC)) stronger than adhesions between a NSC lineage and the CG (A(L^NSC-CG)) would form a physical barrier for the CG, preventing their intercalation in between cells from the NSC lineage and thus leading to their sorting from NSC lineages, which they encapsulate. In this case, a given CG cell would be able to enwrap different NSC lineages.

To discriminate between these 2 hypotheses, we first assessed the result of preventing some CG cells to encase NSC lineages to a normal extent (Fig 3D). We used Coin-FLP clonal analysis [38] to randomly impair the growth of a few CG cells within the entire population (using cyp4g15-FLP; see Methods and S1 Table). A bias in the Coin-FLP system ensures that only rare clones are generated, allowing us to remove adhesions between individual CG and NSC lineages. Overexpression of PTEN in a few CG cells (marked with RFP) before chamber formation resulted in much smaller clones compared to control wild-type clones (Fig 3E and 3F, left graph) and enwrapping less NSCs (Fig 3F, middle graph). However, the CG network itself appeared gapless around PTEN overexpression clones, revealing that CG were able to compensate for the early loss of their neighbours’ membrane to restore NSC chambers (Fig 3E and 3F, right graph). To confirm this result, we used the same approach to express the proapoptotic gene reaper [40] to kill a few CG, this time only once the NSC chambers were already formed (S3C–S3E Fig). Induction of apoptosis led to a near complete loss of RFP-marked clones, only visible through cell remnants. Nevertheless, the CG network around these clones appeared intact. This shows that different CG cells can encase a given NSC lineage and that, in accordance with our previous findings (Fig 3A–3C), they are still able to do so after NSC reactivation. Of note, similar results demonstrating the ability of CG to replace each other were previously obtained for primary neurons [41]. Altogether, our results indicate that NSC lineages can be enwrapped by different CG already encasing other lineages and that specific L^NSC-CG pairings do not occur. This suggests that the same adhesion mechanism might be repeated for each NSC lineage, providing stronger cohesion between cells of the same NSC lineage than between CG and NSC lineages.

Intralineage adherens junctions are present but not absolutely required for individual encasing of NSC lineages

Previous studies had reported the localization within larval NSC lineages of the Drosophila E-cadherin Shotgun (Shg), a component of adherens junctions usually present in epithelia [24,42]. Shg was also shown to localise at the contact between NSC and newborn GMC [43]. We analysed the expression pattern of Shg during larval development using a GFP fusion produced by homologous recombination (shg::GFP [44]; Fig 4A). Shg::GFP was detected from larval hatching, initially present around and between NSCs (ALH24, dashed yellow circle), and also along CG membranes (white arrowhead). At later larval stages, following NSC reactivation and neuronal production (ALH48 to ALH72), Shg::GFP showed a remarkable pattern of expression, with a strong enrichment between cells from the same lineage (yellow star marks the NSC) and decreased staining along the CG membrane (yellow arrowheads). A similar pattern was found for its ß-catenin partner Armadillo (Arm; Fig 4B) through antibody staining. As Shg forms homophilic bonds, this suggests that adherens junctions exist between cells of the same NSC lineage.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Intralineage adherens junctions are present but not absolutely required for individual encasing of NSC lineages.

(A) Representative confocal images of the expression of the Drosophila E-cadherin, Shg, at ALH0, ALH24, ALH48, and ALH72 at 25°C. n ≥ 7 VNCs for all time points. Shg is monitored through a shg::GFP fusion (magenta), CG membrane is visualised by cyp4g15-mtd::Tomato (green), NSCs are labelled with anti-Dpn (grey), and neurons are labelled with anti-ElaV (blue). Yellow stars indicate NSC position, dashed yellow circles indicate Shg::GFP signal between NSCs, and yellow arrowheads indicate Shg::GFP signal between NSC and CG. (B) Representative confocal images of the expression of the Drosophila β-catenin, Arm, at ALH72 at 25°C. n = 6 VNCs. Arm is detected with a specific antibody (magenta), CG membrane is visualised by Nrv2::GFP (green), NSCs are labelled with anti-Dpn (grey), and neurons are labelled with anti-ElaV (blue). Yellow stars indicate NSC position. (C) Representative confocal pictures of the thoracic VNC for control (wor >—(x w¹¹¹⁸)) and shg knockdown by RNAi (wor > shg RNAi, line VDRC 27082) in NSC lineages (driver line Nrv2::GFP, wor-GAL4; tub-GAL80^ts), at the NSC level, at the neuron level, and in orthogonal view. Larvae are dissected after 68 h at 29°C from ALH0. Control, wor >—(x w¹¹¹⁸) (n ≥ 10 VNCs) and wor > shg RNAi (n ≥ 10 VNCs). CG membrane is visualised by Nrv2::GFP (green), and NSCs are labelled with anti-Dpn (grey). (D) Schematic of the respective NSC and CG patterns in shg knockdown in NSC lineages. (E) Representative confocal picture of the thoracic VNC for shg knockdown by RNAi (line VDRC 27082; wor > raeppli-NLS, shg RNAi) in NSC lineages marked with the multicolour lineage tracing Raeppli-NLS (blue, white, orange, and red). Top panel shows the whole thoracic VNC, and bottom panel a close-up of the yellow box. Raeppli-NLS is induced at ALH0 under the control of wor-GAL4 using hs-Flp. Larvae are dissected after 72 h at 29°C from ALH0. See S1 Table for detailed genetics, timing, and conditions of larval rearing. CG membrane is visualised with Nrv2::GFP (green). Two representative CG chambers displaying individual encasing of NSC lineages are delineated with a white dashed line. (F) Quantification of the percentage of NSC lineages non-individually encased from (E). Control, wor > raeppli-NLS (n = 7 VNCs) and wor > raeppli-NLS, shg RNAi (n = 9 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution). Results are presented as box and whisker plots, where whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. ALH, after larval hatching; Arm, Armadillo; CG, cortex glia; NSC, neural stem cell; Shg, Shotgun; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g004

We then wondered whether such adhesion was exclusively found in differentiating neuronal lineages. We first looked at NSC-like Dpn⁺ cells from pros and brat tumours, which are contained clonally within one CG chamber (Fig 2C–2E). A strong Shg staining was detected between NSC-like cells from the same lineage (S4A Fig), but not between NSC and glia. In line with this finding, we observed a strong Shg staining between NSCs of the OL, another type of neural progenitors [45] that are contained within one CG chamber (S4B Fig). Altogether, these results suggest that adherens junctions could be a mean of keeping NSC lineages together by providing differential adhesion.

Previous studies indeed suggested that Shg expression was required in NSC lineages for proper CG structure [24,43]. We first generated NSC lineages mutant for shg by inducing MARCM (Mosaic Analysis with a Repressible Cell Marker) clones [46] during late embryogenesis (see Methods and S1 Table). MARCM allows the generation and labelling of single or clonally related cells homozygous for a mutation in an otherwise unlabelled heterozygous animal. Here, the shg null allele used (shg^R64) is homozygous lethal and requires clonal analysis. To our surprise, Shg-depleted NSC lineages (labelled by an RFP marker) still stayed individually encased within one CG chamber (S4C Fig, upper panel). The same result was obtained when clones were induced at a later time point to prevent potential compensation through the up-regulation of other adhesion molecules (S4C Fig, lower panel). In accordance with these results, we found that RNAi knockdown of shg in NSC lineages (wor > shg RNAi) did not disrupt overall lineage organisation within CG chambers (Fig 4C and 4D), despite successfully decreasing Shg::GFP signal (S4D and S4E Fig, mean of 0.3 normalised to control). To note, we also did not record disruption of CG network when shg was knocked down in the CG themselves (S4F Fig), in contrast to previous findings [24]. Driving Raeppli-NLS (induced at ALH0) along with shg RNAi driven from embryogenesis in the NSC lineage confirmed and quantified the conservation of individual encasing (Fig 4E and 4F) These results argue against the strict requirement of Shg-mediated adhesion for NSC lineage maintenance within one CG chamber [24,43].

We investigated whether intralineage adherens junctions serve as a safety mechanism, ensuring robustness in a system where other strategies would primarily provide intralineage cohesion. As the CG chamber encases NSC at the time they initiate progeny production, timing would first trap these newborn cells together (Fig 2F, Panel I), while adherens junctions would ensure their cohesion if the chamber is affected (such as in Fig 3A–3C). To test this hypothesis, we conditionally blocked CG growth using the Q system, while continuously driving shg RNAi and Raeppli-NLS (induced at ALH0) in NSC lineages under the control of the GAL4/UAS system (Fig 5A). We confirmed the effectiveness of shg knockdown through antibody staining (S5A Fig). If adherens junctions are important for keeping intralineage cohesion in the absence of proper encasing, progeny born before CG growth is reestablished (i.e., before chamber formation) would not be kept together but mixed with other lineages. Conversely, progeny born after the chamber forms would be encased together. Looking at the deeper level in which differentiating progeny reside (Fig 1F”–1G”), we found that most lineages still appeared correctly encased (Fig 5B and 5C) despite efficient shg knockdown and delay in chamber formation (S5A–S5C Fig). We still uncovered a few localised defects in the individual encasing of NSC lineages, with several colours detected within the boundaries of one continuous CG membrane (Fig 5B and 5C). These results indicate that, while adherens junctions are not strictly required for NSC lineage encasing under unchallenged conditions, they may contribute to the robustness of individual NSC lineage encasing when CG are altered, although to a limited extent.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Intralineage adherens junctions do not drive individual encasing yet might provide robustness.

