Differentiation signals from glia are fine-tuned to set neuronal numbers during development

Curation statements for this article:
  • Curated by eLife

    eLife logo

    Evaluation Summary:

    This manuscript will be of interest to cell and developmental biologists and neuroscientists. It addresses the question of how the number of connecting neurons in a circuit is matched whilst maintaining topography. It shows that non-autonomous control of neuronal number involves a relay mechanism through two distinct glial cell types, enabling the specification of distinct neuronal classes.

    This manuscript was co-submitted with: https://www.biorxiv.org/content/10.1101/2022.02.21.481306v1

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

This article has been Reviewed by the following groups

Read the full article See related articles

Abstract

Neural circuit formation and function require that diverse neurons are specified in appropriate numbers. Known strategies for controlling neuronal numbers involve regulating either cell proliferation or survival. We used the Drosophila visual system to probe how neuronal numbers are set. Photoreceptors from the eye-disc induce their target field, the lamina, such that for every unit eye there is a corresponding lamina unit (column). Although each column initially contains ∼6 post-mitotic lamina precursors, only 5 differentiate into neurons, called L1-L5; the ‘extra’ precursor, which is invariantly positioned above the L5 neuron in each column, undergoes apoptosis. Here, we showed that a glial population called the outer chiasm giant glia (xg O ), which resides below the lamina, secretes multiple ligands to induce L5 differentiation in response to EGF from photoreceptors. By forcing neuronal differentiation in the lamina, we uncovered that though fated to die, the ‘extra’ precursor is specified as an L5. Therefore, two precursors are specified as L5s but only one differentiates during normal development. We found that the row of precursors nearest to xg O differentiate into L5s and, in turn, antagonise differentiation signalling to prevent the ‘extra’ precursors from differentiating, resulting in their death. Thus, an intricate interplay of glial signals and feedback from differentiating neurons defines an invariant and stereotyped pattern of neuronal differentiation and programmed cell death to ensure that lamina columns each contain exactly one L5 neuron.

Article activity feed

  1. Author Response

    Reviewer #2 (Public Review):

    The authors argue that xgO secretes Spi and Col4a1 to induce MAPKdependent L5 differentiation. However, no loss-of-function condition for these putative ligands was tested. Since they speculated that expression of Spi and Col4a1 alone may not lead to a sufficient level of MAPK activity, the results of their loss of function conditions have to be included in the paper.

    We agree with Reviewer #2 completely. Our manuscript now includes spi and Col4a1 loss-of-function data specifically in xgO, which has strengthened our manuscript considerably and allows us to draw stronger conclusions as to the roles of Spi and Col4a1.

    The authors found ectopic L5 neurons when apoptosis was repressed (Fig. 1). It is likely that cells that fail to differentiate to L5 are removed by apoptosis, but this link was not clearly demonstrated in the paper. As a result, there is a gap between the data in Fig. 1 (section 1 in the text) and the other part of the paper. The relationship between Fig. 1 and the other data should be carefully discussed. In my opinion, the first section of Results should be moved after the last section so that the results of Fig. 1 are explained as a potential mechanism to remove cells that failed to differentiate to L5.

    We have restructured the manuscript as suggested.

    Reviewer #3 (Public Review):

    There is considerable overlap with Fernandes et al 2017 Science paper: (1) That EGFR signalling is required for L5 neuron survival had been shown in their Fernandes et al 2017 Science paper, as over-expression of p35 rescued apoptosis caused by EGFRDN. Now, using Dronc mutants in the current manuscript is an equivalent experiment. (2) In Fernandes et al 2017 Science, they over-express activated MAPK in lamina neurons (Fig.1G), and in the current, they over-express its target Pnt-P1 (Fig.1I) - equivalent experiment. (3) Figure S1 reports Lamina>MAPKACT rescues Bsh and Spl2 positive neurons. These data are similar to those reported in Fernandes et al 2017 Science, where they showed the rescue of lamina neurons with this same genotype. (4) rho3 mutants cannot secrete Spi and L1-4 cannot differentiate and only a few L5 do (Fernandes et al 2017 Science), they then rescued this phenotype including L5s by over-expressing EGFRACT or Ras in wrapping glia (Figure 2F-I). With the submitted manuscript, they rescue with rho3 overexpression in photoreceptors - genetically different, but rather similar, as together they demonstrate that rescue of L5 requires rho or spi. These close similarities reduce the appeal and novelty of the current manuscript.

