Piwi reduction in the aged niche eliminates germline stem cells via Toll-GSK3 signaling

Transposons are known to participate in tissue aging, but their effects on aged stem cells remain unclear. Here, we report that in the Drosophila ovarian germline stem cell (GSC) niche, aging-related reductions in expression of Piwi (a transposon silencer) derepress retrotransposons and cause GSC loss. Suppression of Piwi expression in the young niche mimics the aged niche, causing retrotransposon depression and coincident activation of Toll-mediated signaling, which promotes Glycogen synthase kinase 3 activity to degrade β-catenin. Disruption of β-catenin-E-cadherin-mediated GSC anchorage then results in GSC loss. Knocking down gypsy (a highly active retrotransposon) or toll, or inhibiting reverse transcription in the piwi-deficient niche, suppresses GSK3 activity and β-catenin degradation, restoring GSC-niche attachment. This retrotransposon-mediated impairment of aged stem cell maintenance may have relevance in many tissues, and could represent a viable therapeutic target for aging-related tissue degeneration.

through the Glycogen synthase kinase-(shaggy in Drosophila)-beta-catenin-(armadillo in Drosophila) E-Catherin pathway. The authors also proved that the Toll mediated immune signalling is activated in piwiRNAi niches, since mislocalization of E-cadherin in the piwiRNAi niche is rescued by Toll depletion. In summary, the authors describe a novel "quality assurance system" for Drosophila adult female germ line stem cells. As symptom of aging, the Piwi transposon regulation system weakens in the stem cell niche. As a consequence, retrotransposons are deregulated and produce virus-like particles capable of infecting the nearby stem cells. De-repression of retrotransposons, however induces a Toll-mediated immune response which activates Glycogen synthase kinase. Glycogen synthase kinase then induces degradation of beta-catenin, which finally results in delocalisation of E-Catherine from CpC-GSC junction. The end result is elimination of GSCs from the stem cell niche infected with the virus-like particles. The focus of the manuscript is the Drosophila female stem cell niche. However, the authors present a series of experiments demonstrating that the immune response to retrotransposon derepression may be a general protective phenomenon. They showed that 1) Piwi and Armadillo expressions are reduced while expression of copia-lacZ is increased in the aged Drosophila male GSC niche. 2) 3TC treatment suppresses GSK3 activity in human cell model of Alzheimer's disease.
3) Using the fly Alzheimer model thy proved that 3TC treatment partially suppresses GSK3 activity induced in the Ab42-overexpressing fly eyes. In summary, the results presented in the manuscript may be of interest to a wide range of readers Critical comments: 1) The authors observed, and quoted the literature, that number of GCS decreased with age. They also showed that piwiRNAi in CpCs was resulting in a decrease in GCS number. However, two weeks were needed for the complete loss of Piwi expression (line 57). The two weeks lag of the effect of RNAi could well be the result of the well-known protein perdurance. In line 61, the authors wrote:" As previously reported12,13, the number of GSCs in control flies decreased with age ( fig. 1d and Supplementary Table 1). This decrease was accelerated in two independent piwiknockdown (KD) lines ( fig. 1d and 64 Supplementary Table 1)." The authors then consistently, but I think wrongly, classify piwi silenced phenotype as an age dependent loss." In my opinion, if RNAi had an immediate effect on Piwi protein depletion, the authors would not write about an increase of age-dependent loss. It would be better to apply the term age-dependent loss only to the phenotypes measured in the aged wild type flies.
2) The authors clearly demonstrated that the piwi mediated protection from the harmful effect of Gypsy retrotransposon is strictly cap cell specific. The other two types of somatic niche cells, the terminal filament cells and the anterior escort cells are not involved in this process. TF cells do not express Piwi at all. Piwi expression is only slightly reduced in the very old, 8-week-old escort cells. To examine which types of retrotransposons are suppressed by Piwi in the niche cells, the authors examined a publicly available transcriptome database for retrotransposon expression levels in a piwi-knock down ovarian somatic cell line (OSC). In line 85 the authors wrote" …...ovarian somatic cells (OSCs), which resemble CpCs to support GSCs 21." According to the citation no 21: "We assume that the somatic component of the cultures (from which the OSC cell lines were sub cloned) is derived from the somatic stem cells (cf. Fig. 1b)". This means that the origin of the OSC cell line is ambiguous and the used cell lines most probably do not represent CpCs properly. Therefore, I suggest to omit the in silico analysis of OSC expression patterns from the manuscript. 