miR-9a modulates maintenance and ageing of Drosophila germline stem cells by limiting N-cadherin expression

Ageing is characterized by a decline in stem cell functionality leading to dampened tissue regeneration. While the expression of microRNAs across multiple species is markedly altered with age, the mechanism by which they govern stem cell-sustained tissue regeneration is unknown. We report that in the Drosophila testis, the conserved miR-9a is expressed in germline stem cells and its levels are significantly elevated during ageing. Transcriptome and functional analyses show that miR-9a directly regulates the expression of the adhesion molecule N-cadherin (N-cad). miR-9a null mutants maintain a higher number of stem cells even in the aged tissue. Remarkably, this rise fails to improve tissue regeneration and results in reduced male fertility. Similarly, overexpression of N-cad also results in elevated stem cell number and decreased regeneration. We propose that miR-9a downregulates N-cad to enable adequate detachment of stem cells toward differentiation, thus providing the necessary directionality toward terminal differentiation and spermatogenesis.

8. Some kind of experiment showing that N-cad knockdown, either with RNAi or genetic mutants, rescues the miR-9a phenotypes is needed in order to convincingly argue that elevated N-cad is responsible for the miR-9a phenotype. 9. A more convincing experiment is needed to support the authors hypothesis that germline stem cells do not proliferate because they remain adhered to the hub. Could the authors generate MARCM-labeled miR-9 mutant germline stem cells. This should show that miR-9a mutant clones remain small and adherent to the hub, while control clones contain progenitor and differentiated cells distributed throughout the testis. Additionally, the authors could use this approach to knockdown N-cad in miR-9a clones to show rescue.
10. Is it possible that the miR-9a phenotype could be the result of elevated expression of other cell adhesion molecules like, for example, the previously identified E-cadherin Fmi? Is Fmi expressed in the male germline?

Minor Comments
Line 17/18: Should be rewritten, since it sounds like miR-9a is germline specific and not expressed in any other tissue.
Line 100/101: Numbers (n=) of samples analyzed should be included in the text here and also later when discussing the analysis of N-cadherin overexpression.
Line 107: Not all miRNA-mediated repression leads to mRNA decay. Fig 2j/Fig 4j: The tables should be relabeled, since it is not clear what is listed on rows 2 and 3 (at least in the PDF version that I received). Fig 2i,2k,3c,4i,4k: This is personal preference, but I'd recommend labeling significance of differences in the histograms using asterisks. They would be easier to see there than buried in the figure legends.
Reviewer #2 (Remarks to the Author) The manuscript by Epstein et al. reports the function of miR-9a in germ cells during Drosophila aging. The authors found that miR-9a is upregulated in germline stem cells (GSCs) and spermatogonia progenitor cells during aging. miR-9a loss-of-function mutants have increased GSCs by staining, but decreased male fertility. The authors further characterized an increase of N-Cad upon miR-9a loss, and showed evidence supporting miR-9a directly targets N-Cad. Overexpression of N-Cad leads to similar phenotype as miR-9a loss-of-function mutant.
Although this story is potentially interesting to define a specific miRNA's function during aging in the germline, a number of deficiencies reduce enthusiasm. Major concerns include: 1. The genetic evidence is far from complete to support a miR-9a-N-Cad pathway in GSC aging. For example, the biological phenotype of miR-9a overexpression line was not described. The function of N-Cad as a downstream mediator of miR-9a activity is missing key rescue experiments. The authors should consider N-Cad knockdown in GSCs to rescue miR-9a E39 line. In addition, cooverexpression of miR-9a and N-Cad can be useful on top of the knockdown experiments. 2. It is evident from Fig 3b that there is increased N-Cad staining in hub cells upon miR-9a loss of function. It is difficult to judge whether N-Cad is increased in GSCs. If the authors believe N-Cad is increased in GSCs, more convincing evidence should be presented. Alternatively, the authors need to figure out why miR-9a loss-of-function mutants will result in increased N-Cad in hub cells, given the claim by the authors that miR-9a is not expressed in hub cells. 3. it is critical that the phenotype of the E39 mutant is indeed due to loss of miR-9a function rather than other alterations at the locus. The original paper that published the E39 line also used another KO line J22. The authors can analyze J22 flies to determine their GSC phenotypes. 4. Although the authors attributed the decreased fertility of miR-9a loss-of-function mutant males to the defect in GSC proliferation, there lacks clear evidence to argue against alternative possibilities. For example, mir-9a is highly expressed in some other tissues beyond GSCs. One may argue that defects in those tissues lead to alterations of mating behavior and hence reduced fertility. I suggest the authors to do two things. First, if the authors' model is correct that the reduced fertility is due to lack of GSC proliferation, one would expect a reduced testis size and/or sperm count in miR-9a E39 aged males. Is this the case? Second, germline specific miR-9a overexpression rescue will be very useful to demonstrate a germ-tissue-specific contribution by miR-9a. 5. The increase of miR-9a in aged male germline is somewhat opposite to the expectation. The data would argue that miR-9a increase is helpful to maintain male fertility during aging. The gainof-function experiments above will help to define the function of mir-9a increase. It could be interesting to discuss on this topic. 3. The p values for comparing the proliferating cells can use Fisher exact test rather than t-test. It is unclear to me how t-test can be used in this case.