(A) Schematic of the timing and genetic conditions used to probe the importance of Shg intralineage adhesion (wor-GAL4 > shg RNAi, VDRC line 27082) on individual encasing of NSC lineages when CG chamber formation around NSC lineages is delayed. CG growth is initially blocked through PTEN overexpression (cyp-QF2 driving QUAS-PTEN and combined with tub-QS under quinic acid–supplemented food). At T1 (72 h at 29°C from ALH0), CG growth is then allowed by preventing PTEN expression (no quinic acid in the food), and CG structure is assessed at T2 (24–30 h at 29°C from T1, so after 100 h at 29°C from ALH0). In the control condition (see Fig 3A), QUAS-PTEN is not present, while the larvae are subjected to the same temperature, time, and quinic acid regimens. See S1 Table for detailed genetics, timing, and conditions of larval rearing. (B) Representative confocal picture at T2 of the extent of individual encasing of NSC lineages by CG after the regimen described in (A). Top panel shows the whole thoracic VNC, and bottom panel a close-up of the yellow box. NSC lineages were marked with the multicolour lineage tracing Raeppli-NLS (blue, white, orange, and red), induced under the control of the GAL4/UAS system at ALH0 using hs-Flp (wor > raeppli-NLS, shg RNAi). Larvae are dissected after 100 h at 29°C. See S1 Table for detailed genetics, timing, and conditions of larval rearing. CG membrane was visualised with Nrv2::GFP (green). Dashed white lines highlight examples of NSC lineages encased individually and dashed yellow lines examples of NSC lineages encased together. (C) Quantification of the percentage of NSC lineages non-individually encased from (B). Control T2 (n = 8 VNCs), Delayed chamber formation T2 (n = 5 VNCs) and Delayed chamber + shg RNAi in NSC lineages T2 (n = 10 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution). p = 2.46 × 10⁻⁹ for the grouped dataset. P values for individual comparisons test are displayed on the graph. Results are presented as box and whisker plots. (D) Representative confocal picture of the thoracic VNC and close-up for shg overexpression in the CG (cyp > shg, driver Nrv2::GFP, tub-GAL80^ts; cyp4g15-GAL4). Larvae are dissected after 68 h at 29°C from ALH0. CG membrane is visualised by Nrv2::GFP (green), NSCs are labelled with anti-Dpn (grey), neurons are labelled with anti-ElaV (blue), and Shg is detected with a specific antibody (magenta). (E) Quantification of the percentage of NSCs non-individually encased from (D). Control, cyp >—(x w¹¹¹⁸) (n = 7 VNCs), cyp > shg (n = 6 VNCs). Data statistics: there is no variance and, thus, statistics cannot be applied (NA, non-applicable). Results are presented as box and whisker plots. For all box and whisker plots: whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. ALH, after larval hatching; CG, cortex glia; NSC, neural stem cell; Shg, Shotgun; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g005

A classical test of the differential adhesion hypothesis is to challenge the balance between the different adhesion complexes. To do so, we overexpressed shg in the CG, with the aim to flatten the Shg-based adhesion difference (with now A(L^NSC-CG) ≈ A(L^NSC) for Shg); Fig 2F, Panel II.2). Lineage-expressed endogenous Shg would have the choice to bind either with itself or with CG-provided Shg. Despite the successful expression of shg in the CG, the usual pattern of CG chambers was nevertheless maintained in this condition (Fig 5D and 5E), suggesting that in this case, A(L^NSC) still stays superior to A(L^NSC-CG) and that other adhesions fulfil this role. These data imply that differential adhesion using adherens junction is not the main driver for ensuring the individual encasing of NSC.

Occluding junction components are expressed in NSC lineages

These findings prompted us to investigate the potential presence and role of other adhesion complexes that could provide intralineage cohesion and differential adhesion to sort NSC lineages from CG.

Occluding junctions (tight junctions in vertebrate and septate junctions in Drosophila) [47,48] primarily perform a permeability barrier function to paracellular diffusion. However, they can also provide some adhesion between the cells they link. Drosophila septate junctions are formed by the assembly of cell surface adhesion molecules that can interact in cis or trans, in an homologous or heterologous fashion, and which are linked to the intracellular milieu by supporting membrane or cytoplasmic molecules [47]. A core, highly conserved tripartite complex of adhesion molecules comprises Neuroglian (Nrg), Contactin (Cont), and Neurexin-IV (Nrx-IV). Nrg, the Drosophila homologue of Neurofascin-155, is an L1-type family transmembrane protein, containing several immunoglobulin domains and is mostly homophilic. Cont, homologous to the human Contactin, also contains immunoglobulin domains, is GPI anchored, and only performs heterophilic interactions. Nrx-IV, homologous to the human Caspr/Paranodin, is a transmembrane protein with a large extracellular domain containing laminin-G domains and EGF repeats [49] and is able to set up heterophilic interactions. Several cytoplasmic or membrane-associated proteins also participate in septate junction formation, such as the FERM-family Coracle (Cora), the MAGUK protein Discs large (Dlg1), and the integral membrane Na K-ATPase pump (ATPα).

We decided to perform a preliminary characterisation of the expression and function of septate junction components in NSC lineages. We found that Nrx-IV, Nrg, Cora, Dlg1, and ATPα are all present in NSC lineages when the chamber is formed (ALH72) and Cora also showed a staining along the CG interface (S6A Fig). We were not able to assess Cont due to lack of access to working reagents. Multiple septate junction components are thus expressed in NSC lineages, localising between cells of the same lineage. We then probed the importance of such expression in the individual encasing of NSC lineages by CG. We knocked down nrx-IV, nrg, dlg1, cont, and ATPα in NSC lineages from ALH0 using specific RNAi lines (S2 Table). Larvae from dlg1, cont, and ATPα knockdown died at early larval stages. From ATPα knockdown, few larvae still reached late larval stage, displaying restricted irregularities in the CG network. In contrast, nrx-IV and nrg knockdowns mostly survived and resulted in altered encasing of individual NSC lineages.

We thus decided to focus on Nrx-IV and Nrg functions in NSC lineages. Interestingly, they both perform nervous system–specific roles outside of the septate junction. Nrx-IV is required in the embryonic CNS for axonal wrapping by the midline glia [50–52]. Neuronal Nrg is important for axonal guidance and dendritic arborization of peripheral neurons [53–55], as well as for the function and axon branching of specific larval CB neurons [56,57]. For both proteins, their role in NSC lineages during larval neurogenesis is, however, poorly known.

Nrx-IV is required in NSC lineages for individual encasing by CG

Using an endogenous GFP fusion (Nrx-IV::GFP; Fig 6A), we first analysed Nrx-IV expression along CG chamber formation. Nrx-IV::GFP was detected from early larval stage in embryonic (primary) neurons, and around NSCs (ALH0), an expression maintained while NSCs proceed through reactivation (ALH24, dashed yellow circle). As NSCs have reactivated and CG grown (ALH48), Nrx-IV::GFP appears expressed at the interface between NSC and CG (yellow arrowhead). Further, accompanying the production of newborn, secondary neurons (ALH72), Nrx-IV::GFP is found enriched at the interface of the cells from the same lineage (NSC, GMC, and neurons), while maintaining a strong expression at the interface with CG.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Nrx-IV is required in NSC lineages for their individual encasing by CG.

(A) Representative confocal images of the expression of Nrx-IV at ALH0, ALH24, ALH48, and ALH72 at 25°C. n ≥ 8 VNCs for all time points. Nrx-IV is monitored through a Nrx-IV::GFP fusion (magenta), CG membrane is visualised by cyp4g15-mtd::Tomato (green), NSCs are labelled with anti-Dpn (grey), and neurons are labelled with anti-ElaV (blue). Yellow stars indicate NSC position, dashed yellow circles indicate Nrx-IV::GFP signal between NSCs, and yellow arrowheads indicate Nrx-IV::GFP signal between NSC and CG. (B) Representative confocal picture of the thoracic VNC for a condition in which nrx-IV is knocked down by RNAi (wor > nrx-IV RNAi, BDSC line 32424) from ALH0 in NSC lineages (driver line Nrv2::GFP, wor-GAL4; tub-GAL80^ts) at the NSC level, at the Neuron level, and in orthogonal view. Larvae are dissected after 68 h at 29°C from ALH0. Control, wor >—(x w¹¹¹⁸) (n ≥ 10 VNCs) and wor > nrx-IV RNAi (n ≥ 10 VNCs). CG membrane is visualised by Nrv2::GFP (green), and NSCs are labelled with anti-Dpn (grey). (C) Schematic of the respective NSC and CG patterns in nrx-IV knockdown in NSC lineages. (D) Representative confocal picture of the thoracic VNC for a condition in which nrx-IV is knocked down by RNAi (BDSC line 32424) from ALH0 in NSC lineages and marked with the multicolour lineage tracing Raeppli-NLS (wor > raeppli-NLS, nrx-IV RNAi). Raeppli-NLS (blue, white, orange, and red) is induced at ALH0 using hs-Flp. Larvae are dissected after 72 h at 29°C from ALH0. See S1 Table for detailed genetics, timing, and conditions of larval rearing. CG membrane was visualised with Nrv2::GFP (green). Dashed white lines highlight examples of NSC lineages encased individually, and dashed yellow lines examples of NSC lineages encased together. (E) Quantification of the percentage of NSC lineages non-individually encased from (D). Control, wor > raeppli-NLS (n = 7 VNCs) and wor > raeppli-NLS, nrx-IV RNAi (n = 11 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution). Results are presented as box and whisker plots, where whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. ALH, after larval hatching; CG, cortex glia; Nrx-IV, Neurexin-IV; NSC, neural stem cell; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g006

We then performed RNAi knockdown, using a line successfully decreasing the levels of Nrx-IV in NSC lineages (S6B and S6C Fig, mean of 0.4 normalised to control), to assess nrx-IV function in NSC lineage encasing. CG pattern first appeared mostly normal when observed at the NSC level, with NSCs seemingly individually separated by CG membranes (compare Fig 6B NSC level with Fig 1F’–1G’). However, we observed a striking, unusual pattern at the level of differentiating progeny, with much larger CG chambers harbouring a clear continuous outline (compare Fig 6B neuron level with Fig 1F”–1G”). A similar result was obtained with another RNAi line against nrx-IV (S6D Fig). We then found that expressing nrx-IV RNAi under the control of the pan-neuronal driver ElaV-GAL4 (S6E Fig), which drives both in immature (secondary here) and mature (primary here) neurons, also resulted in bigger yet well-defined CG chambers. In contrast, no effect was detected under nrx-IV knockdown in mature neurons (with formed or forming synapses; nSyb-GAL4 driver [58]) or in the CG (S6E Fig).

To quantify individual encasing, we further expressed Raeppli-NLS along with nrx-IV RNAi in NSC lineages (both induced at ALH0). It replicated the extensive loss of individual encasing seen in Fig 6B, with multiple NSC lineages not separated by CG membranes but rather clustered together in large, defined chambers (Fig 6D and 6E; 22% of NSC lineages sharing a CG chamber with at least another one).

Altogether, our findings show that the expression of Nrx-IV in whole NSC lineages, including newborn neurons, but not in CG or mature neurons, is required for their individual encasing by CG.

A glia to NSC lineages adhesion through Nrx-IV and Wrapper is required for individual encasing

The dual role of Nrx-IV within and outside septate junctions is sustained by the existence of alternative splicing [51]. Nrx-IV can be produced as a septate junction isoform (Nrx-IV^exon3) and a neuronal isoform outside of SJ (Nrx-IV^exon4). Nrx-IV role within the embryonic CNS is through its recruitment by and binding to its glial partner Wrapper, another member of the immunoglobulin family [50–52]. We thus sought to assess whether the role of Nrx-IV in NSC lineages encasing by CG was dependent on Wrapper.