3) line 94 "Piwi acts in the nucleus through epigenetic mechanisms to suppress transposon transcription22, while cytoplasmic Piwi targets transposon transcripts for degradation23. Ovaries carrying mutant Piwi with a nuclear localization defect (piwiNT) did not accelerate age-dependent GSC loss (Supplementary Fig. 3 and Supplementary Table1), suggesting that Piwi suppresses cytoplasmic retrotransposon transcripts for replication acting as the second defence mechanism to maintain GSCs." Firstly, it seems to me that ref. 23 was wrongly cited. It is a general though in the field, that Piwi exclusively participates in co-transcriptional silencing that happens in the nucleus and does not participate in the cytoplasmic silencing. It is also generally accepted that in the ovarian somatic cells only the co-transcription silencing but not the cytoplasmic silencing is functional. Secondly, The cited ref. 23 says: "Transcript level of the Gypsy and mdg1 retrotransposons increased 10-to 20-fold in the ovaries of homozygous piwiNt and piwi2 or transheterozygous piwi2/piwiNt flies relative to the corresponding heterozygotes (Fig. 3A)." Later: "Here we describe the phenotype of a unique mutation in the piwi gene that leads to the formation of cytoplasmic PiwiNt (i.e.,N-truncated Piwi protein) lacking the NLS. The properties of this mutant made the direct influence of the piRNApathway on GSCmaintenance unlikely, as the piwiNtmutant displayed normal GSC self-renewal ( Fig. 1 C and D) but lost Piwi mediated transposon repression completely in ovarian cells (Fig.3A) including niche cells responsible forGSCself-renewal signalling (Fig. 4B). " There is a plain contradiction between the ref. 23 and the manuscript. According ref. 23, there is an elevated Gypsy level in piwiNt mutant germaria (most probably in young ovaries since the ref. 23 did not focus on the ageing). According the manuscript, however, "Ovaries carrying mutant Piwi with a nuclear localization defect (piwiNT) did not accelerate age-dependent GSC loss." With the other words the elevated Gypsy level in the GSC niche might not be the reason of the age dependent GSC loss. This contradiction must be resolved in every way. 4) line 98: "…..suggesting that Piwi suppresses cytoplasmic retrotransposon transcripts for replication acting as the second defence mechanism to maintain GSCs. To test this hypothesis, we suppressed reverse transcription of replicating retrotransposons by feeding bab1>gfpRNAi-1 and bab1>piwiRNAi flies with (-)-L-2',3'-dideoxy-3'-thiacytidine (3TC), a cytidine analogue that is clinically used to inhibit reverse transcription of human immunodeficiency virus and Hepatitis B virus24,25. If the cytoplasmic function of Piwi in transposon silencing had be true, the logic of the introduction of 3TC experiments would have been still incorrect. Namely, irrespective of whether the tertotransposon de-repression occurs in the nucleus or in the cytoplasm, 3TC experiment may give positive result. That is, by 3TC experiments it is not possible to test the hypothesis about the cytoplasmic activity of Piwi. 5) line 40: "Despite their genomic abundance, transposons are effectively silenced by Piwiinteracting RNAs (piRNAs) at transcriptional and post-transcriptional levels 6,7. However, this process is known to be attenuated in aged tissues 8-10, where stem cells are frequently lost11". In this context, "this process" refers to the piwi pathway. The authors refer three publications (No 8,9,and 10) concerning the age-related attenuation of the Piwi system. Unfortunately, neither of the cited publications is about the Piwi pathway.
Reviewer #3 (Remarks to the Author): This manuscript examines the age-dependent loss of germline stem cells in the Drosophila ovary. The authors combine molecular, imaging, and genomic data to propose a link between piwi and GSC loss. Specifically, they interpret the data to suggest that "retrotransposons in the aged GSC niche generate endogenous virus to activate Toll-mediated immune signaling, which subsequently activates Glycogen synthase kinase 3 (GSK3) to impair b-catenin-E-cadherin-mediated GSC anchorage, finally resulting in GSC loss". This is a very interesting hypothesis that would likely be of interest to a range of researchers studying stem cell biology, ageing, and niche function.
While I appreciate the amount of work presented in this study, and the high technical quality of the experiments, I have several concerns about claims contained in the manuscript. I have detailed my major concerns below.
