Aging shifts mitochondrial dynamics toward fission to promote germline stem cell loss

Abstract Changes in mitochondrial dynamics (fusion and fission) are known to occur during stem cell differentiation; however, the role of this phenomenon in tissue aging remains unclear. Here, we report that mitochondrial dynamics are shifted toward fission during aging of Drosophila ovarian germline stem cells (GSCs), and this shift contributes to aging‐related GSC loss. We found that as GSCs age, mitochondrial fragmentation and expression of the mitochondrial fission regulator, Dynamin‐related protein (Drp1), are both increased, while mitochondrial membrane potential is reduced. Moreover, preventing mitochondrial fusion in GSCs results in highly fragmented depolarized mitochondria, decreased BMP stemness signaling, impaired fatty acid metabolism, and GSC loss. Conversely, forcing mitochondrial elongation promotes GSC attachment to the niche. Importantly, maintenance of aging GSCs can be enhanced by suppressing Drp1 expression to prevent mitochondrial fission or treating with rapamycin, which is known to promote autophagy via TOR inhibition. Overall, our results show that mitochondrial dynamics are altered during physiological aging, affecting stem cell homeostasis via coordinated changes in stemness signaling, niche contact, and cellular metabolism. Such effects may also be highly relevant to other stem cell types and aging‐induced tissue degeneration.


