miR‐31 mutants reveal continuous glial homeostasis in the adult Drosophila brain

Abstract The study of adult neural cell production has concentrated on neurogenesis. The mechanisms controlling adult gliogenesis are still poorly understood. Here, we provide evidence for a homeostatic process that maintains the population of glial cells in the Drosophila adult brain. Flies lacking microRNA miR‐31a start adult life with a normal complement of glia, but transiently lose glia due to apoptosis. miR‐31a expression identifies a subset of predominantly gliogenic adult neural progenitor cells. Failure to limit expression of the predicted E3 ubiquitin ligase, Rchy1, in these cells results in glial loss. After an initial decline in young adults, glial numbers recovered due to compensatory overproduction of new glia by adult progenitor cells, indicating an unexpected plasticity of the Drosophila nervous system. Experimentally induced ablation of glia was also followed by recovery of glia over time. These studies provide evidence for a homeostatic mechanism that maintains the number of glia in the adult fly brain.

foresee a problem in meeting this three-month deadline, please let us know in advance and we may be able to grant an extension.
I look forward to seeing the revised version.

REFEREE REPORTS
Referee #1: The manuscript presents an unexpected role of mir-31a in regulating astrocyte cell number in the adult Drosophila brain. In mir-31a mutants astrocytes undergo apoptosis but are replaced after 3 weeks of life. Based on cell type specific depletion experiments, mir-31a appears to be required in progenitor cells and is dispensable in differentiated glial cells. Gal80ts experiments demonstrate that mir-31a is required in adult progenitor cells to suppress cell death of the differentiated glia suggesting that mir-31a normally prevents an accumulation of a toxic factor that leads to apoptosis of the differentiated cells. In a next step the authors identified a mir-31a target gene (CG16947) that when overexpressed in the the progenitor cells causes a reduction of glial cell number. The authors generated an antibody against the CG16947 protein and report increased expression in a mir-31a knockdown background. The normal function of CG16947 is, however, not revealed. This paper reports two in principle interesting findings. First the authors demonstrate that mir-31a is required in glial progenitor cells to suppress apoptosis of the differentiated astrocytes and second the authors report an apparent reactivation of progenitor proliferation to "regenerate" the missing astrocytes.
The mir-31a mutant phenotype results in a transient loss of 40% of all glial cells. This means that more than 200 glial cells die, an amazing cell number which is not really addressed. I cannot imaging that changes in astrocyte numbers can explain this reduction and unfortunately, Figure 1 gives only relative numbers. The notion that mir-31a is required in the astrocyte progenitors to allow survival of the differentiated cells is not really solved. In fact, the adult depletion of mir-31a in adult progenitor cells also results in the loss of adult glial cells -but at this developmental stage all glial cells have been born already. How can the cell death be explained? The expression of mir-31a is not well documented and the images are not instructive. It would be indeed interesting to see the localization of astrocyte progenitors are located. The generation of adult astrocytes was previously analyzed by Awasaki et al., but this paper is unfortunately not discussed. The present study only uses Repo staining to determine cell numbers and does not address the shape of the astrocytes. In a previous report, Stork et al., (2014) had demonstrated that upon induction of cell death in astrocytes, the surviving cells expand their size. Is this observed in mir-31a mutants as well? And is this process reversed in the regeneration phase?