1. Virus -Toll: In several instances, the manuscript discusses the generation of virus within the stem cell niche. These claims are not supported by the data. Unless the authors show that infectious cell-free virus particles capable of generating an integrated provirus in newly infected cells are produced, they cannot make any statements about virus production. The instances of virus-like particles in some TEMs are certainly interesting, but more work is required. Equally importantly, claims of virus-dependent activation of Toll signaling require more support. To support their claims that the niche is activating a Toll-dependent antiviral response, the authors cite two publications that link Toll signaling to antiviral defenses in Drosophila. However, it is important to note that one of those studies established a role for Toll-7 in antiviral defenses. Importantly, the authors effectively exclude Toll-7 as a receptor. The second cited manuscript implicated Toll in fly antiviral responses. However, the Toll receptor is generally considered to recognize microbial patterns at the plasma membrane. In the context of this manuscript, it is not clear how Toll can detect cytosolic viral particles or genomes, nor do the authors present any data to indicate that is happening. It is possible that the putative antiviral responses under investigation here are being mediated by Toll-5, but there are no data to support an antiviral role for Toll-5, or to suggest that Toll-5 is able to bind viral particles or genomes in the cytosol. Furthermore, there are no data to support activation of the Toll pathway (e.g. phosphorylation and degradation of cactus, nuclear accumulation of Dorsal, involvement of MyD88, etc). Thus, although is possible that Toll-mediated antiviral responses contribute to the phenotypes under investigation, considerably more data is needed to support that contention.
2. Statistical evaluations: The authors have gone to considerable lengths to stain, image, and quantify GSCs in dissected ovaries. This is a truly impressive accomplishment, and has likely generated some very interesting data. However, I am concerned about the statistical methods they have used to quantify their results. In particular, I am concerned about the tests used to evaluate data derived from counting GSCs per ovary. For this part of the study, the authors generated large amounts of categorical data where each ovary was scored as having 0, 1, 2, or 3-4 GSCs. It is not at all clear why 3-4 were grouped in a single bin. Surely, they should also be scored separately? More importantly, it is not clear why the authors chose t-tests for the comparisons between the groups, as the relevant section of the materials and methods is very short. For this reviewer, the t-test is inappropriate for almost all comparisons performed in this manuscript. The authors should be encouraged to re-consider their statistical approaches to all assays presented in the manuscript, relying on more appropriate tests, and writing a much more detailed methods section that allows readers to determine whether the tests are indeed appropriate.
3. Presentation of data: On a minor note, it is not advisable to present multiple data measurements as bar charts with error bars. The authors should be encouraged to use any one of numerous charting methods to properly lay out all data from all assays (box plots/dot plots/violin plots/etc). More importantly, many of the figures do not present data for relevant and important controls. To provide one example, figure 3 compares the effects of piwi-RNAi to a combination of piwi-RNAi and sgg-RNAi. Importantly, the piwi-RNAi flies should express a control RNAi construct (e.g. GFP-RNAi) to confirm that the rescue is not simply a result of diminishing RNAi-dependent depletion of the piwi transcript through the introduction of an extra construct. Similarly, many phenotypes are only reported in combination with the piwi-RNAi phenotypes with no data on the effects of single gene inactivation (e.g. Toll RNAi phenotypes).
4. Phenotypic effect and variability: Much of the quantitative data is interpreted based on p-values. However, for this reviewer the effect of piwi loss appears quite mild in many instances. The authors should be encouraged to include size effects when making claims of significance. Of equal importance, the authors should be asked to address the apparent variability of the piwi-RNAi phenotype in the manuscript. For example, Figure 1d shows an apparently highly significant loss of GSCs in piwi-RNAi flies by two weeks. However, in figure 2e, this effect appears to be gone for piwi-RNAi2. Specifically, a visual inspection of the first column (bab1>GFP-RNAi, no 3-TC), and the fifth column (bab1>piwi-RNAi2, no 3-TC) suggests that there are no differences between the two genotypes. Likewise, the 7th column (bab1>piwi-RNAi2) appears distinct from the 5th. Similar effects are apparent in supplementary figure 7. Specifically, the %GSC remaining data for 2261>GFP raised for 7 weeks in the absence of RU486 (panel d second column), are quite different to the percentage of GSCs remaining in 2261>armS10 raised for 7 weeks in the absence of RU486 (panel e, 2nd column). This variability does not necessarily weaken the claims of the manuscript, and there are several possible explanations (genetic background etc). Nonetheless, it would be helpful for the authors to acknowledge this point in their manuscript, and to provide frequent descriptions of size effects, so that readers can appreciate the extent of variability. 5. Supplementary data: For this reviewer, the possible links to Alzheimer's models is not particularly relevant to the report, and seems distracting and premature. I would encourage the authors to consider removing these supplemental data so that they can focus on an interesting GSC phenotype.