Previous studies had reported that regulatory sequences in the wrapper gene drive in the CG during late larval stages [41,59]. In situ hybridization against wrapper mRNA (HCR RNA-FISH; see Methods) detected wrapper expression in CG throughout larval development and chamber formation (Fig 7A). We then asked whether knocking down wrapper in the CG would recapitulate the encasing phenotype found under nrx-IV loss of function in NSC lineages. We first checked the efficiency of RNAi knockdown in the CG using RNA FISH and found that it was specifically wiping out wrapper signal in the CG (S7A and S7B Fig), while preserving its known expression in the midline glia (see arrowheads in S7A Fig). Driving this RNAi line in the CG reproduced the highly characteristic pattern of large yet defined CG chambers (Fig 7B, observed at the neuron level) found during nrx-IV knockdown in NSC lineages (compare with Fig 6B). A similar result was obtained with a second RNAi line against wrapper (S7C Fig). We further quantified the loss of individual encasing by marking NSC lineages with Raeppli-NLS while driving wrapper RNAi in the CG and found that 25% of the NSC lineages was sharing a CG chamber with at least another one (Fig 7C and 7D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. A CG to lineage interaction through Nrx-IV and Wrapper is required for individual encasing of NSC lineages by the CG.

(A) Representative confocal pictures of the localisation of wrapper mRNA at ALH0, ALH24, ALH48, and ALH72 at 25°C. n ≥ 6 VNCs for all time points. wrapper mRNA (magenta) is detected through RNA FISH. CG membrane is visualised by cyp4g15-mtd::Tomato (green), NSCs are labelled with anti-Dpn (grey), and neurons are labelled with anti-ElaV (blue). Yellow arrowheads indicate examples of colocalisation between CG membrane and wrapper mRNA signal. (B) Representative confocal pictures of the thoracic VNC for a condition in which wrapper is knocked down by RNAi (cyp > wrapper RNAi, BDSC line 29561) in the CG (driver line Nrv2::GFP, tub-GAL80^ts; cyp4g15-GAL4) at the NSC level, at the neuron level, and in orthogonal view. Larvae are dissected after 68 h at 29°C from ALH0. Control, cyp >—(x w¹¹¹⁸) (n ≥ 10 VNCs) and cyp > wrapper RNAi (n ≥ 10 VNCs). CG membrane is visualised by Nrv2::GFP (green), and NSCs are labelled with anti-Dpn (grey). (C) Representative confocal picture of the thoracic VNC for a condition in which wrapper is knocked down by RNAi (BDSC line 29561) from ALH0 in CG, while NSC lineages are marked with the multicolour lineage tracing Raeppli-NLS (blue, white, orange, and red). Raeppli-NLS is induced at ALH0 using hs-Flp and under the control of the LexA/LexAop system (dpn enhancer), while wrapper RNAi is under the control of the GAL4/UAS system (dpn > raeppli-NLS + cyp > wrapper RNAi). Larvae are dissected after 72 h at 29°C from ALH0. See S1 Table for detailed genetics, timing, and conditions of larval rearing. CG membrane was visualised with Nrv2::GFP (green). Dashed white lines highlight examples of NSC lineages encased individually, and dashed yellow lines examples of NSC lineages encased together. (D) Quantification of the percentage of NSC lineages non-individually encased from (C). Control, dpn > raeppli-NLS (n = 10 VNCs) and dpn > raeppli-NLS + cyp > wrapper RNAi (n = 12 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution). Results are presented as box and whisker plots. (E) Representative confocal pictures of thoracic VNCs for which combined NSC lineage and CG drivers (wor + cyp >; driver line Nrv2::GFP, wor-GAL4/CyO; cyp4g15-GAL4, tub-GAL80^ts) were used to perform double RNAi knockdown against nrx-IV and RFP; wrapper and mCherry; and nrx-IV and wrapper. nrx-IV RNAi, BDSC line 32424; wrapper RNAi, VDRC line 105314. Larvae are dissected after 68 h at 29°C. CG membrane is visualised by Nrv2::GFP (green), and NSCs are labelled with anti-Dpn (grey). (F) Quantification of the CG membrane volume per NSC from (E). See Methods for details. nrx-IV + RFP RNAis (n = 8 VNCs), wrapper + mCherry RNAis (n = 9 VNCs), nrx-IV + wrapper RNAis (n = 9 VNCs). Data statistics: Kruskal–Wallis H test with Dunn’s multiple comparisons test. p = 0.0029 for the Kruskal–Wallis H test on grouped dataset. P values from Dunn’s multiple comparisons test are displayed on the graph. Results are presented as box and whisker plots. (G) Representative confocal picture of a thoracic VNC for a condition in which wrapper is overexpressed in NSC lineages from ALH0 (wor > wrapper, driver line Nrv2::GFP, wor-GAL4; tub-GAL80^ts). Larvae are dissected after 68 h at 29°C. CG membrane is visualised by Nrv2::GFP (green), and NSCs are labelled with anti-Dpn (grey). (H) Quantification of the percentage of NSCs non-individually encased from (G). Control, wor >—(x w¹¹¹⁸) (n = 7 VNCs), wor > wrapper (n = 7 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution). Results are presented as box and whisker plots. For all box and whisker plots: whiskers mark the minimum and maximum, the box includes the 25^th–75^th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. ALH, after larval hatching; CG, cortex glia; FISH, fluorescent in situ hydridization; Nrx-IV, Neurexin-IV; NSC, neural stem cell; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g007

These results suggest that Nrx-IV in NSC lineages interact with Wrapper in the CG for ensuring NSC individual encasing. To strengthen this relationship, we performed genetic interactions between nrx-IV and wrapper, comparing CG phenotype between nrx-IV only knowdown, wrapper only knockdown, and double nrx-IV and wrapper knockdown in both NSC lineages and CG (see Methods, S1 Table, and S7D Fig for expression of the combined driver lines; single knockdowns were dose compensated). At a qualitative level, we first observed that all 3 combinations displayed the characteristically large and defined CG chambers associated with nrx-IV loss of function in NSC lineages and wrapper loss of function in CG (Fig 7E). As we could not add a clonal analysis tool to these already complex genotypes, we decided to quantify the volume of CG membrane as a proxy for the density of encasing (i.e., grouped NSC lineages means larger chambers and, conversely, less CG membrane per NSC lineage). We first confirmed the validity of our proxy by assessing whether it was able to detect nrx-IV phenotype. To do so, we measured CG volume between a control condition, shg RNAi in NSC lineages, and nrx-IV RNAi in NSC lineages and found that nrx-IV knockdown resulted in a significantly lower CG volume compared to control, while shg knockdown did not show a decrease (S7E Fig). We then used the same approach to assess NSC encasing upon double knockdown of nrx-IV and wrapper and uncovered that it was significantly lower than upon the individual knockdown of either, which displayed a similar decrease (Fig 7F). These data pinpoint a greater effect of the combined nrx-IV and wrapper knockdowns than their individual contribution on NSC lineage encasing by the CG, showing that the genotype of one affects the phenotype of the other. This demonstrates an epistatic interaction between nrx-IV and wrapper in this context.

We then wondered how the Nrx-IV to Wrapper interaction would fit in the differential adhesion hypothesis. If CG to NSC lineage adhesion is indeed weaker than intralineage adhesion (Fig 2F, panel II.2; A(L^NSC-CG) < A(L^NSC)), overexpression of Wrapper in NSC lineages should not affect the sorting between NSC lineages and CG, since A(L^NSC) would still be superior to A(L^NSC-CG). However, if Nrx-IV to Wrapper interaction is stronger than the sum of intralineage adhesions, then forcing its establishment within the lineage would favour the random grouping of NSC lineages together. We found that misexpressing wrapper in NSC lineages from larval hatching (ALH0), while successful (as confirmed by staining with an antibody that can also detect endogenous Wrapper; see S7F Fig), resulted in very little alteration of CG encasing of individual NSC lineages (Fig 7G and 7H). These data plead in favour of a CG to NSC lineage adhesion through Nrx-IV and Wrapper being weaker than the sum of intralineage adhesions.

Taken together, our results suggest that Nrx-IV in NSC lineages partners with Wrapper in the CG, outside of a septate junction function. This interaction is essential to produce individual encasing of NSC lineages by CG, and in its absence, NSC lineages are randomly grouped in well-defined, larger chambers.

Nrg is required in NSC lineages for integrity of the CG network and individual encasing by CG

We then wondered what could provide strong intralineage adhesion and turned our eyes to Nrg, known to perform homophilic interaction and identified as a candidate of interest in a preliminary screen (S2 Table).

We first assessed Nrg expression in NSC lineages. A protein trap for Nrg (Nrg::GFP, Fig 8A) revealed a strong enrichment between cells of the same NSC lineages following progeny production (ALH72). Moreover, Nrg::GFP was detected at the interface between lineages and CG (yellow arrowhead). However, contrary to what we observed with Nrx-IV::GFP, Nrg::GFP did not appear enriched between NSCs before their encasing by CG (ALH0-24, dashed yellow circle).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Nrg is required in NSC lineages for their individual encasing and CG network integrity.