| INTRODUC TI ON
Stem cells reside in a specialized microenvironment called the niche, which provides both physical contact and stemness factors that ensure and maintain the stem cell fate (Morrison & Spradling, 2008).
While stem cells promote tissue longevity by continually producing differentiated cells, the maintenance and/or function of stem cells often decrease with age, leading to aging-dependent tissue degeneration (Ahmed, Sheng, Wasnik, Baylink, & Lau, 2017;Schultz & Sinclair, 2016). However, the mechanisms by which aging affects stem cells are only partially understood.
Mitochondria frequently undergo coordinated cycles of fusion and fission (known as mitochondrial dynamics) to properly adjust the shape, size, and cellular distribution of the organelle to meet specific cellular requirements (Hoppins, Lackner, & Nunnari, 2007; McQuibban, Lee, Zheng, Juusola, & Freeman, 2006). Fusion produces elongated mitochondria by respectively joining the outer and inner membranes of two mitochondria. The closely related Dynamin-related GTPases, Mitofusin (Mfn) 1 and Mfn2, mediate outer membrane fusion, while optic atrophy (Opa1) is integral for fusion of the inner membrane (Pernas & Scorrano, 2016).
Fzo is exclusively expressed in the testes, while Marf is expressed in the germline and somatic cells (Hwa et al., 2002). Drosophila also has single homologues of Opa1 and Drp1, which have the same names as their mammalian counterparts (Verstreken et al., 2005;Yarosh et al., 2008).
Mitochondrial dynamics are known to influence several mitochondria-dependent biological processes, such as lipid homeostasis, calcium homeostasis, and ATP production (Tilokani, Nagashima, Paupe, & Prudent, 2018). Recent studies have also proposed a role for mitochondrial fusion and fission in regulating stem cell fate (Fu, Liu, & Yin, 2019;Seo, Yoon, & Do, 2018). In one interesting example, murine neural stem cells were shown to exhibit elongated mitochondria, and depletion of Mfn1 or Opa1 impaired their self-renewal (Khacho et al., 2016). Despite tantalizing observations such as these, the overall impact of mitochondrial dynamics in aging stem cells and the mechanisms by which mitochondrial dynamics might affect stem cell function remain unclear.
We used the Drosophila ovary to address the question of how mitochondrial dynamics affect and are affected by stem cell aging, taking advantage of the short lifespan of Drosophila and its amenability to powerful genetic methods. Most importantly, the Drosophila ovary houses well-characterized germline stem cells (GSCs) (Figure 1a) (Kirilly, Spana, Perrimon, Padgett, & Xie, 2005), which gradually escape the niche and become differentiated during aging (Kao et al., 2015). A Drosophila ovary contains 16-20 egg-producing functional units, which are called ovarioles (Spradling, 1993). The germarium is the anterior-most structure of the ovariole, and it houses two to three GSCs at its anterior tip. The terminal filament, cap cells, and anterior escort cells are also located in the anterior tip of the germarium and form the GSC niche (Losick, Morris, Fox, & Spradling, 2011). GSCs directly contact niche cap cells (the major niche component) (Song & Xie, 2002), and each of GSC contains a fusome, an organelle with a membranous-like structure that is juxtaposed to the GSCcap cell interface (Xie & Spradling, 2000). As a single asymmetric GSC division gives rise to a cystoblast (CB), the fusome changes F I G U R E 1 Aged GSCs exhibit increased numbers of fragmented mitochondria. (a) An illustration of the anterior part of the Drosophila germarium. Terminal filament (TF) and cap cells form the GSC niche to house GSCs. Each GSC carries a fusome. The cystoblast (CB), the immediate daughter cell of the GSC, undergoes four rounds of incomplete division to form a 16-cell cyst. Each cyst carries a branched fusome that interconnects germ cells within the cyst. (b) Fusome morphology changes according to the GSC cell cycle phase. During S phase, GSCs display a "plug," "elongated," or "bar" fusome morphology, as a nascent fusome (or plug) is assembled and then fused to the original fusome, thereby connecting the GSC and the cystoblast. During early G2, GSCs exhibit "exclamation point" fusome morphology, as the connection between GSCs and the cystoblast is severed. During late G2 and M, GSC fusomes display a "round" fusome. (c and d) One-(c) and 8-week-old germaria (d) with LamC (red, TF and cap cell nuclear envelopes), 1B1 (red, fusomes), Vasa (blue, germ cells), and ATP5ase (gray, mitochondria). Inserts are higher magnification of GSCs marked by yellow asterisks in E and F, with ATP5ase shown in green. (c′ and d′) are same GSCs shown in the inserts, but with less layers and the surface model of mitochondria from Imaris shown. Mitochondria (mito) forming networks are shown in green; fragmented mitochondria (fragm. mito) are shown in yellow. Asterisks indicate GSC(s) in the germarium; dashed circles outline GSCs. Yellow dashed lines outline the anterior edge of the germarium. Scale bars in c and d are 5 μm, in the insert of c and d are 2 μm, and in c′ and d′ are 1 μm. (e) Percentage (%) of mitochondria with indicated volume in 1-and 8-week-old GSCs. (e′) Number (no) of fragmented mitochondria in 1-and 8-week-old GSCs at S or G2/M phases. (e″) Percentage of mitochondrial content per GSC in 1-and 8-week-day-old GSCs. Numbers of analyzed GSCs are shown above each bar. Error bars, SEM. **p < 0.01; ***p < 0.001. (f and g) Representative electron micrographs of the anterior regions of 1-(f) and 8-week (W)-old germaria (g). f′ and g′ are enlarged views from the areas indicated by squares in f and g. N, Nucleus. CpC, cap cells. Scale bars in e and g are 2 μm, and bars in f′ and g′ are 0.5 μm. (h) The area (μm 2 ) and the width to height (W/H) ratio of individual mitochondria in 1-(red) and 8-week-old GSCs (blue). Mitochondria distributed in blue, green, and pink areas are elongated, medium, and fragmented mitochondria, respectively. Solid and dashed lines represent the mean of W/H ratio and area of mitochondria in 1-week-old GSCs, respectively. n, number of analyzed mitochondria. Percentages of mitochondria in each group (1-versus 8-week-old GSCs) showed significant differences (p < 0.001, Chi-squared test). Representative germaria are shown in 3D-reconstructed images; genotype of flies is yw.
morphology according to the stage of the cell cycle ( Figure 1b).
During G2/M phase, the GSC fusome is round. Then, at G1 and S phases, it grows and fuses with a newly formed fusome destined for the daughter CB, generating an elongated fusome. This elongated fusome is pinched off when the GSC and CB begin to separate during early G2 phase, leading it to regain its round shape in the GSC until the end of M phase (de Cuevas & Spradling, 1998;Kao et al., 2015). After M phase, the daughter CB undergoes four rounds of incomplete division to form a 16-cell cyst; each germ cell within the cyst is interconnected by a branched fusome (Spradling, 1993). Next, the 16-cell cyst is surrounded by a layer of follicle cells, and the whole structure buds off from the germarium, finally developing into a mature egg (Spradling, 1993). Mitochondria are generally found in a big cluster located near the fusome in GSCs.
In contrast, highly fragmented mitochondria are located far from the fusome in 4-and 8-cell cysts, while elongated mitochondria are observed in close proximity to the fusome in 16-cell cysts (see Figure 4b) (Cox & Spradling, 2003).
In this study, we used fluorescence and transmission electron microscopy (TEM) to show that fragmented mitochondria are increased in aged GSCs in fixed ovarian tissues. Our live-imaging data further show that mitochondrial fission is increased in aged GSCs, suggesting a source for the accumulated fragmented mitochondria. GSCs with mitochondrial fragmentation forced by a mutation of marf mimic aged GSCs, which divide slowly, exhibit low Dpp (BMP orthologue) stemness signaling and have a tendency to leave the niche and differentiate (Kao et al., 2015). Furthermore, the fragmented mitochondria exhibit reduced membrane potential