Reviewer #3 (Remarks to the Author)
The manuscript by Epstein et al. explores the function of miR-9 in the Drosophila GSCs. The authors identify miR-9 from an expression screen performed during aging of Drosophila testis, where miR-9 accumulates with age. The authors use an elegant reporter system to unequivocally demonstrate that miR-9 is specifically expressed in GSCs and spermatogonia. The authors then show that loss of miR-9 function result in an increase in GSCs number, their reduced proliferation and ultimately reduced fertility that increases with severity upon age. The authors also identify Ncad as a target, validate it by reporter assays and in vivo both by loss and gain of miR-9 function. Finally, the authors demonstrate that over expression of N-cad in GSCs causes a similar phenotype to miR-9 mutants. In summary, N-cad is a plausible miR-9 target that could be at the source of the phenotype. Overall the manuscript is very well written and approachable to the broad readership of Nature Communications. The data presented are of high quality and the claims are supported by the data provided. From a stem cell, niche and ageing perspective the data are very interesting. I therefore endorse the manuscript for publication but have some concerns and questions that would need to be addressed prior.
Concerns: 1. Is the miR-9[E39] allele on the W111 genetic background? Can the authors confidently compare the impact on GSCs numbers and infertility? 2. Could the authors provide a brief description of the miR-9[E39] allele in the text? Is it a clean null allele and does the deletion affect neighboring loci? 3. I suggest removing the overexpression of miR-9 in the hub cells as the authors state that it was 'difficult and partial'. Given that miR-9 is not normally expressed in these cells it adds little to the manuscript and distracts from the clear phenotype in GSCs. 4. While the authors identify N-cad as a direct and likely important target of miR-9, it is unlikely to be the only one. The authors should present this important point in the discussion.

Response to Reviewers' comments on manuscript NCOMMS-16-18936-T
We would like to thank the Reviewers for their insightful comments and for their patience as we tested a number of reagents and protocols to address their concerns, often in aged genotypes. In this revised manuscript, we address all of the concerns raised by the reviewers, and provide a point-by-point rebuttal as detailed below.

A better control for the sensor experiment (Fig 1e-g) would be the miR-9 sensor in a miR-9 mutant background. The control sensor and miR-9a sensor are from different labs and may have been differently constructed or inserted in different locations and these differences may explain the differences in expression levels. Furthermore, it is possible that the insertion of miR-9 sequences in the 3'UTR may destabilize the GFP transcript in a miR-9a-independent fashion.
This was an excellent suggestion, which we implemented and now appears in Supplementary  Fig. 1a-b of the manuscript. Indeed expressing the miR-9a sensor in a miR-9a mutant background resulted in GFP expression in GSCs and spermatogonia cells, indicating that the sensor is specifically repressed by miR-9a in these cells. The miR-9a sensor result was further confirmed by miR-9a Fluorescence In Situ Hybridization (FISH) (micrury LNA detection probe xtr-miR-9a-5p 5' and 3' DIG labeled from Exiqon), which showed expression in GSCs and spermatogonia germ cells and not in the hub. Furthermore, there was no FISH signal in miR-9a[E39] mutants (Supplementary Fig. 1c-e). This information was added to the text in the last paragraph of page 3.