Referee #2: In "Defects in astrocyte production in mir-31a mutants unveil glial homeostasis in the adult Drosophila brain," Foo and colleagues describe a mechanism for continued glial turnover in the mature fly brain by newly identified mir-31a-positive neuroglial progenitors. The authors demonstrate that reduced mir-31a function in global mutants and by RNAi KD in progenitor cells resulted in transient loss of brain glia. While mir-31a was not required for initial production of glial cells, they found that glial cells were lost over time due to apoptosis, and that cell death occurred due to impaired post-transcriptional regulation of the E3 ubiquitin ligase Rchyl1 by mir-31a. Interestingly, glial cell loss was followed by a period of biased glial cell production from mir-31apositive cells, suggesting a homeostatic mechanism by which glial cell numbers are closely monitored and maintained within the adult brain.
Overall, the data that the authors present is clear, and the importance of the work is quite high. The authors should be commended for the depth of their analysis. Yet there are several important changes/additions necessary before it is suitable for publication in EMBO J.
Major comments: 1. What is the "cell of origin" for the replacement glia?-glia? neurons? progenitors? other? The evidence that they come from mir-31a-gal4 positive cells is not sufficient to identify the cell type that produces these late-born glia. For example, on page 16 it says "production of new neurons and new glia from progenitor cells" -what is the evidence that these new cell come from neuroblasts (or any known progenitor)?
2. In general, the authors provide many graphs to support their conclusions, but not enough representative images of these data. To feel confident in their conclusions, representative images for the following experiments should be displayed. If space limitations are a problem, some of the images could be included as supplemental figures. a) Figure 1C. The paper highlights that these data suggest a mechanism for adult astrocyte homeostasis that hasn't been clearly demonstrated before, so it is essential to show that astrocytes are specifically affected, using an astrocyte specific marker. b) For images comparing the mutant phenotype to loss of glia due to use of mir031a sponge in insc-Gal4 (Fig. 2D) and mir-31a-Gal4 (Fig. 2F) animals, as they are used as proxies for the global mutant. Because the data is always presented as % of the control, it is not possible to determine how well the cell-specific KDs reflect the mutant phenotype. c) For experiments where the authors suppress the mutant phenotypes either by blocking apoptosis (Fig. 2D) or by providing exogenous mir-31a ( Fig. 2H) d) Fig. 3A, C to definitively show mutant phenotype in response to CG16947 manipulations 3. The authors should use a "progenitor specific" gal4 line, such as wor-gal4 in addition to insc-gal4 (which is expressed in some glia; Omoto et al., 2015). Showing both give the same phenotype, for at least a few experiments, would help prove that the effects are neuroblast specific.
Minor issues -please provide more information in Table S1 (e.g. p-values, primers used for all genes) and in the methods regarding their qPCR conditions, standardizations and results, which should follow MIQE guidelines (PMID 19246619) -authors should not use the term 'astrocytes' unless cell identity is confirmed with an astrocyte marker (e.g. alrm-gal4) -The authors show that overexpression of CG16947 results in reduced glial cell number, but they don't directly show that this is due to apoptosis (as was shown for loss of mir-31a). Please stain for a cell death marker to show that overexpression of Rachyl mimics the mir-31a mutant phenotype. -Whenever presenting data, please include +/-SD or SEM. For example, on page 5, "dropped to approximately 62% of Canton S controls" how many n's and what is the range of glial numbers are not shown. -Throughout the text and figure legends, please be more clear about whether cells were counted by reporter expression or by antibody stain. For example, the authors use both repo-Gal4 and anti-Repo, and it's not always clear what method was used for counting - Fig. 4E, please state what animals and what time point was used for this quantification -Please label each figure panel with the time point analyzed -The wording in the discussion is at times too strong based on the data the authors provide. For example, on page 17, the authors intimate that the mir-31a progenitors are specifically astrogenic. Although the data that they provide shows that glial survival is specifically affected by loss of mir-31a, these data do not preclude that mir-31a progenitors also give rise to cortex and ensheathing glia and that they simply do not have a survival phenotype. 2) For example, on page 19, glial turnover has been shown in the healthy brain during learning/memory tasks in rodents (reviewed in PMID 4634839).
