Flies were cultured following standard procedures at 25°C except for RNAi experiments at 29°C. All strains were obtained from the Bloomington Drosophila Stock Center (BDSC), National Institute of Genetics Stock Center (NIG), and Vienna Drosophila RNAi Center (VDRC). Stocks used were: UAS-mito-GFP (8443), UAS-Lifeact-GFP (35544), UAS-RhoA V14 (7330), UAS-ROCK.CAT (6667), UAS-Mfn RNAi (55189 and 31157), UAS-Mfn (67157), UAS-Drp1 (51647), Sqh-GFP (57144), Drp1 ¹ (24885) and Drp1 ² (24899) from BDSC, UAS-Drp1 RNAi (3210R-3 and 3210R-4) from NIG, UAS-Mfn RNAi (40478) from VDRC. Mutant FRT clones were induced using hs-FLP. Flies were heat shocked for 1 h per day at 37°C for 3 days before eclosion. The newly eclosed females were raised on fresh food with yeast paste for 3 days prior to dissection.

Fly ovaries were dissected in phosphate-buffered saline (PBS), and then fixed in 4% formaldehyde for 10 min. After washes in PBST (PBS with 0.3% Triton X-100), ovaries were incubated with blocking solution (PBST containing 10% goat serum) for 30 min and then stained overnight at 4°C. The primary antibodies used were mouse anti-ATP5A (ab14748). After overnight staining, ovaries were incubated with secondary antibodies (1:200, Jackson) for 2 h at room temperature. For S2 cell immunostaining, 48 h after transfection, cells were harvested and washed with PBS. Cells were fixed in 4% formaldehyde in PBS buffer for 20 min at room temperature, treated with PBST for 20 min, and washed with PBS for 20 min three times. Mitotracker was used to label mitochondria. F-actin was labeled by Rhodamine-phalloidin (P1951, Sigma). Image of cells were acquired by Zeiss 880 confocal microscope (with Airyscan).

Egg chambers were dissected from ovaries for live imaging as described previously (Bianco et al., 2007). Egg chambers were dissected in live imaging medium and then transferred to an 8 chamber (No.155411, Thermo Fisher, United States), with each chamber containing 200 μl medium. For drug treatment, egg chambers were continuously imaged for 20–30 min by Zeiss 880 confocal microscope, before adding Oligomycin (5, 50 and 500 nM) or DMSO to the media and further live imaging.

For S2 cells, mitochondria were stained with Mitotracker for 30 min to detect their morphologies. Mitochondrial morphology was analyzed with Mitochondria Analyzer (Chaudhry et al., 2020), a plugin of Image J software widely used for characterizing mitochondrial morphology in cultured cells (Mao et al., 2021; Stifel et al., 2022).

S2 cells were maintained at 25°C in Schneider’s Drosophila medium (S9895, Sigma) supplemented with 10% FBS (F0718, Gibco), 100 U/ml penicillin (Life Technologies), and 100 ug/ml streptomycin. Drosophila S2 cells were transfected using PEI (Polyethylenimine, 1 μg/μl) with final DNA concentration of 1 μg/ml and final PEI concentration of 3 μg/ml in total culture media. Drp1-HA, Mfn-HA, RhoA-HA, ROCK-HA and ROCK-Fg were subcloned into pUAST-3xFg or pUAST-3xHA vectors and used for transfection of S2 cells.

The western blot and Immunoprecipitation experiments were performed as previously described (Kang et al., 2018; Qu et al., 2019). Briefly, cell lysates were incubated with anti-HA antibody (Santa Cruz) overnight at 4°C with end-over-end rotation. The samples were further incubated with protein A/G agarose beads (Santa Cruz) for 1 h and rinsed twice with IP wash buffer. The beads were resuspended in ×2 loading buffer for 10 min at 95°C before loading onto SDS-PAGE gels and immunoblotting. The following antibodies were used for immunoprecipitation and western blot: mouse anti-HA (1:2000); mouse anti-Fg (1:3000); mouse anti-Actin (1:5000) and goat anti-mouse HRP (1:10000). The blot was visualized using a chemiluminescent detection kit.