    We agree with the reviewer that our previous work established that MAPK signalling was necessary and sufficient to drive premature neuronal differentiation in the lamina. Therefore, we have removed the data related to this point, which were previously contained in Figure S3A-C of our prior submission; namely laminats>AopACT and DroncI24; UAS-AopACT MARCM clones.

    However, this manuscript makes substantially different points from the previous paper regarding the roles of EGFR activity and survival. Although Fernandes et al., (2017) did show that lamina neurons differentiated prematurely in lamina>MAPKACT, here we evaluate apoptosis and lamina neuron sub-type identities and show that the ‘extra’ LPCs do not die but differentiate into L5s under these conditions. This is a key message of our manuscript and was not evaluated nor reported before. Additionally, the Dronc mutants used here reveal that preventing apoptosis is not sufficient to drive differentiation of the additional LPC in each column, addressing a different point and not simply reproducing prior data showing the EGFR promotes LPC survival.

    Similarly, we previously established that photoreceptor-derived Spi was received by wrapping glia, the involvement of photoreceptor-Spi and L5 differentiation had not been thoroughly explored and the involvement of xgO is novel.

    Establishing the cells expressing spi, argos, Col41a and Ddr is key to supporting the hypothesis. The authors claim that they confirmed the best screen candidates by testing their expression using enhancer trap lines. What is the evidence that these enhancer trap reporters reproduce the endogenous expression patterns of these genes? A description of their location in the loci and potential drawbacks should be provided and discussed.

    We now clarify whether enhancer traps used in our study were validated previously and provide in situ hybridization chain reaction data where enhancer traps were not previously validated.

    Fig.4A and Fig.S3K do not demonstrate that aos-lacZ and Ddr-lacZ are in L5 neurons, and showing this with Bsh and Spl2 as they do for other data would support the claim that L5 neurons receive Col4a1 and distal L5 neurons can receive aos.

    We use L5 specific markers with aoslacZ. For Ddr-Gal4>UAS-lacZ the entire lamina was labelled, and we provide new data showing Ddr expression by in situ hybridization chain reaction to show that it is expressed throughout the lamina.

    Fig.S3M uses HCR in situ to show that spi mRNA is found in xg{degree sign} glia. However, the given images are not convincing. Since in situs detect mRNA, wouldn't the nuclear signal correspond to two sites of transcription, whereas a more abundant signal would be expected in the cytoplasm? Instead, the nucleus contains as many spots as the surrounding background and there is no clear signal in the cytoplasm. The authors must provide separate channels and convincing evidence that spi mRNA is present in xg{degree sign} glia or remove/weaken the claim (ie use only the GAL4 evidence).

    We have understood that the main concern around the spi HCR included in our manuscript relates to the fact that the signal detected in the nucleus was more abundant than just two puncta as would be expected from two sites of transcription.

    The reviewers are correct that only two puncta corresponding to active sites of transcription would be expected in the nucleus when detected by single molecule FISH (smFISH). However, here we are not using smFISH but HCR with maximal amplification. This results in signal proportional to the relative abundance of transcripts (Choi et al., 2018; Trivedi et al., 2018) and as such all transcripts, including those moving away from the transcription site in the nucleus, are also detected by this method. Other groups who have used this method also report the same (Andrews et al., 2020; Duckhorn et al., 2022; Schwarzkopf et al., 2020; Zhuang et al., 2020). We used this form of HCR over single molecule HCR (smHCR or digital-HCR), which uses limited amplification (Trivedi et al., 2018), as these other methods require diffraction-limited spot detection, which would be very challenging in our system.

    We apologise for not explaining the HCR protocol sufficiently and have included more details in the Materials and Methods.