The miR-9a phenotypes need to be rescued, either with a genomic fragment or a UAS-miR-9a transgene.
We thank the Reviewer for suggesting this critical experiment. Overexpressing UAS-miR-9a-DsRed (Bejarano et al, 2010) in GSCs and spermatogonia cells of miR-9a[E39] mutants (nosGal4,UAS-miR-9a-DsRed; miR-9a [E39]) was sufficient to return the number of GSCs associated with the hub back to normal both in young and aged adults, increase the division rate of GSCs in aged males, and rescue the age-related sterility of miR-9a[E39] mutants. These results are depicted in Fig. 2 and described in the last paragraph of page 4 of the manuscript.

Ideally, the authors would also show that additional miR-9a alleles also show the same germline stem cell phenotypes. Otherwise, these phenotypes could simply be due to background mutations in the control of miR-9 strains.
Similar to miR-9a[E39] that was reported in the first version, we tested a second miR-9a null allele, miR-9a[j22] (Li et al., 2006), which we also found to display a high average number of GSCs in the niche of young (10.6±0.7, n=29) and aged males (8.4±0.5, n=17). Moreover, GSCs of aged miR-9a [E39] and miR-9a[J22] null mutants completely arrested division in aged flies. These results were added to the manuscript on page 4.

Also, the authors mention that strains were outcrossed, but this should be more clearly described (how many backcrossings?).
miR-9a[E39] was obtained from Prof. Gao's lab as a backcrossed line. These flies were first outcrossed in our lab to w 1118 . The siblings obtained (miR-9a[E39]/+) were then crossed again to obtain the miR-9a [E39]. Homozygocity was confirmed by the wing phenotype (Li et al., 2006) and PCR. This explanation was added to the Methods, page 1.

The description of the gene expression analysis (e.g. lines 113-118) is confusing and needs to be described more clearly. First, is it really the case that 17,477 are differentially expressed? According to Flybase, the fly genome contains 17,728 genes, but it seems unlikely that essentially all genes are differentially expressed. Furthermore, it is also unclear whether this differential expression is between time points or between genotypes. Finally, the "additional filtration" (line 115) needs to be much more fully described to explain how the 17,477 were whittled down to 231.
We apologize for the confusing description provided in the original version and have now provided the following explanation on page 5 of the revised version: "Reads were aligned to the Drosophila genome and gene expression levels were quantified using Htseq-count. This provided a list of 11,416 genes that are expressed in the testis ( Supplementary Fig. 2). Differential gene analysis using the edgeR-classic method provided count per million (CPM) values and p-values. After filtration based on log Fold Change (logFC ≥ 0.9), significance cutoff (p value ≤ 0.05) and minimal CPM per each gene ( ≥ 1), we obtained a group of 450 genes that showed higher expression in young mir-9a[E39] mutant versus control, and 446 genes that were increased in old mutants versus control. Of these, 231 genes showed higher expression in miR-9a[E39] mutant versus control in both young and aged testis (Fig. 3a). A comparison of this list to the 194 in-silico predicted miR-9a targets (Targetscan Fly) yielded six potential direct targets, one of which was senseless, a previously characterized target of miR-9a, confirming library reliability ( Fig.3a and Table 1)". The software and filtration parameters are described in the Methods on page 3.

The elevated N-cad expression in miR-9a mutants looks like it is also found between hub cells. How do the authors explain that if miR-9a is not expressed in the hub. To specifically label the elevated germline N-cad expression, could hub N-cad be knocked down in a miR-9a mutant?
We thank the Reviewer for this comment, as it clarified that we failed to properly explain the niche architecture. The hub is a spherical 3-dimensional (3D) structure of approximately 12 cells, the great majority of which (~9-10) are associated on several planes with all the surrounding GSCs (~8). Furthermore, the size of the hub cells is considerably smaller than that of the GSCs. However, depiction of this data in a 2D image may appear as though many more hub cells are in contact only among themselves than in reality. Therefore, what may appear as an increase in N-cad among hub cells following miR-9a loss is in fact an increase that occurs between hub and GSCs. To depict this point we added a series of Z-stack images of the image presented in Fig. 3b, which shows that GSCs pile around the hub sphere on several plans (Supplementary Fig. 3a-a''). We also presented a 3D projection of 10 Z-stacks showing the spherical structure of the niche (Supplementary Fig. 3b). To further strengthen our hypothesis that N-cad adherent junctions are mostly found in GSCs-hub boundaries, we reduced its expression in the GSCs of the miR-9a[E39] mutants (nosGal4,UAS-Ncad RNAi ;miR-9a[E39]). As shown in Supplementary Fig. 3c-d, this resulted in overall reduction of N-cad staining. This experiment is described on the first paragraph of page 6. Additional phenotypes of this experiment are entailed in Section 8 of this rebuttal. An explanation about hub architecture was added to the Introduction on page 2. Fig 3c quantification,