The mitochondria were isolated using the Mitochondria Isolation Kit (Best Bio, Shanghai, China) for ovaries and S2 cells, according to the manufacturer’s protocol. ATP concentrations were measured with enhanced ATP assay kit (Beyotime, Shanghai, China) according to the manufacturer’s protocol. Luminescence was measured on an Infinite M200Pro multifunction reader. The relative ATP levels were calculated by dividing the luminescence by total protein concentration, which was determined by Bradford method.

For TEM sample preparation, adult ovaries were dissected in PBS, and then fixed in 2.5% Glutaraldehyde for more than 48 h at 4°C. Samples were then washed three times with phosphate buffer (PB, 0.1M Na₂HPO₄ and 0.1M NaH₂PO₄, PH7.4), and post-fixed with 1% osmium tetroxide for about 2 h at 4°C. Samples were dehydrated with 2% uranyl acetate buffer overnight at 4°C, and then gradually dehydrated through graded ethanols (30%, 50%, 70%, 90%, and 100%, respectively). Samples were finally dehydrated in acetone for two times with 15 min each time, before being embedded in Spurr low viscosity embedding agent resin (SPI, United States) and cured in 70°C for 24 h. Ultrathin sections were cut on a Leica microtome and imaging was performed using a HITACHI HT7800 TEM.

In order to determine the roles of mitochondrial dynamics in border cell migration, we first examined whether Drp1 and Mfn are essential for border cell migration. Two to three different RNAi lines for Drp1 or Mfn were used to confirm their phenotypes. slbo-Gal4, a border cell-specific Gal4 driver was used to drive expression of various RNAi transgenes in border cells. Drp1 RNAi and Mfn overexpression both resulted in phenotype of strong migration delay as compared with the wild-type control (Figures 1A–C). The migration index (M.I.) of the control was 0.97 ± 0.01 (maximum value being 1), whereas those of Drp1 RNAi and Mfn overexpression were 0.64 ± 0.03 and 0.75 ± 0.01, respectively, indicating strong migration defects. The completion index (C.I.) of the control was 0.94 ± 0.02 (maximum value being 1), whereas those of Drp1 RNAi and Mfn overexpression were 0.53 ± 0.02 and 0.59 ± 0.02, respectively. The M.I. and C.I. of Drp1 overexpression was 0.82 ± 0.03 and 0.72 ± 0.02, respectively, displaying moderate migration defects (Figures 1A–C). However, there was no significant difference between M.I. and C.I. of Mfn RNAi as compared with the wild-type control (Figures 1A–C). Together, these results indicate that Drp1 but not Mfn is genetically required for border cell migration. Consistently, border cell clusters containing homozygous Drp1 ¹ or Drp1 ² mutant clones also caused migration defects (Figure 1E and data not shown). Specifically, 100% of clusters containing only Drp ¹ mutant border cells displayed migration delay (n = 15), while 26% of mosaic clusters containing both wildtype and mutant border cells displayed migration delay (n = 19) (Figure 1G).

Expression of Mito-GFP, a widely used mitochondrial marker (Chen, 2020), in border cells revealed that both Drp1 RNAi and Mfn overexpression (OE) resulted in a morphology of increased mitochondrial fusion as compared with the wildtype control (Figure 1D). The Mfn OE or Drp1 RNAi border cell cluster contains a number of large patches of bright Mito-GFP signals in the focal plane that reflect a highly fused or clustered mitochondrial network (with Mfn OE more severe than Drp1 RNAi), whereas the control or Drp1 OE border cell cluster contains mostly scattered small-sized dots of Mito-GFP signals that indicate a more fragmented mitochondrial network. In addition, likely due to their small size, mitochondria in individual border cells from the control or Drp1 OE are much more uniformly distributed around nucleus and throughout cytoplasm than those in the Mfn OE or Drp1 RNAi border cells. In contrary, the large patches of Mito-GFP in the Mfn OE or Drp1 RNAi border cells are often asymmetrically distributed or concentrated in a corner of individual cell (i.e., on one side of the nucleus and not around nucleus). This phenomenon is likely resulted from most mitochondria within a cell organizing themselves into one or two highly-fused networks, whose large size prevent them from distributing evenly throughout the cytoplasm of the cell. Similarly, clones of Drp1 ¹ mutant follicle cells also resulted in increased mitochondrial fusion as compared with the control (Figure 1F). Taken together, these data indicate that Drp1-mediated mitochondrial fission is required for border cell migration. In comparison, Mfn-mediated mitochondrial fusion does not seem to be as critical during border cell migration.