    In addition to using HCR to detect spi expression in xgO in controls and when EGFR signalling is blocked in xgO, we now also provide new data to show Col4a1 and Ddr expression using HCR, to lend support to enhancer traps that were not validated previously. We found that both spi and Col4a1 expression in xgO decreased when EGFR signalling was blocked in xgO and provide single channel images in Figure 3 – figure supplement 1.

    With this clarification, we hope the reviewers will reconsider the inclusion of these data as we feel it is important to show that xgO express these ligands in an EGFR signalling-dependent manner, especially in light of the spi and Col4a1 loss-of-function data detailed above. Nonetheless, if the reviewers still feel that these data should be removed from the manuscript, we will be happy to do so.

    Involvement of Spi does not seem to have been entirely unresolved. They show that over-expression of rho3 in photoreceptors in rho 3 mutants rescued L5 neurons, suggesting that Spi from photoreceptors can rescue L5 neurons. As this is slightly different from what they saw before, what is the penetrance of these phenotypes? These phenotypes have not been quantified (other than providing sample size) and the incomplete penetrance of phenotypes could explain both observations.

    Spi secreted from photoreceptor axons is insufficient to induce L5 neuronal differentiation directly as it is unable to do so when EGFR signalling is blocked in xgO (Figure 1F,H, Figure 1 – figure supplement 1N). Therefore our results argue that xgO are a critical mediator of photoreceptor signals. Since restoring rho3 expression in photoreceptors in rho3 background rescues neuronal differentiation of all lamina neurons, these results imply that the signalling relays through both wrapping glia and xgO have been reactivated.

    We have quantified of the number of L5s per column in rho3 heterozygotes, rho3 homozygotes and in rho3 homozygotes when rho3 expression was restored in photoreceptors only (Figure 3C). Importantly, compared to rho3 heterozygotes, the number of L5s per column in rho3 homozygotes was significantly reduced (Figure 3C; one-way ANOVA with Dunn’s multiple comparisons test with rho3/-; GMR as control; P****<0.0001), whereas they were fully rescued in rho3; GMR>rho3 (Figure 3C; one-way ANOVA with Dunn’s multiple comparisons test with rho3/-; GMR as control; P>0.05).

    They claim that whereas L5 neurons are lost in xg{degree sign}>EGFRDN over-expressing glia, concomitant over-expression of Spi rescues L5 neurons. Also, over-expression of spi with xg{degree sign}>spi clearly results in ectopic L5 neurons. However, in Fig.3P they show rescue with membrane-tethered m.spi and not secreted s.spi. Why was secreted s.spi not used instead? How does membrane-tethered spi from glia reach to rescue distal L5 neurons?

    Spi is initially produced as an inactive transmembrane precursor (mSpi) that needs to be cleaved into its active secreted form (sSpi) (Tsruya et al., 2002). This requires the intracellular trafficking protein Star and Rhomboid proteases (Tsruya et al., 2002; Urban et al., 2002; Yogev et al., 2008). mSpi thus represents wild-type (unprocessed) Spi. Whereas misexpression of sSpi results in secretion of active Spi from any cell type, misexpression of mSpi results in secretion of active Spi only from cells capable of processing mSpi to sSpi.

    Thus, mis-expressing mSpi to rescue L5 neurons in the xgO>EGFRDN background also demonstrates that xgO are capable of processing mSpi into sSpi, which is a more stringent experimental condition and gives us more confidence in our results. We also performed these experiments with sSpi and observed an equivalent and statistically significant rescue (included in the quantifications in Figure 3 – figure supplement 1C). We have also clarified the use of these reagents in the text as follows:

    Page 6, lines 166-168:

    “Spi is initially produced as an inactive transmembrane precursor (mSpi) that needs to be cleaved into its active secreted form (sSpi) (Tsruya et al., 2002). This requires the intracellular trafficking protein Star and Rhomboid proteases (Tsruya et al., 2002; Urban et al., 2002; Yogev et al., 2008).”