| Aging increases fragmented mitochondria in GSCs
To understand whether aging affects mitochondrial morphology, we first labeled germaria with antibodies for LamC, 1B1, and Vasa to delineate the various cell types. We also labeled ATP synthase 5 α subunit (ATP5ase) in the mitochondria of young (1-week-old) and aged ovaries (8-week-old) to analyze mitochondrial location and size. In young GSCs (indicated by an asterisk in Figure 1c), mitochondria formed a big cluster near the fusome, an observation that is in agreement with a previous report (Cox & Spradling, 2003); this location was not changed in aged GSCs (indicated by an asterisk in Figure 1d). The volume of individual mitochondria in GSCs was about 0.4 ± 0.14 μm 3 (n = 20 GSCs) on average, but the size ranged up to 48 μm 3 (data not shown). Notably, the sizes of individual mitochondria and numbers of mitochondria in a cluster could not be accurately assessed due to limitations in resolution.
However, we could find that mitochondria with sizes smaller than 0.05 μm 3 were increased in aged GSCs compared to young GSCs ( Figure 1e,e′, and see yellow signals in Figure 1c′,d′); we defined these small mitochondria as fragmented mitochondria. This difference was not correlated with the phase of the GSC cell cycle ( Figure 1e′). In addition, total mitochondrial content (ratio of total mitochondrial volume to the GSC volume) was significantly lower in aged GSCs than in young GSCs ( Figure 1e″). The immediate daughter cells of GSCs (CBs) also displayed a wide range of mitochondrial sizes, as well as age-associated increases in fragmented mitochondria and decreases in mitochondrial content ( Figure  we reasoned that the stochastic nature of mitochondrial orientation should still allow us to see a difference between experimental groups if mitochondria were more fragmented in aged GSCs. Indeed, averages of mitochondrial area and length were larger in young GSCs (area: 0.11 ± 0.11 μm 2 ; W/H ratio: 2.34 ± 1.9, 177 mitochondria from 6 GSCs), as compared to aged GSCs (area: 0.06 ± 0.04 μm 2 , p < 0.05; W/H ratio: 1.92 ± 1.4, p < 0.001, 129 mitochondria from 5 GSCs) (Figure 1h). For further analysis, mitochondria were classified into three groups according to the mean area and W/H (of mitochondria analyzed in young GSCs: mitochondria with areas bigger than 0.11 μm 2 and W/H ratio larger than 2.34 were considered "elongated mitochondria," mitochondria with areas bigger than 0.11 μm 2 and W/H ratio smaller than 2.34 were counted as "medium mitochondria," and mitochondria with areas smaller than 0.11 μm 2 were called "fragmented mitochondria" (Figure 1h). The respective percentages of elongated, medium, and fragmented mitochondria in young GSCs were 19.7%, 15.3%, and 65% versus 5.4%, 0.8%, and 93.8% (p < 0.001) in aged GSCs ( Figure 1h). Thus, we found that aged GSCs exhibited high levels of fragmented mitochondria, consistent with our conclusion from fluorescence microcopy experiments. Because the mitochondria in aged GSCs were smaller than those in young cells, we suspected that mitochondria in aged cells may either undergo less fusion or more fission to yield fragmented mitochondria.

| Aged GSCs display a preference for mitochondrial fission
To distinguish between the two possibilities described above, we first labeled mitochondria in live young (1-week-old) and aged ovaries (7-week-old) by MitoTracker, a fluorescent dye ( Figure S1D and E). We found that mitochondria in aged GSCs were more fragmented than those in young GSCs (insets in Figure 1c,d); however, we sometimes could not distinguish if analyzed mitochondria were from GSCs or somatic cells. We, therefore, expressed mito-gfp [a marker for mitochondria (Cox & Spradling, 2003)] in the germline using nos-GAL4. We made live recordings of nos>mito-gfp GSCs for 10 min (300 time points with intervals of about 2 s to generate 300 stacks, each stack containing 100 slices along the z-axis) using lattice light-sheet microscopy (Movie S1 and Movie S2), which allows us to capture images with high resolution and fast acquisition speed . Hoechst staining was used to identify cap cells according to their small oval-shaped nuclei and anterior-most location in the germaria; GSCs were identified by their direct contact with cap cells (left panel in Figure 2a). We found that over the 10-min

| Preventing mitochondrial fusion decreases GSC division and maintenance
GSC division and maintenance are reduced during aging (Kao et al., 2015;Pan et al., 2007;Zhao, Xuan, Li, & Xi, 2008). To understand whether mitochondrial fission is involved in this process, we introduced a mutation in the fusion regulator, of control and mutant cystoblasts and cysts in marf germaria containing at least one control and one mutant GSC. The relative numbers of wild-type and mutant CBs or cysts were unaffected by early germline death ( Figure S4), and the numbers of progeny derived from control GSCs were approximately equal to those without GFP in mock mosaic germaria (relative division rate equal to approximately 1.0) at 1, 2, and 3 weeks after clone induction (ACI) (Figure 3f). In contrast, division was significantly reduced in GSCs homozygous for marf E , a hypomorphic allele, and for marf B , a null allele (Yarosh et al., 2008) (Figure 3f). We next asked whether diminished fusion reduces GSC maintenance by counting the number of germaria carrying marf mutant GSCs over time ( Figure 3g and Table S1). At 3 weeks ACI, about 93 ± 4% of FRT19A control germaria (n = 336) retained at least one GFP-negative control GSC from the first week, indicating that up to ~7% of GSCs may be naturally turned over. In 3-week ACI mutant germaria, only 68 ± 15% (marf E , n = 251) and 38 ± 9% (marf B , n = 310) of mutant GSCs were maintained. Since we did not detect any apoptotic marf mutant GSCs ( Figure S4), we suspect that marf mutant GSCs leave the niche and undergo differentiation. These results show that mitochondrial fragmentation is detrimental to GSC division and maintenance.

