Abstract

   Mitochondria are maternally inherited, but the mechanisms underlying
   paternal mitochondrial elimination after fertilization are far less
   clear. Using Drosophila, we show that special egg-derived
   multivesicular body vesicles promote paternal mitochondrial elimination
   by activating an LC3-associated phagocytosis-like pathway, a cellular
   defense pathway commonly employed against invading microbes. Upon
   fertilization, these egg-derived vesicles form extended vesicular
   sheaths around the sperm flagellum, promoting degradation of the sperm
   mitochondrial derivative and plasma membrane. LC3-associated
   phagocytosis cascade of events, including recruitment of a
   Rubicon-based class III PI(3)K complex to the flagellum vesicular
   sheaths, its activation, and consequent recruitment of Atg8/LC3, are
   all required for paternal mitochondrial elimination. Finally, lysosomes
   fuse with strings of large vesicles derived from the flagellum
   vesicular sheaths and contain degrading fragments of the paternal
   mitochondrial derivative. Given reports showing that in some mammals,
   the paternal mitochondria are also decorated with Atg8/LC3 and
   surrounded by multivesicular bodies upon fertilization, our findings
   suggest that a similar pathway also mediates paternal mitochondrial
   elimination in other flagellated sperm-producing organisms.

   Subject terms: Embryonic induction, Mitochondria, Mitophagy
     __________________________________________________________________

   Mitochondria are maternally inherited, but the mechanisms underlying
   paternal mitochondrial elimination remain unclear. Here, the authors
   show that Drosophila eggs rapidly degrade paternal mitochondria after
   fertilization by triggering an innate immunity pathway commonly
   employed against invading microbes.

Introduction

   Organisms as diverse as plants, fungi, protists, and animals differ
   greatly in external morphology, internal structure, and physiology.
   However, when it comes to sexual mode of reproduction, they share
   multiple similar cellular and molecular patterns, such as meiosis,
   merger of the gametes, and a uniparental mode of mitochondrial
   inheritance^[52]1–[53]9. In many organisms, including mammals and
   Drosophila, the union of the gametes during fertilization occurs
   between a pair of highly distinct cells, the elongated male sperm, and
   the spherical female egg, each contributing half of the parental
   genomic material to the zygote^[54]10–[55]12. In contrast, only one
   parent, generally the mother, passes on the mitochondria to the
   offspring^[56]8,[57]13,[58]14. Although this uniparental mode of
   inheritance is ubiquitous among eukaryotes, there are still abundant
   gaps in our knowledge concerning the molecular mechanisms underlying
   maternal mitochondrial inheritance, the fate of the paternal
   mitochondria after fertilization, and the rationale behind this highly
   conserved phenomenon^[59]8,[60]9,[61]13–[62]15. These ambiguities may
   be attributed to various factors, including the widely held
   misconception that upon fertilization, sperm cells discard both the
   flagellum and the mitochondria, and that only the sperm head enters the
   egg^[63]16. In addition, it is difficult to exclude technical artefacts
   and false positive or negative findings due to assay sensitivity
   limits, such as assays for detection of scarce paternal mitochondrial
   (mt)DNA copies among the vast copy number of maternal mtDNA
   molecules^[64]17. Moreover, such assays may be highly misleading, as in
   some organisms, including Drosophila and perhaps also in humans,
   elimination of the paternal mtDNA has been reported to be uncoupled
   from, and already occurring prior to, degradation of the paternal
   vacuolar mitochondria (i.e., mtDNA-free mitochondria)^[65]18–[66]24.
   The anatomical divergence in male gametes and their mitochondria among
   different organisms is another impeding factor, as this often hinders
   cross-study comparisons.

   Two general theories have emerged to explain the mechanisms underlying
   paternal mitochondrial elimination (PME) after fertilization: A passive
   model of simple dilution of the paternal mitochondria by the excess
   copy number of egg mitochondria; and an active model, in which
   egg-derived mechanisms specifically target the paternal mitochondria
   for degradation. Both models have their drawbacks. The passive model is
   primarily based on two studies using mainly, albeit not exclusively,
   interspecific mouse crosses^[67]21,[68]25. This experimental setup
   could lead to paternal mitochondrial leakage in the progeny, as has
   been previously documented with murine and bovine hybrids, presumably
   due to failure of the egg to recognize and target the paternal
   mitochondria^[69]26–[70]28. Another reason to be cautious about this
   model is that proteins which may imply involvement of active mechanisms
   in mediating PME, were shown to decorate the paternal mitochondria
   after fertilization in mice and other mammals^[71]21,[72]29–[73]31.
   Most relevant are proteins involved in recycling of damaged
   mitochondria by selective autophagy (mitophagy), including ubiquitin,
   p62, and Atg8/LC3. Genetic studies in the roundworm Caenorhabditis
   elegans, which constitute the basis for the active model, suggest that
   PME is mediated by mechanisms reminiscent of
   mitophagy^[74]18,[75]30,[76]32–[77]37. However, as opposed to
   flagellated sperm in mammals, in which the mitochondria undergo
   extensive structural organization and become part of the flagellum
   midpiece^[78]6,[79]38, the C. elegans amoeboid sperm lacks a flagellum
   and contains mitochondria of simple morphology, which can be
   accommodated by mitophagy^[80]39.

   Unlike C. elegans, Drosophila melanogaster flies produce unusually long
   sperm, consisting of an approximately 10 µm long needle-shaped nucleus
   and a ~ 1.8 mm long flagellum ( ~35 times the length of human sperm).
   The flagellum is mainly occupied by two major organelles: a single
   cylindrical mitochondrion, termed the mitochondrial derivative (MD),
   which extends from the posterior edge of the nucleus and along the
   entire length of the flagellum, and an intimately associated axoneme, a
   microtubule cytoskeleton-based structure which generates the propulsive
   force for sperm cell movement. The extraordinary size of the MD in
   Drosophila, which completely penetrates the egg, facilitates high
   spatiotemporal resolution of the PME process, with possible
   implications for PME in mammals^[81]5,[82]40,[83]41. Indeed, we
   previously demonstrated that in Drosophila, PME is an active process
   mediated by two types of egg-derived vesicles: multivesicular bodies
   (MVBs) of the endocytic pathway, which are the first to engage and
   target large segments of the MD for degradation; and autophagosomes of
   the autophagy pathway, that recycle small remnants of the degraded
   MD^[84]42. However, the pathway by which the MVBs recognize the MD and
   mediate PME remains unknown.