    And Page 8, lines 221-223:

    “Note that expressing either sSpi or wild-type (unprocessed) mSpi (referred to as Spiwt) in xgO rescued L5 numbers (Figure 3 – Figure supplement 1C), indicating that xgO are capable of processing mSpi into the active form (sSpi).”

    To support the involvement of spi in promoting survival of proximal L5 in wildtype, a loss of function experiment would be required e.g. xg{degree sign}>spi-RNAi, and visualise apoptosis with Dcp1 and remaining L5 neurons.

    We knocked down spi and Col4a1 simultaneously in xgO and observed a statistically significant decrease in the number of L5 neurons relative to controls (Figure 3T-W and Figure 3 – figure supplement 2A-B). Under these conditions we also observed Dcp1 positive cells in the most proximal row of the lamina, which were never observed in controls. Thus, suggesting that Spi and Col4a1 promote L5 neuronal differentiation and survival.

    Quantifications are incomplete in places and statistical analysis is incorrect in places. For genotypes that are not quantified in graphs (ie cell number), sample sizes have been provided, but phenotypic penetrance has not (Fig.1F dronc-/-; Fig.2K, L rho3 and rescue) and this is required to report variability.

    We apologise for these omissions. We have quantified the rho3 mutant and rescue phenotypes. The Dronc mutant phenotype was fully penetrant and we have stated this explicitly in the text.

    Fig.2I, J: A quantification is provided within the text for apoptosis caused by xg{degree sign}>EGFRDN, with 5.93{plus minus}0.18 Dcp1 cells per column (N=19). However, this number alone does not mean much unless it is compared to Dcp1 in wild-type. Apoptosis in wild-type is shown but not quantified in Fig.2I. A comparison of Dcp1 counts in control and xg{degree sign}>EGFRDN is required and validated with statistical analysis.

    We thank the reviewer for pointing out this mistake. We have now added the graph to the figure (Figure 2D) and have stated this explicitly in the text as follows:

    Page 5-6, line 151-156 (Figure 2D):

    “We used an antibody against the cleaved form of Death Caspase-1 (Dcp-1), an effector caspase, to detect apoptotic cells (Akagawa et al., 2015) and, indeed, observed a significant increase in the number of Dcp-1 positive cells in the lamina when EGFR signalling was blocked in the xgO (132.8 cells/unit volume ± 19.48 standard error of the mean) compared to controls (49.14 cells/unit volume ± 4.53) (Figure 2A-B, 2D, P<0.0005, Mann-Whitney U Test).”

    Fig.S3L, P: authors claim that over-expression of spi in xg{degree sign}>EGFRDN does not rescue nuclear dpMAPK in xg{degree sign}, but it does in L5 neurons. However, the quantification of these data in Fig.S3L shows that nuclear:cytopl dpMAPK levels are not statistically significantly different from xg{degree sign}>EGFRDN. No evidence has been provided of how this single piece of data supports both contradictory claims. The authors must either quantify accurately and separately dpMAPK in xg{degree sign} glia and L5 neurons - it is unclear how this could be done from the data provided - or remove or modify the claim to adjust accurately to the data.

    We have now quantified dpMAPK levels in both xgO and L5s in these conditions.

    Statistical analysis needs revising. It is unclear why they use non-parametric tests throughout, are data always not normally distributed? The use of bar charts, means, and s.e.m. combined with non-parametric tests does not faithfully represent the data, and box plots or other displays (eg volcano or dot plots, etc) that show the distribution would be more appropriate. And multiple comparison corrections are required. For example, if Fig.S3F is a Kurskal Wallis ANOVA (should be, but it is not stated explicitly), then this requires multiple comparison tests to a fixed control (post hoc Dunn test), and the figure legend should provide the p-value for the ANOVA. Fig.3K, P use Mann Whitney test, whereas these graphs have both more than 2 sample types and therefore should be Kruskal Wallis ANOVA (if distributions are not normal, if they are normal they should be One Way ANOVA), and Dunn post hoc comparison to fixed control, box plots, and no s.e.m as above.