| Forcing mitochondrial elongation promotes GSC competitiveness for niche occupancy
We also examined the impact of promoting elongated mitochondria on GSC homeostasis by generating GSCs with a mutation of mitochondrial fission regulator, drp1 (indicated by the absence of GFP). GSCs homozygous for drp1 1 or drp1 2 null alleles (Yarosh et al., 2008) displayed large mitochondrial clusters that occupied a major portion of the cell ( Figure 3h,h′,h″,i,i′). However, drp1 mutant GSCs exhibited comparable division and maintenance rates compared to control GSCs (Figure 3g,j).
These results indicate that preventing mitochondrial fission neither decreases GSC division nor maintenance. Interestingly, we found that the proportions of mutant germaria carrying at least one GFP-positive GSC (partial GSC clone) decreased from 75% to 34% (drp1 2 ) and 70% to 14% (drp1 1 ), while the proportion of germaria in which all GSCs were mutant (full GSC clone) increased from 25% to 66% (drp1 2 , 41% increase) and 30% to 86% (drp1 1 , 56% increase) by 3-week ACI (Figure 4k). In FRT40A mock mosaic germaria, only a 25% increase was observed, due to the natural loss of neighboring GFP-positive GSCs (Figure 3k). These results indicate that GSCs with drp1 mutation (shifting mitochondrial dynamics balance toward fusion) tend to push away neighboring control GSCs to dominate niche occupancy.
However, pMad expression was significantly reduced in marf E or marf B mutant GSCs compared to neighboring GFP-positive control GSCs (Figure 4b

| Impaired mitochondrial fusion reduces egg laying
To determine whether mitochondrial dynamics affect egg production, we also knocked down marf and drp1 specifically in the adult germline. To this end, we used nos-GAL4 and cultured flies at 18°C before eclosion then switched the flies to 29°C after eclosion. One- week-old nos>marf RNAi were smaller than nos>drp1 RNAi (control) and nos>drp1 RNAi ovaries ( Figure S5A-C). Compared to control GSCs, nos>marf RNAi GSCs had highly fragmented mitochondria, while mitochondria were elongated in nos>drp1 RNAi GSCs ( Figure S5D-E), indicating the RNAi lines we used could recapitulate the mutant phenotypes. Consistently, nos>marf RNAi GSCs were more quickly lost from the germaria with age than the controls ( Figure S5G). Although our clonal analysis showed that Drp1 is not required for GSC maintenance (see Figure 3), 3-week-old nos>drp1 RNAi GSCs were lost faster than control GSCs (Figure 5g), possibly because varied expression of nos-GAL4 among aged GSCs in the niche created a competitive environment (Tseng et al., 2014). As a consequence, nos>marf RNAi ovaries had a dramatic reduction of egg production compared to nos>drp1 RNAi and control ovaries ( Figure 5h). Thus, oogenesis appears to be disturbed when mitochondria are fragmented in the germline.