   Although originating in the endocytic pathway and expressing late
   endosomal markers, such as Rab7, the MVBs that associate with the MD
   also display Atg8, a ubiquitin-like protein of the autophagy system, on
   their limiting membrane^[85]42. In recent years, several related
   noncanonical autophagy pathways, which mediate conjugation of Atg8 to
   single membranes (CASM), have emerged. As opposed to the double
   membrane autophagosomes that mediate classical autophagy, CASM pathways
   commonly utilize a subset of the autophagy machinery for Atg8
   lipidation onto single-membrane vesicles^[86]43–[87]47. The most
   extensively investigated CASM pathway is LC3-associated phagocytosis
   (LAP), which is involved in the clearance of different extracellular
   cargos by phagocytosis, including a variety of pathogenic microbes, but
   also dying cells, axon debris, soluble ligands, and protein aggregates,
   in different tissues and cell types^[88]48–[89]60. LAP is activated
   upon engagement of extracellular cargos with specific surface
   receptors, mainly but not exclusively Toll-like receptors (TLRs),
   triggering inward invagination of the plasma membrane and engulfment of
   the cargo into a single-membrane phagosome (also termed
   LAPosome)^[90]49,[91]59. A LAP-specific class III PI(3)K complex
   (PI3KC3; also known as complex III), composed of at least five proteins
   (Rubicon, Uvrag, Pi3k/Vps34, Atg6/Beclin1, and Vps15)^[92]61,[93]62, is
   recruited to the LAPosome membrane, where it generates the phospholipid
   phosphatidylinositol-3-phosphate (PtdIns(3)P). Consequently, via a yet
   unknown mechanism, PtdIns(3)P recruits the Atg8/LC3 conjugating system,
   which in turn promotes recruitment of Atg8/LC3 to the LAPosomal
   membrane in a manner that requires the NADPH oxidase-2 (Nox2)-derived
   reactive oxygen species (ROS). Atg8/LC3 lipidation then facilitates
   fusion of the LAPosomes with lysosomes, promoting cargo
   digestion^[94]46,[95]51,[96]52,[97]61,[98]63,[99]64.

   Here, to uncover the sperm MD cellular degradative pathway mediated by
   the MVBs, we isolated these vesicles from early fertilized Drosophila
   eggs, using a subcellular fractionation procedure. Mass spectrometry
   (MS) analysis of the MVB-enriched fractions, followed by pathway
   enrichment analysis of the protein composition data, suggested the
   involvement of an innate immunity pathway with prevailing elements of
   phagocytosis. This, and the observation that Atg8 is expressed on the
   limiting membrane of the MVBs that associate with the MD^[100]42,
   prompted us to explore the idea that the MVBs might mediate PME by
   activating a LAP-like CASM pathway. We demonstrate that mutants and
   RNAi knockdown of the LAP-specific PI3KC3 component, Rubicon, as well
   as knockdown of three other main components of the complex, all
   significantly attenuate degradation of the MD. Maternal expression of a
   rubicon-eGFP rescue transgene, and of tdTomato-tagged endogenous and
   transgenic Rubicon in the early fertilized egg, revealed the presence
   of multiple MVBs expressing Rubicon mainly on their limiting membrane.
   We then show that immediately after fertilization, many of the
   Rubicon-positive MVBs engage the sperm flagellum, creating extended
   vesicular sheaths around large flagellar segments. The flagellum
   vesicular sheaths (FVS) mediate rapid degradation of the sperm plasma
   membrane in a Rubicon-dependent manner, exposing the MD and axoneme to
   the luminal contents of these vesicular coats. Other LAP pathway
   events, such as Rubicon-dependent production of PtdIns(3)P and
   association of ROS with the sperm flagellum, also unfold during this
   process. Subsequently, Atg8 is recruited to the FVS in a Rubicon- and
   Atg7-dependent manner, ultimately leading to degradation of the MD, but
   not the axoneme, within strings of large Atg8 positive vesicles that
   originate in the FVS and contain lysosomal contents. These findings
   indicate that the egg activates a LAP-like pathway to target the MD for
   degradation. Given previous observations of MVBs in the vicinity of the
   paternal mitochondria in both bovine and hamster early
   embryos^[101]27,[102]65, as well as scattered indications of the
   presence of LC3 decorated paternal mitochondria in early embryos of
   some mammalian organisms^[103]21,[104]29–[105]31, these findings
   suggest that a similar pathway might also mediate PME in mammals.

Results

MVBs contain elements of innate immunity and phagocytosis

   In a previous study, we reported that degradation of the paternal
   mitochondria after fertilization in Drosophila is an efficient and
   rapid process, such that within 1.5-h after egg laying (AEL), the
   entire ~1.8 mm long mitochondrial derivative (MD) is
   eliminated^[106]42. To further optimize this system for live imaging
   and minimize photobleaching, we generated a transgenic line that
   produces sperm with red fluorescently labeled MD (red-MD). This
   transgene consists of a fusion of a mitochondrial targeting sequence
   (MTS) with the red fluorescent protein tdTomato, which is both one of
   the brightest and most resistant to photobleaching among the DsRed
   derivatives^[107]66. Monitoring PME in eggs laid by wild-type (WT)
   females and fertilized by males producing red-MD sperm, revealed that
   this process is even faster, and that the MD is eliminated within only
   1-hour AEL (Supplementary Movie [108]1). Note that time zero AEL is not
   equivalent to time zero after fertilization, as the latter occurs prior
   to egg laying, inside the female reproductive tract.

   A key step in PME is the association of egg-derived MVBs with the sperm
   flagellum, an event that gradually and asynchronously culminates in the
   degradation of large MD segments (illustrated in Fig. [109]1a;
   Supplementary Fig. [110]1a–c)^[111]42. Accordingly, inactivation of
   genes required for proper biogenesis of the intraluminal vesicles
   (ILVs), the functional units of the MVBs, strongly attenuate
   PME^[112]42. Reasoning that the MVBs trigger a cellular degradation
   pathway to eliminate the MD, we set out to isolate the MVBs from early
   fertilized eggs and determine their proteome by MS. To label the MVBs,
   we generated a transgenic fly line containing an eGFP-tagged human CD63
   (hCD63) gene, a member of the tetraspanin superfamily and a general
   marker of MVBs in human and Drosophila, which mainly localizes to the
   ILV compartment of these vesicles^[113]67. This transgene is expressed
   under the control of the UASp regulatory region, which allows for
   efficient expression of transgenes in the female germline by the
   maternal driver matα-GAL4. Note that some of the maternal transgenes
   were placed under UASz, an equivalent regulatory
   region^[114]68,[115]69. Eggs laid by females maternally expressing the
   eGFP-hCD63 transgene and fertilized by males producing red-MD sperm
   displayed multiple eGFP-hCD63 labeled vesicles, including vesicles
   associated with large segments of the MD (Fig. [116]1b).

Fig. 1. Rubicon is required for PME after fertilization.