We next investigated the underlying cause of migration defects by Drp1 RNAi, Mfn or Drp1 OE, which supposedly caused disruption of mitochondrial dynamics. We used Lifeact-GFP to label the F-actin and actin cytoskeleton during live imaging of border cell migration. The wild-type border cell cluster utilizes a long leading protrusion enriched with F-actin at the leading edge of the cluster to power their collective migration (Figure 2A; Supplementary Video S1), reducing protrusion formation would lead to strong migration defects. Drp1 knock down, Mfn OE, and Drp1 OE each led to shortened leading protrusions, as shown by live imaging of Lifeact-GFP (Figures 2A,B; Supplementary Videos S1–S4), and similar results was obtained by phalloidin staining of fixed samples (Supplementary Figure S5). Results from loss-of-function experiments indicate that Drp1-mediated mitochondrial fission is required for formation of long leading protrusion. On the other hand, gain-of-function Mfn and Drp1 results suggest that over-activation of fusion or fission also impairs protrusion formation. Formation of lamellipodial protrusion depends heavily on actin polymerization, an ATP-dependent process. We then sought to determine whether Drp1-mediated mitochondrial dynamics promote protrusion formation through regulating ATP levels during border cell migration. The small number of border cells for each egg chamber precludes direct ATP measurement in border cells. Instead, we proceeded to measure the ATP levels of ovaries that contain egg chambers at the early and middle stages of oogenesis. Each mid-stage egg chamber is composed of the somatic tissue of 650 follicle cells and the germline tissue of 15 nurse cells and one oocyte (Montell, 2003). Since border cells originate from the epithelial layer of follicle cells, results from follicle cells would closely reflect what would happen in border cells. Therefore, we drove the expression of Drp RNAi, Drp1 and Mfn specifically in the follicle cells using the Gr1-Gal4 driver, and found that loss-of-function of Drp1 or gain-of-function of Mfn resulted in increased mitochondrial fusion as compared to the control (Figure 2C). The Mfn OE or Drp1 RNAi follicle cells contains large patches of bright Mito-GFP signals (with Mfn OE more severe than Drp1 RNAi) that reflect a highly clustered mitochondrial morphology and are often asymmetrically distributed or concentrated in a corner of individual follicle cell (i.e., on one side of nucleus), whereas the control or Drp1 OE follicle cells exhibit a uniform distribution pattern of mitochondria around the central nucleus, likely due to their small size. Together, these results are similar to what we have observed in the border cells (Figures 1D, 2C), validating the use of follicles cells and thus egg chambers for ATP measurement.