(A) Representative confocal images of the expression of Nrg at ALH0, ALH24, ALH48, and ALH72 at 25°C. n ≥ 7 VNCs for all time points. Nrg is monitored through a Nrg::GFP fusion (magenta), CG membrane is visualised by cyp4g15-mtd::Tomato (green), NSCs are labelled with anti-Dpn (grey), and neurons are labelled with anti-ElaV (blue). Yellow stars indicate NSC position, dashed yellow circles indicate the lack of Nrg::GFP signal between NSCs, and yellow arrowheads indicate Nrg::GFP signal between NSC and CG. (B) Representative confocal pictures of the thoracic VNC for a condition in which nrg is knocked down by RNAi (wor > nrg RNAi, BDSC line 37496) in NSC lineages (driver line Nrv2::GFP, wor-GAL4; tub-GAL80^ts) at the NSC level, at the neuron level, and in orthogonal view. Larvae are dissected after 24 h at 18°C followed by 54 h at 29°C. Control, wor >—(x w¹¹¹⁸) (n ≥ 10 VNCs) and wor > nrg RNAi (n ≥ 10 VNCs). CG membrane is visualised by Nrv2::GFP (green) and NSCs are labelled with anti-Dpn (grey). White arrows illustrate gaps in CG membranes. This phenotype is seen in 7/10 cases, 3/10 show a milder phenotype. (C) Schematic of the respective NSC and CG patterns in nrg knockdown in NSC lineages. (D) Representative confocal pictures of the thoracic VNC for a condition in which nrg is knocked down by RNAi (BDSC line 37496) in NSC lineages marked with the multicolour lineage tracing Raeppli-NLS (wor > raeppli-NLS, nrg RNAi). Raeppli-NLS (blue, white, orange, and red) is induced at ALH0 using hs-Flp, and RNAi after 24 h at 18°C. Larvae are dissected 60 h after RNAi induction. See S1 Table for detailed genetics, timing, and conditions of larval rearing. CG membrane was visualised with Nrv2::GFP (green). Dashed white lines highlight examples of NSC lineages encased individually, and dashed yellow lines examples of NSC lineages encased together. (E) Quantification of the percentage of NSC lineages non-individually encased from (D). Control, wor > raeppli-NLS (n = 8 VNCs) and wor > raeppli-NLS, nrg RNAi (n = 10 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution). Results are presented as box and whisker plots, where whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. ALH, after larval hatching; CG, cortex glia; Nrg, Neuroglian; NSC, neural stem cell; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g008

We then assessed the impact of nrg knockdown in NSC lineages, using an RNAi line able to significantly decrease Nrg::GFP signal (S8A and S8B Fig, mean of 0.5 normalised to control). Driving nrg knockdown from larval hatching (ALH0) led to few larvae of the right genotype, in which CG displayed some restricted defects in individual encasing of NSC lineages (S2 Table). We thought these animals might have survived due to a weak phenotype and decided to delay the RNAi knockdown, started after 1 day at 18°C (to allow its repression by the GAL80^ts), and then induced for 2 to 3 days (at 29°C). In this case, we obtained more surviving larvae, which displayed strong alterations of NSC individual encasing by CG. NSCs were indeed clumped together, seemingly touching each other (Fig 8B NSC level, compare with Fig 1F’–1G’). Analysis at the level of the neuronal progeny also revealed larger zones devoided of CG membrane (Fig 8B neuron level, compare with Fig 1F”–1G”). In addition, we noticed that, in contrast to nrx-IV knockdown in NSC lineages, CG membranes rather seemed irregular, exhibited gaps and broken/stunted ends, under nrg knockdown (see white arrows for examples in Fig 8B). Driving nrg RNAi in all neurons (ElaV >) also led to defects in the encasing of NSC lineages (S8C Fig). In contrast, nrg knockdown in mature neurons only (nSyb >) or in CG (cyp >) did not lead to observable CG alteration (S8C Fig). We further quantified nrg phenotype by expressing Raeppli-NLS (induced at ALH0) along with nrg RNAi in NSC lineages (wor >). It confirmed the loss of individual encasing, with multiple NSC lineages clumped together and not separated anymore by CG membrane, in a seemingly random fashion (Fig 8D and 8E; mean of 40% of NSC lineages grouped with at least another one). The expressivity of the phenotype was variable, with a class showing nearly no individual encasing (around 100% of NSC lineages grouped with at least another one; Fig 8E) and a class with a range of milder alterations (less than 50% of NSC lineages grouped with at least another one; Fig 8E), something we also noticed qualitatively in the conditions of Fig 8B. In addition, we noticed rare occurrences of CG chambers containing only neurons (S8D Fig).

Altogether, our findings suggest that the expression of Nrg in NSC lineages is required for correct CG network architecture and the individual encasing of NSC lineages.

Intralineage adhesion through Nrg drives cell sorting and individual encasing by CG

Like Nrx-IV, the dual role of Nrg in and outside of septate junction comes from differential splicing [60]. Nrg comes in 2 isoforms, with the same extracellular domain but different intracellular parts (Fig 9A). While the short isoform, Nrg¹⁶⁷, localises in the septate junction of epithelial tissues, the long isoform, Nrg¹⁸⁰, is expressed in neurons of the developing central and peripheral nervous systems [55,56,60].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Individual encasing of NSC lineages relies on strong intralineage adhesion through Nrg.

(A) Schematic depicting the 2 isoforms for Nrg, Nrg¹⁶⁷, and Nrg¹⁸⁰. Only the intracellular C-terminal part differs and will modulate downstream signalling. (B) Representative confocal close-up picture of the localisation of the Nrg¹⁸⁰ isoform in NSC lineages, in a thoracic VNC at ALH72 at 25°C. n = 8 VNCs. All Nrg isoforms are monitored through a Nrg::GFP protein trap (magenta) and the Nrg¹⁸⁰ isoform is detected with a specific antibody (BP104, light blue). (C) Representative confocal close-up picture of the respective localisations of the Nrg¹⁶⁷ and Nrg¹⁸⁰ isoforms in NSC lineages, in a thoracic VNC at ALH72 at 25°C. n = 6 VNCs. The Nrg¹⁶⁷ isoform is visualised by a protein trap in the nrg gene leading to the preferential expression of this isoform (Nrg¹⁶⁷::GFP, yellow). The Nrg¹⁸⁰ isoform is detected with a specific antibody (BP104, light blue). The dashed white line highlights the perimeter of the NSC devoid of BP104 signal. (D) Representative confocal picture of a thoracic VNC and close-up for nrg¹⁸⁰ overexpression in the CG from ALH0 (cyp > nrg¹⁸⁰, driver Nrv2::GFP, tub-GAL80^ts; cyp4g15-GAL4). Larvae are dissected after 68 h at 29°C. CG membrane is visualised by Nrv2::GFP (green), NSCs are labelled with an anti-Dpn (grey), and Nrg¹⁸⁰ is detected with a specific antibody (BP104, magenta). (E) Representative confocal pictures of thoracic VNCs for nrg¹⁶⁷ and nrg^GPI overexpression in the CG from ALH0 (cyp > nrg¹⁶⁷ and cyp > nrg^GPI, respectively, driver Nrv2::GFP, tub-GAL80^ts; cyp4g15-GAL4). Larvae are dissected after 68 h at 29°C. CG membrane is visualised by Nrv2::GFP (green), and NSCs are labelled with an anti-Dpn (grey). (F) Quantification of the percentage of NSCs non-individually encased from (D, E). Control, cyp >—(x w¹¹¹⁸) (n = 10 VNCs), cyp > nrg¹⁸⁰ (n = 11 VNCs), cyp > nrg¹⁶⁷ (n = 7 VNCs), and cyp > nrg^GPI (n = 6 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution) for nrg¹⁸⁰. For nrg¹⁶⁷ and nrg^GPI, there is no variance and, thus, statistics cannot be applied (NA, non-applicable). Results are presented as box and whisker plots. (G) Representative confocal picture of a thoracic VNC and close-up for Nrg¹⁸⁰ overexpression in the NSC lineages from ALH0 (wor > nrg¹⁸⁰, driver Nrv2::GFP, worniu-GAL4; tub-GAL80^ts). Larvae are dissected after 68 h at 29°C. CG membrane is visualised by Nrv2::GFP (green), NSCs are labelled with an anti-Dpn (grey), and Nrg¹⁸⁰ is detected with a specific antibody (BP104, magenta). (H) Representative confocal pictures of thoracic VNCs for nrg¹⁶⁷ and nrg^GPI overexpressed from ALH0 in NSC lineages (wor > nrg¹⁶⁷ and wor > nrg^GPI, respectively, driver line Nrv2::GFP, wor-GAL4; tub-GAL80^ts). Larvae are dissected after 68 h at 29°C. CG membrane is visualised by Nrv2::GFP (green), and NSCs are labelled with anti-Dpn (grey). (I) Quantification of the percentage of NSCs non-individually encased from (G, H). Control, wor >—(x w¹¹¹⁸) (n = 8 VNCs), wor > nrg¹⁸⁰ (n = 8 VNCs), wor > nrg¹⁶⁷ (n = 7 VNCs), and wor > nrg^GPI (n = 6 VNCs). Data statistics: generalised linear model (Binomial regression with a Bernoulli distribution). p = 1.93 × 10⁻¹¹³ for the grouped dataset. P values for individual comparisons test are displayed on the graph. Results are presented as box and whisker plots. For all box and whisker plots: whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. ALH, after larval hatching; CG, cortex glia; Nrg, Neuroglian; NSC, neural stem cell; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g009

We first determined which isoform is expressed in NSC lineages during the larval stage, taking advantage of isoform-specific tools. Staining of Nrg::GFP CNS (ALH72) with an antibody (BP104) specifically recognising the Nrg¹⁸⁰ isoform [60] revealed that Nrg¹⁸⁰ localises in the membranes of all cells from NSC lineages, but not in known septate junctions within the tissue (yellow arrowheads) (Figs 9B and S9A). In neurons, Nrg¹⁸⁰ was not only found in the cell body but appeared also enriched in their axonal bundle, a localisation reported previously [61]. We then took advantage of an Nrg::GFP fusion shown in other tissues to preferentially target the Nrg¹⁶⁷ isoform (called Nrg¹⁶⁷::GFP; [53,55]). Nrg¹⁶⁷::GFP also appeared enriched between cells of the same NSC lineage, where it colocalised with BP104 staining, except on the NSC perimeter, devoided of BP104 (Figs 9C and S9B; see dashed white line for lack of BP104). In contrast, only Nrg¹⁶⁷ is detected in septate junctions. We then wondered whether nrg knockdown was able to lower the levels of both isoforms. nrg knockdown in NSC lineages completely depleted the BP104 signal (S9C and S9D Fig; mean of 0.1 normalised to control). Nrg¹⁶⁷::GFP levels also were strongly decreased upon nrg knockdown (S9E and S9F Fig; mean of 0.25 normalised to control). While it appears in a lower fashion than Nrg¹⁸⁰, it might be due to a higher stability of the Nrg¹⁶⁷::GFP fusion, while endogenous Nrg¹⁸⁰ was detected with an antibody. Taken together, these data suggest that the 2 isoforms of Nrg are expressed, and can be efficiently knocked down, in NSC lineages.

Mostly homophilic interactions (between the same or different isoforms) have been reported for Nrg. We thus wondered whether an Nrg to Nrg interaction within the NSC lineages could fulfil the role of an intralineage adhesion stronger than a CG to NSC adhesion.

