In vivo AGO-APP identifies a module of microRNAs cooperatively controlling exit from neural stem cell state

  1. Karine Narbonne-Reveau
  2. Andrea Erni
  3. Norbert Eichner
  4. Surbhi Surbhi
  5. Gunter Meister
  6. Christophe Beclin
  7. Cédric Maurange
  8. Harold Cremer

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

MicroRNAs (miRNAs) are essential regulators of all developmental processes. Their function is particularly important during neurogenesis, when the production of large numbers of neurons from a limited number of neural stem cells depends on the precise control of determination, proliferation and differentiation. However, miRNA regulation of target mRNAs is highly promiscuous, one miRNA can target many mRNAs and vice versa, raising the question of how specificity is achieved to elicit a precise regulatory response.

Here we introduce AGO-APP, a novel approach to purify Argonaute-bound miRNAs directly from cells and tissues in vivo, to isolate actively inhibiting miRNAs from different neural cell populations in the larval Drosophila central nervous system. We identify a defined group of miRNAs that redundantly target all iconic genes known to control the transition from neuroblasts to neurons. In vivo functional studies demonstrate that knockdown of individual miRNAs does not induce detectable cellular phenotypes. However, simultaneous knockdown of multiple miRNAs leads to precocious stem cell differentiation, demonstrating functional interdependence. Thus, miRNAs cooperate within a regulatory module to specify the targeted gene network.

Article activity feed

  1. Review Commons

    Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    __Reviewer #1 (Evidence, reproducibility and clarity (Required)): __ Summary In this manuscript the authors address the largely unexplored role of micro RNAs (miRNAS) in Drosophila melanogaster brain development, in particular in neural stem cell lineages. The authors for the first time adapt the Ago protein Affinity Purification by Peptides (AGO-APP) technology for Drosophila. They show that this technique works efficiently in neural stem cell lineages and identify several cell type specific active miRNAs. Through a series of bioinformatic analysis the authors identify candidate mRNA targets for these miRNAs. The authors then functionally analyse the role of some of the identified miRNAs, focusing on miRNAs significantly over-represented in neuroblasts.

    By overexpressing Mir-1, the authors demonstrate that this miRNA effectively targets the UTR of Prospero, resulting in the overproliferation of neuroblasts. In a parallel experiment, overexpression of Mir-9c causes neuroblast differentiation defects, similar to the phenotype caused by nerfin-1 mutants, a previously validated target. Loss of function analyses show that knock down of single miRNAs has little functional effects in neuroblast size, showing that the individual effect caused by miRNAs knock down is likely compensated. In contrast, a sponge against a selected group of miRNAs leads to a reduction in poxn positive neuroblasts. Overall these results validate the approach and support the theory that miRNAs cooperate in functional modules during stem cell differentiation.

    We thank Reviewer 1 for its overall positive review. We are grateful for the useful suggestions and we believe the additional experiments we have performed and added strongly improve the quality of the study and will hopefully satisfy the reviewer's concerns.

    Comments

    Title: As the authors do not really explore exit from neural stem cell state this should be altered. The authors do not assess for the levels of any temporal genes, nor other markers of neural stem cell state exit (e.g. nuclear Pros).

    We now have further evidence that the identified microRNA module preserves neuroblasts, in particular in the optic lobe. We have modified the title accordingly: "In vivo AGO-APP identifies a module of microRNAs cooperatively preserving neural progenitors"

    The observed effects, with the available experiments, rather say that neural stem cell state is not maintained in general, not being clear what mechanistically happens to these cells expressing Cluster 2 sponges. The described phenotype caused by the expression of sponges against individual miRNAs also rather shows a blockage in differentiation.

    -The miRNAs analysed were found in Ago-APP to be predominantly active in neuroblasts, but was there any phenotypes of OE or KD in neurons or glial cells?

    Since the analyzed miRNAs were either not or poorly expressed in neurons or glia overall, it seemed less essential to investigate potential phenotypes in these cells. However, we did mis-expressed miR-cluster1sponge and miR-cluster2sponge in neurons and in glial cells (using elav-GAL4 and Repo-GAL4, respectively) throughout development, and did not observe any major impact on viability. All pupae were able to hatch.

    In addition, we show now that mis-expression of the miR-*cluster2sponge *(that induces strong phenotypes in neuroblasts) specifically in the wing pouch throughout development did not lead to any phenotype in the adult (e.g. wing size (tissue growth), patterning defects (cell differentiation)) (Fig6K,L). Importantly, this experiment rules out unspecific effects of the sponge construct on cell fitness, and highlight the tissue-specificity of the phenotype.

    • The authors obtained a phenotype when using a sponge against Cluster 2 in poxn neuroblasts. Is this specific for these 6 neuroblasts? What happens if this sponge is expressed with a pan-neuroblast driver in central brain/VNC/optic lobe? These experiments should be included as they would show if these are conserved effects for all neuroblasts.

    We already showed in Fig.4B of the first version of the manuscript (using a flip-out approach in clones) that miR-cluster1sponge or miR-cluster2sponge expression leads to an overall reduction in the neuroblast size in the VNC and CB.

    We have now added four more experiments, all suggesting that these sponges specifically affect type I neuroblasts:

    • using the pan-neuroblast driver nab-GAL4, we show that neuroblasts in the VNC and CB expressing these sponges are significantly smaller in late L3. Also, their number is reduced, indicated that some neuroblasts are eliminated (Fig.4C-G).
    • Using pox-GAL4 (already in first version) and eagle-GAL4, we show that different subset of type I neuroblasts in the VNC exhibit different sensitivities to the sponges (from light/medium - neuroblast shrinkage, to high - neuroblast elimination) (Fig.4H-J, S6C-E)
    • using the *dpnOL-GAL4 *driver, that is specific and strongly active in medulla neuroblasts in the optic lobe, we demonstrate that both, miR-cluster1sponge and miR-cluster2sponge, induce neuroblast shrinking. In addition, we find that the width of the medulla neuroblast stripe is strongly reduced when using the miR-cluster2sponge, providing further evidence for precocious neuroblast elimination (6C,D). Importantly, this leads to a smaller medulla in late L3 (Fig 6F), implying that in these conditions, medulla neuroblasts produce fewer neuronal progeny. Because medulla neuroblasts generate GMCs that undergo a single division, they are also considered as type I neuroblasts
    • using a worniu-GAL4, ase-GAL80 driver, that is specifically active in type II neuroblasts, we show that expression of miR-cluster1sponge and miR-cluster2sponge does not affect neuroblast size and the number of intermediate progenitors (Fig 6H-J). Together, these additional experiments in different types of neuroblasts and in non-neural tissue (the wing pouch, see above) demonstrate a type I neuroblast-specific effect. Our new results also imply that the microRNA module is active in most, if not all type I neuroblasts. In contrast, it is not present or not affecting differentiation genes in type II neuroblasts. Importantly, in Type II lineages, intermediate progenitors produced by neuroblasts undergo themselves a few rounds of divisions before differentiating, unlike GMCs that give rise to two differentiated progeny after a single division. Therefore, the dynamics of differentiation is different in the two lineages, involving a distinct sequential expression of differentiation factors, and possibly different miRNAs.