| Fragmented mitochondria in GSCs display low membrane potential and disrupted lipid homeostasis
The mitochondrial membrane potential results from a proton gradient across the inner membrane, which is generated by oxidative phosphorylation complexes and is thought to reflect mitochondrial functional output. To know whether altered mitochondrial dynamics affects mitochondrial membrane potential, we used the probe, TMRE (Perry, Norman, Barbieri, Brown, & Gelbard, 2011), in isolated GSCs and analyzed TMRE signals by flow cytometry. GSCs were isolated from the germaria with nos-GAL4 driven gfp RNAi ,marf RNAi or drp1 RNAi along with vasa-gfp (for GSC isolation) and a mutation of bag of marbles (bam, encodes a master differentiation factor) to increase GSC number (Kao et al., 2015). Isolated 2-week-old gfp RNAi -knockdown (KD) GSCs were treated with FCCP, a potent mitochondrial oxidative phosphorylation uncoupler, which served as a positive control (Heytler, 1979). These FCCP-treated GSCs showed dramatically reduced TMRE signals (Figure 5a). Compared to nos>gfp RNAi control GSCs, drp1-KD GSCs exhibited similar levels of TMRE signal, while 2-week-old and 4-week-old marf-KD GSCs displayed 19% and 71% reductions of TMRE signal, respectively (Figure 6a). This result suggests that mitochondria with impaired fusion are less functional. Furthermore, cellular ROS levels, as detected by DHE (Benov, Sztejnberg, & Fridovich, 1998), in 2-week-old marf-KD GSCs were reduced by 16% (Figure 5b), suggesting that oxidative phosphorylation activity in fusion-defective mitochondria is attenuated. To our surprise, like gfp-KD GSCs treated with paraquat to increase ROS (Ali, Jain, Abdulla, & Athar, 1996), drp1-KD GSCs displayed slightly increased ROS levels (Figure 5b), indicating that oxidative phosphorylation activity is promoted in fission-defective mitochondria.
Similar reductions of membrane potential and cellular ROS were also observed in aged GSCs (8-week-old), as compared to young GSCs (1-week-old) (Figure 5a′,b′), in line with our observation that aged GSCs carry more fragmented mitochondria.
To further explore the dysfunction of fragmented mitochondria, Thus, the nature of mitochondrial dysfunction caused by excessive fragmentation during aging appears to be complex and multifaceted, and it is unlikely that lipid accumulation alone can explain the loss of GSCs during aging.

| Reduced fragmented mitochondria promotes maintenance of aged GSCs
Because we saw that the balance of mitochondrial dynamics in aged GSCs shifts toward fission, we further asked whether this switch is associated with changes in Marf or Drp1 expression levels. We found that according to expression of a genomic construct, marf-gfp (Zhang, Mishra, Hay, Chan, & Guo, 2017), Marf levels in young and aged GSCs were comparable (Figure 6a,b,e). However, Drp1 expression was significantly increased in aged GSCs and their progeny, as compared to young GSCs (Figure 6c,d,f), suggesting a role for increased Drp1 in aging-induced mitochondrial fragmentation.
To test this idea, we used the flip-out system, in which a transcriptional stop sequence flanked by two FRT sites was inserted between the nanos (nos) promoter and GAL4 (nos>STOP>GAL4; Figure 6g) (Ma et al., 2014). We used heat shock to express drp1 RNAi or mCD8-gfp (control) expression along with a GFP reporter specifically in germ cells of 4-week-old females. Strikingly, drp1-KD mosaic ovaries of 8-week-old flies looked younger than mCD8-gfp-expressing ovaries in flies at the same age (Figure 6h,i), as they carried many vitellogenic egg chambers (arrow heads in Figure 6i). In addition, consistent to our previous results (see Figure 3h,i), mitochondria in drp1-KD GSCs formed a big cluster as compared to control GSCs of 8-week-old flies ( Figure S7), indicating a disruption of mitochondrial fission.
F I G U R E 3 Disrupting mitochondrial fusion decreases GSC division and maintenance while disrupting mitochondrial fission promotes GSC-niche occupancy. (a) Mitotic recombination was used to generate GSC mutants for marf or drp1. Females carrying a wild-type (wt, +) allele linked to a marker gene (gfp) in trans with a mutant (mut) allele were generated. FLP-mediated recombination between FRT sites during mitotic division generates a homozygous mutant cell, identifiable by the absence of marker expression. (b-e, h and i) FRT19A control (ctrl) (b), marf E (c), marf B mutant mosaic germaria (d and e), drp1 2 (h) and drp1 1 mutant mosaic germaria (i) with GFP (green, wt cells), 1B1 (red, fusomes), and ATP5ase (gray, mitochondria) at 1 (b-d), 2 (e and f), and 3 weeks (w) after clone induction (ACI) (i). Wt GSCs are indicated by asterisks; GFP-negative GSCs and their daughter cells are outlined by yellow and white dashed lines, respectively. Inserts in b and c show GFP-negative (−) GSCs, in d and h show GFP-positive (+) GSCs at different focal planes. b′ and b″, c′, d′, d″, h′ and h″ show higher magnifications of GFP + or GFP − GSCs with 1B1 and ATP5ase staining. In the germarium (d), GFP − GSCs only produce one GFP − GSC daughter cell, indicating a low rate of GSC division; the germarium (e) carries GFP − germ cells but not GFP − GSCs, indicating the loss of GFP − GSCs. Germaria carrying GSCs that are not all GFP-negative are referred to as partial GSC clones, while germaria carrying GSCs that are all GFP-negative are referred to as full GSC clones. Scale bar, 10 µm. (f and j) GSC relative division rates (ratio of GFP-negative to GFPpositive GSC progeny) in mosaic germaria. The number of GSCs analyzed is shown above each bar. (g) Relative percentage of GSC clones (as a proportion of total GSCs) at 1, 2, and 3 W ACI. (k) Relative percentage of germaria with partial GSC clones versus germaria with full GSC clones at 1, 2, and 3 W ACI. Numbers of germaria analyzed are shown above each bar. *p < 0.05; **p < 0.01; ***p < 0.001. Error bars, mean ± SEM.
Five-and 8-week-old mCD8-gfp-expressing mosaic germaria carried 52% and 38% GFP-positive GSCs, respectively (Figure 6j,k,m), indicating that 14% of GSCs were naturally lost. However, 5-and 8-week-old drp1-KD mosaic germaria, respectively, carried 49% and 61% GFP-positive GSCs (Figure 6l,m), showing a net increase of drp1-KD GSCs. In fact, 19% of 8-week-old drp1-KD mosaic germaria carried at least one GFP-positive GSC (partial GSC clone) and 81% carried only GFP-positive GSCs (full GSC clone) (Figure 6m). In contrast, 59% of 8-week-old mCD8gfp-expressing mosaic germaria were partial clones, and 41% were full GSC clones (Figure 6m). This  numbers of GSCs (Figure 6n,o,r), while 9-week-old flies treated with rapamycin showed significantly higher GSC numbers, compared to flies at the same age without rapamycin feeding (Figure 6p,q,r). These results were reminiscent of previous studies that showed autophagy promotes stem cell fate (Boya, Codogno, & Rodriguez-Muela, 2018;Zhao, Fortier, & Baehrecke, 2018), and aging slows autophagy activity (Cuervo, 2008). Together, these results suggest that preventing production of fragmented mitochondria or clearing dysfunctional mitochondria during aging may promote GSC maintenance. The interplay between aging and mitochondrial dynamics has been previously studied in various laboratory models. For example, in Drosophila, promoting Drp1 expression or mitophagy in midlife flies extends lifespan, with mitochondria in the muscles of those flies appearing more fragmented compared to controls (Aparicio, Rana, & Walker, 2019;Rana et al., 2017). In a C. elegans study (Morsci, Hall, Driscoll, & Sheng, 2016), it was revealed that mitochondria size and density in neurons are increased in midlife but decreased to early-life levels in aged worms, while promotion of mitochondrial fusion by AMPK or dietary restriction extends lifespan (Weir et al., 2017). Similarly, in fungal models, reducing mitochondrial fission also extends lifespan (Scheckhuber et al., 2007). Thus, manipulation of mitochondrial dynamics appears to be globally related to aging in tissues across many organisms.