We found that ATP levels were significantly decreased in Drp1 RNAi and Mfn OE as compared to that of the control (Figure 2D). It should be noted that if we can measure the ATP levels of only the follicle cells and not the nurse cells and oocyte (where Drp1 and Mfn levels are not altered), the extent of ATP level decrease would have been even greater than what we have observed. This result indicates that misregulation of mitochondrial fission results in decrease of ATP levels in vivo. To confirm this result, we utilized the S2 cells as an in vitro system to test the effects on ATP levels by misregulation of mitochondrial dynamics. dsDrp1 as well as Drp1 and Mfn were expressed in S2 cells to achieve Drp1 knock down, Drp1 OE, and Mfn OE respectively. We found that Drp1 loss-of-function and Mfn gain-of-function each resulted in significant increase of mitochondrial fusion accompanied by significant reduction of ATP levels (Figures 2E–H; Supplementary Figures S1, S2), similar to those observed in the follicle cells. This result indicates that Drp1-mediated mitochondrial fission is required for optimal ATP levels both in vivo and in vitro. Furthermore, overactivation of fission by Drp1 OE was shown to significantly increase mitochondrial fission or reduce mitochondria size (Supplementary Figure S1). Interestingly, Drp1 OE led to moderate increase of ATP levels 48 h after transfection but subsequent decrease of ATP levels 60 h after transfection. This result implies that moderate increase of Drp1 levels (at 48 h, Figure 2F), or moderate increase of fission is beneficial for ATP production. In contrast, strong increase of Drp1 levels (at 60 h, Figure 2F), or too much increase of fission is detrimental for ATP production. Taken together, the above results indicate that the reduced leading protrusions correlates with the reduction of ATP levels as resulted from the misregulation of mitochondrial dynamics, suggesting that during border cell migration Drp1 promotes formation of long leading protrusion through regulating mitochondrial fission and optimal ATP production.

We next sought to investigate the effect of ATP reduction on protrusion behavior during border cell migration. We treated the stage 9 egg chambers with Oligomycin, a drug that blocks ATP generation from mitochondria by inhibiting the proton channel of ATP synthase F0 subunit (Symersky et al., 2012), at different concentrations. Oligomycin treatment at 5 nM resulted in significant reduction of ATP levels in the ovaries, while higher concentrations of 50 and 500 nM resulted in strong reduction of ATP levels, indicating that the drug treatment is effective (Figure 3D). Live imaging revealed that treatment of 5 nM Oligomycin causes border cell cluster to retract leading protrusion immediately after drug treatment (Figure 3A; Supplementary Videos S5, S6). This phenotype is similar to what was observed for the Drp1 RNAi border cell cluster, which mostly exhibits a short leading protrusion and occasionally extends a long protrusion that was then quickly retracted (Supplementary Video S2). Treatment of 50 and 500 nM Oligomycin both resulted in similar retraction of long protrusions (Figures 3B,C; Supplementary Videos S7–S10). Furthermore, the highly dynamic F-actin patches labeled by Lifeact-GFP in the border cells before drug treatment exhibited significant reduction of dynamics after treatment, which is more severe in 50 and 500 nM treatment than that in 5 nM treatment. This dramatic reduction of dynamics of F-actin patches was not apparent in Drp1 RNAi border cells, likely due to the fact that drug treatment causes an immediate and drastic reduction in ATP levels whereas the effect on ATP levels from the genetic knockdown is over a longer time course and is less severe. We have recently found that a dynamic supracellular actomyosin network spanning throughout the outer surface of border cell cluster promotes a front polarized cluster morphology and collective migration (Wang et al., 2020). We therefore examined the effect of ATP reduction on Sqh (Myosin II’s regulatory light chain) organization and dynamics. Live imaging revealed that the assembly of Sqh-GFP into large patches as well as their dynamic movement on both the cluster surface and within polar cells (cluster interior) are significantly affected, with the 50 and 500 nM treatments being more severe than the 5 nM treatment (Figures 3E–G; Supplementary Videos S11–S16). Taken together, these results indicate that strong ATP reduction by Oligomycin treatment affects actin and myosin organization and dynamics, suggesting that reduction of ATP levels by Drp1 RNAi may result in similar (albeit milder) effects in border cells, negatively affecting protrusion behaviors.