    The authors do different analyses in different brain regions, making also a hard to conclude if all brain regions behave the same way. As authors show that some miRNAs are only expressed in sub-sets of cells, this becomes particularly relevant.

    The new set of experiments in different types of type I neuroblasts and in type II neuroblasts, presented above, addresses the points on the specificity of the microRNA module.

    Could sponge of cluster 1 cause a phenotype if it had been expressed in other neuroblast lineages?

    Yes, it can. See our new experiments discussed above.

    __ __In addition, a discussion of the results obtained from sponge 1 should be included and put in context with miRNA function, technical limitations, levels/cell, targets, pitfalls of analyses, sponges, etc.

    We have more carefully acknowledge that sponge mediated knock-down is not very efficient and dose-dependent. We also clarified that other approaches will be required in the future to rigorously assess the specificity of each miRNA/mRNA interaction as well as their cooperativity.

    For example: "In contrast to genetic miR-1KO (Fig. 3O), we found that sponge mediated knock-down of this miRNA, or of other individual miRNAs in the module, had never a significant effect on neuroblast size (Fig. 4B), likely because the inhibition induced by sponges is incomplete. However, expression of either multi-sponge 1 or multi-sponge 2 significantly reduced neuroblast size in a dose dependent manner - two copies of the transgene exacerbate the phenotype (Fig. 4B)."

    We also state at the end of the discussion: "In the future, the combination of Ago-APP with complementary genetic strategies will be required to rigorously assess the specificity of each miRNA/mRNA interaction as well as their cooperativity."

    It would also be interesting to further explore the phenotypes caused by Mir-1 sp expression - are there any milder lineage defects?

    We observed an increase in Prospero expression and a decrease of the neuroblast size in miR-1null mutant neuroblast clones (Fig.3L-O). These phenotypes are not observed when miR-1sponge is mis-expressed. This is probably due to the fact that miR-1sponge expression leads to only a partial knock -down of miR-1. Moreover, we have added data about the expression of miR-1sponge in medulla neuroblasts in the optic lobe, showing an absence of obvious phenotype when assessing neuroblast size and neuroblast maintenance. This contrasts with expression of miR-cluster1sponge and miR-cluster2sponge (Fig. 4F,G). This new data is in line with our hypothesis that the knockdown of miRNAs of a common module synergize/cooperate to produce the phenotype expected from the deregulation of their common target mRNAs.

    Any defects in other brain regions/lineages, like in type 2 neuroblasts that usually do not express Pros?

    As suggested by the reviewer, and discussed above, we tested expression of miR-cluster1sponge and miR-cluster2sponge in type-II neuroblasts using the worniu-GAL4, asense-GAL80 driver (Neumüller et al., 2011). Interestingly, in contrast to type I neuroblasts in VNC, CB and OL regions, we did not observe neuroblast shrinking or changes in INP numbers. This suggests that either the self-renewing state is more robust in Type II than in Type I neuroblasts, or that that the uncovered miRNA module is more specific to type I neuroblasts than to type II. We have added and discussed these important data in Fig 6H-J in the revised version.

    Ago-APP identifies cell type specific miRNAs in larval neurogenesis section:

    • "...29oC... allows Gal4-dependent expression (Fig.1B,C)" - this description of Gal80ts/Gal4 works is not correct, expression is not prevented.

    Gal80 directly binds to Gal4 carboxy terminus and prevents Gal4-mediated transcriptional activation.

    We have tried to clarify this point in the revised version.

    "Thus, when x-GAL4, tub-GAL80ts, UAS-T6B animals are maintained at 18{degree sign}C (restrictive temperature), GAL80 binds to Gal4 and inhibits its activity. *Switching to 29{degree sign}C (permissive temperature) for 24 hours inactivates GAL80, allowing for GAL4-mediated transcriptional activation of UAS-T6B" *

    • Fig S1 - nab-Gal4 also drives expression in GMCs and neurons, rephrase text. Is nab-Gal4 expressed in optic lobe, VNC and central brain neuroblasts?

    nab-GAL4 drives UAS-T6B expression in neuroblasts (in the VNC and in the CB), but also at lower levels in the medulla neuroblasts of the OL.

    We now describe this expression more precisely in the text and in Fig.S1C:

    "nab-GAL4 was used for T6B expression in all neuroblasts. However, because GAL4 is inherited by neuroblast progeny, T6B will also be present in GMCs and a few immature neurons (Fig.S1A,C)24. Of note, nab-GAL4 is highly expressed in the neuroblasts of the ventral nerve cord (VNC) and of the central brain (CB), and weaker in the neuroblasts of the optic lobe (OL) (Fig. S1C)".

    • "20 late larval CNS" - mention the exact stage

    We mention now the precise stage: the wandering stage.

    • Providing a more detailed and interpretive description of Figures 1D and 1E would greatly enhance their clarity. Currently, the descriptions of these pannels resemble typical figure legends.

    We now provide a more detailed description of the data, emphasizing that they are consistent with previous studies on specific miRNAs.

    • Fig. 1F,G,H - It is not clear why the authors sometimes use the optic lobe, other ventral nerve cord as both regions have both neuroblasts, neurons and glia. Are the drivers used for Ago-APP not expressed in all brain regions?

    We now document the activity of the GAL4 drivers used for AGO-APP throughout the entire larval central nervous system in Fig.S1B-D. We also show images of the entire larval central nervous system for the different reporter lines (Fig S1E-K) and focus on regions of interest in the main Fig 1F-M with quantitative measurement of reporter gene expression.

    • Show "data not shown" for 1H.

    It is now shown in Fig. 1M'.

    • Fig. 1F, G, H - Please quantify intensity levels in the different cell types to facilitate comparison with Ago-APP graphs. Include in figure legend what is "cpm".

    Quantification of intensity levels is now represented in Fig. 1F,I and L. Cpm means "counts per millions". We added this in the figure legend.

    A regulatory module controlling neuroblast-to-neuron transition section:

    • Fig. 2C - A more detailed explanation in text is required in addition to what is mentioned in the figure legend. Including a brief summary/conclusion of the results would be helpful. If possible, add in X-axis 1, 2, 3.

    We clarified this point in the text:

    "We used the Targetscan algorithm1* to determine the predicted target genes of each neuroblast-enriched miRNA. Next, we investigated the correlation between the identified miRNAs and the presence of their targets, based on independently generated mRNA expression data44.*

    *This analysis showed that neuroblast-enriched miRNAs predominantly target mRNAs that are normally highly expressed in neurons (Fig. 2C), consistent with a differentiation inhibiting function." *

    • Figure S2B - as mentioned in the text elav is expressed from the neuroblast, although this is not represented in the figure.

    I In this scheme, we depict the expression of proteins, not the presence of mRNAs. elav mRNA is indeed present at low levels in neuroblasts but the protein is absent from both neuroblasts and GMCs (as shown by all our immunostainings against Elav). This fact strongly suggests post-transcriptional repression of elav mRNA (possibly by miRNAs). This likely explains why the elav-GAL4 is also active in neuroblasts. It also suggests some post-transcriptional mechanisms to silence elav in the neuroblasts/GMCs (miRNAs?)

    It is hard to tell what are young vs maturing neurons in the cartoon, pls add a label/legend.

    We added new labels in Fig S2B to uncouple neuronal maturation from temporal identity. We hope it is clearer now.