how was hub cell N-cad distinguished from germline N-cad?
Quantification of N-cad expression was done by defining the entire hub domain. As explained in the previous item, the great majority of the hub cells are in contact with GSCs (i.e. maintain hub-GSC and hub-hub boundary) whereas only a 2-3 cells have only hub-hub contacts. Therefore, most of the observed increase in N-cad levels is related to the loss of miR-9a in the germline.

Does overexpression of miR-9a in cell culture affect cell viability. This could explain the reduction in GFP expression shown in Fig 3.
The Reviewer is correct in that overexpression may sometimes affect cell viability. However, this is not the case with miR-9a overexpression in S2R+ cell culture as no exceptional amounts of dead cells were observed compared to the controls in either the flow experiments or under the microscope. Moreover, as depicted in Fig. 3e, miR-9a is also overexpressed together with N-cad 3'UTR Mut (lane 4), which did not affect the expression of GFP. The transfection efficiency of miR-9a co-expression with GFP-N-cad-3'UTR WT or GFP-N-cad-3'UTR Mut reporters was also measured by qRT-PCR for mature miR-9a, and confirmed similar expression levels. A comment regarding this issue was added to the manuscript on page 6.

The change in N-cad expression in 4c vs 4d is not entirely convincing, since the background looks higher in c than in d. The reduction could be clearer if miR-9a was driven with both upd and nos-gal4's. How many representatives are these images of miR-9a mutant testes?
Overexpressing UAS-miR-9a-DsRed (Bejarano et al, 2010) in the hub (updGal4,UAS-miR-9a-DsRed) was partial in the sense that we could clearly detect a DsRed signal in only a few samples (8/21). However, in those samples where DsRed expression was detected N-cad expression decreased dramatically. We now provide two representative images (Fig. 4d-e) relative to control (Fig. 4c), all taken at the same exposure time. Additionally, we performed the experiment requested by the Reviewer where UAS-miR-9a-DsRed was simultaneously overexpressed in both the hub and GSCs (updGal4;nosGal4,UAS-miR-9a-DsRed). This caused a dramatic reduction in N-cad expression (Fig. 4g). However, here too clear detection of the DsRed signal in both hub and GSCs was only apparent in part of the samples (14/39, Fig. 4g) while the rest of the samples expressed miR-9a-DsRed only in the germline (25/39, Fig. 4f). These results are described in page 6 of the manuscript and in Fig. 4.

Some kind of experiment showing that N-cad knockdown, either with RNAi or genetic mutants, rescues the miR-9a phenotypes is needed in order to convincingly argue that elevated N-cad is responsible for the miR-9a phenotype.
Again, we thank the Reviewer for suggesting this critical experiment. Reducing N-cad levels with UAS-N-cad RNAi in GSCs and spermatogonia cells of miR-9a [E39] mutants (nosGal4,) was sufficient to restore a normal average number of GSCs, division rate and fertility. These results are described in page 7 of the manuscript and in Fig 5.

A more convincing experiment is needed to support the authors hypothesis that germline stem cells do not proliferate because they remain adhered to the hub. Could the authors generate MARCM-labeled miR-9 mutant germline stem cells. This should show that miR-9a mutant clones remain small and adherent to the hub, while control clones contain progenitor and differentiated cells distributed throughout the testis. Additionally, the authors could use this approach to knockdown N-cad in miR-9a clones to show rescue.
The Reviewer is correct in his/her comment that MARCM miR-9a mutant clones could further support our hypothesis and we have attempted to perform the suggested experiment. However, we encountered significant technical issues in obtaining the triple recombinant flies that we designed for the MARCM crosses (miR-9a[E39], FRT2A, nanos-GAL4). However, in this revised manuscript we used the other critical experiments suggested by the Reviewers to strengthen our hypothesis by showing that GSCs of miR-9a mutants arrest division in aged flies (PHH3 staining and fertility assay) and that these phenotypes are completely rescued either by N-cad RNAi or miR-9a overexpression in GSCs.