   [117]Fig. 1
   [118]Open in a new tab

   a Illustration of the onset of PME in Drosophila based on^[119]42.
   Left, an early fertilized egg. Egg-derived MVBs associate with the
   1.8 mm long sperm flagellum. Middle, a cross-section through the sperm
   flagellum illustrating a single cylindrical MD (red) and an intimately
   associated axoneme (green), both coated by a common plasma membrane
   (gray) and extend along the entire length of the flagellum. Right,
   enlargement of an MVB (shown in a cross-section displaying the ILVs and
   the encapsulating limiting membrane) that encapsulates the flagellum.
   Atg8/LC3 (cyan) is loaded on the MVB limiting membrane. b A
   representative confocal image of the anterior (top) region of an egg
   maternally expressing the MVB transgenic construct UASp-hCD63-eGFP
   (green), and fertilized by a coiled red-MD sperm cell (magenta;
   dj-(MTS)tdTomato). MVBs that coat large flagellar segments are in white
   (arrowhead). Scale bar, 10 μm. c Schematic view of three main features
   shared by the egg MVBs and the LAP pathway. d Schematic view of the LAP
   pathway as defined in mammalian cells (created with BioRender.com). e
   Live imaging of the anterior regions of early fertilized eggs laid by
   females of the indicated genotypes and fertilized by red-MD sperm.
   Early fertilized eggs carrying a single copy of the rubicon gRNA served
   as control. Representative time-lapse images of each fertilized egg
   were taken at 0, 1, 2, and 3 h AEL. Brightfield and fluorescence
   channel image overlay reveals the egg (gray) and the red-MD sperm
   flagellum (magenta), respectively. Scale bar, 100 μm. f Quantifications
   of normalized red-MD fluorescence intensities [arbitrary units (a.u.)]
   in fertilized eggs represented in e. Error bars indicate standard error
   of the mean (SEM). The respective numbers of scored early fertilized
   eggs (n) laid by females of the control, rubicon^Δ10, and rubicon^Δ10;
   rubicon-eGFP genotypes are 61, 32, and 34. **P  <   0.01 and ****P  < 
    0.0001. Two-way repeated measures ANOVA, followed by Dunnett’s
   multiple comparisons test. The P values are as follows: control vs.
   rubicon^Δ10, P < 0.0001 (all tested hours AEL); control vs.
   rubicon^Δ10; rubicon-eGFP, P = 0.0070 (0 h AEL). Source data are
   provided as a Source Data file.

   To isolate the MVBs, extracts prepared from eGFP-hCD63-expressing early
   fertilized eggs were subjected to a modified OptiPrep ^TM density
   gradient procedure of subcellular fractionation, and the MVB-enriched
   fractions were detected using Western blotting with anti-GFP antibodies
   (Supplementary Fig. [120]1d). The enrichment of largely intact MVBs and
   their ILVs in the peak fractions was confirmed by transmission electron
   microscopy (TEM; Supplementary Fig. [121]1e), and the composition of
   the proteins in these fractions was revealed by MS (Supplementary
   Data [122]1). Pathway enrichment analysis of the data revealed six
   major pathways, five of which were somewhat trivial or unrelated
   directly to cellular degradation processes (i.e., mitochondria,
   metabolism, endomembrane system, glycosylation, and transport), but one
   pathway, innate immunity with elements of phagocytosis, seemed
   particularly intriguing, as it implied possible involvement of a
   targeted cellular destruction pathway in PME (Supplementary
   Fig. [123]1f).

Components of the LAP-specific PI3KC3 are required for PME

   In the past decade, an unconventional phagocytosis pathway termed
   LC3-associated phagocytosis (LAP) has emerged as an important defense
   pathway against invading microbes. LAP combines the pathways of
   endocytosis and autophagy, mediating the delivery of pathogens and
   other extracellular elements to lysosomes by
   phagocytosis^[124]51,[125]52,[126]57–[127]59,[128]63,[129]64. The
   finding that the MVBs, which are known to originate in the endocytic
   pathway, contain elements of phagocytosis, as well as the observed
   association of Atg8/LC3 with the limiting membrane of the MVBs that
   engage the flagellum^[130]42, raised the hypothesis that the MVBs
   mediate PME by activating a LAP-like pathway (Fig. [131]1c).
   Conventional autophagy and LAP are functionally and mechanistically
   distinct, as the autophagosome is a double-membrane structure, while
   the LAPosome, the characteristic vesicle mediating LAP phagocytosis, is
   a single membrane structure^[132]49. Furthermore, although LAP involves
   a subset of the autophagic machinery, it does not require the very
   first autophagy-specific complex. Also known as the pre-initiation
   complex, the first autophagy complex translocates to autophagy
   initiation sites and regulates the recruitment of the next autophagy
   complex^[133]57,[134]60,[135]70,[136]71. Moreover, as opposed to the
   PI3KC3s complexes I and II that promote autophagy, the PI3KC3 complex
   III that is essential for LAP, contains a specific protein, Rubicon,
   which functions to inhibit autophagy
   (Fig. [137]1d)^[138]61,[139]62,[140]72–[141]75. Significantly, MVBs are
   also vesicles with a single limiting membrane, and as in LAP,
   inactivation of several components of the autophagy pre-initiation
   complex, including Atg1/Ulk1, Fip200, and Atg13, had either weak or no
   effect on PME^[142]42,[143]76. In contrast, downregulation of uvrag,
   which is a component of the PI3KC3 complexes II and III, significantly
   attenuates PME^[144]42.

   Given the anatomical and molecular similarities between the LAP and PME
   events, we set out to explore the possibility that a LAP-like pathway
   underlies PME in fertilized Drosophila eggs. For this, we first asked
   whether Rubicon, the main protein that distinguishes between the
   LAP-specific and the autophagy-specific PI3KC3s, is involved in this
   process^[145]57,[146]61,[147]62,[148]77. Using CRISPR/Cas9, we
   generated four rubicon mutant alleles, all containing small
   frameshifting deletions at the beginning of the second exon, leading to
   premature stop codons and putative truncated proteins that lack the
   highly conserved RUN domain and the Rubicon homology (RH) domain
   (Supplementary Fig. [149]2a). At the RNA level, two rubicon mutant
   alleles, rubicon^Δ10 and rubicon^Δ13, display a significant reduction
   in the expression levels of maternal rubicon mRNA, whereas a third
   mutant allele, rubicon^Δ196, containing a deletion encompassing the
   first splice-acceptor site, generates a truncated mRNA lacking part of
   the translated 5′ region (Supplementary Fig. [150]2b).

   To test for possible effects of loss of Rubicon on PME, females
   homozygous for each of the four rubicon mutant alleles were crossed
   with males producing red-MD sperm. To ensure that effects on PME are
   not a consequence of failed embryonic development, we monitored normal
   progression in development of the early fertilized eggs through the
   process of blastoderm cellularization and omitted from our analysis the
   early fertilized eggs that failed to develop. Critically, all rubicon
   mutant fertilized eggs displayed significant attenuation in the
   degradation of the MD, as revealed by the persistence of the red-MD in
   the developing mutant fertilized eggs, even at 3 h AEL (Fig. [151]1e, f
   and Supplementary Fig. [152]2c–i). Furthermore, we generated a
   rubicon-eGFP transgene, which upon maternal expression restores
   degradation of the MD in rubicon mutant eggs (Fig. [153]1e, f and
   Supplementary Fig. [154]2d, e).