    • Fig.3I - please shown a control brain. The merge images are not easy to see. I think it would be nicer to change the figures to be color-blind friendly.

    We added the control brain in Fig 3I for VNC clones, and Fig S3A for OL clones.

    We also changed all the figures to be color-blind friendly.

    • Fig. 3K,L - why is this now done in the VNC?

    We now focus on the VNC in the main Figure 3 (Fig.3I,J,K,L,N), and show similar phenotypes in the OL in the Supplemental Figure S3 (Fig.S3A-C).

    • Are there any lineage defects when Mir-1 sp is expressed?

    See previous comment on miR-1sponge.

    • Based on which parameters/variables of the predicted targets was the Hierarchical clustering done? A brief explanation would help the interpretation of the results and of the choice of the clusters that were further analysed.

    Hierarchical clustering is now explained in the "Bioinformatics analysis" section of the Material & Methods section with an additional matrix available in Table S1.

    • "revealed the presence of three main groups" - this should be rephrased as this "grouping" was done arbitrarily by the authors and not by hclust. Hclust is set to merge individual clusters/sub-trees up to 1. Furthermore, a more detailed explanation that supported this decision of choosing this 3 large clusters should be included.

    See previous question.

    • Fig. 4B, S4B - please include in legend how were these clones generated. S4B - scale bars missing.

    We included the missing information and added the missing scale bars.

    • Fig. 4H - was the ratio of UAS/Gal4 kept in both experimental conditions? Increasing the number of UAS/Gal4 leads to weaker expression of UAS and thus could lead to a weaker phenotype. Including in legends genotype details would help.

    This is a very good point as the number of copies of the UAS and/or GAL4 can influence transgene expression and consequently the phenotype observed. We indeed kept the ratio of UAS/GAL4 in both experimental conditions. The exact genotypes for the experiments are:

    Hs-FLP/+; act>stop>Gal4, UAS-GFP/+; UAS-RFP/UAS-miR-1

    Hs-FLP/+; act>stop>Gal4, UAS-GFP/UAS-cluster2sp; UAS-miR-1/+.

    To address this important issue in the manuscript, we added a table (Table S3) listing the precise genotypes for each experiment.

    Minor

    • Abstract: "a defined group of miRNAs that are predicted to redundantly target all..." This is only predicted, not experimentally shown, this should be modified accordingly.

    Although the request here is not clear to us, we made a few minor changes to the abstract that we hope will satisfy the reviewer.

    • Intro: "Elav, an RNA binding protein, is expressed as soon as post-mitotic neurons..." - Elav is expressed already in neuroblasts, as also mentioned by the authors in the result section. Correct, add references.

    elav is indeed already transcribed in neuroblasts and GMCs. However, the protein is absent in the two cell types (as shown by all our immunostainings), and only present in neurons. Thus, there is a level of post-transcriptional regulation that prevents elav mRNA translation in neuroblasts and GMCs (likely at least partly mediated by miRNAs). This also explains why in elav-GAL4; UAS-T6B brains T6B is expressed in neuroblasts and GMCs, as the GAL4 mRNA transgene is not submitted to the same post-transcriptional regulation.

    • Last paragraph of Intro (Bioinformatic analyses...) - it is not easy to understand the content of this paragraph. Rewrite to improve clarity.

    The paragraph has been rewritten for more clarity with the addition of Table S1

    • All legends: Please mention which developmental stage is being analysed in each panel (i.e. wandering 3IL, hours After Larval Hatching, hours After Puparium Formation, or other), in which brain region the analyses/images are being done.

    The CNS regions are now systematically annotated in the figures. All experiments have been done in wandering L3 (except for the new Fig.6 K,L, where the experiment is done in the adult wing). We now systematically mention in the text and legend the developmental stage at which the experiment is performed.

    Please include more detailed information about the genetics in figure legends.

    We added Table S3 that describes the exact genotype of all crosses done in this study.

    • Please include brief explanation of the genetics of miR-10KOGal4 line.

    This is now also explained in the new Table S3.

    • Why are miRNAs sometimes referred as (e.g.) "miR-1" and others "miR-1-3p"?

    The miRNA found enriched (and thus active) in the neuroblast is the miR-1-3p strand. The UAS-miR-1-sponge has been designed to be complementary to the miR-1-3p strand, and is then referred as miR-1-3psp in the text and figure legend. The miR-1 null clones have been made using the miR-1KO allele, which inactivates the entire locus and therefore both, the miR-1-3p and miR-1-5p strands. This is referred to as miR-1KO or miR-1 in the text. Finally, constructions used to mis-expressed miR-1 and other miRNAs are made with the pre-miRNA, meaning that both strands of the miRNA are mis-expressed. This is then referred as miR-1 in the text.

    • Fig. 3I-M - stage of the animal? 3M - in which brain region is this?

    We have systematically mentioned the brain region on panels on all figures.

    • Fig. 3N - can actual sizes be additionally shown, or at least averages mentioned in text?

    Average sizes are indicated in the legend of new Fig. 4F.

    • If non differentially expressed miRNAs, or miRNA with other expression patterns, had been analysed to determine their targets in the sub-set of genes expressed in neuroblasts (from the transcriptome) would different targets been found? Meaning, how specific are these binding patterns for the selected miRNA?

    This is an interesting and important point. To answer, we added a new analysis (Fig.S2C), where the total number of target sites in the 3'UTR of the pro-differentiation/temporal network genes are shown for different categories of miRNAs: neuroblast-enriched miRNAs (analysed in this study), neuron-enriched miRNAs, glia-enriched miRNAs, and random miRNAs not expressed in the brain. This analysis shows that neuroblast-enriched miRNAs exhibits a higher level of promiscuity with the iconic pro-differentiation/temporal genes than other identified or random miRNAs, arguing for functional relevance.

    **Referees cross-commenting**

    *think this study is very interesting as it optimizes a novel technique in Drosophila for the investigation of cell-specific active miRNAs, and it globally addresses the role of miRNAs in neural stem cell lineages. Although the authors do not explore deeply the biological effect of these miRNAs in neural lineages, I think that the technical contribution and the identification of some miRNA targets is relevant on its own. The authors use Prospero as an example, which is very interesting, as this gene is required to be lowly expressed in Neuroblasts and then upregulated during differentiation. Which the authors propose can be regulated by miRNAs, identifying a novel player in this differentiation mechanism. I do not feel the authors need to perform additional experiments to corroborate their findings, as they are well supported by the experiments presented. I do agree that the authors did not explore deeply the biological effect in neural lineages, and the claims regarding premature terminal differentiation, nerfin, etc need to be toned down accordingly.

    Reviewer #1 (Significance (Required)):

    This study is both a technical and conceptual advance. It is very interesting as it optimizes a novel technique in Drosophila for the investigation of cell-specific active miRNAs, and it globally addresses the role of miRNAs in neural stem cell lineages. However, the text, especially in the results section, could benefit from increased detail to enhance the comprehension of the experiments, results, and conclusions. Given that the functional analyses were not conducted at a very detailed level, there exist certain instances of over-interpretation, which could be easily addressed either by revising the text or by incorporating additional experiments, as elaborated upon below. This manuscript will be interesting for research fields interested in stem cell differentiation, brain development, micro RNAs, both for Drosophilists and scientists working with other animal models. I am an expert in Drosophila brain development.