    Thank you for flagging that we had not reported our statistical analyses appropriately. We apologise for this and have made sure to explicitly state the statistical test performed for multiple and pairwise comparisons with the Pvalues as detailed by Reviewer 3. These are highlighted throughout the text with track-changes. As well, we have changed all our graphs to box and whisker plots showing the entire distribution of the data as well as the interquartile range, as recommended.

    Much of the data in our manuscript are proportions generated from cell counts and, by definition, are limited to numerical values between 0 and 1 (inclusive). As such, as with count data (i.e. discrete numbers such as from cell counts), parametric statistics are generally inappropriate for proportion data because the data violate assumptions about normality (Douma and Weedon, 2019). Therefore, we used non-parametric tests throughout the manuscript except for Figure 1- Figure Supplement 1R where appropriate assumptions were met..

  2. Evaluation Summary:

    This manuscript will be of interest to cell and developmental biologists and neuroscientists. It addresses the question of how the number of connecting neurons in a circuit is matched whilst maintaining topography. It shows that non-autonomous control of neuronal number involves a relay mechanism through two distinct glial cell types, enabling the specification of distinct neuronal classes.

    This manuscript was co-submitted with: https://www.biorxiv.org/content/10.1101/2022.02.21.481306v1

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

  3. Reviewer #1 (Public Review):

    Sensory information from photoreceptors (PRs) is first conveyed to the lamina neuropil of the optic lobe, which consists of 5 neuron types - L1-5. Signals from the developing PRs - and the glia that ensheathes them - are intricately involved during the development of the lamina to ensure that retinotopy is accurately established. This includes multiple processes such as induction of the precursors in the lamina, their terminal divisions, and organisation into columns.

    In this manuscript, the authors identify another series of non-cell-autonomous processes that ensure the survival and differentiation of the appropriate number of L5 neurons - one of the 5 neuron types in the lamina. They use a variety of genetic manipulations and screens to show that:
    PR-derived EGF is sensed by a population of glial cells (the outer chiasm glia) that then secrete signalling molecules such as Spi and a type IV collagen. These signals in turn induce differentiation and survival of the L5 lamina precursor via MAPK. Finally, the authors show that the differentiated L5s inhibit the survival of an adjacent L5 lamina precursor by secreting Argos that eventually inhibits MAPK activity in it.

    The manuscript is well written and clearly presented.

  4. Reviewer #2 (Public Review):

    Precise control of neuronal diversity is an important issue in developmental neurobiology. The lamina in the Drosophila visual system is an essential model system to deal with this issue. The inductive action of Hedgehog and EGF signals from the retina triggers the differentiation of L1-L5 lamina neurons that are precisely arranged to form the lamina column (or cartridge) structure. The authors previously revealed that the wrapping glia secretes insulin ligands in response to EGF and induces the differentiation of L1-L4 lamina neurons. However, the mechanism that induces L5 differentiation was not known.

    In this study, the authors focus on the role of the outer chiasm giant glia (xgO) to induce L5 differentiation. They argue that xgO produces multiple ligands in response to the EGF signal to induce MAPK-dependent L5 differentiation. They also show that the differentiating L5 produces Argos, a secreted EGF antagonist, to repress extra L5 differentiation. Those extra L5 cells may be removed by apoptosis. As a result, each lamina column contains one L5 neuron.

    The study is based on the authors' previous study and is significantly extended by integrating the roles of xgO glia to induce L5 differentiation. The authors identified multiple candidate ligands such as Spi and Col4a1 that may induce L5 differentiation.

    The results solely depend on the ectopic expression of Spi and Col4a1. Although the authors performed rescue experiments, they depend on EGFR[DN] and non-physiological levels of ligand expression under the control of the Gal4/UAS system. To meet the standard of Drosophila genetics, the authors have to demonstrate the loss of function experiments for Spi and Col4a1 in xgO.

    Major points:

    1. The authors argue that xgO secretes Spi and Col4a1 to induce MAPK-dependent L5 differentiation. However, no loss-of-function condition for these putative ligands was tested. Since they speculated that expression of Spi and Col4a1 alone may not lead to a sufficient level of MAPK activity, the results of their loss of function conditions have to be included in the paper.