   Flies homozygous for the rubicon mutant alleles are viable, but display
   varying levels of female and male fertility. Whereas homozygous females
   for the four rubicon mutant alleles show a significant reduction in
   fertility, with rubicon^Δ13 and rubicon^Δ196 causing the most dramatic
   effect, a milder but still significant reduction in male fertility is
   only detected in flies homozygous for these two alleles (Supplementary
   Fig. [155]3a, b). Accordingly, maternal but not male germline knockdown
   of rubicon, leads to a significant reduction in female fertility
   (albeit the effect is less pronounced than in the mutants) with no
   change in male fertility, respectively (Supplementary Fig. [156]3c, d).
   However, whereas these findings indicate that maternal rubicon is
   required for normal development, this effect likely occurs at
   post-embryonic cellularization stages, as both maternal rubicon mutants
   and knockdown progressed to the embryonic cellularization stage with
   similar kinetics as their respective control early fertilized eggs
   (Supplementary Fig. [157]3e, f). In conclusion, these findings suggest
   that PME and maybe other maternal Rubicon-dependent processes, are
   highly important for later organismal development.

   In addition to Rubicon, the LAP-specific PI3KC3 protein complex also
   contains Uvrag, Pi3k/Vps34, and Atg6/Beclin1 (Fig. [158]1d). Consistent
   with our previous report^[159]42, knocking down uvrag by maternal
   expression of specific shRNA transgenes (ShR), significantly attenuated
   PME also in the current setup (Fig. [160]2d, e). Likewise, knockdowns
   of pi3k and atg6 attenuated PME, although the effect of atg6 knockdown
   was less pronounced than that of the uvrag, pi3k and rubicon knockdowns
   (Fig. [161]2a–i and Supplementary Fig. [162]4a–d). In contrast, no
   effect on PME was detected upon knockdown of atg14, a key component of
   the autophagy-specific PI3KC3 (complex I)^[163]78,[164]79
   (Fig. [165]2j, k and Supplementary Fig. [166]4e). Taken together, these
   findings suggest that a LAP-specific PI3KC3 is assembled in fertilized
   eggs and mediates degradation of the MD.

Fig. 2. Main members of the LAP-specific, but not the autophagy-specific
PI3KC3 complex are required for PME.

   [167]Fig. 2
   [168]Open in a new tab

   a, b, d, f, h, j PME kinetic assays in control (expressing the maternal
   driver only) (a); and shRNA (ShR) mediated maternal knockdown in early
   fertilized eggs of rubicon (b), uvrag (d), pi3k (f), atg6 (h), and
   atg14 (j). A live imaging assay performed and presented as in
   Fig. [169]1e. Scale bar, 100 μm. c, e, g, i, k Quantifications of
   normalized red-MD fluorescence intensities in the fertilized eggs
   [magenta; arbitrary units (a.u.)] of the genotypes indicated on the
   left. Error bars indicate SEM. The respective numbers of scored
   fertilized eggs (n) laid by females of the control, rubicon^ShR (c),
   uvrag^ShR (e), pi3k^ShR (g), and atg6^ShR (h), are 68, 60, 24, 22, and
   24. For atg14 knockdown and its control, 30 and 51 fertilized eggs were
   scored, respectively (k). Note that atg14 has its own control group, as
   this experiment was conducted at a later time. *P  <   0.05, **P  < 
    0.01, ***P  <   0.001, and ****P  <   0.0001. Two-way repeated
   measures ANOVA, followed by Šídák’s multiple comparisons test. P values
   in c, P < 0.0001 (1–3 h AEL); e, P < 0.0001 (1–2 h AEL), P = 0.0446
   (3 h AEL); g, P = 0.0436 (0 h AEL), P < 0.0001 (1 h AEL), P = 0.0011
   (2 h AEL); i, P = 0.0244 (1 h AEL); k, P = 0.0005, (0 h AEL),
   P = 0.0122 (1 h AEL). MD mitochondrial derivative. Source data are
   provided as a Source Data file.

Rubicon vesicles associate with the sperm and degrade the MD

   To examine whether the LAP-specific PI3KC3 assembles on egg vesicles,
   live fertilized eggs, laid by females maternally expressing the
   rubicon-eGFP rescuing transgene, and fertilized by males producing
   red-MD sperm, were collected 0–15 min (min) AEL and imaged for 45 min.
   Remarkably, Rubicon-eGFP specifically labeled three elements in the
   egg; numerous vesicles scattered throughout the ooplasm, large sperm
   flagellar segments, and large vesicles, 1–1.5 µm in diameter on
   average, that bud from, but remaining in the vicinity of, the
   flagellum, and contain MD pieces and debris (Fig. [170]3a–c and
   Supplementary Movie [171]2). Interestingly, within the large vesicles
   that bud from the flagellum, Rubicon-eGFP primarily localizes to the
   limiting membrane, reminiscent of the assembly of the LAP-specific
   PI3KC3 on the LAPosome limiting membrane (Fig. [172]3c). It is
   important to note that, consistent with the requirement of PI3KC3 for
   PME, the appearance of Rubicon on the flagellum always precedes the
   elimination of the MD fluorescent signal (Supplementary Movie [173]2).

Fig. 3. Rubicon is localized to the limiting membrane of multiple egg
vesicles which generate extended vesicular sheaths that enwrap the sperm
flagellum.

   [174]Fig. 3
   [175]Open in a new tab

   a, d Representative confocal images of the anterior (top) region of
   early fertilized eggs. a An egg maternally expressing the
   UASz-rubicon-eGFP transgene (green) and fertilized by a red-MD sperm
   cell (magenta). Numerous egg vesicles display Rubicon-eGFP, many of
   which associate with large segments of the sperm flagellum (arrows),
   whereas other flagellar segments show only a few vesicles or are still
   free of vesicles (arrowheads), due to the asynchronous manner of the
   PME process along the sperm flagellum. Scale bar, 10 μm. b
   Magnification of the area outlined by a dashed rectangle in (a). Two
   types of Rubicon positive vesicles are detected: Numerous small
   vesicles that densely associate with flagellar segments in a rather
   unified order (arrowhead); Large Rubicon positive vesicles (green) that
   appear to bud from the flagellum, encapsulating MD pieces and debris
   (magenta; arrows). Scale bar, 2 μm. c Magnification of a projection of
   a few Z-sections in the area outlined by a dashed rectangle in (b).
   Rubicon (green) remains localized on the limiting membrane of the large
   vesicles that contain MD fragments and bud off from the flagellum.
   Scale bar, 1 μm. d An egg expressing tdTomato-tagged endogenous Rubicon
   (magenta) and fertilized by green-MD sperm (green). Rubicon is
   localized on multiple egg vesicles, many of which associate with large
   flagellar segments (arrowheads). Scale bar, 10 μm. e, f Shown are
   super-resolution images of flagellar regions in eggs maternally
   expressing the UASz-rubicon-tdTomato transgene (magenta), fertilized by
   green-MD sperm (green; dj-DJ-GFP), and prepared for ExM. The images
   feature large MD segments that are readily enwrapped within Rubicon
   positive FVS (arrow), as well as large Rubicon positive vesicles that
   contain MD fragments and bud off from the flagellum (arrowhead). Note
   that the image in (e) presents a single non-computed optical section,
   while the image in (f) presents a 3D computer rendering of segmented
   surfaces of this early fertilized egg. MD, mitochondrial derivative.