    __Reviewer #2 (Evidence, reproducibility and clarity (Required)): __ Summary MicroRNAs (miRNAs) have a well-established role in fine-tuning gene expression. Because the mechanisms by which miRNAs recognize specific target transcripts are poorly understood, their functionally relevant targets in the physiological context are mostly poorly defined. Studies in vertebrates have suggested that miRNAs play a prominent role in regulating cell type specification during brain development. Insight into miRNA regulation of target selection will improve our understanding of neural development. Cell type-specific gene expression patterns and functions in the neural stem cell (neuroblast) lineage in the fly larval brain are well characterized. The fly genome is compact, and gene redundancy including miRNAs is significantly less than vertebrates. For these reasons, the authors chose to investigate how miRNAs regulate cell-state transitions by first establishing a comprehensive miRNA expression profile for major cell types in the fly larval brain. They combined the AGO-APP strategy and the GAL4-UAS inducible expression system to pull-down cell type-specific miRNAs from fly larval brain. The authors focused on miRNAs that are enriched in neuroblasts and examine how multi-miRNA modules regulate the maintenance of an undifferentiated state in neuroblasts. The cell type-specific inducible AGO-APP system introduced in this study is innovative and allows for systematic identification of miRNAs that most standard RNA-sequencing techniques missed in previously published datasets. The technological note sets high promise for this study, but the findings appear tame. It is my opinion that there are a number of shortcomings that can improve the rigor of this study. For example, strategies used to determine spatial expression patterns of miRNAs as well as to validate miRNA target genes are indirect with high likelihood of caveats. The choices of candidate target genes to assess the function of miRNAs in the cell state transition appear counterintuitive.

    We thank the reviewer for qualifying our study as "technologically excellent" and for emphasizing the "innovative character of AGO-APP" and the potential of such studies to "be hugely significant to the general audience".

    We are aware that there could be ways to more rigorously and systematically investigate the interactions between miRNAs and their targets and assess their cooperativity. Beyond in vitro luciferase assays (an approach we have used in this study), this would ideally involve multiple new transgenic assays, with point mutations in various miRNA sites in the 3'UTR of predicted target genes as proposed by Reviewer 2. Also, measuring the direct effect of miRNA knockdown on its target is notoriously difficult as it can be modest (and only be revealed through the cooperative action with other miRNAs, as proposed in this study), and sometimes not detected by measuring mRNA levels (e.g. by transcriptomic approaches or FISH).

    One of our aims in the future is to develop such non-trivial approaches, which will take a considerable amount of time and work. At this stage we believe that it would go beyond the scope of the present study which aims at illustrating how introducing a new technology for miRNA isolation (AGO-APP) can help to reassess important questions on miRNA biology and function (e.g. miRNA cooperation within in the context of developmental transitions). We discuss this point now in the last paragraph of the discussion in the revised version.

    Our unbiased AGO-APP results reveal a group of neuroblast enriched miRNAs that are predicted to target multiple times pro-differentiation genes (prospero, elav, nerfin-1, brat) while not targeting stemness genes such as miranda, worniu, inscuteable, deadpan, grainyhead. Mutation in pro-differentiation genes are known to either promote neuroblast tumors (prospero, nerfin-1, brat ) (https://doi.org/10.1016/j.cell.2006.01.03; 10.1101/gad.250282.114) or perturb neuronal differentiation (elav) (https://doi.org/10.1002/neu.480240604). On the other hand, mis-expression of these genes in neuroblasts often promotes shrinkage, precocious differentiation and /or cell cycle-exit (10.1016/j.cell.2008.03.034 ; 10.7554/eLife.03363 ; 10.1101/gad.250282.114). Therefore, bioinformatic prediction and previous studies made it likely that GOF of the neuroblast-enriched miRNAs would lead to neuroblast expansion or differentiation defects, and that LOF would lead to neuroblast shrinkage, cell cycle exit or differentiation. All these predictions are experimentally validated in our study. To reinforce our data, we have performed a number of additional experiments that are described below.

    Furthermore, the authors provided no rationale as to why they chose cell types that are not in the brain (such as wing cells and cells in the optic lobe) to assess the phenotypic effect of manipulating miRNAs.

    All our analysis were done either in the different types of neuroblasts found in the central nervous system (CNS) composed of the ventral nerve cord (VNC) (equivalent to vertebrate spinal cord) and brain (comprising the central brain (CB) and the optic lobes (OL) (10.1016/j.neuron.2013.12.017) - not to be confused with eye imaginal discs that produce the retina but do not contain neuroblasts. We tested the role of the neuroblast-enriched miRNAs in all neuroblasts of the CNS based on the pan-neuroblast activity of the nab-GAL4 driver used for the AGO-APP experiment. We then focused on different types of neuroblasts using lineage specific GAL4 drivers (poxn-GAL4, eagle-GAL4, dpnOL-GAL4, type II-GAL4). This is shown in the entirely revisited last paragraph of the results (Fig 4, 5, 6, S6 and S7). These experiments demonstrate that sponges simultaneously targeting several miRNAs of the module only affect type I neuroblasts but not type II neuroblasts.

    To investigate whether miR-1 directly regulates prospero mRNA in vivo, we used a tissue where prospero is not normally expressed (the wing pouch of the wing imaginal disc in late l3 larvae), allowing us to test how over-expressing miR-1 post-transcriptionally affects versions of prospero mRNAs that either possess or not its endogenous 3'UTR. The obtained results are consistent with in vitro luciferase assays, and miR-1 gain-of function in neuroblasts and GMCs, supporting the hypothesis that prospero mRNA is a direct target of miR-1 via its 3'UTR. We have clarified these points in the revised version of the manuscript.

    Using solely a reduced cell size as the functional readout for "precocious differentiation" is not rigorous and should be complemented with additional measures.

    Reduced neuroblast size always precedes neuroblast differentiation and has been widely used as functional readout of precocious differentiation (this is more clearly emphasized and referenced in the revised version). We have now also observed this phenotype in the neuroblasts of the optic lobe (Fig 6), together with precocious "plunging" of old neuroblasts in the deep layer of the medulla (Fig S7G), another sign of differentiation. These experiments show that the shrinkage phenotype is robust to all type I neuroblasts (medulla neuroblasts of the optic lobe can also be considered as type I neuroblasts because they generate GMCs that undergo a single division).

    Moreover, opposite to precocious differentiation induced by the simultaneous knockdown of multiple miRNAs of the neuroblast module, we now show that mis-expression of many of the miRNAs of the module prevents proper neuronal differentiation (miR-1, miR-9, miR-92a, miR-8) (Fig S5). Taken together, these experiments strongly suggest that the miRNAs of the module have the ability to block neuronal differentiation and that they represent a functional module in type I neuroblasts.

    Major concern:

    1. The authors should use a direct method to confirm the expression pattern of identified miRNAs such as miRNA scope (ACD) in the whole mount brain instead of indirect methods such as reporters.

    Such techniques are not trivial and do not represent a standard in Drosophila. Instead, the reporter genes we have used in our study have been already validated in other studies to reflect the expression of particular miRNAs in different tissues. We thus have taken advantages of these available lines to correlate expression patterns as reflected by transgenics with our AGO-APP experiment. All reporter lines tested quantitatively support the AGO-APP data as now shown in the revised Fig 1F,I,L.