<Response to reviewers>
We thank all of the reviewers for all of their critical and constructive comments.

<Response to reviewer>
We thank the reviewer for these questions. The gyspy retrotransposon is a highly We also attempted to rescue the piwi-KD phenotype by knocking down another retrotransposon, copia; however, copia knockdown was unable to rescue the GSC-loss phenotype. As for the other retrotransposons increased in piwi-KD OSCs, we do not have suitable RNAi lines to test their roles. In our genomic sequence analysis, we did identify new insertion sites generated by DNA transposons and retrotransposons; however, we believe that testing whether these insertions disrupt gene function and the effects on GSC maintenance are beyond the scope of our current study. Overall, we show that gypsy upregulation is at least largely responsible for the GSC-loss phenotype, but our results do not completely rule out roles for other retrotransposons on the GSC-loss phenotype.

<Response to reviewer>
We thank the reviewer for these comments pointing out our potential mis/over-interpretation of data from previous studies. We have completely removed the piwi NT data and conclusion regarding the cytoplasmic role of Piwi in retrotransposon silencing, as it is not our major focus (also see our response to reviewer 2, comment 3).

It helps if authors can describe levels of F-actin, toll signaling and pSer9-GSK9 in
aged cap cells, in addition to the KD model, to argue that they occur during aging.

<Response to reviewer>
We thank the reviewer for these suggestions. In this revised manuscript, show that Cactus is also dramatically reduced in aged cap cells; however, we cannot clam this reduction is due to activation of Toll signaling.

It would make the manuscript much stronger if authors do RNA-seq in sorted niche
cells. OSC can serve as a reference, but it describes other cell types like follicular cells.

<Response to reviewer>
We understand the reviewer's concern. While OSCs express follicle cell markers, these cells also exhibit niche characteristics by expressing Dpp, which can support cultured ovarian germline stem cells (Niki et al., PNAS, 2006). Therefore, the transcriptome analysis in piwi-KD OSCs can provide useful information for our study.
However, we also noticed that the transcriptome analysis in OSCs was derived from only one replicate; therefore, we have completely removed the OSC data from the current version of the manuscript.
As suggested by the reviewer, we isolated niche cells from bab1>gfp & piwi RNAi Since GSK3 is regulated by phosphorylation, which is carried out by different kinases, and Toll signaling is activated by bacteria and fungi, we believe GSK3 phosphorylation state and Toll signaling are also regulated by Piwi-independent mechanisms. We have added a brief discussion regarding the complexity of these aging-related phenotypes in the main text, pg. 11, line 7 and supplementary Fig. 10.
6. Can 3TC prevent RT in GSCs as well to at least in part explain the rescue?