   To confirm that the expression dynamics and localization pattern of the
   UASz-based rubicon-eGFP transgene in early fertilized eggs reliably
   reflect those of the endogenous rubicon locus and encoded protein, we
   used CRISPR/Cas9 to tag the endogenous rubicon gene with tdTomato. Live
   imaging of eggs laid by Rubicon-tdTomato females and fertilized by
   males producing green-MD sperm (flies expressing the DJ-GFP transgenic
   protein under the regulation the dj regulatory
   regions^[176]42,[177]80), revealed that endogenous Rubicon is expressed
   in essentially identical spatiotemporal patterns as those of the
   rubicon-eGFP transgene (Fig. [178]3d).

   To visualize the anatomical details of the PME process in high
   resolution, we adopted an expansion microscopy (ExM) protocol to
   visualize Rubicon and the sperm MD in super resolution^[179]81,[180]82.
   This protocol, which expands the samples by a factor of ~4 in all three
   dimensions, was applied to fertilized eggs laid by females maternally
   expressing a rubicon-tdTomato transgene, and fertilized by males
   producing green-MD sperm. Importantly, as shown in Fig. [181]3e, the MD
   resides within Rubicon positive extended vesicular sheaths. Moreover,
   segmentation of the surfaces of the expanded fertilized egg at the area
   of the sperm flagellum and subsequent 3D surface rendering, clearly
   revealed numerous Rubicon-positive vesicles densely arranged one next
   to the other, generating flagellum vesicular sheaths (FVS) that
   completely enwrap the degrading MD (Fig. [182]3f and Supplementary
   Movie [183]3).

Breakdown of the sperm membrane precedes MD degradation

   In many insects, the fertilizing sperm cell enters the egg through an
   opening in the chorion-containing eggshell termed the
   micropyle^[184]83. In Drosophila, the sperm cell enters through the
   micropyle while coated by its own plasma
   membrane^[185]40,[186]84,[187]85. This means that to degrade the MD,
   the sperm plasma membrane is likely to break down soon after
   fertilization, before engagement of the egg vesicles with the MD, or
   that the egg vesicles engage and form the FVS over the intact sperm
   plasma membrane. To determine which of these possibilities prevails, we
   first generated a transgenic marker for the sperm plasma membrane,
   composed of the mouse CD8 transmembrane domain fused to a Venus
   fluorescent protein, and expressed under the control of the late
   spermatid promoter and regulatory sequences of the dj gene. We then
   applied the ExM protocol to eggs maternally expressing the
   rubicon-tdTomato transgene and fertilized by sperm expressing the
   CD8-Venus marker of the plasma membrane. As a control for a
   non-degradable sperm organelle, the early fertilized eggs were also
   stained with the anti-polyglycylated Tubulin antibody to label the
   axoneme^[188]42,[189]86,[190]87. Significantly, optical sections
   through the sperm flagellum revealed the sperm plasma membrane breaking
   down over an intact axoneme, and both are entrapped within
   Rubicon-positive FVS (upper panels in Fig. [191]4a). Furthermore,
   surface segmentation and subsequent 3D rendering of these fertilized
   eggs, readily indicated the dense arrangement of the Rubicon vesicles
   that form FVS around the sperm plasma membrane (lower panels in
   Fig. [192]4a).

Fig. 4. Sperm plasma membrane breakdown after fertilization occurs within the
FVS at a much faster rate than that of MD degradation.

   [193]Fig. 4
   [194]Open in a new tab

   a, f Super-resolution images of flagellar regions in an egg maternally
   expressing the UASz-rubicon-tdTomato transgenic construct (magenta;
   immunostained with anti-RFP antibody), fertilized by sperm cells
   expressing the dj-CD8-Venus transgene that labels the plasma membrane
   (a) or the dj-DJ-GFP transgene that labels the MD (f, a green-MD sperm
   cell), both immunostained with an anti-GFP antibody (cyan),
   immunostained to visualize the axoneme (yellow), and prepared for ExM.
   The images at the top depict single optical sections before computing
   [scale bar (left), 1 μm; scale bar (right), 0.5 μm], while the images
   at the bottom depict 3D computer rendering of segmented surfaces in
   the early fertilized eggs (scale bar, 2 μm). Note the narrow vesicular
   tube within the FVS that confines the axoneme (arrowheads). b, c
   Representative confocal images of the anterior regions in early
   fertilized eggs laid by control (b) and maternal rubicon knockdown (c)
   mothers. The eggs were fertilized by sperm cells expressing the
   dj-CD8-Venus transgene (cyan; immunostained with anti-GFP antibody) and
   immunostained to visualize the axoneme (yellow). DAPI stained the DNA
   (magenta). Scale bar, 20 μm. d The graphs depict staining volumes of
   the sperm plasma membrane relative to the axoneme volume in each
   fertilized egg, corresponding to the fertilized eggs in (b,c). The
   respective numbers of examined fertilized eggs (n) laid by control and
   rubicon^ShR females are 22 and 18. Two-tailed unpaired Student’s t
   test. ****P  <  0.0001. Summary of the statistics reported in the
   violin plots: the center line represents the median of the frequency
   distribution of the data. Quartiles are represented by a dashed black
   line. Each dot corresponds to the relative sperm plasma
   membrane/Axoneme volume in a single fertilized egg. e, Live confocal
   imaging of the anterior region of an early fertilized egg laid by a
   female maternally expressing the rubicon-tdTomato transgene (magenta)
   and fertilized by a sperm cell expressing the CD8-Venus transgenic
   marker of the plasma membrane (green). Shown are representative images
   from Supplementary Movie [195]10 at the indicated times (min AEL).
   Scale bar, 3 μm. Yolk autofluorescence is indicated by arrowheads. MD
   mitochondrial derivative. Source data are provided as a Source Data
   file.

   We then examined whether, similar to MD degradation, sperm plasma
   membrane breakdown is also mediated by Rubicon. Monitoring maternal
   rubicon knockdown eggs fertilized by sperm expressing the CD8-Venus
   plasma membrane marker, and stained to visualize the axoneme, detected
   a highly significant attenuation in sperm plasma membrane breakdown
   upon maternal rubicon knockdown, as compared with counterpart control
   eggs (Fig. [196]4b–d). Note that the examined early fertilized eggs
   were all at a similar stage (syncytial early fertilized eggs up to
   nuclear division cycle two).