Since nrg knockdown in CG (S8C Fig) did not recapitulate nrg knockdown in NSC lineages, the homophilic Nrg interactions between CG and NSC lineages, if existing, are not involved in individual encasing. We then assessed the relevance of intralineage Nrg interactions in the differential adhesion hypothesis (Fig 2F, panel II.2). If such adhesion is stronger than the CG to NSC lineage interaction, then expressing Nrg in CG would force CG to interact with each other. Strikingly, misexpressing Nrg¹⁸⁰ in CG from larval hatching (ALH0) resulted in altered CG morphology and loss of individual encasing of NSC lineages (Fig 9D). CG membranes displayed local accumulation as well as unusual curvature, and NSCs were not separated from each other by CG anymore but were rather found grouped close to each other. Overexpressing Nrg¹⁶⁷ in CG (from ALH0) produced an even more dramatic phenotype, with localised, compact globules of CG membranes and the complete lack of individual encasing of NSC lineages (Fig 9E). Interestingly, overexpression of an Nrg^GPI construct in which the transmembrane and cytoplasmic domains are replaced by a GPI anchor signal [62] also resulted in aggregated CG and clustered NSC lineages (Fig 9E). This shows that intracellular signalling through the divergent C-terminal domain is not required for this sorting of CG and NSC lineages, but rather that adhesion through the extracellular part mediates this effect. The quantification of NSC encasing upon expression of the different Nrg isoforms in CG confirmed our interpretation (Fig 9F).

Altogether, these results demonstrate that providing Nrg homophilic interactions in the CG is sufficient to segregate them from the whole population of NSC lineages, which they normally bind to through a weaker Nrx-IV to Wrapper interaction. This further suggests that Nrg homophilic adhesions between cells of the same NSC lineage are responsible for keeping these cells together and excluding the CG.

If Nrg interactions are indeed responsible for providing binding between cells of the same NSC lineage, including the stem cell, one consequence is that NSCs could bind to each other. Interestingly, Nrg appears expressed in NSCs only after their encasing (see Fig 8A). This fits the idea that early on, when NSCs are not encased yet and separated from other NSCs by the CG, A (NSC-NSC) is kept low. As such, a precocious expression of Nrg in NSCs would be predicted to lead to their grouping (and further the grouping of their neuronal lineages) in a CG chamber. Strikingly, expressing either of the 3 Nrg isoforms from ALH0 resulted in multiple, larger, and well-defined CG chambers containing several NSCs, all in a similar fashion (Fig 9G–9I). As expressing Nrg^GPI also led to the grouping of NSC lineages, it implies that the adhesive role of Nrg is responsible for such effect. This contrasts with the lack of effect of misexpressing Wrapper in NSC lineages (also from ALH0; see Fig 7G and 7H), showing that not all adhesion complexes can lead to A(NSC-NSC) high enough to group NSCs together. These results suggest that a proper timing in establishing intralineage adhesion through Nrg is instrumental in ensuring the individual encasing of NSC lineages by CG.

Nrx-IV and Nrg adhesions are required in NSC lineages for correct axonal path during development

So far, our data show that Nrx-IV- and Nrg-mediated adhesions in NSC lineages are both important for the individual encasing of NSC lineages by CG. We further sought to assess the functional relevance of such adhesions for the cells of the developing NSC lineages themselves. These functions, if any, could be linked to or independent from their role in niche architecture.

We first turned our eyes to the NSCs. We counted their numbers in the VNC and found no significant difference between shg, nrx-IV, and nrg knockdown in NSC lineages and a control condition (S10A and S10B Fig). As NSC core function is dividing to produce differentiated progeny, we assessed NSC proliferation under shg, nrg, and nrx-IV knockdown in NSC lineages, using phospho-histone 3 (PH3) to mark mitotic DNA. We found that both mitotic indexes and phase distribution in mitosis were similar between these conditions and control (S10C–S10E Fig). We further checked whether NSC were still performing asymmetric division, by staining for the asymmetric determinant aPKC, whose localization at the cell cortex is strongly polarised during mitosis along the division and spindle axis [63]. We first recorded whether aPKC was polarised during metaphase and found no significant difference between control and shg, nrg, and nrx-IV knockdown in NSC lineages (S10F and S10G Fig). We also found no significant difference in the alignment between aPKC polarisation and the axis of the mitotic spindle, detected through α-tubulin staining (S10F–S10H Fig). Taken together, these results show that Nrg and Nrx-IV adhesions in the niche are not critical for NSC survival and asymmetric division.

We then wondered whether neurons were affected upon the loss of Nrg and Nrx-IV adhesions.

Following previous studies showing that impaired CG growth and proliferation can lead to neuronal apoptosis [23,27], we first stained for Dcp-1 (apoptotic marker Drosophila cleaved caspase 1) and found no significant difference in the number of positive cells between shg, nrg, and nrx-IV knockdown in NSC lineages and control (S10I–S10L Fig). This implies that the loss of Nrg and Nrx-IV adhesions does not affect neuronal survival at this stage.

During development, immature secondary neurons start sending axons to establish synaptic connections with proper partners in the neuropile, with axons from the same lineage grouped as 1 or 2 tight bundles following the same path [28,61] (Figs 10A and S1A). This axonal fascicle shows a well-defined tract for each lineage, with stereotyped entry in and path within the neuropile. We thus wondered whether disruption of adhesion in NSC lineages could translate into an altered pattern of axonal projections. To assess this possibility, we first marked NSC lineages in a multicolour clonal fashion, this time using a membrane version of Raeppli (CAAX tag) [37] to label both the cell body as well as the extending axons. Due to the high density of marked lineages, we focused on 1 out of the 4 possible fluorophores for display and quantification (Fig 10B and 10C). We first found that the organisation in bundles of axons from neurons of the same lineage appeared preserved in shg, nrx-IV, and nrg knockdowns in NSC lineages compared to control and that most of them still found their way to the neuropile. We, however, noticed a less regular pattern in their path to the neuropile, drifting from the classic boat shape seen from the antero-posterior view (Fig 10A–10C, view 1) and appearing less aligned in a longitudinal view (Fig 10A–10C, view 2). We calculated the angle of axonal extension to the antero-posterior axis of the VNC (Fig 10D). We found that, compared to a control condition at the same stage, the angles were less stereotyped in overall (Fig 10E and 10F), with a broader distribution and slightly shifted, being either more closed (nrx-IV RNAi) or more open (nrg RNAi). We did not detect any significant changes upon shg knockdown in NSC lineages. These data show that Nrx-IV and Nrg adhesions in the NSC lineages influence the extension of axonal tracts from newborn neurons in the VNC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. Loss of Nrx-IV and Nrg adhesions in NSC lineages during development alters axonal projection from newborn secondary neurons.

(A) Schematic of the axonal projections (grey) coming from secondary, newborn neurons generated by NSCs during larval neurogenesis. Only the VNC region (third instar larva) is depicted, and not all lineages are represented. The neuropile is shown in a pink shade. (1) Antero-posterior view. (2) Longitudinal view. (B) 3D reconstruction of a group of NSC lineages visualised with a membrane marker (mTFP1-CAAX) in antero-posterior (1) and longitudinal (2) views for a control condition (wor > raeppli-CAAX), for shg knockdown in NSC lineages (wor > raeppli-CAAX, shg RNAi, VDRC line 27082) and for nrx-IV knockdown (wor > raeppli-CAAX, nrx-IV RNAi, BDSC line 32424) in NSC lineages. Clonal labelling was obtained through the induction of Raeppli-CAAX in NSC lineages at ALH0. Larvae were dissected after 72 h at 29°C. See S1 Table for detailed genetics, timing, and conditions of larval rearing. (C) 3D reconstruction of a group of NSC lineages visualised with a membrane marker (mTFP1-CAAX) in antero-posterior (1) and longitudinal (2) views for a control condition (wor > raeppli-CAAX) and nrg knockdown (wor > raeppli-CAAX, nrg RNAi, BDSC line 37496) in NSC lineages. Clonal labelling was obtained through the induction of Raeppli-CAAX in NSC lineages at ALH0. Larvae are dissected after 24 h at 18°C followed by 60 h at 29°C. See S1 Table for genetics, timing, and conditions of larval rearing. (D) Schematic of the angle (α_L) between the main axonal tract projecting from secondary newborn neuron and the antero-posterior axis. (E) Quantification of the angle α_L depicted in (D) in VNCs for control, shg knockdown, and nrx-IV knockdown in NSC lineages, in the same conditions than shown in (B). Control, wor > raeppli-CAAX (n = 167 axonal projections from 8 VNCs), wor > raeppli-CAAX, shg RNAi (n = 110 axonal projections from 7 VNCs), and wor > raeppli-CAAX, nrx-IV RNAi (n = 143 axonal projections from 8 VNCs). Data statistics: Kruskal–Wallis H test with Dunn’s multiple comparisons test. p < 0.0001 for the Kruskal–Wallis H test on grouped dataset. P values from Dunn’s multiple comparisons test are displayed on the graph. Results are presented as individual values; the line represents the median. (F) Quantification of the angle α_L depicted in (D) in VNCs for control and nrg knockdown in NSC lineages, in the same conditions than shown in (C). Control, wor > raeppli-CAAX (n = 144 axonal projections from 7 VNCs) and wor > raeppli-CAAX, nrg RNAi (n = 129 axonal projections from 6 VNCs). Data statistics: Mann–Whitney U test. Results are presented as individual values; the line represents the median. (G) Schematic of the axonal projections (green) coming from secondary, newborn neurons generated by FasIII-positive NSCs in the third instar larval VNC in a (3) ventral top view. At this stage, FasIII-positive lineages appear as 2 symmetric groups of 4 lineages each side of the midline; pink stars mark the extremities and middle for each of the 4 groups. FasIII-positive axons connect in a ladder pattern in the neuropile (shown in a pink shade). Not all other NSC lineages (grey) are represented. (H) 3D reconstruction of FasIII-positive NSC lineages (stained with an anti-FasIII) in ventral top (3), antero-posterior (1), and longitudinal (2) views for a control condition (wor >—(x w¹¹¹⁸)), for shg knockdown (wor > shg RNAi, VDRC line 27082) in NSC lineages, and for nrx-IV knockdown (wor > nrx-IV RNAi, BDSC line 32424) in NSC lineages. Larvae are dissected after 68 h at 29°C. Pink stars mark the extremities and middle for each of the 4 groups of FasIII-positive NSC lineages. See S1 Table for genetics, timing, and conditions of larval rearing. (I) 3D reconstruction of FasIII-positive NSC lineages (stained with an anti-FasIII) in ventral top (3), antero-posterior (1), and longitudinal (2) views for a control condition (wor >—(x w¹¹¹⁸)) and for nrg knockdown (wor > nrg RNAi, BDSC line 37496) in NSC lineages. Larvae are dissected after 24 h at 18°C followed by 54 h at 29°C. Pink stars mark the extremities and middle for each of the 4 groups of FasIII-positive NSC lineages. See S1 Table for genetics, timing, and conditions of larval rearing. (J) Quantification of the volume of FasIII signal in VNCs for control, for shg knockdown, and for nrx-IV knockdown in NSC lineages, in the same conditions shown in (H). See Methods for volume measure. Volume is normalised to the mean of control. wor >—(x w¹¹¹⁸) (n = 20 VNCs), wor > shg RNAi (n = 10 VNCs), and wor > nrx-IV RNAi (n = 12 VNCs). Data statistics: one-way ANOVA with Tukey’s multiple comparisons test. p < 0.0001 for the one-way ANOVA test on grouped dataset. P values from Tukey’s multiple comparisons test are displayed on the graph. Results are presented as box and whisker plots. (K) Quantification of the volume of FasIII signal in VNCs for control and for nrg knockdown in NSC lineages, in the same conditions shown in (I). See Methods for volume measure. Volume is normalised to the mean of control. wor >—(x w¹¹¹⁸) (n = 9 VNCs) and wor > nrg RNAi (n = 10 VNCs). Data statistics: unpaired Student t test. Results are presented as box and whisker plots. For all box and whisker plots: whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. Nrg, Neuroglian; Nrx-IV, Neurexin-IV; NSC, neural stem cell; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3002352.g010