<Response to reviewers>
We thank the reviewer for this question. It is possible, but less likely, that 3TC prevents reverse transcription (RT) in GSCs, unless viral materials encoded by retrotransposons are transported/or present in GSCs. However, we did not find virus-like particles in GSCs by TEM (see Fig. 5).
7. AD data in the end seem largely out of place. Adding these weakens the manuscript.

<Response to reviewers>
We understand the reviewer's concern, which is why we present these results as supplementary information. We would like to use these data to demonstrate the relevance of retrotransposon-GSK actions to other aging-related conditions.
Although the results are not highly related to GSCs, we believe this finding "retrotransposon-GSK3 regulation" could benefit research on aging-associated tissue degeneration (e.g., AD). In this revised manuscript, we therefore still kept these findings as supplementary information.

<Response to reviewers>
We thank the reviewer for this suggestion. We have modified the text to specifically mention cap cells when the analysis was performed in cap cells.
10. Line 178, will EC (in addition to TF+CpC) also be GFP-positive?

<Response to reviewers>
We thank the reviewer for this question. Only the anterior-most ECs, which contact GSCs, express very weak GFP (Supplementary fig. 3a).

Method of retrotransposon analysis is poorly documented. For instance, which
annotation is used? e.g. springer cannot be found in either RepBase of FlyBase.

<Response to reviewers>
We thank the reviewer for this comment. We have added more information to our description of retrotransposon analysis in the Materials and Methods, pg. 29, line 5.
We used the "Transposable elements (canonical set)" database downloaded from FlyBase for transposon annotation; the ID of Springer in FlyBase is FBte0000333.
12. Fig 2c, pie chart is an ineffective way to visualize data, where one cannot appreciate a de-repression of transposon without reading the numbers.

<Response to reviewers>
We thank the reviewer for this comment. We have changed the pie chart to a scatter plot to better show TE expression levels in piwi-KD niche cells (Fig. 2c).
13. Nearly all "enlarged" images lack scales. Also, in Fig S7b, outline in enlarged image seems highly inconsistent with the outline in the original image.

<Response to reviewers>
We thank the reviewer for this comment. We have added scale bars in the enlarged images and mentioned these scale bar sizes in the legends of figures 2, 3 and 4. We also modified Fig. S12b (original S7b) to make the enlarged image and original image consistent.

<Response to reviewers>
We thank the reviewer for this comment. We have replaced the original image with a more representative one.

While cap cells survive by 2W
, there is statistically significant increase of DNA damage (Fig S4), which should be described in text.

<Response to reviewers>
We thank the reviewer for this comment. We have added a brief statement that DNA damage is significantly increased in 2-and 5-week-old cap cells, in the main text, pg.

<Response to reviewers>
We understand the reviewer's comment. However, we measured GSCs in flies at ages: day 1, 2 weeks, and 5 weeks. GSC numbers in control flies from 2 weeks to 5 weeks of age dropped from 83% to 69% of day one levels (a 14% reduction). Regardless of whether GSCs were maintained by Piwi perdurance before 2 weeks of age, flies with depleted Piwi expression in the niche showed a 21-23% reduction of GSCs from 2 to 5 weeks using either of two different piwi RNAi lines. These results clearly show that piwi depletion in cap cells accelerates the age-dependent GSC loss. We, therefore, think that our conclusion is accurate and have kept our description.
2) The authors clearly demonstrated that the piwi mediated protection from the  . 1 C and D) but lost Piwi mediated transposon repression completely in ovarian cells (Fig.3A) including niche cells responsible for GSC self-renewal signalling (Fig. 4B). "

<Response to reviewer>
We thank the reviewer for these comments and understand their meaning. We have completely removed the piwi NT data and any mention of the cytoplasmic role of Piwi

<Response to reviewer>
We agree with the reviewer's comment. We have removed our hypothesis regarding the cytoplasmic role of Piwi in targeting transposon transcripts for degradation.
Instead, we ask about the involvement of retrotransposons in the GSC loss phenotype by treating piwi-KD flies with 3TC; pg. 6, line 6.
However, this process is known to be attenuated in aged tissues 8-10, where stem cells are frequently lost11". In this context, "this process" refers to the piwi pathway.