   Given that the MD is still coated by the sperm plasma membrane after
   fertilization, the findings that the breakdown of the sperm plasma
   membrane and degradation of the MD both occur within the FVS, raise the
   hypothesis that in order to degrade the MD, the FVS first break down
   the sperm plasma membrane. To test this possibility, we used live
   imaging to monitor sperm plasma membrane breakdown kinetics after
   fertilization in eggs maternally expressing the rubicon-tdTomato
   transgene and fertilized by sperm expressing the plasma membrane
   CD8-Venus marker. As shown in Supplementary Movie [197]4 and
   representative still images in Fig. [198]4e, complete breakdown of the
   sperm plasma membrane already occurs within 8–10 min AEL, which is much
   faster than the time it takes to degrade the MD (40–60 min AEL),
   consistent with a timeline where sperm plasma membrane breakdown
   precedes MD degradation.

   Since the axoneme and the MD are intimately associated, an intriguing
   question is how the axoneme, which is also initially embedded within
   the FVS, is spared from degradation. Unexpectedly, examining ExM images
   of the axoneme in eggs maternally expressing the rubicon-tdTomato
   transgene and fertilized by the green-MD sperm, revealed that whereas
   both the MD and the axoneme are enwrapped by the FVS, they occupy
   different compartments, such that the axoneme resides in a narrow
   vesicular tube generated by the FVS (Fig. [199]4f). This finding
   suggests that the FVS actively separate between the MD and the axoneme,
   likely protecting the axoneme from sharing the same fate as the MD and
   the sperm plasma membrane.

Rubicon localizes on MVBs

   Given the unique FVS formation generated by the Rubicon vesicles, we
   asked whether these vesicles are the MVBs detected by TEM
   (Supplementary Fig. [200]1a–c)^[201]42. For this, we generated flies
   maternally co-expressing the Rubicon-eGFP and the MVB marker protein
   hCD63 fused to tdTomato (i.e., UASz-hCD63-tdTomato). Eggs laid by these
   females and fertilized by WT males displayed numerous vesicles labeled
   by either Rubicon-eGFP, hCD63-tdTomato, or both (Fig. [202]5a). Whereas
   multiple vesicles labeled only by hCD63-tdTomato and located in the egg
   periphery were not associated with the MD, 85% of the Rubicon-eGFP
   positive vesicles were also positive for hCD63-tdTomato, including
   large vesicles associated with the flagellum (Fig. [203]5a–c; note that
   the large flagellar segments can be detected merely by virtue of being
   enwrapped by Rubicon positive FVS). Closer examination of the dually
   labeled vesicles revealed that the two tagged proteins occupy different
   compartments in these vesicles: Rubicon-eGFP is primarily located at
   the limiting membrane of the MVBs, while hCD63-tdTomato resides in the
   ILV compartment (Fig. [204]5b). Interestingly, traces of hCD63-tdTomato
   are detected along the flagellar segments within the Rubicon-eGFP FVS
   (Fig. [205]5a), which is consistent with our ultrastructural studies
   suggesting that upon engagement with the flagellum, the MVBs release
   their ILVs in the vicinity of the MD (Supplementary Fig. [206]1c
   and^[207]42).

Fig. 5. Rubicon-coated MVBs express active PI3KC3.

   [208]Fig. 5
   [209]Open in a new tab

   a, b A live imaging projection of the anterior region of an egg
   maternally expressing both a UASz-rubicon-eGFP (green) and the
   UASz-hCD63-tdTomato (magenta; MVBs) transgenic constructs, and
   fertilized by a WT (non-fluorescent) sperm cell. Magnification of the
   area outlined by a dashed rectangle in (a) is shown in (b). Rubicon
   positive vesicles and FVS also display hCD63-tdTomato (arrow in a).
   Scale bars in (a) 5 μm; (b) 0.5 μm. c Most (~85%) of the Rubicon-eGFP
   positive vesicles also display the MVB marker hCD63-tdTomato. Measure
   of center is the mean percentage. Error bar indicates standard
   deviation (SD). The pulled Pearson’s Correlation Coefficient (PCC) is
   0.3402292 (P-value < 2.2e–16). The pulled Manders’ Colocalization
   Coefficient is 0.9984238. n = 4. d, e, h Representative confocal images
   of the anterior region of early fertilized eggs. d An egg maternally
   expressing both the rubicon-eGFP transgene (green) and the PtdIns(3)P
   reporter TagRFPt-2xFYVE (magenta), and fertilized by a WT
   (non-fluorescent) sperm cell. The egg contains multiple PtdIns(3)P
   positive vesicles (arrowhead), including large Rubicon-eGFP positive
   vesicles (arrows). Scale bar, 6 μm. e An egg maternally expressing the
   PtdIns(3)P reporter (magenta) and fertilized by green-MD sperm (green).
   In addition to the large PtdIns(3)P positive vesicles derived from the
   flagellum (arrowheads), flagellar PtdIns(3)P is also readily detected
   (arrows). Scale bar, 10 μm. f Representative confocal images of control
   (top) and maternal rubicon knockdown (bottom) early fertilized eggs at
   0–15 min AEL, maternally expressing the PtdIns(3)P reporter (magenta)
   and fertilized by green-MD sperm (green). Arrows indicate flagellar
   areas. Scale bar, 2 μm. g Quantification of the number of early
   fertilized eggs (corresponding to the eggs in f) that display flagellar
   PtdIns(3)P. Two-sided Fisher’s exact test. ***P = 0.0002. The number of
   examined fertilized eggs (n) laid by females of both genotypes is 23. h
   A WT egg fertilized by red-MD sperm (magenta) and stained with the
   fluorescent ROS indicator H2DCFDA (green). Scale bar, 10 μm. MD
   mitochondrial derivative. Source data are provided as a Source Data
   file.

   To test whether Rubicon might function in biogenesis and/or flagellar
   recruitment of the MVBs, we took advantage of the eGFP-hCD63 transgenic
   marker of the ILV compartment in MVBs. We monitored both the percentage
   of maternal rubicon knockdown early fertilized eggs that exhibit
   flagellar positive eGFP-hCD63, as well as the intensity of this
   fluorescent marker. Whereas ~95% of control early fertilized eggs
   display flagellar eGFP-hCD63, about 30% of the maternal
   rubicon-deficient eggs were completely devoid of flagellar eGFP-hCD63
   (Supplementary Fig. [210]5a, b). Moreover, the fluorescence intensity
   of flagellar eGFP-hCD63 that is associated with the sperm MD was
   significantly reduced in the maternal rubicon-deficient eggs, as
   compared with control eggs (Supplementary Fig. [211]5a, c). Whereas
   these findings indicate that Rubicon is required for the proper
   function of the egg MVBs in the context of PME, we cannot differentiate
   between an effect on MVB biogenesis or on the recruitment of the MVBs
   to the FVS. Given that maternal inactivation of Uvrag and Atg7 resulted
   in deformed ILV formation^[212]42, it is likely that rubicon knockdown
   might also affect MVB biogenesis in a similar manner. Collectively,
   these observations indicate that the egg contains numerous MVBs
   pre-loaded with Rubicon, which densely enwrap the sperm flagellum
   immediately after fertilization, generating extended FVS that mediate
   PME.