    The entire figure 3 aims to provide evidence to support that prospero mRNA is a direct target of miR-1-3p. These convoluted experiments with significant caveats should be replaced with mutating the endogenous miR-1-3p binding sites in the 3'UTR of the prospero reading frame, and demonstrate that the endogenous prospero transcript level is increased by sm-FISH. The authors could also use this novel allele to assess the phenotypic effect of "unregulated prospero" in the larval brain.

    It would indeed be an interesting experiment to perform to show that miR-1 directly regulates pros RNA in vivo. However, our miR-1 mutant clones suggests that miR-1 on its own has only a small contribution to prospero mRNA regulation during the neuroblast-to-neuron transition. This could be due to the low physiological levels of miR-1-3p in neuroblasts and to the fact that several miRNAs of the module may act partly redundantly and collaboratively to maintain the correct level of prospero mRNA. Thus, in this case, it is well possible that changes in the endogenous prospero mRNA transcript may not be significant and detected by smFISH, unless more miRNA sites are mutated. Such an experiment would involve the generation of several new transgenic lines using the CRISPR technology, which represents a long-term project.

    Again, these approaches are powerful and we agree that they would represent a more rigorous assessment of miRNA cooperation. But we feel that it goes beyond the scope of this article, as mentioned above.

    The effect of overexpressing mir-1 on the prospero transgene with its 3'UTR vs without 3'UTR cannot easily compared since the UTR might be regulated by other regulatory mechanisms in addition to mir-1.

    To minimize the potential effect of other regulators, we only compare conditions where the only difference is the presence or absence of miR-1. We do not directly compare levels of Prospero with its 3'UTR vs without 3'UTR. However, there is indeed still the possibility that miR-1 overexpression would change the expression of a protein that regulates prospero mRNA via its 3'UTR.

    Considering this we have tuned-down our conclusion concerning this part in the revised version of the manuscript and now used the sentence:

    "These experiments performed in two different cellular contexts strongly suggest that prospero mRNA is a direct target of miR-1-3p."

    How could the author use evidence-based strategy to demonstrate that massive amplification of Mira-expressing cells induced by overexpressing mir-1 in the optic lobe is indeed due to mis-regulation of prospero instead of mimicking the prospero-mutant phenotype?

    First, we noted that miR-1 overexpression in neuroblast clones causes neuroblast amplification in all regions of the CNS (not only in the optic lobe) at the expense of neuronal differentiation. This is now shown in Fig 3 and S3.

    Second, multiple chemical or genome-wide RNAi screens have been performed (Gould lab, Chia lab, Knoblich lab, etc) to identify genes whose downregulation causes efficient neuroblast amplification (10.1186/1471-2156-7-33 ; 10.1016/j.stem.2011.02.022). In VNC type I neuroblasts, only inactivation of prospero or miranda can lead to efficient neuroblast amplification in late larvae, generating tumour-like structures devoid of neurons. We find that while Miranda is highly expressed in neuroblast clones overexpressing miR-1 (Fig 3J), Prospero is completely absent, suggesting that it is efficiently silenced by miR-1 overexpression, and therefore responsible for the observed phenotype. This new result is now added in Fig.S3D. It is very unlikely that the down-regulation of another gene is responsible for this phenotype. However, we cannot exclude that other genes are deregulated that contribute to this phenotype in addition to prospero knockdown.

    Similarly, what is the evidence that the phenotype associated with mir-9a knockout is due to mis-regulation of nerfin-1?

    In contrast to prosperoKD clones that are devoid of neurons, nerfin-1 mutant clones are known to be composed of a mix of neuroblasts and neurons (Fig S4E,G) (10.1101/gad.250282.114 ). When over-expressing miR-9 in neuroblast clones in the VNC, we observed a strong downregulation of nerfin-1 (Fig S4A, C) showing that nerfin-1 is a likely target of miR-9. However, downregulation is not complete which could explain why we do not see neuroblast amplification in the VNC (Fig 4F). Together with the significant up-regulation of nerfin-1 upon miR-9sponge expression, and the results of our luciferase assays, these data are consistent with nerfin-1 being a direct target of miR-9. Finally, the fact that overexpression of miR-9 in the optic lobes triggers phenotypes very similar to loss of function of nerfin-1 (but different from loss of function of prospero which is upstream of Nerfin-1 in epistatic tests) suggests that down-regulation of nerfin-1 is at least partially responsible for the phenotype (Fig S4D,E).

    Again, we cannot exclude that other deregulated targets contribute to the phenotype.

    Most of look-alike mutant phenotypes presented by the authors appear to occur in the OL. Is there any reason why cells in the visual center, which is not a part of the brain, appears to be more suspectable to loss of function of miRNAs? This is particularly important when manipulating the same miRNAs appear to have very subtle effects on VNC neuroblasts.

    Optic lobes (OL) are a part of the brain (10.1016/j.neuron.2013.12.017). Indeed, each OL constitutes a large region located on both sides of the central brain that integrates signals from retinal photoreceptors coming from the retina in the eyes. Moreover, medulla neuroblasts in the OL can be considered as type I neuroblasts because they generate GMCs that undergo a single division, in contrast to intermediate progenitors (INPs) produced by type II neuroblasts.

    In the original version of our manuscript, we mainly showed gain-of-function in the OL , as for some of the miRNAs the phenotypes were more striking than elsewhere. We have now more systematically tested our gain-of-function and loss-of-function in both the VNC (type-I neuroblasts) (Fig 3, 4, 5, S3, S4, S6) and in the OL (medulla neuroblasts) (Fig 6, S4, S5, S7).

    Results in the VNC are presented generally in the main figures, while results in the OL are presented mainly in supplemental figures; but phenotypes obtained in both parts are now clearly described in the text of the revised version.

    How do the authors know that multi-sponge 2 expression leads to loss of stemness potential in neuroblasts? Any additional evidence that supports precocious differentiation but not death or cell cycle exit?

    This is indeed an important point which we have investigated further in the new version. We now show that inhibiting apoptosis partially rescues neuroblast elimination but not shrinkage when miR-cluster2sponge is expressed in the poxn lineage in the VNC (Fig.4L,M). This shows that VNC neuroblast can disappear by apoptosis upon miR-cluster2sponge, but that shrinkage precedes apoptosis. We also show that optic lobe neuroblasts also shrink upon miR-cluster2sponge and are precociously eliminated as indicated by the thinner neuroblast stripe, by a mechanism independent of apoptosis (Fig 6C,D, S7F). Indeed, the neuroblast stripe in the optic lobe remains free of anti-activated caspase 1 (Dcp1), a widely used label of apoptotic cells, upon miR-cluster2sponge (Fig S7F). Finally, we also show precocious "plunging" of the old OL neuroblasts deep in the medulla, another sign of precocious differentiation (Fig S7G).

    Therefore, these experiments reinforce the conclusion that the neuroblast-enriched miRNA module is involved in neuroblast maintenance and that down-regulation of this module leads to the progressive loss of the neuroblast state.

    Lastly, we show that miR-cluster2sponge has no effect on type II neuroblasts or wing imaginal discs arguing for a specific type I neuroblast effect (including VNC, CB and medulla neuroblasts).