<Response to reviewer>
We thank the reviewer for this critical comment. In mammals, TLR7 and TLR9 are located in the endosome and directly detect foreign DNA/RNA and proteins in the cytoplasm (Mogensen, 2009, Clin Microbiol Rev). Interestingly, Drosophila Toll is the homologue of TLR7/9 (supplementary table 4), as predicted by DIPOT (Hu, et al., 2011, BMC Bioinformatics). Toll is present in the plasma membrane and cytoplasm, and it requires the endocytic pathway for activation (Huang, et al., 2010, PNAS). Thus, Toll might behave similar to TLR7/9 in detecting viral genetic materials to activate immune signaling. On the other hand, it has been reported that Toll-5 interacts with Toll and participates in activation of Toll signaling (Luo, et al., 2001, Insect Mol Biol.).
Toll-5 is the homologue of TLR4, which detects viral proteins (Mogensen, 2009, Clin Microbiol Rev); however, it is not clear if Toll-5 also detects viral proteins and its role in activating Toll signaling in cap cells is not confirmed. We have added a brief discussion regarding the possible mechanism for Toll/Toll-5 detection of viral materials on pg. 15, line 10.
Although we do not conclusively know that Toll/Toll-5 detect viral genetic material (which is beyond our scope at the moment), our data nevertheless show that Toll or Toll-5 knockdown in piwi-KD cap cells are able to rescue GSC number to the wild-type levels.
Furthermore, there are no data to support activation of the Toll pathway (e.g. phosphorylation and degradation of cactus, nuclear accumulation of Dorsal, involvement of MyD88, etc).

<Response to reviewer>
We thank the reviewer for this comment. As suggested, we have performed IHC for were below detectable levels. In addition, decreased pSer9GSK3 in piwi-KD via Toll may be independent of MyD88. We, therefore, did not test if MyD88 is involved in Toll signaling.
Thus, although is possible that Toll-mediated antiviral responses contribute to the phenotypes under investigation, considerably more data is needed to support that contention.

<Response to reviewer>
We thank the reviewer for this evaluation. In the revised manuscript, we have softened the description regarding the link between Toll and virus; however, we believe that our data can fully support a model where retrotransposon derepression in the aged niche requires Toll to reduce GSCs. <Response to reviewer>

Statistical evaluations
We thank the reviewer for this question. The number of germaria carrying 4 GSCs is very low in most groups (less than 10% of the total). Separating the groups of germaria carrying 4 GSCs from germaria carrying 3 GSCs does not change our conclusions. Thus, to simplify the graph we grouped germaria carrying 3 and 4 GSCs in the same category.

More importantly, it is not clear why the authors chose t-tests for the comparisons between the groups, as the relevant section of the materials and methods is very short. For this reviewer, the t-test is inappropriate for almost all comparisons
performed in this manuscript. The authors should be encouraged to re-consider their statistical approaches to all assays presented in the manuscript, relying on more appropriate tests, and writing a much more detailed methods section that allows readers to determine whether the tests are indeed appropriate.