Rubicon positive MVBs and the FVS exhibit PI3KC3 activity

   Active Rubicon PI3KC3 facilitates sustained PtdIns(3)P production by
   Pi3k/Vps34, which is required for recruitment of Atg8/LC3 to LAPosomes
   (Fig. [213]1d). We therefore tested whether the Rubicon-positive MVBs
   are also associated with PtdIns(3)P. For this, we first maternally
   co-expressed the rubicon-eGFP transgene together with TagRFPt-2xFYVE, a
   genetically encoded reporter for PtdIns(3)P based on the
   PtdIns(3)P-binding peptide motif FYVE, fused to a red fluorescent
   protein^[214]88,[215]89. Multiple PtdIns(3)P positive vesicles were
   detected by live imaging of eggs laid by these females and fertilized
   with WT sperm, while a subset of these vesicles were also positive for
   Rubicon-eGFP (Fig. [216]5d). Note that at least some of the vesicles
   labeled by PtdIns(3)P alone may correspond to other endosomes, such as
   late endosomes that enwrap yolk granules during the cellularization
   stage^[217]90. Significantly, the presence of flagellar PtdIns(3)P is
   detected in eggs maternally expressing the PtdIns(3)P reporter and
   fertilized by green-MD sperm, indicating that the FVS PI3KC3 is active
   (Fig. [218]5e and Supplementary Movie [219]5).

   To examine whether the flagellar PtdIns(3)P is produced by the Rubicon
   PI3KC3, we first wanted to track the flagellum association kinetics of
   Rubicon and PtdIns(3)P by time-lapse confocal microscopy. However,
   already at 0 h AEL, which- as previously mentioned- is a later
   timepoint than actual fertilization, both Rubicon and PtdIns(3)P were
   associated with the sperm flagellum, precluding the ability to
   determine which of these events comes first (Supplementary
   Movie [220]6). Nevertheless, monitoring the presence of flagellar
   PtdIns(3)P in maternal rubicon knockdown early fertilized eggs,
   revealed no flagellar PtdIns(3)P in 100% of the fertilized eggs, as
   compared with equivalent controls where flagellar PtdIns(3)P was
   present in ~50% of the early fertilized eggs (Fig. [221]5f, g). We
   therefore conclude that the Rubicon PI3KC3 on the FVS is indeed the
   complex that produces the observed flagellar PtdIns(3)P.

ROS are involved in PME

   In addition to PtdIns(3)P, recruitment of the Atg8/LC3 to the maturing
   LAPosome was reported to also require ROS production, which in
   phagocytes is generated by the NOX2 multiprotein
   complex^[222]51,[223]64,[224]75,[225]77,[226]91,[227]92. WT eggs
   fertilized by red-MD sperm and labeled with the cell-permeant,
   fluorescent ROS indicator H2DCFDA^[228]93,[229]94, revealed high levels
   of ROS on the MD (Fig. [230]5h). Two NADPH oxidase (NOX) homologs, Nox
   and Duox, are present in Drosophila^[231]95. Quantitative reverse
   transcription PCR (RT-qPCR) analyses of early fertilized egg
   transcripts indicated that of the two, duox is essentially the sole
   maternally expressed gene (Supplementary Fig. [232]5d), in agreement
   with omics data in two public databases^[233]96,[234]97. Duox is also
   enriched in the egg MVB proteome (Supplementary Data [235]1).
   Accordingly, knockdown of duox, but not nox, in eggs fertilized by
   red-MD sperm, moderately attenuated PME, suggesting that Duox might be
   responsible for a portion of the ROS required for PME (Supplementary
   Fig. [236]5e–i and Supplementary Fig. [237]4f, g). In addition, since
   mitochondria are known as a main source of ROS, we also tested if ROS
   is already present in the mature sperm flagellum prior to
   fertilization. Indeed, mature sperm in the seminal vesicle, the male
   sperm storage organ, stained with H2DCFDA, revealed pronounced ROS
   expression in the sperm flagellum (Supplementary Fig. [238]5j). Taken
   together, we conclude that the two mechanistic requirements for
   recruitment of Atg8/LC3 to the LAPosome, production of PtdIns(3)P and
   ROS, are also fulfilled during the PME process.

Atg8/LC3 is recruited to the FVS and is required for PME

   In a previous study, we showed that transgenes expressing fusions of
   the key autophagic protein Atg8 with fluorescent proteins are recruited
   to the vicinity of the MD after fertilization^[239]42. However, several
   questions remain unanswered regarding the significance of this
   association. First, we asked whether endogenous Atg8a is also recruited
   to the vicinity of the MD, and if so, where it is localized and at
   which stage. To address these questions, we immunostained ~50 WT eggs
   fertilized by the green-MD sperm and collected at 0–1-hour AEL, to
   visualize the MD (anti-GFP antibody), endogenous Atg8a (anti-Atg8a
   antibody), and the axoneme (anti-polyglycylated Tubulin antibody).
   Three consecutive and distinct PME phases can be clearly detected,
   often within a single fertilized egg: (1) Intact MD segments that are
   Atg8a negative and usually still reside in the vicinity of parallel
   axonemal segments (Fig. [240]6a, I); (2) FVS that become positive for
   Atg8a, which often display strings of large Atg8a positive vesicles
   that encapsulate degrading MD fragments (Fig. [241]6a, IIa), resulting
   in MD-free axonemal segments (Fig. [242]6a, IIb); (3) MD-free axonemal
   segments displaying neither MD traces nor Atg8a positive FVS,
   indicative of completion of the PME process (Fig. [243]6a, III). Taken
   together, these observations indicate that endogenous Atg8a is
   recruited to the FVS at a relatively late PME stage, during which the
   MD is readily degraded.

Fig. 6. Atg8a recruitment to the FVS coincides with advanced MD degradation
stages.

   [244]Fig. 6
   [245]Open in a new tab

   a A representative confocal image of a WT egg fertilized by green-MD
   sperm (magenta; stained with an anti-GFP antibody) and immunostained to
   visualize the axoneme (yellow) and Atg8a (cyan). Shown is a sperm
   flagellar region exhibiting at least three defined consecutive PME
   stages: Intact MD segments that are Atg8a negative and usually still
   reside in the vicinity of parallel axonemal segments (I); Atg8a
   positive FVS and derived strings of large vesicles that encapsulate
   degrading MD fragments (IIa) and hence result in MD-free axonemal
   segments (IIb); MD-free axonemal segments displaying neither MD traces
   nor Atg8a positive FVS, indicative of completion of the PME process
   (III). Scale bar, 4 μm. b, c Atg8 is required for efficient MD
   degradation. A live imaging assay performed and presented as in
   Fig. [246]1e. Early fertilized eggs expressing the maternal driver
   alone served as control (b, upper panel). Scale bar, 100 μm. c
   Quantifications of normalized red-MD fluorescence intensities
   [arbitrary units (a.u.)] in early fertilized eggs represented in (b),
   indicating that although single maternal knockdowns of atg8a and atg8b
   have no effect on PME kinetics (Supplementary Fig. [247]6b–e), double
   maternal knockdown fertilized eggs exhibit significant PME attenuation.
   Error bars indicate SEM. The respective numbers of examined fertilized
   eggs (n) laid by control and atg8a^ShR; atg8b^ShR females are 51 and
   76. ****P <  0.0001. Two-way repeated measures ANOVA, followed by
   Šídák’s multiple comparisons test. MD, mitochondrial derivative. d–g
   Live confocal imaging of the anterior region of an early fertilized egg
   laid by a female maternally expressing both the UASz-rubicon-eGFP
   (green) and UASp-mCherry-atg8a (magenta) transgenic constructs. Shown
   are representative images from Supplementary Movie [248]7 at the
   indicated times (min AEL). Note the 6–12 min lag between the formation
   of Rubicon-eGFP positive FVS and the subsequent recruitment of
   mCherry-Atg8a. Scale bar, 10 μm. h–k Enlargement of the areas confined
   by dashed rectangles in (d–g), presenting the same flagellar region at
   different timepoints AEL. Scale bars, 4 μm. Source data are provided as
   a Source Data file.