-the authors should carefully review and edit genetic nomenclature throughout the text and figures (e.g. use of lowercase italics whenever genes and transgenes are used for Drosophila versus rodent genes) Very minor issues: -Consistency with labeling in the figures -Make sure to define transgenes when they're first presented. For example, repo-Gal4 is first defined on page 7, but introduced on page 6 -Typo on page 11 "labeled after 1 day with anti-Repo to mark neurons ( Figure 4A) or glia ( Figure  4B, S2A). The authors used Elav to mark neurons, Repo to mark glia - Fig This paper by Foo and colleagues describes a novel progenitor pool in the young adult Drosophila brain that is identifiable by microRNA mir-31a expression. These progenitors give rise largely to glial cells, likely astrocytes. In the absence of mir-31a, additional adult progenitors restore glial cells to normal numbers over the course of weeks. This provides the first description of a homeostatic mechanism in the adult Drosophila brain to maintain sufficient number of glia. These results are novel and very exciting and will provide important new insight into microRNA and ubiquitin ligasemediated regulation of glial cell numbers in adults, which will be of broad interest to many in the field of glial cell biology. Overall, the manuscript is well written and the conclusions largely supported. Some thoughts and comments to improve the paper are provided below: Major points: 1 . Figure 1A, the loss of Repo-positive cells is apparent. However, because astrocyte cell bodies are specifically located at the edge of neuropil regions, it's a little surprising that the Repo reduction appears consistent (broad) throughout the entire brain. Perhaps due to the particular image that was selected? It would be nice if confocal images were provided for some additional key experiments throughout the paper. For example, images corresponding to the Fig 2E-H. This would provide another opportunity to potentially observe the spatial distribution of changes in Repo expression throughout the central brain.
2 . As written, the manuscript seems to make an abrupt transition from using alrm-Gal4 to Insc-Gal4 to visualize/manipulate glia. Is there experimental evidence showing that alrm+ positive cells do not give rise the same progenitors that Insc+ (proliferative) cells do? Insc may not be specific to astrocytes. It would nice if a subset of key expts were repeated co-labeling with the GABA transporter Gat -or the excitatory amino acid transporter EAAT1 (aka GLAST) -to show that the newly produced cells that contribute to recovery of the glial population in mir31a mutants are indeed astrocytes.
3 . Repo-driven expression of the mir31a sponge did not alter the number of glial cells. Repo appears to co-express with mir31a drivers quite early (during the first week). How do the authors reconcile the fact that there was no phenotype with RepoGal4 -> mir31a sponge expression? 4 . Using heterozygote H99 mutant animals may influence apoptosis during development. In Figure  2D, the authors show that the reduction of Repo-positive cells in 31a sponge flies is rescued in H99 heterozygotes. Might the H99 hets already have higher levels of glia (and perhaps other cells?) Including an H99 heterozygotes is an essential control in this experiment. Also, confirming that apoptosis is altered with an activated caspase-3 stain would be nice addition to round out this part of the model. 5 . The model of astrocyte homeostasis is intriguing one, and this paper presents the first description of this mechanism in adult Drosophila. The authors state that "turnover of astrocytes has not yet been observed in healthy brains of adult mammals", but I don't believe this is true.  Referee #1:

The manuscript presents an unexpected role of mir-31a in regulating astrocyte cell number in the adult Drosophila brain. In mir-31a mutants astrocytes undergo apoptosis but are replaced after 3 weeks of life. Based on cell type specific depletion experiments, mir-31a appears to be required in progenitor cells and is dispensable in differentiated glial cells. Gal80ts experiments demonstrate that mir-31a is required in adult progenitor cells to suppress cell death of the differentiated glia suggesting that mir-31a normally prevents an accumulation of a toxic factor that leads to apoptosis of the differentiated cells.
In a next step the authors identified a mir-31a target gene (CG16947) that when overexpressed in the the progenitor cells causes a reduction of glial cell number. The authors generated an antibody against the CG16947 protein and report increased expression in a mir-31a knockdown background. The normal function of CG16947 is, however, not revealed.
This paper reports two in principle interesting findings. First the authors demonstrate that mir-31a is required in glial progenitor cells to suppress apoptosis of the differentiated astrocytes and second the authors report an apparent reactivation of progenitor proliferation to "regenerate" the missing astrocytes.
The mir-31a mutant phenotype results in a transient loss of 40% of all glial cells. This means that more than 200 glial cells die, an amazing cell number which is not really addressed. I cannot imaging that changes in astrocyte numbers can explain this reduction and unfortunately, Figure 1 gives only relative numbers.
We have now included the raw numbers of astrocytes, ensheathing glia and cortex glia as a graph (Appendix Figure S2A). The number of glia observed in 7d old Canton S controls is 688±28 anti-Repo-positive cells and in the mir-31a mutant is 417±37 anti-Repo-positive cells. At 7d, there are 482±19 alrm-Gal4>UAS-Histone-RFP+ cells in the control animals (n=53) and 348±11 alrm-Gal4>UAS-Histone-RFP+ cells in the mir-31a mutants (n=35). Thus, astrocytes account for a large number of the cells that are lost. We do not exclude that ensheathing, cortex, subperineurial and perineurial glia might be affected to some extent, but loss of astrocytes appears to account for the majority of the total glial loss observed. Note that the number of astrocytes is not significantly different at 21d post-eclosion between mir-31a mutant animals and control animals ( Figure EV1D and Appendix Figure S2B) The notion that mir-31a is required in the astrocyte progenitors to allow survival of the differentiated cells is not really solved. In fact, the adult depletion of mir-31a in adult progenitor cells also results in the loss of adult glial cells -but at this developmental stage all glial cells have been born already. How can the cell death be explained?
The finding that depletion of mir-31a leads to transient loss of adult glia shows that the population of glia is not static. The number of glia does not change much under normal conditions, so it may appear that all the glia are cells "that have been born already".
Our findings indicate that this apparently stable situation reflects a dynamic process in which cell number is actively maintained and new cells are produced to maintain the steady state. We suggest that loss of glia in the mutant reflects a normal process of turnover (cell death and replacement). In other words the loss of cells seen in the mutant results from failure to replace glia at the normal rate.
The mature glial cells are not dying due to absence of mir-31a: reducing the amount of mir-31a in glia with the sponge does not change glial numbers (Figure 2A). Rather, in the absence of miR-31a the replacement of glia by the progenitors is inefficient, because more of their progeny die (due to aberrant inheritance of Rchy1 from the progenitors). Lower than normal replacement rate leads to (transient) net loss of differentiated glia.
The expression of mir-31a is not well documented and the images are not instructive. It would be indeed interesting to see the localization of astrocyte progenitors are located.
To address this, we looked at the expression of mir-31a in 1d old adult brains using mir-31a sensor flies, which carry two copies of the reverse complement of the mature mir-31a sequence downstream of GFP expressed under the control of the tubulin promoter. In these flies, cells that lack mir-31a are GFP-positive and cells that express mir-31a are GFP-negative. We crossed the sensor into the Insc-Gal4>UAS-Histone-RFP background. Cells that were GFP-negative and RFPpositive indicate the position of the progenitors that are largely astrogenic ( Figure EV 1F).
We also looked at the position of the Alrm-Gal4>UAS-Histone-RFP cells in the mir-31a sensor background. We identified RFP-expressing and mir-31a-expressing cells (Appendix Figure S6 B, C). We note that the position of both the Insc-Gal4>UAS-Histone-RFP-expressing, GFP-negative and Alrm-Gal4>UAS-Histone-RFP, GFP-negative cells are in approximately the same location within the brain.
The generation of adult astrocytes was previously analyzed by Awasaki et al., but this paper is unfortunately not discussed.
Thanks for pointing this out. We have added a comment on Pages 4, 5 and 10.
The present study only uses Repo staining to determine cell numbers and does not address the shape of the astrocytes. In a previous report, Stork et al., (2014) had demonstrated that upon induction of cell death in astrocytes, the surviving cells expand their size. Is this observed in mir-31a mutants as well? And is this process reversed in the regeneration phase?
We appreciate the reviewer's interest in this, but think this addresses an issue beyond the scope of this study.