    2. The authors found ectopic L5 neurons when apoptosis was repressed (Fig. 1). It is likely that cells that fail to differentiate to L5 are removed by apoptosis, but this link was not clearly demonstrated in the paper. As a result, there is a gap between the data in Fig. 1 (section 1 in the text) and the other part of the paper. The relationship between Fig. 1 and the other data should be carefully discussed. In my opinion, the first section of Results should be moved after the last section so that the results of Fig. 1 are explained as a potential mechanism to remove cells that failed to differentiate to L5.

  5. Reviewer #3 (Public Review):

    This manuscript addresses the question of how neuronal numbers are determined and whether invariant cell patterns reflect deterministic developmental programmes or the reliability of cell interactions and non-autonomous processes. This question has long been addressed and there is previous evidence that non-autonomous, trophic mechanisms - including by EGF ligands - maintain cell survival in insects, just as they do in mammals, but further evidence is necessary to reaffirm the shift in the wider understanding of fundamental principles of development. This work is a beautiful contribution in support of the above notion. This work shows that non-autonomous control of neuronal survival and differentiation involves a relay mechanism through two distinct glial cell types, enabling the specification of distinct neuronal classes. This is a very nice demonstration of the importance of interactions and coupling between interacting cell populations. The work builds on the previous publication by this author, Fernandes et al in Science 2017 "Glia relay differentiation cues to coordinate neuronal development in Drosophila", where they showed that photoreceptors produce the EGFR ligand Spi, which is received by wrapping glia, which then secretes insulin to promote differentiation of L1-L4 neurons. There, they showed that some of the lamina neurons die and EGFR is required to maintain their survival. With the current manuscript, the authors show that a second class of glia, the outer chiasm giant glia (xg{degree sign}), receive the EGF ligand Spi from photoreceptors, in turn, produce Spi and Col4a1 to activate proximal L5 neuron differentiation from lamina precursor cells, L5 neurons, in turn, produce argos, an antagonist of EGFR signalling, which by preventing survival signalling in distal L5 neurons cause their apoptosis.

    The strength of this manuscript is its beautiful cell biology and microscopy data, where cellular events can be seen with single-cell resolution. The images and data are of excellent quality. The clear narrative supports the concept that stereotypy arises from reliable non-autonomous trophic interactions. The weakness is that novelty is limited as the overall idea was already presented in the Science paper, the current work completes the details of the original model and there is data overlap. There are also technical issues, which if solved, would strengthen the evidence in support of the claims, and the quantitative analysis would benefit from improvement.

    1. There is considerable overlap with Fernandes et al 2017 Science paper: (1) That EGFR signalling is required for L5 neuron survival had been shown in their Fernandes et al 2017 Science paper, as over-expression of p35 rescued apoptosis caused by EGFRDN. Now, using Dronc mutants in the current manuscript is an equivalent experiment. (2) In Fernandes et al 2017 Science, they over-express activated MAPK in lamina neurons (Fig.1G), and in the current, they over-express its target Pnt-P1 (Fig.1I) - equivalent experiment. (3) Figure S1 reports Lamina>MAPKACT rescues Bsh and Spl2 positive neurons. These data are similar to those reported in Fernandes et al 2017 Science, where they showed the rescue of lamina neurons with this same genotype. (4) rho3 mutants cannot secrete Spi and L1-4 cannot differentiate and only a few L5 do (Fernandes et al 2017 Science), they then rescued this phenotype including L5s by over-expressing EGFRACT or Ras in wrapping glia (Figure 2F-I). With the submitted manuscript, they rescue with rho3 over-expression in photoreceptors - genetically different, but rather similar, as together they demonstrate that rescue of L5 requires rho or spi. These close similarities reduce the appeal and novelty of the current manuscript.

    2. Establishing the cells expressing spi, argos, Col41a and Ddr is key to supporting the hypothesis. The authors claim that they confirmed the best screen candidates by testing their expression using enhancer trap lines. What is the evidence that these enhancer trap reporters reproduce the endogenous expression patterns of these genes? A description of their location in the loci and potential drawbacks should be provided and discussed.