We then focused on one specific subset of neurons, with stereotyped projections we could track and measure, and identified by the expression of the cell adhesion protein FasIII [64]. FasIII-positive neurons include motor neurons, RP1, RP3 (also called VL3/4), and RP4 [64–66], born from the activity of specific NSCs (neuroblasts NB3-1 [67]) already at embryonic stages. While these lineages have been mostly characterised during embryogenesis with the formation of primary axonal tracts [64,65,68,69], further neurons and corresponding secondary tracts are added during larval development, as with other lineages [70,71]. At larval stage, FasIII-positive lineages appear as 2 symmetric rows of NSC lineages (cell bodies and axons) on each side of the midline (Fig 10G and controls in Fig 10H and 10I, green tracts and pink stars). FasIII-positive axons connect and generate a ladder pattern at the neuropile level. We stained for FasIII upon shg, nrx-IV, and nrg knockdowns compared to a control condition and observed a strong decrease in FasIII signal when Nrx-IV and Nrg adhesions were disrupted (Fig 10H–10K). In contrast, we did not detect significant changes upon shg knockdown in NSC lineages. These data indicate that FasIII-positive motoneurons are altered upon loss of Nrx-IV and Nrg adhesions, displaying decreased expression or localisation of FasIII in their axonal projection.

Altogether, our results show that adhesions within NSC lineages and between NSC lineages and the niche are important for axonal features and path of newborn, still immature neurons.

The function of Nrx-IV and Nrg adhesions in NSC lineages during development shapes adult locomotor behaviour

Our data show that a balance of specific adhesions is required in the neurogenic niche both for its architecture and for developing neuronal circuits. Whether it has any physiological relevance for the health of the resulting mature organism or is a transient phenomenon whose impact is later resolved remain unknown. Beyond selected cases with dramatic outcomes, causally linking defined developmental defects to specific functions of the adult CNS is rarely achieved. Our biological context and findings are particularly relevant to test causality between specific neurodevelopmental processes and adult neurological functions. While secondary neurons are generated during larval development, they are not functional at this stage, but in adults, following their maturation and integration into remodelled circuits during the pupal period [72]. The impact of developmental parameters on their function thus will be detectable, and has to be assessed, at the adult stage.

We decided to determine whether the loss of Nrx-IV and Nrg adhesions in NSC lineages during development impairs adult neurological function. As our analyses of CG and axonal tract phenotypes have focused on the VNC, in which motor neurons are produced (such as the FasIII-positive neurons), we focused on locomotor parameters in the adult. To do so, we took advantage of an ethoscope-based tracking system [73] to record locomotion metrics such as fraction of time moving, velocity, and circadian activity (Fig 11A). This high-throughput platform relies on video acquisition to record positional data in real time for multiple individual flies. Several behavioural parameters can be extracted by calculating the position of the fly overtime, including the fraction of time moving, locomotion speed (velocity), and circadian activity. Statistics on several flies draw an average behaviour for the population.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. Loss of Nrx-IV and Nrg adhesions in NSC lineages during development results in locomotor hyperactivity in the resulting adults.

(A) Schematics of the behavioural analysis performed on adult flies in which alterations on different adhesion complexes have been performed during the larval stage. The activity of individual flies is recorded by video tracking in ethoscopes. Statistical analysis on the populations allows to infer an average behaviour. (B) Schematics of the rearing regimen for non-induced and induced conditions. In non-induced conditions, animals are constantly kept at 18°C before the recordings, a temperature allowing the repression of the GAL4/UAS system by the thermosensitive GAL80^ts and thus blocking expression of the RNAi under the control of the worniu driver (wor-GAL4, tub-Gal80^ts). In induced conditions, the animals are transiently shifted to 29°C from early larval stage to late larval stage/early pupal stage, a temperature allowing the expression of GAL4/UAS system and thus of the RNAi. Animals were kept at 18°C both during embryonic development and from early pupal stage. Adult flies were assessed 7–10 days after eclosion. See S1 Table for genetics, timing, and conditions of larval rearing. (C-E) For all metrics. Control (wor-GAL4, tub-Gal80^ts x w¹¹¹⁸), n = 35 non-induced adult males and n = 55 induced adult males. shg RNAi (wor-GAL4, tub-Gal80^ts x shg RNAi^VDRC27082), n = 35 non-induced adult males and n = 55 induced adult males. nrx-IV RNAi (wor-GAL4, tub-Gal80^ts x nrx-IV RNAI^BDSC32424), n = 35 non-induced adult males and n = 55 induced adult males. nrg RNAi (wor-GAL4, tub-Gal80^ts x nrg RNAI^BDSC37496), n = 35 non-induced adult males and n = 55 induced adult males. (C) Plot representing the percentage of global time moving (measured as the fraction of time moving within 30-min intervals), in non-induced and induced conditions. (D) Fraction (%) of the time moving across light/dark cycles (% moving, ratio between total moving time and total time) in non-induced and induced conditions. Black arrows indicate anticipation of the evening peak for nrx-IV and nrg knockdowns. Data statistics: Kruskal–Wallis H test with Dunn’s multiple comparisons test for both non-induced and induced conditions. p (non-induced) < 0.0001 and p (induced) < 0.0001 for the Kruskal–Wallis H test on grouped dataset. P values from Dunn’s multiple comparisons test are displayed on the graph. Results are presented as box and whisker plots. (E) Mean velocity (in relative units) across light/dark cycles in non-induced and induced conditions. Data statistics: Kruskal–Wallis H test with Dunn’s multiple comparisons test for both non-induced and induced conditions. p (non-induced) = 0.0115 and p (induced) < 0.0001 for the Kruskal–Wallis H test on grouped dataset. P values from Dunn’s multiple comparisons test are displayed on the graph. Results are presented as box and whisker plots. For all box and whisker plots: whiskers mark the minimum and maximum, the box includes the 25th–75th percentile, and the line in the box is the median. Individual values are superimposed. The data underlying this figure’s quantifications can be found in S1 Data. Nrg, Neuroglian; Nrx-IV, Neurexin-IV; NSC, neural stem cell.

https://doi.org/10.1371/journal.pbio.3002352.g011

We recorded locomotion metrics for shg, nrx-IV, and nrg knockdowns in NSC lineages, as well as for a control line, in 2 conditions. First, RNAi expression was only allowed during larval phase and prevented shortly after pupariation (see Methods and Fig 11B, “induced” condition). Second, gene knockdowns were never activated (same genetic background, but RNAi always off; see Methods and Fig 11B, “non-induced” condition).

We first look at the overall pattern of activity through a 24-h light:dark (LD) cycle. In LD 12 h:12 h conditions, Drosophila indeed displays a characteristic rest/activity pattern where they become highly active in anticipation of the transitions between light and dark periods. Rest/sleep takes place mostly during the night and in the middle of light and dark periods. We found that this pattern of activity was kept in the different lines in induced condition, with 2 main peaks of activity (morning and evening, Fig 11C). In addition, we noticed a slightly higher anticipation for the evening peak in the case of nrx-IV and nrg knockdowns (black arrows), as well as wider peaks for nrg knockdown. We then quantified activity metrics by measuring the fraction (%) of time flies spent moving (Fig 11D). We first found a stunning change in the behaviour of nrg RNAi flies, which spent 53% of their time moving, while control flies only spent 19% of their time moving (Fig 11D, induced). This dramatic locomotor hyperactivity was apparent throughout both light and dark periods (S11A and S11B Fig). nrx-IV RNAi flies also spend significantly more time moving, which was increased to 27% of their time (Fig 11D, induced). In contrast, shg RNAi flies appeared similar to control, spending 21% of their time moving. In the non-induced condition, nrx-IV RNAI, nrg RNAi, and control mostly behaved in a similar fashion (Fig 11D; ctrl = 23%; nrx-IV = 27%; nrg = 28% of time sleeping), yet with shg RNAi displaying increased moving (shg = 32%).

We wondered whether this locomotor hyperactivity was only visible as the time flies spent moving, or also in the way they were moving. We then determined the speed of locomotion for the different lines (Fig 11E). In induced conditions, we found that the mean velocity throughout the cycle was significantly increased in nrg (3.5, in relative unit (see Methods)) and nrx-IV (2.8) but not shg (2.1) knockdowns compared to control condition (2.3). In non-induced conditions, all lines exhibited similar values of velocity (ctrl = 2.0; shg = 2.2; nrx-IV = 2.3; nrg = 2.1).

These data show that the loss of Nrx-IV- and Nrg-based adhesions in NSC lineages specifically during development affects adult locomotor behaviour. During the same period, these complexes are also important for building correct niche architecture. We wondered whether these 2 functions were linked or distinct. Nrg perform homophilic interactions within NSC lineages, what makes identifying the contribution of CG-dependent mechanisms challenging. In contrast, Nrx-IV in the NSC lineages works with Wrapper in the CG to set up specific adhesion required for correct niche architecture around NSC lineages. As such, assessing locomotion metrics of wrapper knockdown in the larval CG would determine the contribution of such NSC lineage to niche interaction to adult behaviour, beyond potential NSC lineage-intrinsic roles of Nrx-IV. As wrapper function is required in the CG for its architecture, it would by itself reveal the dependency of adult locomotion on the developing niche.