    Again, how do the authors know that mir-1 overexpression efficiently silenced prospero mRNA in neuroblasts and GMCs in Fig. 4F?

    This relevant question is addressed in our response to questions 2 and 3.

    Have the authors considered other targets to better assess the function of these miRNAs enriched in neuroblasts. For example, could these miRNAs function to dampen the expression of genes that are required for maintaining these cells in an undifferentiated state? Several studies using the neuroblast model suggest that the expression of these genes needs to be downregulated at the transcriptional and post-transcriptional levels. Perhaps, these miRNAs might target these "stemness" transcripts instead of "differentiation" transcripts. Is there evidence for or against this possibility?

    This is definitely a good point that we have now discussed in the revised version. We found that neuroblast identity genes (e.g. Mira, Dpn, Insc, etc) are not targeted by the miRNA module. However, the module of miRNA in late L3 neuroblasts also appears to target the early temporal genes (Chinmo, Imp), that are strongly oncogenic and stemness promoting. These need to be silenced in late L3 to ensure that neuroblasts stop dividing during metamorphosis ( 10.7554/eLife.13463). Therefore, there is indeed a strong possibility that the miRNA module we have identified in late L3 both maintains stemness by inhibiting differentiation genes and dampens stemness by silencing early temporal genes ensuring timely elimination in pupal stage. We are actively working on the regulation of temporal genes by microRNAs along development and will describe this in details in another study.

    This point was clarified in the discussion as followed:

    "In this context it is interesting to note that, in addition to differentiation factors, the early temporal factors Chinmo and Imp are predicted to be highly targeted by the neuroblast-enriched miRNA module. Given the strong oncogenic potential of these genes30*, it possible that the microRNA module not only protects neuroblasts against precocious differentiation but also protects against uncontrolled self-renewal. Therefore, in principle the same miRNA module could control neuroblast activity through the control of both self-renewal and differentiation, two seemingly opposing biological activities." *

    Minor point

    1. There are a number of mis-leading statements throughout the manuscript. -In the abstract, the authors indicated "isolate actively inhibiting miRNAs from different neural cell populations in the larval Drosophila central nervous system". For example, the expression patterns of Nub-Gal4 an Elav-Gal4 drivers appear to be partially overlapping in multiple cell types and might be active in the visual center (optic lobe). If true, it was unclear to me what neural cell types were actually used in their analyses and how they could confidently indicate that cell types in the central nervous system were used in their study. Aren't there more specific Gal4 drivers or more sophisticated genetic tools available to increase the purity of cell types? If not, the alternative could be a much more precise secondary screening step to directly determine where these miRNAs are actually detected instead of relying on indirect readouts of where they might be expressed.

    The expression patterns with additional figures are now more clearly described in the main text and in Fig.S1C,D.

    We are in the process of using other GAL4 drivers that target more specific populations of neurons. But this is beyond the scope of this first study and will be published later.

    -The statement "GMCs lacking Prospero, Nerfin-1 or Brat fail to differentiate and reacquire a neuroblast identity" is very problematic. Nerfin-1 does not appear to be expressed in GMCs according to Fig. S2B. Furthermore, Froldi et al., 2015 suggested that Nerfin-1 appears to prevent activated Notch from reverting neurons to ectopic neuroblasts.

    Indeed, Nerfin-1 is not expressed in GMCs but in immature neurons to stabilize neuronal identity and prevent reversion as shown by Froldi et al. and other studies (DOI: 10.1101/gad.250282.114 ; https://doi.org/10.1242/dev.141341). We have now clarified this point in the introduction: "This process involves the sequential activity of key cell fate determinants such as the transcription factor Prospero and the RNA-binding protein Brat in the GMCs followed by the transcription factor Nerfin-1 and the RNA-binding protein Elav in the maturing neurons20-23. GMCs lacking Prospero, or immature post-mitotic progeny lacking Nerfin-1, fail to initiate or maintain differentiation respectively, and progressively reacquire a neuroblast identity, leading to neuroblast amplification 21,23-25."

    -The statement on page 6 "Strikingly, the group of genes ... contained all iconic genes known to induce neuron differentiation after neuroblast asymmetric division, including nerfin-1, prospero, elav and brat" is problematic. Again, Nerfin-1 probably functions to maintain a neuronal state rather to induce differentiation. Is there evidence that Elav induces neuron differentiation after neuroblast asymmetric division? Brat seems to downregulate Notch signaling in neuroblast progeny rather than instructing neuron differentiation. Furthermore, previous studies suggested that loss of brat function does not affect identity of GMCs and their symmetric division to generate neurons. A similar statement is used at the end of this same paragraph to reiterates mis-leading messages.

    Prospero and Nerfin-1 are sequentially expressed in maturing neurons. Nerfin-1 shares many similar targets as Prospero. It has been proposed that Nerfin-1 prolonged the action of Prospero, allowing stabilisation/maintenance of the differentiated neuronal state (10.1101/gad.250282.114 ; 10.1016/j.celrep.2018.10.038)

    Brat is also involved in the sequence of events needed to produce neurons upon neuroblast asymmetric division. However, the mode of action of Brat in GMCs from type-I neuroblasts and in INPs from type-II neuroblasts is unclear. It was shown that Brat is an RNA-binding protein that has multiple targets. For example, it can bind and silence Myc, Zelda and Deadpan, and promote neuroblast-to-INP differentiation. It may also inhibit Notch signaling which is required for neuroblast-to-INP differentiation (https://doi.org/10.1016/j.devcel.2006.01.017; 10.1016/j.devcel.2008.03.004 ; https://doi.org/10.15252/embr.201744188; https://doi.org/10.1158/0008-5472.CAN-15-2299)

    We have clarified the difference between Type I and Type II neuroblasts in the introduction: "A sparse subset of neuroblasts (Type II) generate intermediate progenitors (INPs) that can undergo a few more asymmetric divisions, allowing for larger lineages to be produced. The neuroblast-to-neuron process in Type II lineages involves a slightly different sequential expression of differentiation factors21,24."

    We have also added a new reference describing that neuronal differentiation and maintenance are severely affected upon elav loss of function:

    Yao, K.-M., Samson, M.-L., Reeves, R. & White, K. Gene elav of Drosophila melanogaster: A prototype for neuronal-specific RNA binding protein gene family that is conserved in flies and humans. J. Neurobiol. 24, 723-739 (1993).

    **Referees cross-commenting**

    My main concern about data in this study remains direct vs. indirect effects of manipulating miRNA functions and the corresponding phenotype in various cell types in flies. The authors focused most of their effort on using genes that promote GMC differentiation in order to establish the role of neuroblast-specific miRNAs. Most of the experiments were not rigorously performed to the level that eliminates obvious caveats and suggests their interpretation is the most likely possibility. It is a technologically excellent study but lacks in-depth analyses in biological effects.