    Fig.4A and Fig.S3K do not demonstrate that aos-lacZ and Ddr-lacZ are in L5 neurons, and showing this with Bsh and Spl2 as they do for other data would support the claim that L5 neurons receive Col4a1 and distal L5 neurons can receive aos.

    Fig.S3M uses HCR in situ to show that spi mRNA is found in xg{degree sign} glia. However, the given images are not convincing. Since in situs detect mRNA, wouldn't the nuclear signal correspond to two sites of transcription, whereas a more abundant signal would be expected in the cytoplasm? Instead, the nucleus contains as many spots as the surrounding background and there is no clear signal in the cytoplasm. The authors must provide separate channels and convincing evidence that spi mRNA is present in xg{degree sign} glia or remove/weaken the claim (ie use only the GAL4 evidence).

    3. Involvement of Spi does not seem to have been entirely unresolved. They show that over-expression of rho3 in photoreceptors in rho 3 mutants rescued L5 neurons, suggesting that Spi from photoreceptors can rescue L5 neurons. As this is slightly different from what they saw before, what is the penetrance of these phenotypes? These phenotypes have not been quantified (other than providing sample size) and the incomplete penetrance of phenotypes could explain both observations.

    They claim that whereas L5 neurons are lost in xg{degree sign}>EGFRDN over-expressing glia, concomitant over-expression of Spi rescues L5 neurons. Also, over-expression of spi with xg{degree sign}>spi clearly results in ectopic L5 neurons. However, in Fig.3P they show rescue with membrane-tethered m.spi and not secreted s.spi. Why was secreted s.spi not used instead? How does membrane-tethered spi from glia reach to rescue distal L5 neurons?

    To support the involvement of spi in promoting survival of proximal L5 in wild-type, a loss of function experiment would be required e.g. xg{degree sign}>spi-RNAi, and visualise apoptosis with Dcp1 and remaining L5 neurons.

    4. Quantifications are incomplete in places and statistical analysis is incorrect in places. For genotypes that are not quantified in graphs (ie cell number), sample sizes have been provided, but phenotypic penetrance has not (Fig.1F dronc-/-; Fig.2K, L rho3 and rescue) and this is required to report variability.

    Fig.2I, J: A quantification is provided within the text for apoptosis caused by xg{degree sign}>EGFRDN, with 5.93{plus minus}0.18 Dcp1 cells per column (N=19). However, this number alone does not mean much unless it is compared to Dcp1 in wild-type. Apoptosis in wild-type is shown but not quantified in Fig.2I. A comparison of Dcp1 counts in control and xg{degree sign}>EGFRDN is required and validated with statistical analysis.
    Fig.S3L, P: authors claim that over-expression of spi in xg{degree sign}>EGFRDN does not rescue nuclear dpMAPK in xg{degree sign}, but it does in L5 neurons. However, the quantification of these data in Fig.S3L shows that nuclear:cytopl dpMAPK levels are not statistically significantly different from xg{degree sign}>EGFRDN. No evidence has been provided of how this single piece of data supports both contradictory claims. The authors must either quantify accurately and separately dpMAPK in xg{degree sign} glia and L5 neurons - it is unclear how this could be done from the data provided - or remove or modify the claim to adjust accurately to the data.

    Statistical analysis needs revising. It is unclear why they use non-parametric tests throughout, are data always not normally distributed? The use of bar charts, means, and s.e.m. combined with non-parametric tests does not faithfully represent the data, and box plots or other displays (eg volcano or dot plots, etc) that show the distribution would be more appropriate. And multiple comparison corrections are required. For example, if Fig.S3F is a Kurskal Wallis ANOVA (should be, but it is not stated explicitly), then this requires multiple comparison tests to a fixed control (post hoc Dunn test), and the figure legend should provide the p-value for the ANOVA. Fig.3K, P use Mann Whitney test, whereas these graphs have both more than 2 sample types and therefore should be Kruskal Wallis ANOVA (if distributions are not normal, if they are normal they should be One Way ANOVA), and Dunn post hoc comparison to fixed control, box plots, and no s.e.m as above.