However, it is not clear whether all tissues show similar shifts
in preference of mitochondrial dynamics during aging, and it is also unknown whether age-related phasic patterns of mitochondrial dynamics happen in all cell types and in higher organisms.
In mammals, this issue appears to be somewhat complicated. In studies on skeletal muscle, both fragmented mitochondria and large, less circular mitochondria have been reported in aged tissues (Beregi, Regius, Huttl, & Gobl, 1988;Iqbal, Ostojic, Singh, Joseph, & Hood, 2013;Leduc-Gaudet et al., 2015). Furthermore, reducing mitochondrial fission partially reduces neurodegeneration in mouse neuronal disease models (Manczak, Kandimalla, Fry, Sesaki, & Reddy, 2016), while promoting mitochondrial fission reverses certain degenerative phenotypes in Drosophila parkin and park mutants (Deng, Dodson, Huang, & Guo, 2008;Yang et al., 2008). These species and tissue-dependent results illustrate the importance of studies that experimentally delineate the mitochondrial profile during aging in various cellular contexts, which may aid the discovery of factors that influence long-term mitochondrial maintenance.

| Age-dependent GSC loss is not due to defective fatty acid metabolism or increased ROS levels
An important function of mitochondria is to carry out long-chain fatty acid oxidation for energy production. Forcing mitochondrial According to the free radical theory of aging, ROS induces oxidative damage and leads to cellular dysfunction and aging (Barja, 2013). In support of this idea, damaged mitochondria and ROS levels are known to be increased in aged fly and mammalian brains (Chakrabarti et al., 2011;Scialo et al., 2016). Surprisingly, in aged GSCs from female Drosophila, there is an approximately 14% reduction in ROS levels compared to young GSCs. Nevertheless, constitutive overexpression of superoxide dismutase (SOD; helps remove ROS) delays GSC loss during aging (Pan et al., 2007). These observations might be reconciled if SOD expression keeps ROS levels low throughout the entire lifespan of the GSC. Further analysis of ROS levels in aged GSCs from male flies or GSCs in other species will help to determine whether and how the free radical theory of aging may be applicable in a cell type-specific manner.  (Khacho et al., 2016).