Referee #2:
In "Defects in astrocyte production in mir-31a mutants unveil glial homeostasis in the adult Drosophila brain," Foo and colleagues describe a mechanism for continued glial turnover in the mature fly brain by newly identified mir-31a-positive neuroglial progenitors. The authors demonstrate that reduced mir-31a function in global mutants and by RNAi KD in progenitor cells resulted in transient loss of brain glia. While mir-31a was not required for initial production of glial cells, they found that glial cells were lost over time due to apoptosis, and that cell death occurred due to impaired post-transcriptional regulation of the E3 ubiquitin ligase Rchyl1 by mir-31a. Interestingly, glial cell loss was followed by a period of biased glial cell production from mir-31apositive cells, suggesting a homeostatic mechanism by which glial cell numbers are closely monitored and maintained within the adult brain.

Overall, the data that the authors present is clear, and the importance of the work is quite high. The authors should be commended for the depth of their analysis. Yet there are several important changes/additions necessary before it is suitable for publication in EMBO J.
Major comments:

What is the "cell of origin" for the replacement glia?-glia? neurons? progenitors? other? The evidence that they come from mir-31a-gal4 positive cells is not sufficient to identify the cell type that produces these late-born glia. For example, on page 16 it says "production of new neurons and new glia from progenitor cells" -what is the evidence that these new cell come from neuroblasts (or any known progenitor)?
We think the issue here is mainly a sematic one. The clonal analysis in Figure EV3D shows that Insc-Gal4 expressing cells make neurons and glia. Likewise, cells that express mir-31a-Gal4 divide to make glia and (to a lesser extent) neurons ( Figure 4C,D). In other words, by virtue of the fact that they divide to make new progeny in the adult, both the Insc-expressing and the mir-31a-expressing cells are functionally defined as progenitor cells.
The Insc-Gal4-expressing cells are mainly neurogenic, but also make some glia. The mir-31a cells are mainly gliogenic, but can make some neurons. In our view, there is no better way to define these cells as progenitors than to show that they divide to make glial and/or neuronal progeny in the adult. We don't know of other markers that would help to better define the identity of the adult progenitor cells that make neurons and glia.

In general, the authors provide many graphs to support their conclusions, but not enough representative images of these data. To feel confident in their conclusions, representative images for the following experiments should be displayed. If space limitations are a problem, some of the images could be included as supplemental figures.
We have now included representative pictures for all genotypes used in the Appendix Supplemental Figures S1, S2, S4, S5, S6, S8, S9 and S10.
a) Figure 1C. The paper highlights that these data suggest a mechanism for adult astrocyte homeostasis that hasn't been clearly demonstrated before, so it is essential to show that astrocytes are specifically affected, using an astrocyte specific marker.
It is clear from the numbers that loss of astrocytes can account for the majority of the total glial loss observed, but we do not exclude that ensheathing, cortex, subperineurial and perineurial glia might be affected to some extent that is not significantly different. We have modified the text on Page 6 to emphasize this.
b) For images comparing the mutant phenotype to loss of glia due to use of mir031a sponge in insc-Gal4 (Fig. 2D) and mir-31a-Gal4 (Fig. 2F) animals, as they are used as proxies for the global mutant. Because the data is always presented as % of the control, it is not possible to determine how well the cell-specific KDs reflect the mutant phenotype.
Raw numbers are now provided as graphs in the Appendix Supplemental Figures. Specifically, the raw data graph for Fig2D is Appendix Figure 3D and representative images are in Appendix Figure  S4 and S5. For Fig 2F the raw data graph is Appendix Figure 3F and the representative images are in Appendix Figure S5 and S6. (Fig. 2D) or by providing exogenous mir-31a (Fig. 2H) The raw data graph for Fig2D is Appendix Figure 3D and representative images are in Appendix Figure S4 and S5. For Fig 2H, the raw data graph is Appendix Figure 3H and the representative images are in Appendix Figure S5 and S6.

d) Fig. 3A, C to definitively show mutant phenotype in response to CG16947 manipulations
The raw data graph for Fig 3A is in Appendix Figure S7B and the representative images in Appendix Figure S8.
The raw data graph for Figure 3C is in Appendix Figure S7D and the representative images in Appendix Figure S8.
3. The authors should use a "progenitor specific" gal4 line, such as wor-gal4 in addition to insc-gal4 (which is expressed in some glia; Omoto et al., 2015). Showing both give the same phenotype, for at least a few experiments, would help prove that the effects are neuroblast specific.

page 16, are the EdU+ glia endoreplicating or forming de novo?
Glial cells are lost in the mutant, and the number of cells subsequently recovers. The MARCM clonal analysis also shows that new cells are made by division of progenitors during the recovery process. The EdU incorporation presumably reflects the production of these new glia. We do not exclude that there might also be some endoreplication (for example in perineural glia).

Minor issues -please provide more information in Table S1 (e.g. p-values, primers used for all genes) and in the methods regarding their qPCR conditions, standardizations and results, which should follow MIQE guidelines (PMID 19246619)
We have included information on qPCR conditions in the methods section on Page 24 in accordance with the MIQE guidelines. The sequences of the primers are in Table S1.
-authors should not use the term 'astrocytes' unless cell identity is confirmed with an astrocyte marker (e.g. alrm-gal4) We have changed the manuscript accordingly.
-The authors show that overexpression of CG16947 results in reduced glial cell number, but they don't directly show that this is due to apoptosis (as was shown for loss of mir-31a). Please stain for a cell death marker to show that overexpression of Rachyl mimics the mir-31a mutant phenotype.
We have overexpressed UAS-CG16947 using Repo-Gal4 and find the presence of activated caspase 9, anti-Repo-expressing cells in 2d old adult brains ( Figure EV2 E).
-Whenever presenting data, please include +/-SD or SEM. For example, on page 5, "dropped to approximately 62% of Canton S controls" how many n's and what is the range of glial numbers are not shown.
We used scatter plots so that "n" would be evident for each sample. Error bars in the graphs show SEM. Phrasing in the text sometimes used less precise language. To address the reviewer's request we have compiled a table with SEM and sample size in numerical form (Table EV2).