    Reviewer #2 (Significance (Required)):

    I believe there is a strong general interest in better appreciating how miRNAs regulate precise gene expression. Deriving some sort of rules such as the specificity of target selection or the efficiency of downregulating gene expression will be hugely significant to the general audience

  2. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    Summary

    MicroRNAs (miRNAs) have a well-established role in fine-tuning gene expression. Because the mechanisms by which miRNAs recognize specific target transcripts are poorly understood, their functionally relevant targets in the physiological context are mostly poorly defined. Studies in vertebrates have suggested that miRNAs play a prominent role in regulating cell type specification during brain development. Insight into miRNA regulation of target selection will improve our understanding of neural development. Cell type-specific gene expression patterns and functions in the neural stem cell (neuroblast) lineage in the fly larval brain are well characterized. The fly genome is compact, and gene redundancy including miRNAs is significantly less than vertebrates. For these reasons, the authors chose to investigate how miRNAs regulate cell-state transitions by first establishing a comprehensive miRNA expression profile for major cell types in the fly larval brain. They combined the AGO-APP strategy and the GAL4-UAS inducible expression system to pull-down cell type-specific miRNAs from fly larval brain. The authors focused on miRNAs that are enriched in neuroblasts and examine how multi-miRNA modules regulate the maintenance of an undifferentiated state in neuroblasts.

    The cell type-specific inducible AGO-APP system introduced in this study is innovative and allows for systematic identification of miRNAs that most standard RNA-sequencing techniques missed in previously published datasets. The technological note sets high promise for this study, but the findings appear tame. It is my opinion that there are a number of shortcomings that can improve the rigor of this study. For example, strategies used to determine spatial expression patterns of miRNAs as well as to validate miRNA target genes are indirect with high likelihood of caveats. The choices of candidate target genes to assess the function of miRNAs in the cell state transition appear counterintuitive. Furthermore, the authors provided no rationale as to why they chose cell types that are not in the brain (such as wing cells and cells in the optic lobe) to assess the phenotypic effect of manipulating miRNAs. Using solely a reduced cell size as the functional readout for "precocious differentiation" is not rigorous and should be complemented with additional measures.

    Major concern:

    1. The authors should use a direct method to confirm the expression pattern of identified miRNAs such as miRNA scope (ACD) in the whole mount brain instead of indirect methods such as reporters.
    2. The entire figure 3 aims to provide evidence to support that prospero mRNA is a direct target of miR-1-3p. These convoluted experiments with significant caveats should be replaced with mutating the endogenous miR-1-3p binding sites in the 3'UTR of the prospero reading frame, and demonstrate that the endogenous prospero transcript level is increased by sm-FISH. The authors could also use this novel allele to assess the phenotypic effect of "unregulated prospero" in the larval brain. The effect of overexpressing mir-1 on the prospero transgene with its 3'UTR vs without 3'UTR cannot easily compared since the UTR might be regulated by other regulatory mechanisms in addition to mir-1.
    3. How could the author use evidence-based strategy to demonstrate that massive amplification of Mira-expressing cells induced by overexpressing mir-1 in the optic lobe is indeed due to mis-regulation of prospero instead of mimicking the prospero-mutant phenotype? Similarly, what is the evidence that the phenotype associated with mir-9a knockout is due to mis-regulation of nerfin-1?
    4. Most of look-alike mutant phenotypes presented by the authors appear to occur in the OL. Is there any reason why cells in the visual center, which is not a part of the brain, appears to be more suspectable to loss of function of miRNAs? This is particularly important when manipulating the same miRNAs appear to have very subtle effects on VNC neuroblasts.
    5. How do the authors know that multi-sponge 2 expression leads to loss of stemness potential in neuroblasts? Any additional evidence that supports precocious differentiation but not death or cell cycle exit?
    6. Again, how do the authors know that mir-1 overexpression efficiently silenced prospero mRNA in neuroblasts and GMCs in Fig. 4F?
    7. Have the authors considered other targets to better assess the function of these miRNAs enriched in neuroblasts. For example, could these miRNAs function to dampen the expression of genes that are required for maintaining these cells in an undifferentiated state? Several studies using the neuroblast model suggest that the expression of these genes needs to be downregulated at the transcriptional and post-transcriptional levels. Perhaps, these miRNAs might target these "stemness" transcripts instead of "differentiation" transcripts. Is there evidence for or against this possibility?

    Minor point

    1. There are a number of mis-leading statements throughout the manuscript. -In the abstract, the authors indicated "isolate actively inhibiting miRNAs from different neural cell populations in the larval Drosophila central nervous system". For example, the expression patterns of Nub-Gal4 an Elav-Gal4 drivers appear to be partially overlapping in multiple cell types and might be active in the visual center (optic lobe). If true, it was unclear to me what neural cell types were actually used in their analyses and how they could confidently indicate that cell types in the central nervous system were used in their study. Aren't there more specific Gal4 drivers or more sophisticated genetic tools available to increase the purity of cell types? If not, the alternative could be a much more precise secondary screening step to directly determine where these miRNAs are actually detected instead of relying on indirect readouts of where they might be expressed. -The statement "GMCs lacking Prospero, Nerfin-1 or Brat fail to differentiate and reacquire a neuroblast identity" is very problematic. Nerfin-1 does not appear to be expressed in GMCs according to Fig. S2B. Furthermore, Froldi et al., 2015 suggested that Nerfin-1 appears to prevent activated Notch from reverting neurons to ectopic neuroblasts. -The statement on page 6 "Strikingly, the group of genes ... contained all iconic genes known to induce neuron differentiation after neuroblast asymmetric division, including nerfin-1, prospero, elav and brat" is problematic. Again, Nerfin-1 probably functions to maintain a neuronal state rather to induce differentiation. Is there evidence that Elav induces neuron differentiation after neuroblast asymmetric division? Brat seems to downregulate Notch signaling in neuroblast progeny rather than instructing neuron differentiation. Furthermore, previous studies suggested that loss of brat function does not affect identity of GMCs and their symmetric division to generate neurons. A similar statement is used at the end of this same paragraph to reiterates mis-leading messages.

    Referees cross-commenting

    My main concern about data in this study remains direct vs. indirect effects of manipulating miRNA functions and the corresponding phenotype in various cell types in flies. The authors focused most of their effort on using genes that promote GMC differentiation in order to establish the role of neuroblast-specific miRNAs. Most of the experiments were not rigorously performed to the level that eliminates obvious caveats and suggests their interpretation is the most likely possibility. It is a technologically excellent study but lacks in-depth analyses in biological effects.

    Significance

    I believe there is a strong general interest in better appreciating how miRNAs regulate precise gene expression. Deriving some sort of rules such as the specificity of target selection or the efficiency of downregulating gene expression will be hugely significant to the general audience

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Summary

    In this manuscript the authors address the largely unexplored role of micro RNAs (miRNAS) in Drosophila melanogaster brain development, in particular in neural stem cell lineages. The authors for the first time adapt the Ago protein Affinity Purification by Peptides (AGO-APP) technology for Drosophila. They show that this technique works efficiently in neural stem cell lineages and identify several cell type specific active miRNAs. Through a series of bioinformatic analysis the authors identify candidate mRNA targets for these miRNAs. The authors then functionally analyse the role of some of the identified miRNAs, focusing on miRNAs significantly over-represented in neuroblasts.

    By overexpressing Mir-1, the authors demonstrate that this miRNA effectively targets the UTR of Prospero, resulting in the overproliferation of neuroblasts. In a parallel experiment, overexpression of Mir-9c causes neuroblast differentiation defects, similar to the phenotype caused by nerfin-1 mutants, a previously validated target. Loss of function analyses show that knock down of single miRNAs has little functional effects in neuroblast size, showing that the individual effect caused by miRNAs knock down is likely compensated. In contrast, a sponge against a selected group of miRNAs leads to a reduction in poxn positive neuroblasts. Overall these results validate the approach and support the theory that miRNAs cooperate in functional modules during stem cell differentiation.