| Mitochondrial dynamics control germ cell differentiation
However, the role of mitochondrial dynamics in germ cell differentiation is less clear.
In Drosophila female GSCs, mitochondria are elongated and form clusters near the fusome (see Figure 1); similar mitochondrial morphology is also observed in the immediate daughter cells, CBs. Mitochondria are highly fragmented in 4-and 8-cell cysts (see Figure 4), suggesting that fission is preferred during differentiation; meanwhile, mitochondria are elongated again in 16-cell cysts (see Figure 4), which are undergoing meiosis (Lin & Spradling, 1993).
These observations are in agreement with a previous study (Cox & Spradling, 2003). In this study, we show that forcing mitochondria fragmentation in a marf mutant promotes GSC differentiation, and aging-associated mitochondrial fragmentation contributes to age-dependent GSC loss, at least in part via the upregulation of Drp1.
Although Marf is also required for Drosophila male GSC maintenance

| Drosophila strains and culture
Drosophila stocks were maintained at 22-25°C on standard medium, unless otherwise indicated. For aging experiments, flies were fed with normal diet plus a paste of wet yeast and food was changed daily. yw was used as a wild-type control. The following fly strains were used in this study: Hypomorphic marf E and null bam △86 , bam 1 , marf B , drp1 1 , and drp1 2 alleles have been described previously were described previously (Deng et al., 2016;Smith et al., 2019) and were tested again in this study ( Figure S4). The nos-Vp16-GAL4 line was used to drive RNAi expression in the germline line (Doren, Williamson, & Lehmann, 1998;Tseng et al., 2014

| Immunostaining and fluorescence microscopy
Ovaries were dissected, fixed, and immunostained as described (Yang et al., 2013). In brief, ovaries were dissected in pre-warmed Grace's insect medium (GIM, Lonza) and fixed with 5.3% paraformaldehyde/GIM for 13 min with gentle agitation at room temperature. Ovaries were washed in PBST (0.1% Triton X-100 in 1X PBS) for 20 min three times and teased apart in PBST before incubating with blocking solution (5% bovine serum albumin (BSA) and 0.05% normal goat serum in PBST) for 3 hr at room temperature or 4°C overnight. Ovaries were incubated with primary antibodies (diluted in blocking solution) for 3 hr at room temperature or 4°C overnight, followed by three or four 30-min washes with PBST. Next, ovaries were incubated with secondary antibodies (diluted in blocking solution) for 3 hr at room temperature or 4°C overnight, followed by three or four 30-min washes with PBST.
Apoptag ® Fluorescein In Situ Apoptosis Detection Kit (cat#S7110, Merck) was used to detect apoptotic cells following the instruction manual with slight modifications (Tseng et al., 2014). In brief, ovaries were fixed, teased apart, and incubated with 300 µl of equilibration buffer for 5 min on rotator twice at room tempera- (g) A flip-out system was used for nos-gal4-driven drp1 knockdown in aged GSCs; knockdowns are identified by the presence of GFP expression. In females carrying a nos promoter-driven FRT-flanked flip-in GAL4/VP16 construct (nos > STOP > GAL4), GAL4 is not expressed, preventing expression of UAS-gfp and UAS-drp1 RNAi . GAL4 expression is turned on by removing the stop cassette through Flippase-mediated recombination, which in turn activates expression of UAS-gfp and UAS-drp1 RNAi . (g and i) 8-week-old nos>gfp &mCD8gfp (g) and nos>gfp & drp1 RNAi mosaic ovaries (i) were heat-shocked at 4 weeks old for 3 days to activate nos-GAL4. Red arrows point to previtellogenic egg chambers. (j-l) 5-week-old nos>gfp &mCD8gfp (i), nos>gfp and drp1 RNAi (j) and 8-week-old nos>gfp and drp1 RNAi mosaic germaria (l) heat-shocked at 4 weeks old for 3 days with LamC (red), 1B1 (red), and GFP (green, cloned cells) labeling. (m) Percentage (%) of 5-or 8-week-old germaria (Left y-axis) carrying partial or full GSC clones (containing 1 or ≥2 GSCs) from flies with indicated genotypes heat-shocked at 4 weeks old. Right y-axis shows percentage of GFP-positive GSCs in flies with indicated genotypes. (n-q) Two-and 9-week-old germaria with or without rapamycin treatment for 1 week labeled with LamC (red) and 1B1 (red). (r) Percentage of germaria with indicated GSC number of 2-and 9-week-old flies fed with or without rapamycin, beginning at 1 and 8 weeks after eclosion, respectively. Solid lines outline GSCs and dashed lines outline GSC progeny. Numbers of analyzed GSCs are shown above each bar. *p < 0.05; **p < 0.01; ***p < 0.001. Error bars, mean ± SEM. Scale bars in a, c, j and n are 10 μm; bar in h is 0.5 mm.
Images of fixed ovaries were obtained using a Zeiss LSM 700 Laser Scanning confocal microscopes.