-Throughout the text and figure legends, please be more clear about whether cells were counted by reporter expression or by antibody stain. For example, the authors use both repo-Gal4 and anti-Repo, and it's not always clear what method was used for counting
We have changed the text and figure legends to indicate 'anti-Repo' where we counted cells labeled by antibody. Where we counted cells visualized with a Gal4 driving the expression of UAS-Histone-RFP, we have indicated as such.

please state what animals and what time point was used for this quantification
We have added this information in the figure legend of Fig 4E, Page 37.

-Please label each figure panel with the time point analyzed
Done -The wording in the discussion is at times too strong based on the data the authors provide. For example, on page 17, the authors intimate that the mir-31a progenitors are specifically astrogenic. Although the data that they provide shows that glial survival is specifically affected by loss of mir-31a, these data do not preclude that mir-31a progenitors also give rise to cortex and ensheathing glia and that they simply do not have a survival phenotype.
We counted the number of cortex glia, ensheathing glia and astrocytes and only observed a significant defect in the number of astrocytes in the central brain. This suggests that mir-31a progenitors affect astrocyte survival. We cannot exclude the possibility that mir-31a-expressing progenitors also give rise to cortex and ensheathing glia, but if so, it appears that inheritance of rchy1 by these cells does not lead to loss of their progeny. We have added a comment to the text to note the reviewer's caveat on Page 17.

2) For example, on page 19, glial turnover has been shown in the healthy brain during learning/memory tasks in rodents (reviewed in PMID 4634839).
Thanks for pointing this out. It has been noted in the discussion on Page 20.. This paper by Foo and colleagues describes a novel progenitor pool in the young adult Drosophila brain that is identifiable by microRNA mir-31a expression. These progenitors give rise largely to glial cells, likely astrocytes. In the absence of mir-31a, additional adult progenitors restore glial cells to normal numbers over the course of weeks. This provides the first description of a homeostatic mechanism in the adult Drosophila brain to maintain sufficient number of glia. These results are novel and very exciting and will provide important new insight into microRNA and ubiquitin ligase-mediated regulation of glial cell numbers in adults, which will be of broad interest to many in the field of glial cell biology. Overall, the manuscript is well written and the conclusions largely supported. Some thoughts and comments to improve the paper are provided below: Major points: 1 . Figure 1A, the loss of Repo-positive cells is apparent. However, because astrocyte cell bodies are specifically located at the edge of neuropil regions, it's a little surprising that the Repo reduction appears consistent (broad) throughout the entire brain. Perhaps due to the particular image that was selected? It would be nice if confocal images were provided for some additional key experiments throughout the paper. For example, images corresponding to the Fig 2E-H. This would provide another opportunity to potentially observe the spatial distribution of changes in Repo expression throughout the central brain.
2 . As written, the manuscript seems to make an abrupt transition from using alrm-Gal4 to Insc-Gal4 to visualize/manipulate glia. Is there experimental evidence showing that alrm+ positive cells do not give rise the same progenitors that Insc+ (proliferative) cells do? Insc may not be specific to astrocytes.
We used alrm-G4 to label astrocytes in Fig 1C. We used alrm-G4 as a marker to compare astrocyte number in mutant and wild-type flies. Subsequent experiments used Insc-Gal4 or mir-31a-G4 to direct marker or sponge expression. The MARCM clonal experiments shows that both Insc-Gal4 and mir-31a-G4 cells divide to give progeny in the adult -identifying them as functional progenitor cells in the adult.
We did not suggest that alrm-G4 cells are progenitor cells, or that they give rise to them.
Note also, that we did not suggest that InscG4 was specific to astrocytes. Quite the opposite. We provide MARCM clonal data showing that InscG4 cells serve as progenitors for both neurons and glia.
It would nice if a subset of key expts were repeated co-labeling with the GABA transporter Gat -or the excitatory amino acid transporter EAAT1 (aka GLAST) -to show that the newly produced cells that contribute to recovery of the glial population in mir31a mutants are indeed astrocytes.
The data provided in expanded view figure E1 addressed this point. We used Alrm-Gal4 to drive UAS-Histone-RFP to label astrocytes. The number of astrocytes was significantly reduced in the mutant at day 7 (compared to control) but recovered by day 21. This provides evidence that loss and replacement of astrocytes contributes to the mutant phenotype.
The antibodies recommended by the reviewer would not be suitable for cell counting, because they label the whole cells, not just the cell body. Using whole cell markers (such as UAS-CD8-GFP) makes it very difficult to get accurate cell counts, compared to use of a nuclear marker, like UAShistone-RFP. In addition the quality of the antibodies available to us too poor to given reliably quantifiable data. We cannot do the specific experiment, but we suggest that the point is adequately addressed by the alrm-G4 marker experiment.
3 . Repo-driven expression of the mir31a sponge did not alter the number of glial cells. Repo appears to co-express with mir31a drivers quite early (during the first week). How do the authors reconcile the fact that there was no phenotype with RepoGal4 -> mir31a sponge expression?
As the reviewer points out, removing mir-31a expression with the sponge in mature glia has no effect ( Figure 2A). We think that the co-expression of Repo with RFP may reflect perdurance of Gal4 and Histone-RFP proteins expressed in the progenitors (mir-31a-Gal4 drives UAS-Histone RFP). Using a mir-31a sensor transgene (GFP expression indicated absence of the microRNA), we did not see evidence of the miRNA in mature glia (Expanded view Figure EV1E). Figure  2D,