Is it possible that the miR-9a phenotype could be the result of elevated expression of other cell adhesion molecules like, for example, the previously identified E-cadherin Fmi? Is Fmi expressed in the male germline?
Our transcriptome analysis of testes from wild-type and miR-9a[E39] mutants shows that Fmi levels are low and are not further increased in the mutant flies (See Reviewer table 1 below). Moreover, staining the testis with anti-Fmi Ab (Hybridoma Bank) did not reveal any signal, although staining embryos in the same sample showed Fmi expression in the CNS (See Reviewer Fig. 1 below). These results suggest that in the testis, the miR-9a phenotype is not due to elevated expression of Fmi.

Line 17/18: Should be rewritten, since it sounds like miR-9a is germline specific and not expressed in any other tissue.
The sentence was revised to show that the described phenomenon pertains to the Drosophila testis: "Here we report that in the Drosophila testis, the conserved miR-9a is expressed in germline stem and progenitor cells and its levels are significantly elevated during ageing."

Line 100/101: Numbers (n=) of samples analyzed should be included in the text here and also later when discussing the analysis of N-cadherin overexpression.
Reviewer Fig. 1. Apical tip of the testis (left) and embryo stage 14 (right, arrow marks anterior) stained with Fmi (red), Vasa (green) and DAPI (blue). Fmi is expressed in the embryo CNS but not in the testis. Scalebars 20µm.
Reviewer Table 1: Transcriptome analysis of N-cadherin and Fmi in testis of young and aged wild-type and miR-9a[E39] mutants. Fmi levels are low and not increase in miR-9a mutants.
The number of GSCs that were scored and the percentage of PHH3 positive GSCs were added to the text on pages 4 and 7.

Line 107: Not all miRNA-mediated repression leads to mRNA decay.
The sentence was changed to: "miRNAs repress mRNA translation, which is often followed by the mRNA deadenylation and decay. "   Fig 2j/Fig 4j: The tables should

be relabeled, since it is not clear what is listed on rows 2 and 3 (at least in the PDF version that I received).
We apologize for the quality of the PDF in general and particularly for the data that was presented in the Tables in Fig. 2 and 4. We realize that presentation of the data in this manner was cumbersome. We therefore removed these Tables from the revised version and include the data regarding mitotic GCSs and the overall GSCs that were scored in the text on pages 4 and 7.

Fig 2i,2k, 3c, 3f-3i, 4i, 4k: This is personal preference, but I'd recommend labeling significance of differences in the histograms using asterisks. They would be easier to see there than buried in the figure legends.
The Reviewer is correct and asterisks were added to the histograms.

1.The genetic evidence is far from complete to support a miR-9a-N-cad pathway in GSC aging. For example, the biological phenotype of miR-9a overexpression line was not described. The function of N-cad as a downstream mediator of miR-9a activity is missing key rescue experiments. The authors should consider N-cad knockdown in GSCs to rescue miR-9a E39 line. In addition, co-overexpression of miR-9a and N-cad can be useful on top of the knockdown experiments.
We thank the Reviewer for this insightful comment. While overexpression of miR-9a in the germline of WT flies (nos-GAL4;UAS-miR-9a-DsRed) reduces N-cad expression (Fig. 4), it does not show a significant effect on the number of GSCs in the aged fly. On the other hand, overexpression of miR-9a on a mutant background (nosGal4, ) restores the average stem cell number back to their original numbers in the aged fly (Fig. 2). The reason for which the miR-9a overexpression in the WT does not show any effect probably stems from the fact that aged animals already express very high amounts of miR-9a (up to ~1% of the entire miRNA in the testis; a comment was added to the results on page 3), this may generate a 'ceiling effect' where additional overexpression cannot enhance the effect further. Having said that, we performed the key experiments suggested by the Reviewer, which considerably strengthen the genetic evidence for miR-9a -N-cad axis in aging. We now show that reducing N-cad levels in GSCs and spermatogonia cells of miR-9a [E39] mutants (nosGal4,) was sufficient to: a) return the number of GSCs associated with the hub back to normal numbers in both young and aged adults b) increase the division rate of GSCs in aged males, and c) rescue the age-related sterility of miR-9a [E39] mutants (See Fig. 5 and page 7 in the manuscript).