<Response to reviewer>
We thank the reviewer for this comment. In comparisons of two groups, we chose the t-test. We also performed chi-square tests to compare frequencies of events between two genotypes, but we omitted the results as they are similar to the t-test analyses. Nevertheless, as suggested by the reviewer, we have changed most of the statistical comparisons to One-way ANOVA (Fig. 3g, 3h,

<Response to reviewer>
We thank the reviewer for this comment. We have changed bar charts to Box plots or dot plots in Fig. 3 and 4 and supplementary fig. 1 , 7, 8, 9 ,10, 12 and 15. We still kept the original graphs showing GSCs and cap cell numbers, as it clearly shows the difference among germaria in response to the genetic manipulation.
We have also added proper controls for each experiment. For the rescue experiments using double knockdown (Fig. 3h, j, g and Fig. 4a), we performed co-knockdown of piwi and gfp simultaneously as the control. In this revised manuscript, we have added the genotype to each panel.
We only used piwi RNAi alone as the control for piwi  co-knockdown experiments (Fig. 2e, Fig. 3h  We also added results of GSC analysis in single gene knockdown flies, including sgg-, toll-, toll-5, and toll-7-KD (Supplementary Table 1). We found that depletion of Toll-5 in the niche for 2 weeks reduced GSC number by 17% compared to gfp-KD control, while depletion of sgg, toll and toll-7 did not affect GSC number at 2 weeks.
We do not know why knockdown toll-5 alone caused a reduction of GSCs.

Phenotypic effect and variability: Much of the quantitative data is interpreted
based on p-values. However, for this reviewer the effect of piwi loss appears quite mild in many instances. The authors should be encouraged to include size effects when making claims of significance.

<Response to reviewer>
We thank the reviewer for this comment. In the control group, the normal rate of GSC loss from flies at Day 1 to 2 weeks old was approximately 17-20%. This loss rate was increased to 25-34% in flies carrying piwi-KD cap cells (either using piwi  or piwi    Figure 1d shows an apparently highly significant loss of GSCs in piwi-RNAi flies by two weeks. However, in figure 2e, this effect appears to be gone for piwi-RNAi2. Specifically, a visual inspection of the first column , and the fifth column  suggests that there are no differences between the two genotypes. Likewise, the 7th column (bab1>piwi-RNAi2) appears distinct from the 5th.

<Response to reviewer>
We thank the reviewer for this comment. piwi RNAi-2 did give us a milder phenotype, in terms of GSC loss. We now have mentioned this point in the main text, pg. 6, line 13.

Supplementary data: For this reviewer, the possible links to Alzheimer's models is
not particularly relevant to the report, and seems distracting and premature. I would encourage the authors to consider removing these supplemental data so that they can focus on an interesting GSC phenotype.

<Response to reviewer>
We understand the reviewer's concern. Although these results are not highly related to GSCs, we believe this finding "retrotransposon-GSK3 regulation" may benefit researchers interested in aging-associated tissue degeneration (e.g. AD). In the revised manuscript, we therefore still kept these findings in the supplementary information.
Major concerns:

A direct link between Toll and transposon remains weak.
In rebuttal letter the authors admit that they are not able to provide molecular evidence for a direct link between retrotransposons and Toll signaling and write that they 'soften' the title and changed the abstract accordingly.  4d. and described the results in pg. 14, line 15.
However, niche-specific co-knockdown of piwi with either gypsy, toll or toll5, or suppression of reverse transcription in the piwi-KD niche by 3TC treatment were all able to reduce GSK3 activity, prevent -catenin degradation, restore E-cadherin expression in the GSC-niche junction, and rescue GSC number. These results clearly show that retrotransposon cDNA, gyspy and Toll-GSK3 signaling are all required to produce the major piwi-KD phenotype.
In addition, the data for Toll-KD alone is shown in the Supplementary Table 1.
We added a brief description in the main text, pg.13, line 15. Knockdown of toll or toll7 alone in the niche does not affect GSC maintenance as compared to controls, but knockdown of toll5 alone significantly enhances GSC loss from D1 to 2 weeks after eclosion (33% loss), as compared to controls (12-17% loss, P < 0.001). However, co-knockdown of piwi with toll or toll5 significantly restores GSC maintenance rate to control levels. These results indicate that Toll-mediated signaling is involved in the GSC loss induced by Piwi depletion in the niche.
In the original fig. 4d, for the grouping '% of germaria carrying indicated GSC number' we only showed the results at 2 weeks. To avoid confusion, we have now also added the results from newly enclosed flies (D1) and show these data in Fig. 4a' of the revised manuscript. The results show that knockdown of piwi in the niche significantly accelerates GSC loss from D1 to 2 weeks after eclosure, while co-knockdown of piwi with toll or toll5 in the niche rescues the increased rate of GSC loss induced by piwi depletion.
These genetic results clearly support our conclusion that Toll is necessary for the GSC loss phenotype. However, as the reviewer has commented, we did not overexpress retrotransposons in the niche and examine Toll-GSK3 signaling to show a direct connection between these two determinants. We have now changed our Title from the original one implicating retrotransposons to "Piwi Reduction in the Aged For the original 8 th column, the experiment was not done in parallel with co-KD copia/gyspy. We have now added the results of piwiRNAi-2 performed in parallel with co-KD gypsy experiments. Both of the results (bab1>piwiRNAi-2 alone in Fig. 1d and Fig. 2e This concern is enhanced by the fact that, in Fig.2e, double KD of gypsy and piwi appears to have similar numbers of GSC with piwi-RNAi-2 alone (7th vs. 9th column) indicating that gypsy knockdown has no effect. The difference in two metrices authors present (% of remaining GSC vs. unnormalized number of GSC) hampers claimed the importance of gypsy.