    Comments

    Title: As the authors do not really explore exit from neural stem cell state this should be altered. The authors do not assess for the levels of any temporal genes, nor other markers of neural stem cell state exit (e.g. nuclear Pros). The observed effects, with the available experiments, rather say that neural stem cell state is not maintained in general, not being clear what mechanistically happens to these cells expressing Cluster 2 sponges. The described phenotype caused by the expression of sponges against individual miRNAs also rather shows a blockage in differentiation.

    • The miRNAs analysed were found in Ago-APP to be predominantly active in neuroblasts, but was there any phenotypes of OE or KD in neurons or glial cells?
    • The authors obtained a phenotype when using a sponge against Cluster 2 in poxn neuroblasts. Is this specific for these 6 neuroblasts? What happens if this sponge is expressed with a pan-neuroblast driver in central brain/VNC/optic lobe? These experiments should be included as they would show if these are conserved effects for all neuroblasts. The authors do different analyses in different brain regions, making also a hard to conclude if all brain regions behave the same way. As authors show that some miRNAs are only expressed in sub-sets of cells, this becomes particularly relevant. Could sponge of cluster 1 cause a phenotype if it had been expressed in other neuroblast lineages? In addition, a discussion of the results obtained from sponge 1 should be included and put in context with miRNA function, technical limitations, levels/cell, targets, pitfalls of analyses, sponges, etc. It would also be interesting to further explore the phenotypes caused by Mir-1 sp expression - are there any milder lineage defects? Any defects in other brain regions/lineages, like in type 2 neuroblasts that usually do not express Pros?

    Ago-APP identifies cell type specific miRNAs in larval neurogenesis section:

    • "...29oC... allows Gal4-dependent expression (Fig.1B,C)" - this description of Gal80ts/Gal4 works is not correct, expression is not prevented.
    • Fig S1 - nab-Gal4 also drives expression in GMCs and neurons, rephrase text. Is nab-Gal4 expressed in optic lobe, VNC and central brain neuroblasts?
    • "20 late larval CNS" - mention the exact stage
    • Providing a more detailed and interpretive description of Figures 1D and 1E would greatly enhance their clarity. Currently, the descriptions of these pannels resemble typical figure legends.
    • Fig. 1F,G,H - It is not clear why the authors sometimes use the optic lobe, other ventral nerve cord as both regions have both neuroblasts, neurons and glia. Are the drivers used for Ago-APP not expressed in all brain regions?
    • Show "data not shown" for 1H.
    • Fig. 1F, G, H - Please quantify intensity levels in the different cell types to facilitate comparison with Ago-APP graphs. Include in figure legend what is "cpm".

    A regulatory module controlling neuroblast-to-neuron transition section:

    • Fig. 2C - A more detailed explanation in text is required in addition to what is mentioned in the figure legend. Including a brief summary/conclusion of the results would be helpful. If possible, add in X-axis 1, 2, 3.
    • Figure S2B - as mentioned in the text elav is expressed from the neuroblast, although this is not represented in the figure. It is hard to tell what are young vs maturing neurons in the cartoon, pls add a label/legend.
    • Fig.3I - please shown a control brain. The merge images are not easy to see. I think it would be nicer to change the figures to be color-blind friendly.
    • Fig. 3K,L - why is this now done in the VNC?
    • Are there any lineage defects when Mir-1 sp is expressed?
    • Based on which parameters/variables of the predicted targets was the Hierarchical clustering done? A brief explanation would help the interpretation of the results and of the choice of the clusters that were further analysed.
    • "revealed the presence of three main groups" - this should be rephrased as this "grouping" was done arbitrarily by the authors and not by hclust. Hclust is set to merge individual clusters/sub-trees up to 1. Furthermore, a more detailed explanation that supported this decision of choosing this 3 large clusters should be included.
    • Fig. 4B, S4B - please include in legend how were these clones generated. S4B - scale bars missing.
    • Fig. 4H - was the ratio of UAS/Gal4 kept in both experimental conditions? Increasing the number of UAS/Gal4 leads to weaker expression of UAS and thus could lead to a weaker phenotype. Including in legends genotype details would help.

    Minor

    • Abstract: "a defined group of miRNAs that are predicted to redundantly target all..." This is only predicted, not experimentally shown, this should be modified accordingly.
    • Intro: "Elav, an RNA binding protein, is expressed as soon as post-mitotic neurons..." - Elav is expressed already in neuroblasts, as also mentioned by the authors in the result section. Correct, add references.
    • Last paragraph of Intro (Bioinformatic analyses...) - it is not easy to understand the content of this paragraph. Rewrite to improve clarity.
    • All legends: Please mention which developmental stage is being analysed in each panel (i.e. wandering 3IL, hours After Larval Hatching, hours After Puparium Formation, or other), in which brain region the analyses/images are being done. Please include more detailed information about the genetics in figure legends.
    • Please include brief explanation of the genetics of miR-10KOGal4 line.
    • Why are miRNAs sometimes referred as (e.g.) "miR-1" and others "miR-1-3p"?
    • Fig. 3I-M - stage of the animal? 3M - in which brain region is this?
    • Fig. 3N - can actual sizes be additionally shown, or at least averages mentioned in text?
    • If non differentially expressed miRNAs, or miRNA with other expression patterns, had been analysed to determine their targets in the sub-set of genes expressed in neuroblasts (from the transcriptome) would different targets been found? Meaning, how specific are these binding patterns for the selected miRNA?

    Referees cross-commenting

    I think this study is very interesting as it optimizes a novel technique in Drosophila for the investigation of cell-specific active miRNAs, and it globally addresses the role of miRNAs in neural stem cell lineages. Although the authors do not explore deeply the biological effect of these miRNAs in neural lineages, I think that the technical contribution and the identification of some miRNA targets is relevant on its own. The authors use Prospero as an example, which is very interesting, as this gene is required to be lowly expressed in Neuroblasts and then upregulated during differentiation. Which the authors propose can be regulated by miRNAs, identifying a novel player in this differentiation mechanism.

    I do not feel the authors need to perform additional experiments to corroborate their findings, as they are well supported by the experiments presented.

    I do agree that the authors did not explore deeply the biological effect in neural lineages, and the claims regarding premature terminal differentiation, nerfin, etc need to be toned down accordingly.

    Significance

    This study is both a technical and conceptual advance. It is very interesting as it optimizes a novel technique in Drosophila for the investigation of cell-specific active miRNAs, and it globally addresses the role of miRNAs in neural stem cell lineages. However, the text, especially in the results section, could benefit from increased detail to enhance the comprehension of the experiments, results, and conclusions. Given that the functional analyses were not conducted at a very detailed level, there exist certain instances of over-interpretation, which could be easily addressed either by revising the text or by incorporating additional experiments, as elaborated upon below.

    This manuscript will be interesting for research fields interested in stem cell differentiation, brain development, micro RNAs, both for Drosophilists and scientists working with other animal models. I am an expert in Drosophila brain development.

  4. Version published to 10.1101/2023.10.05.560992v1 on bioRxiv