It is evident from Fig 3b that there is increased N-cad staining in hub cells upon miR-9a loss of function. It is difficult to judge whether N-cad is increased in GSCs. If the authors believe N-cad is increased in GSCs, more convincing evidence should be presented. Alternatively, the authors need to figure out why miR-9a loss-of-function mutants will result in increased N-cad in hub cells given the claim by the authors that miR-9a is not expressed in hub cells.
This is an important point that we now clarify. First, by adding a FISH experiment for miR-9a, we strengthen the data obtained from the miR-9a-GFP sensor (Fig. S1) that indeed miR-9a is expressed in GSCs and spermatogonia but not in hub cells. Second, the hub is a spherical 3dimensional (3D) structure of approximately 12 cells, the great majority of which (~9-10) are associated on several planes with all the surrounding GSCs (~8). Furthermore, the size of the hub cells is considerably smaller than that of the GSCs. However, depiction of this data in a 2D image may appear as though many more hub cells are in contact only among themselves than in reality. Therefore, what may appear as an increase in N-cad among hub cells following miR-9a loss is in fact an increase that occurs between hub and GSCs. To depict this point we added a series of Z-stack images of the image presented in Fig. 3b, which shows that GSCs pile around the hub sphere in several plans ( Supplementary Fig. 3a-a''). We also presented a 3D projection of 10 Z-stacks showing the spherical structure of the niche ( Supplementary Fig. 3b). To further strengthen our hypothesis that the major role of N-cad is to adhere stem and niche, we stained the described above N-cad RNAi in the miR-9a [E39] mutants (nosGal4,UAS-N-cad RNAi ;miR-9a[E39]) with anti-N-cad antibodies. As shown in Supplementary Fig. 3c-d, this resulted in an overall reduction of N-cad staining. This experiment is described on the first paragraph of page 6 of the manuscript. An explanation about hub architecture was added to the Introduction on page 2.

It is critical that the phenotype of the E39 mutant is indeed due to loss of miR-9a function rather than other alterations at the locus. The original paper that published the E39 line also used another KO line J22. The authors can analyze J22 flies to determine their GSC phenotypes.
We thank the Reviewer for suggesting these experiments to further strengthen our data.
To verify that the phenotypes are due to miR-9a loss we did the following: 1) miR-9a[E39] was obtained from Prof. Gao's lab as a backcrossed line. These flies were then outcrossed in our lab first to w 1118 . The siblings obtained (miR-9a[E39]/+) were then crossed again to obtain the miR-9a [E39]. Homozygocity was confirmed by the wing phenotype 2 3 and PCR. This explanation was added to Methods, page 1.

Although the authors attributed the decreased fertility of miR-9a loss-of-function mutant males to the defect in GSC proliferation, there lacks clear evidence to argue against alternative possibilities. For example, mir-9a is highly expressed in some other tissues beyond GSCs. One may argue that defects in those tissues lead to alterations of mating behavior and hence reduced fertility. I suggest the authors to do two things. First, if the authors' model is correct that the reduced fertility is due to lack of GSC proliferation, one would expect a reduced testis size and/or sperm count in miR-9a E39 aged males. Is this the case? Second, germline specific miR-9a overexpression rescue will be very useful to demonstrate a germ-tissue-specific contribution by miR-9a.
The Reviewer is correct in his/her comment that lack of fertility in aged miR-9a mutants can result from additional factors besides a GSC proliferation defect. Since the sperm cells in the Drosophila testis are long, convoluted and attached to one another, we were unable to obtain a reliable measurement. We therefore took the Reviewers second advice and selectively overexpressed miR-9a in the GSCs and spermatogonia cells of the miR-9a[E39] mutants. The results, which were added to Fig. 2, clearly indicate that this was sufficient to completely rescue the age-related sterility.

The increase of miR-9a in aged male germline is somewhat opposite to the expectation.
The data would argue that miR-9a increase is helpful to maintain male fertility during aging. The gain-of-function experiments above will help to define the function of mir-9a increase. It could be interesting to discuss on this topic.