<Response to Reviewer>
We understand the reviewer's concern. As we describe in the main text (pg. 6, line 5), the niche cells we isolated for RNA sequencing contained cap cells and terminal filament (TF) cells, which express very low levels of Piwi but high levels of

<Response to Reviewer>
We thank the reviewer for raising this point and agree with the comment that although VLPs accumulate in piwi-KD niches, those VLPs may not be generated by gypsy. piwi-KD niches treated with 3TC exhibit decreased VLPs when compared to piwi-KD niche without 3TC treatment, suggesting that VLPs may be related to retrotransposon cDNA made via reverse transcription. Notably, VLPs can also be directly generated from retrotransposon transcripts present in the cytoplasm.
Because gypsy transposition was not found in our genomic sequencing results, the VLPs we observed may not be formed by gypsy. Indeed, our new experiments show that co-knockdown of gypsy and piwi cannot reverse VLP accumulation. We have now added this data to Fig. 5a, and shortly described it in the main text, pg. 15, line 6.
We also cannot rule out the possibility that 3TC may have effects on other cellular functions, or somehow act on GSC directly. However, GSC numbers are similar in the gfp-KD flies with or without 3TC treatment, indicating that GSC maintenance is not affected by 3TC treatment for at least the two-week experimental period. However, same treatment in flies with piwi-KD niche reverses the GSC loss phenotype, including effects on GSK3 activity, beta-catenin degradation and E-cadherin expression in the GSC-niche junction that are observed in the piwi-KD niche. Overall our experiments, including fly and human cell models for Alzheimer's disease, suggest retrotransposon derepression is highly relevant to age-related diseases, and 3TC treatment may be beneficial for these diseases.

Alzheimer
's data appears out of place and carries weight heavier than authors' evidence. As mentioned by reviewer 3, AD data seems premature. While it ok to keep these results as supplementary figures they should not be used for strong claims in the Abstract.

<Response to Reviewer>
We thank the reviewer and agree with the comment. We have now completely removed AD information from the Abstract.
6. gypsy mRNA increased 1.2 fold in line 98 but was wrongly colored to have more than 1.25-fold increase in Fig.2c.

<Response to Reviewer>
We thank the reviewer for pointing out this mistake; we have now corrected it from 1.2-fold change to 1.26-fold change in pg. 6, line 13.
7. Constitutive over-expression of Piwi in niche cells slowed, but did not prevent, GSC loss (supp fig 11). It should be described in text, possibly around line 187 or 200, that there is Piwi-independent maintenance of GSC.

<Response to Reviewer>
We thank the reviewer for this comment. We have modified our sentence and added short note that GSC loss during aging also occurs via Piwi-independent mechanisms, pg. 11, line 9 and pg. 12, line 4.