We recorded locomotion during a 24-h LD cycle and found that wrapper RNAi flies had increased locomotor activity levels, especially for the evening peak (S12A and S12B Fig). They spent more time moving compared to control (28% compared to 23% of their time, S12C Fig) in induced conditions, a difference we did not detect in non-induced condition (26% compared to 26% of their time, S12C Fig). wrapper RNAi flies also displayed increased velocity (ctrl = 1.9 and wrapper = 2.4, S12D Fig). These results show that wrapper knockdown in the CG during larval development leads to locomotor hyperactivity and thus that its function in the developing niche is required for proper locomotor behaviour in the adult.

Altogether, our results show that the functions of Nrg and Nrx-IV adhesions in the NSC lineages during development are necessary for proper locomotor activity in the adult. This causally links specific processes during CNS development to adult neurological functions. In addition, the results obtained with wrapper knockdown in the CG first show that properties of the developing niche shapes adult locomotor behaviour and also suggest that it might be linked to the function of Nrx-IV to Wrapper adhesion in niche architecture. We could not determine whether nrg role in adult locomotion was linked to its role in niche formation or rather the output of other NSC lineage-intrinsic functions during development.

Fly lines and husbandry

Drosophila melanogaster lines were raised on standard cornmeal food at 25°C. Lines used in this study are listed in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Drosophila transgenic lines used in this study.

https://doi.org/10.1371/journal.pbio.3002352.t002

Larval culture and staging

Embryos were collected within 2 to 4 h window on grape juice agar plates and kept at 25°C for 20 to 24 h. Freshly hatched larvae were collected within a 1-h time window (defined as 0 h after larval hatching, ALH0), transferred to fresh yeast paste on a standard cornmeal food plate and staged to late first instar (ALH24), late second instar (ALH48), mid third instar (ALH72), and late third instar (ALH96).

For growth on quinic acid, food plates were prepared by mixing 250 mg/ml stock solution of quinic acid (dissolved in sterile water) into melted food at 50°C for a final concentration of 20 mg/ml of quinic acid.

For D. melanogaster, 25°C was used as the normal developmental temperature. To keep the thermosensitive allele of GAL80 (GAL80^ts), a repressor of GAL4, active (so to switch off GAL4 expression), 18°C was used, while 29°C was used for its inactivation (so to switch on GAL4 expression). For all experiments using the GAL80^ts, embryogenesis happened at 18°C, unless noted otherwise. This was to prevent a contribution of embryonic function to the larval phenotype.

Detailed genotypes, crosses, and culture regimens are listed in S1 Table.

DNA cloning and Drosophila transgenics

A portion of the cyp4g15 enhancer (GMR55B12, Flybase ID FBsf0000165617), which drives in the CG and (some) astrocyte-like glia, was amplified from genomic DNA extracted from cyp4g15-GAL4 adult flies, with a minimal Drosophila synthetic core promoter [DSCP] [81] fused in C-terminal. For creating cyp4g15-FRT-STOP-FRT-GAL4, an FRT STOP cassette was amplified from a UAS-FRT.STOP-Bxb1 plasmid (gift from MK. Mazouni), and the GAL4 sequence was amplified from the entry vector pENTR L2-GAL4::p65-L5 (gift from M. Landgraf). The 2 amplicons were joined together by overlapping PCRs. This FRT-STOP-FRT-GAL4 amplicon together with the cyp4g15^DSCP enhancer were inserted in the destination vector pDESThaw sv40 (gift from S. Stowers) using the Multisite gateway system [82] to generate a cyp4g15^DSCP-FRT-STOP-FRT-GAL4 construct. The construct was integrated in the fly genome at an attP2 or attP40 docking sites through PhiC31 integrase-mediated transgenesis (BestGene). Several independent transgenic lines were generated and tested, and one was kept for each docking site.

The grainyhead (grh) D4 enhancer (4 kb from the second intron of the grainyhead gene), which has been reported to drive in postembryonic NSCs [8,42], was amplified from genomic DNA extracted from grh (NB)-GAL4 adult flies, with an hsp70 promoter fused in C-terminal. For creating grh^D4-FLP, the FLP DNA, which codes for the flippase enzyme, was amplified from the plasmid pMH596 (Addgene 52531) and was joined to the grh^hsp70 enhancer using the Multisite gateway system in the destination vector pDESThaw sv40. The construct was integrated in the fly genome at an attP18 docking site through PhiC31 integrase-mediated transgenesis (BestGene). Several independent transgenic lines were generated and tested, and one was kept (grh^D4-FLP).

For creating QUAS-PTEN, the PTEN coding sequence was amplified from genomic DNA extracted from UAS-PTEN [83] adult flies, as described in [84]. This amplicon together with the QUAS sequence (pENTRY L1-QUAS-R5, gift from S.Stowers) were joined using the Multisite gateway system [82] in the destination vector pDESThaw sv40. The construct was integrated in the fly genome at an attP40 docking site through PhiC31 integrase-mediated transgenesis (BestGene). Several independent transgenic lines were generated and tested, and one was kept (QUAS-PTEN).

Generation of UAS-Raeppli-NLS 89A and LexAOp-Raeppli-NLS 89A

The original construct (BDSC 55088), placing Raeppli NLS under the control of both UAS and LexAOp sequences, was crossed to a Cre recombinase line (BDSC 851) to randomly excise one of the 2 control sequences. The resulting lines were checked by PCR to determine whether they carried the UAS or LexAop version.

Fixed tissue immunohistochemistry

For immunohistochemistry, CNS from staged larvae were dissected in PBS, fixed for 20 min in 4% formaldehyde diluted in PBS, washed 3 times in PBS-T (PBS+0.3% Triton X-100), and incubated 2 nights at 4°C with primary antibodies diluted in blocking solution (PBS-T, 5% bovine serum albumin, 2% normal goat serum). After washing 3 times in PBS-T, CNS were incubated overnight at 4°C or 3 to 4 h at room temperature with secondary antibodies (dilution 1:200) diluted in blocking solution. Brains were washed 3 times in PBS-T and mounted in Mowiol mounting medium on a borosilicate glass side (number 1.5; VWR International). For the Nrx-IV antibody, CNS were fixed for 3 min in Bouin’s fixative solution (Sigma Aldrich, HT10132), and the rest of the protocol was identical. Primary antibodies used were as follows: guinea pig anti-Dpn (1:5,000; [27]), chicken anti-GFP (1:2,000, Abcam ab13970), rat anti-ELAV (1:100, DSHB 7E8A10-c), mouse anti-ELAV (1:100, DSHB 9F8A9-c), rat anti-dE-cadherin (1:50, DSHB DCAD2), mouse anti-Armadillo (1:50, DSHB N2 7A1), rabbit anti-Repo (1:10,000, kind gift from B. Altenheim), mouse anti-Repo 1:100 (DSHB 8D12-c), mouse anti-Prospero (1:100, DSHB MR1A), rabbit anti-Asense (1:3,000, kind gift from the Yan lab), rabbit anti-Phospho-histone H3 (1:100, Millipore 06–570), rabbit anti-Nrx-IV (1:1,000; [51]), mouse anti-wrapper (1:20, DSHB 10D3, supernatant), mouse anti-Nrg¹⁸⁰ (1:50, DSHB BP104, supernatant), rabbit anti-Dcp-1 (1/100, Cell Signalling 9578S), rabbit anti-aPKc (1/500, Santa Cruz Biotechnology sc-17781), mouse anti-α-tubulin (1/200, clone DM1A, Cell Signalling Technology #3873), mouse anti-FasIII (1/50, DSHB 7G10, concentrate). Fluorescently conjugated secondary antibodies Alexa Fluor 405, Alexa Fluor 488, Alexa Fluor 546, and Alexa Fluor 633 (Thermo Fisher Scientific) were used at a 1:200 dilution. DAPI (4′,6-diamidino-2-phenylindole, Thermo Fisher Scientific 62247) was used to counterstain the nuclei.

RNA FISH (HCR in situ hybridization)

We used the Multiplexed HCR RNA-FISH technique [85,86], using reagents and adapted protocols from Molecular Instruments. First, wrapper probe set was designed using software from Molecular Instruments (https://www.molecularinstruments.com/). RNA FISH was then performed as follows. First, CNS from staged larvae were dissected in PBS, fixed for 20 min in 4% formaldehyde diluted in PBS, washed 3 times 5 min in PBS and prehybridized in 200 μl of hybridization buffer for 30 min at 37°C. Meanwhile, 0.8 pmol (0.8 μl of 1 μM stock) of the probe set was added to 200 μl of probe hybridisation buffer and prehybridised for 30 min at 37°C. The prehybridization solution was removed from the fixed samples, which were then incubated with the probe solution at 37°C overnight. Samples were washed 4 × 15 min with 500 μl warmed (37°C) wash buffer, followed by 5 min with 500 μl of 50% wash buffer/50% 5XSSC-0.1% Tween and, finally, by 2 × 5 min with 500 μl of 5XSSC-0.1% Tween at room temperature. Samples were further preamplified with 100 μl of amplification buffer for 10 min at room temperature. In the meantime, 6 pmol of hairpin h1 and 6 pmol of hairpin h2 (2 μL of 3 μM stock) were individually prepared by heating at 95°C for 90 s followed by cooling to room temperature for 30 min in the dark. Snap-cooled hairpins h1 and h2 are then added to 100 μL of amplification buffer at room temperature. The preamplification solution was removed from the samples, which were incubated in the hairpin solution overnight (>12 h) in the dark at room temperature. The samples were washed 2 × 5 min, 2 × 30 min, and 1 × 5 min with 5XSSC-0.1% Tween before mounting on slides.

For the ALH0 time point, 4 pmol (4 μl of 1 μM stock) of the probe set was used instead of 0.8 pmol.

Image acquisition and processing

Confocal images were acquired using a laser scanning confocal microscope (Zeiss LSM 880, Zen software (2012 S4)) with a Plan-Apochromat 40×/1.3 oil objective. All brains were imaged as z-stacks with each section corresponding to 0.3 to 0.5 μm. The spectral mode was used for acquiring pictures of Raeppli clones. Images were subsequently analysed and processed using Fiji (Schindelin, J. 2012), Volocity (6.3 Quorum Technologies), and the Open-Source software Icy v2.1.4.0 (Institut Pasteur and France Bioimaging, licence GPLv3). Denoising was used for some images using the Remove noise function (Fine filter) in Volocity. Images were assembled using Adobe Illustrator 25.4.6.

shg^null MARCM clones

shg^64R, FRT42B; cyp4g15-mtd::Tomato flies were crossed to the TUG13 MARCM line (y,w, hs-FLP; FRTG13, tubP-GAL80[LL2]/ (CyO, act-GFP[JMR1]); tubP-GAL4[LL7], UAS-mCD8-GFP[LL6]/TM6B). The resulting progeny was let to develop at 25°C, then subjected to 37°C heatshock either at 14 to 18 h after egg laying for 2 h or at ALH48 for 30 min, and dissected at ALH72 (S4C Fig).