the authors show that the reduction of Repo-positive cells in 31a sponge flies is rescued in H99 heterozygotes. Might the H99 hets already have higher levels of glia (and perhaps other cells?)
Including an H99 heterozygotes is an essential control in this experiment. Also, confirming that apoptosis is altered with an activated caspase-3 stain would be nice addition to round out this part of the model.
We have included the control genotype of Insc-Gal4>GFP, Df(3L)H99 and see no appreciable change in glia number in these flies ( Figure 2D).
We observe apoptotic glial cells (activated caspase 9-expressing, anti-Repo-expressing cells) in the brains of 4d post-eclosion mir-31a mutants ( Figure 1E). Thanks for pointing this out. We have modified the text on Page 20 accordingly.

. The model
Minor points:

. State N values for experiments.
We used scatter plots for the quantitative data so that "N" would be evident for each sample. Error bars in the graphs show SEM. Table S2 contains the N values for all experiments in Table S2.
Thanks for pointing this out. We have modified the text on Page 7 accordingly.
3 . Change sentence: "Adult flies were heat-shocked the day of eclosion to induce clones and the brains were labeled after 1 day with anti-Elav or anti-Repo to mark neurons ( Figure 4A) or glia ( Figure 4B, S2A)." Fig 3G,H  Thanks for submitting your revised manuscript to The EMBO Journal. Your manuscript has now been re-reviewed by the three referees.

. Most of the images in
As you can see below, the referees appreciate the introduced revisions and support publication here. They have a number of relative minor points that I would you to sort in a final revision.

REFEREE REPORTS
Referee #1: The authors have addressed most of the comments but neglected one, which in my view is a must before publication. This is only about 50% of the total glial cell loss (271 or 261). A statement that loss of astrocytes accounts for the majority of the total glial loss is thus not correct and should be removed. Since no changes in the number of ensheathing glia and cortex glia were noted, it is likely that the number of the perineurial glial cells is affected. Since these cells are generated by different means compared to astrocytes, this should be stated and discussed.
Referee #2: The revised manuscript addresses all my concerns and is suitable for publication.
Referee #3: The authors appear to have addressed the concerns of this reviewer in good faith, and the manuscript is suitable for publication in EMBO.
Two minor points: 1. It is unclear what the "heterozygote" versus "mutant" animals reflect in Figure 4C and D? 2. Arrowhead is missing from Figure EV4 (repo EDU merge panel).
2nd Revision -authors' response 09 February 2017 Referee 1: The authors have addressed most of the comments but neglected one, which in my view is a must before publication.