RhoA and its downstream effector Rho kinase (ROCK) play critical roles in regulation of actin and Myosin II dynamics during border cell migration (Zhang et al., 2011; Montell et al., 2012; Wang et al., 2020). Recent studies in mammalian cells and animal models have also revealed that ROCK could directly phosphorylate and interact with Drp1 to either help recruit Drp1 to mitochondria to promote fission or prevent Drp1 from binding to mitochondria to suppress fission (Wang et al., 2012; Hu et al., 2020). We then sought to determine whether Drosophila RhoA and ROCK can regulate mitochondrial fission through affecting Drp1’s mitochondria binding ability. We overexpressed RhoA and ROCK or their activated forms in the S2 cells, follicle cells and border cells to test whether signaling from Rho A and ROCK affects Drp1’s recruitment to the mitochondria in vitro and in vivo. First, we found that overexpressing RhoA and ROCK in S2 cells did not alter the total levels of Drp1 level as compared to the control, suggesting that signaling from Rho A or ROCK does not change Drp1’s expression levels (Supplementary Figures S4A,B). Interestingly, RhoA and ROCK expression in S2 cells each caused significantly increased recruitment of Drp1 to the mitochondria, and treatment of ROCK inhibitor Y27632 abolished Rho’s positive effect on Drp1 recruitment, indicating that Rho A promotes Drp1 recruitment to the mitochondria through ROCK (Supplementary Figures S4C,D). Furthermore, Rho A and ROCK expression in S2 cells each resulted in significantly increased mitochondrial fission (Supplementary Figures S3A–D), as indicated by the significant reduction of four parameters of mitochondrial morphology. And treatment of ROCK inhibitor Y27632 also abolished Rho’s effect on increased fission, indicating that RhoA promotes mitochondrial fission through ROCK (Supplementary Figures S3A–D). Importantly, we also found that ROCK was able to physically interact with Drp1 (Figure 4G). We generated an antibody against Drosophila Drp1 and performed co-IP (Co-immunoprecipitation) assay using the newly generated Drp1 antibody. We found that the endogenous Drp1 in the S2 cells could be co-immunoprecipitated together with ROCK-Fg, and conversely ROCK-Fg could be co-immunoprecipitated together with Drp1-HA (Figure 4G). This result provides the mechanistic detail that ROCK regulates Drp1’s mitochondrial recruitment by binding to Drp1. Moreover, the effect of increased mitochondrial fission as a result of ROCK transfection was abolished by Drp1 knockdown (dsDrp1) (Supplementary Figure S3E–H), indicating that Drp1 is required for ROCK-induced mitochondrial fission in S2 cells. Together, these results demonstrate that Rho signaling promotes Drp1’s mitochondrial recruitment and mitochondrial fission in S2 cells through ROCK and its physical interaction with Drp1.

Second, we found that expression of RhoA V14 and ROCK.CAT, the activated forms of RhoA and ROCK respectively, in the follicle cells each resulted in significant increase of Drp1 recruitment onto mitochondria without affecting Drp1’s total levels (Figures 4E,F; Supplementary Figures S4E,F), consistent with the in vitro results. Furthermore, TEM (transmission electron microscopy) analysis on stage 9 or 10 egg chambers confirmed that ROCK.CAT and Drp1 expressing follicle cells contain significantly smaller mitochondria than the wildtype follicle cells, whereas Drp1 RNAi expressing cells display significantly larger mitochondria (Figures 4B–D). Therefore, both in vivo and in vitro results support the conclusion that RhoA signaling induces mitochondrial fission through ROCK. Interestingly, expression of RhoA V14 or ROCK.CAT in the follicle cells results in higher ATP levels in the egg chambers as compared with the control (Supplementary Figure S6), further supporting the notion that increased mitochondrial fission (as caused by elevated Rho or ROCK activity) promotes higher ATP levels. Lastly, confocal microscopy demonstrates that increasing ROCK or Rho activity in the border cells (by ROCK.CAT or RhoA V14 expression) could increase the staining of Drp1 that colocalizes with mitochondria as compared with that of the wildtype control (Figure 4A). This result is consistent with the data that Rho and ROCK increase the recruitment of Drp1 to the mitochondria in the follicle cells (Figures 4E,F) and S2 cells (Supplementary Figures S4A–D).