Quantification of Type II chambers and clones

Each CB in which either control or brat tumour Type II lineages were marked with Raeppli-NLS was manually scored throughout the entire dorso-ventral z-stack for the number of CG chambers containing marked lineages. A CG chamber was defined by the existence of a continuous Nrv2::GFP signal around a volume of adjacent cells. In the brat condition, some chambers were very big and convoluted and, thus, difficult to score. We decided to eliminate the corresponding CBs. The number of differently coloured clones and the total number of colours (for scattered cells) within one chamber (and within the entire CB) were also recorded manually.

Quantification of NSC (Dpn staining) and NSC lineage encasing (Raeppli-NLS)

NSCs are identified by Dpn staining and NSC lineages by the use of the multicolour clonal tool Raeppli-NLS, whose early induction under the control of worniu-GAL4 leads to the labelling of clonally related cells with the same colour. CG are tracked by a membrane marker (either Nrv2::GFP or cyp4g15-myr::mtd-Tomato). A CG chamber is defined as a continuous layer of membranes around a group of cells, regardless of their identity. The continuity is manually assessed by visualisation through the entire z-stack, also making use of orthogonal views.

One NSC is scored as individually encased if no other NSC (Dpn⁺ cell) can be detected within the same CG chamber. Conversely, one NSC is scored as non-individually encased if at least one other NSC (Dpn⁺ cell) can be detected within the same CG chamber. The percentage of non-individually encased NSCs is calculated by dividing the count of non-individually encased NSCs by the total count of NSCs in each VNC.

One NSC lineage is scored as individually encased if no more than one colour of Raeppli-NLS can be detected within the same CG chamber. Conversely, one NSC lineage is scored as non-individually encased if at least one other colour for Raeppli-NLS can be detected within the same CG chamber (i.e., 2 different colurs are not fully separated by CG membrane). The percentage of non-individually encased NSC lineages is calculated by dividing the count of non-individually encased NSC lineages by the total count of NSC lineages in each VNC.

Clonal analyses using CoinFLP in the CG

The Coin-FLP method [38] was used to induce rare clones of PTEN- or reaper-overexpressing CG cells, by crossing cyp4g15-FLP; CoinFLP GAL4::LexA; UAS-mCD8::RFP females to Nrv2::GFP/CyO; UAS-PTEN or Nrv2::GFP, tub-GAL80^ts/CyO; UAS-reaper males. Crosses to Nrv2::GFP/CyO and Nrv2::GFP, tub-GAL80^ts/CyO males were respectively used for control. For PTEN overexpression clones, larvae were maintained at 25°C and staged to ALH48-ALH72 for dissection. For reaper overexpression clones, larvae were maintained at 18°C for 96 h after larval hatching then shifted to 29°C for 48 h and dissected.

Clone volume (using the mCD8::RFP signal) was segmented and measured in Volocity 6.3 (Quorum Technologies) using adjusted protocols for intensity thresholding. Threshold were corrected on each sample to exclude background fluorescence. NSC number per clone was scored manually by counting the number of NSCs encased in RFP⁺ membrane. Number of non-encased NSCs around clones was scored manually by counting the NSCs, which were not individually encased by the CG (marked with Nrv2::GFP).

Clonal analyses using CoinFLP in Type II NSC lineages

To generate brat RNAi clones in Type II NSC lineages, grh-FLP^D4; CoinFLP GAL4::LexA females were crossed to Nrv2::GFP, UAS-mCD8::RFP/CyO; UAS-brat RNAi males. A cross to Nrv2::GFP, UAS-mCD8::RFP/CyO males was used for control. Larvae were maintained at 25°C and staged to ALH72 to ALH96 for dissection. RFP⁺ clones were manually assessed for complete enclosure by CG membrane (marked with Nrv2::GFP).

Cortex glial membrane volume measurements

Nrv2::GFP signal was used as a proxy for CG membrane signal. Each VNC was sampled with 2 rectangular prisms (x = 100 μm; y = 50 μm; z = 3 μm) on each part of the midline, centred on the neuron level of the chamber (where nrx-IV and wrapper phenotypes are the most visible and devoid of trachea or nerve signal). The total volume of CG membrane for these areas was segmented and measured in Volocity 6.3 (Quorum Technologies) using adjusted protocols for intensity thresholding. Threshold were corrected on each sample to exclude background fluorescence.

Quantification of levels of RNAi knockdown

For each gene, one reporter of expression was chosen and the sum of its own volume in a defined volume within the VNC was measured under control and RNAi genotypes. Reporter volume was measured in Volocity 6.3 (Quorum Technologies) using adjusted protocols for intensity thresholding. Threshold were corrected on each sample to exclude background fluorescence. Each value was then normalised to the mean of the control. The conditions for each gene were as follows (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Choices for the reporter and tissue volume used to quantify RNAi knockdowns corresponding to the different adhesion molecules.

https://doi.org/10.1371/journal.pbio.3002352.t003

Quantification of NSC number and mitotic index

CNSs of the chosen genotypes were stained with Dpn and phospho-histone H3 antibodies to detect NSC identity and mitosis, respectively. Quantification was performed on NSCs from the thoracic part of the VNC. Mitotic phases (Prophase, Prometaphase/metaphase, Anaphase, and Telophase) were manually determined by the localization and pattern of PH3⁺ DNA and Dpn staining. Normalised mitotic index corresponds to the ratio between mitotic NSCs over all NSCs, then divided by the mean of this ratio for the control sample.

Quantification of aPKC polarisation

Quantification was performed on NSCs from the thoracic part of the VNC. aPKC was detected using a specific antibody, mitotic spindles were labelled with anti-α-tubulin, and DNA was counterstained with DAPI. NSCs were recognised by their size. aPKc localization was assessed during metaphase, recognised through its stereotypic DNA staining (metaphasic plate), and scored as polarised when a clear crescent was seen restricted to part of the NSC. The mitotic spindle was considered aligned when perpendicular to the main axis of aPKC crescent and of the metaphasic plate and considered misaligned when titled compared to this axis. The mitotic spindle was finally scored as undetermined when we could not detect it on all its length, often when not parallel to the plan of confocal acquisition. Percentages per VNC were calculated by dividing the total count of each of the 3 categories by the sum of the counts.

Quantification of apoptotic (Dcp-1⁺) cells

CNSs of the chosen genotypes were stained with anti-Dcp-1 antibody to detect apoptotic cells. Quantification was performed on the thoracic part of the VNC. Dcp-1⁺ cells were detected in Volocity 6.3 (Quorum Technologies) using adjusted protocols for intensity thresholding and object detection. Threshold was corrected on each sample to exclude background fluorescence.

Quantification of axonal angle

Z stacks of Raeppli CAAX clones induced in NSC lineages were visualised in Volocity (6.3 Quorum Technologies), in a lateral view along the antero-posterior axis (xz axis). The brightest colour (mTFP1) was chosen and lines were drawn parallel to and following the main axonal projection for each clone. Pictures of the line were recorded through snapshots and imported into Icy v2.1.4.0 where the Angle Helper plugin was used to measure the angle formed with the intersection with the AP axis (Fig 10D).

Quantification of FasIII signal

CNSs of the chosen genotypes were stained with a specific anti-FasIII antibody. Quantification was performed on a standardised section of the thoracic VNC spanning the 3 central FasIII⁺ NSC lineages (Fig 10G), cut left of the most anterior lineage and right of 2 commissural axons after the most posterior lineage. Total FasIII volume was detected in Volocity 6.3 (Quorum Technologies) using adjusted protocols for intensity thresholding. Threshold was corrected on each sample to exclude background fluorescence. Each value was normalised to the mean of the respective control sample.

Behavioural analysis

The locomotor activity of individual flies was measured with the Drosophila ethoscope activity Open Source system [73] at 23°C. Young adult males (7 to 10 days) were individually placed in 6.5 cm transparent tubes containing 1.5 cm nutrient medium (agarose 2% (p/v), sucrose 5% (p/v)) closed with wax at one end and cotton at the other. A total of 20 tubes were positioned in each ethoscope, and flies were first entrained to 12 h:12 h LD cycles for 3 days and their activity was recorded for 3 more days. The activity data analysis was done with the R software, using Rethomics packages [87]. The percentage of time moving across LD cycles was measured as the fraction of time moving within 30-min intervals, whereas global time moving per fly corresponds to a total moving over total time ratio. Velocity was measured using a previously described tracking algorithm [87] and is expressed in relative units. Genotypes positions in 20 tubes arenas were changed from one experiment to another to avoid positional bias.

The genotypes and sample sizes were the following:

• Figs 11 and S11: wor-GAL4, tub-GAL80^ts x w¹¹¹⁸; wor-GAL4, tub-GAL80^ts x shg RNAi^VDRC27082; wor-GAL4, tub-GAL80^ts x nrx-IV RNAi^BDSC32424; wor-GAL4, tub-GAL80^ts x nrg RNAi ^BDSC37496 for knockdown in NSC lineages. A total of 35 non-induced flies and 55 induced flies, coming from 8 independent experiments, were analysed for each genotype.
• S12 Fig: tub-GAL80^ts; cyp4g15-GAL4 x w¹¹¹⁸; tub-GAL80^ts; cyp4g15-GAL4 x wrapper RNAi ^BDSC29561 for knockdown in CG. A total of 20 non-induced flies and 20 induced flies, coming from 3 independent experiments, were analysed for each genotype.

In non-induced flies, RNAi expression was always kept off by maintaining the progeny issued from the cross at 18°C at all times (ALH0 to assay time). In induced flies, RNAi expression was only activated during larval stage: The progeny was kept at 18°C from egg laying to ALH24, then shift to 29°C from ALH24 to just after pupariation/early pupae (shifted back to 18°C within 1 day at 29°C from wandering L3 larvae), and finally maintained at 18°C until the behavioural assay).

Statistics and reproducibility

Statistical tests used for each experiment are stated in the figure legends. Statistical tests were performed using GraphPad Prism 7.0a or in R for the generalised linear model (Binomial regression with a Bernoulli distribution). For all box and whisker plots, whiskers mark the minimum and maximum, the box includes the 25th to 75th percentile, and the line in the box is the median. Individual values are superimposed.

For all knockdown (RNAi) and misexpression experiments, penetrance was always 100% (n numbers are indicated in the figure legends), and expressivity was represented through sample distribution for the parameter assessed.