| BODIPY 493/503 staining
Ovaries were dissected, and immunostaining was performed as described above. After immunostaining, ovaries were incubated with 50 μM BODIPI 493/503 (D3922, Thermo Fisher) in 0.1% PBST in the dark at RT for 20 min. Samples were washed 3 times with 0.1% PBST, stained with 0.5 μg/ml DAPI for 5 min, and mounted in mounting solution, as described above.

| L-Carnitine and rapamycin treatment
L-carnitine (C0283-5G, Thermo Fisher, final concentration of 25 mg/ml) or rapamycin (R0395, Sigma-Aldrich, final concentration of 200 μM) were added to wet yeast (yeast to water ratio was 1.8 g:1 ml). One-week-old flies were fed wet yeast with or without L-carnitine (or rapamycin) in vials containing molasses food for 1 week. Food was changed every day until dissection. For mitochondria analysis, images of germaria labeled for ATP5ase

| Image analysis
were deconvoluted using MetaMorph (Molecular Devices) and analyzed by Imaris.

| Live imaging and image processing
Ovaries of 1-and 8-week-old nos>mito-gfp flies were dissected in pre-warmed GIM and stained with Hoechst (5 μg/ml) for 10 min at room temperature. Anterior ovarioles were dissected from Hoechststained ovaries, amounted with CellTak (Corning) on 5-mm-diameter pre-cleaned cover slips (Warner Instruments, 64-0700), immersed in a PBS-filled chamber, and imaged with lattice light-sheet microscopy . Images were scanned with 100 z-sections every ~2 s for 300 frames with a detection objective (Nikon, CFI Apo LWD 25XW, 1.1 NA, 2 mm WD) at a speed of 10 ms exposure per plane.
Raw images taken from the light-sheet microscope were deconvoluted by Amira (version 6.4, Thermo Fisher) with the PSF kernel acquired under the same optical condition. The deconvoluted images taken at the first 10 time points were stacked and corrected for bleaching to decrease background using ImageJ (Fiji, NIH). For each stack, the anterior of the germarium was cropped and subjected to Amira (Thermo Fisher) segmentation in order to define mitochondria by "hysteresis thresholding." Each stack with defined mitochondria was analyzed with Imaris (Bitplane) to track mitochondrial dynamics within time points 1-4, 4-7, and 7-10. Mitochondrial fusion/fission was tracked using the connected components model (Dillencourt, Samet, & Tamminen, 1992).

| Transmission electron microscopy (TEM) of adult germaria
Ovaries of were dissected in 2.5% glutaraldehyde/2% paraformalde- (Gatan) at an accelerating voltage of 120 kV. Finally, images were subjected to analysis of mitochondria using ImageJ and Amira software or movie production by Tomographic 3D Image Reconstruction.
Seven to twelve pairs of ovaries from each genotype were dissected in pre-warmed GIM plus 10% FBS (GIM-FBS) and were subsequently incubated with 0.45% Trypsin (Solution 10X, cat# 9002077, Sigma-Aldrich) and 2.5 mg/ml collagenase (cat#17018-029, Gibco) on a rotator at 25°C for 25 min with vigorous shaking; samples were vortexed every 5 min. Digested ovaries were filtered through a 40-µm nylon mesh and then centrifuged at 1,000 g for 7 min to harvest the cell pellet. The pellets were resuspended in 500 µl GIM-FBS containing 10 nM membrane potential probe TMRE (cat#T669, Thermo Fisher)/ or 30 µM ROS probe DHE (cat#D11347, Invitrogen) and 0.5 µg/ml of DAPI with vigorous shaking at RT for 10 min in a dark chamber. For a positive control of membrane potential measurement, cells dissociated from bam △86 vasa-gfp/nos-GAL4bam 1 ovaries were co-treated with 10 nM TMRE and 10 µM FCCP (to depolarize the mitochondrial membrane, C2920, Sigma-Aldrich) for 10 min under the same conditions as described above. For a positive control of ROS measurement, cells dissociated from bam △86 vasa-gfp/nos-GAL4bam 1 ovaries were treated with 30 µM DHE and 100 µM paraquat (to induce cellular ROS, cat#3752782, Sigma-Aldrich) for 10 min under the same conditions as described above.

ACK N OWLED G EM ENTS
We thank A. C. Spradling, H. J. Bellen, H. Imamura, H.-C. Chen, the Bloomington Drosophila Stock Center, the VDRC Stock Center and the DSHB for Drosophila stocks and antibodies. We also thank the Taiwan fly core for ordering fly lines and reagents, core facilities in the Institute of Molecular Biology, and the Institute of Cellular and Organismic Biology, Academia Sinica for assistance with EM and image analysis, Chen-Hui Chen and Yi-Ching Lee for valuable comments, and Marcus Calkins for English editing. This work was supported by two thematic grants of Academia Sinica.

CO N FLI C T O F I NTE R E S T
The authors declare that there is no conflict of interest regarding the publication of this article.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available upon request.