   Given the pronounced association of Atg8a with the strings of large
   vesicles that contains degrading MD fragments, we next asked whether
   Atg8a is required for efficient PME. However, maternal knockdown of
   atg8a, the major fly atg8 gene, had no significant effect on PME
   kinetics (Supplementary Fig. [249]6b, c and Supplementary
   Fig. [250]4h). Drosophila has another closely related but distinct atg8
   gene, atg8b, although its expression is mainly restricted to the testis
   and is barely detected in other tissues^[251]98–[252]100. Accordingly,
   maternal knockdown of atg8b also had no effect on PME kinetics
   (Supplementary Fig. [253]6d, e).

   It was previously reported that atg8b was elevated in Drosophila eye
   imaginal discs homozygous for a hypomorphic atg8a mutant
   allele^[254]101. Likewise, RT-qPCR analysis of early fertilized eggs
   also revealed a significant increase in the level of atg8b transcripts
   upon maternal knockdown of atg8a, which was abrogated in maternal
   double knockdown of both atg8a and atg8b (Supplementary Fig. [255]4i,
   j). Importantly, monitoring early fertilized eggs with the double
   maternal knockdowns revealed a significant attenuation in PME,
   demonstrating that Atg8 is required for an efficient PME process
   (Fig. [256]6b, c).

   To evaluate the time elapsed between coating of the flagellum by the
   Rubicon MVBs and recruitment of Atg8a to the FVS, we maternally
   expressed rubicon-eGFP and mCherry-atg8a transgenes, and used live
   imaging to monitor their relative spatiotemporal patterns of
   expression. Large Rubicon-eGFP positive FVS that are mCherry-Atg8a
   negative were readily detected at time 0 AEL, while recruitment of
   mCherry-Atg8a to the FVS occurred 6–12 min later (Fig. [257]6d–k and
   Supplementary Movie [258]7). Altogether, these findings are consistent
   with the LAP pathway sequence of events, in which loading of the
   Rubicon PI3KC3 on the FVS precedes recruitment of Atg8/LC3.

Atg8/LC3 recruitment to the FVS requires Rubicon

   To examine whether early loading of Rubicon PI3KC3 is also required for
   Atg8a association with the FVS, we next monitored the staining volume
   of Atg8a in 0–1-hour AEL fertilized eggs maternally expressing an shRNA
   transgene against rubicon. Maternal expression of an atg7 shRNA
   transgene served as positive control, whereas down regulation of atg8a
   was used to validate the specificity of the anti-Atg8a antibody
   (Supplementary Fig. [259]4h, k). To avoid possible unrelated Atg8a
   positive vesicles, such as autophagosomes, we restricted the
   measurements to an area surrounding the axoneme. In addition, to ensure
   that the developmental stages of the early fertilized eggs were highly
   synchronized, we monitored syncytial early fertilized eggs up to
   nuclear division cycle six. The Atg8a staining volume in a 0–10 µm
   radius around the axoneme, corresponding to the FVS and its immediate
   surrounding, was significantly reduced upon down regulation of rubicon,
   atg7, and atg8a, as compared with counterpart WT eggs, indicating that
   Atg8a recruitment to the FVS requires both Rubicon and Atg7
   (Fig. [260]7).

Fig. 7. Atg8 recruitment to FVS requires Rubicon.

   [261]Fig. 7
   [262]Open in a new tab

   a–d Representative confocal images of the anterior regions in early
   fertilized eggs laid by control (a) and maternal atg8a (b), atg7 (c),
   and rubicon (d) knockdown mothers. The eggs were fertilized by WT sperm
   cells, and immunostained to visualize Atg8a (cyan) and the axoneme
   (yellow). The 0–1-hour AEL fertilized eggs were further staged by the
   number of nuclei, as revealed by DAPI staining of the DNA (magenta).
   Scale bar, 20 μm. e The graphs depict staining volumes of Atg8a in
   early fertilized eggs corresponding to the fertilized eggs in (a–d).
   Calculations were performed in intervals of 1 μm in 10 μm radius area
   around the axoneme. The respective numbers of scored early fertilized
   eggs (n) laid by females of the control, atg8a^ShR, atg7^ShR, and
   rubicon^ShR, are 20, 12, 11, and 15. *P  <  0.05, **P <  0.01, and
   ***P <  0.001. P values are as follows: control vs. atg8a^ShR,
   P = 0.0003, 0.0004, 0.0009, 0.0019, 0.0019, 0.0063, 0.0029, 0.0047, and
   0.0035, for the distances 1, 2, 4, 5, 6, 7, 8, 9, and 10 μm^3 from
   axoneme, respectively; control vs. atg7^ShR: P = 0.0013, 0.0013,
   0.0013, 0.0042, 0.0059, 0.0138, 0.0076, 0.0111, and 0.0098, for the
   distances 1, 2, 4, 5, 6, 7, 8, 9, and 10 μm^3 from axoneme,
   respectively; control vs. rubicon^ShR: P = 0.0154, 0.0236, 0.0128,
   0.0348, 0.0379, 0.0371, 0.0294, 0.025, and 0.0203, for the distances 1,
   2, 4, 5, 6, 7, 8, 9, and 10 μm^3 from axoneme, respectively. Two-way
   repeated measures ANOVA, followed by Holm-Šídák’s multiple comparisons
   test. Error bars indicate SEM. Source data are provided as a Source
   Data file.

   The observed reduced level of flagellar Atg8a in rubicon knockdown
   early fertilized eggs might be the consequence of inefficient Atg8a
   recruitment to the FVS and/or delayed recruitment. To determine which
   of the possibilities is correct, we used live imaging to monitor WT and
   rubicon knockdown eggs maternally expressing the mCherry-atg8a
   transgene and fertilized by the green-MD sperm. In an otherwise WT
   background, most of the flagellar segments recruit Atg8a and contain
   degraded MD within 10–20 min AEL, such that only a very few short MD
   segments are still detected at 50 min AEL, by virtue of their
   mCherry-Atg8a positive FVS (Supplementary Fig. [263]7a and
   Supplementary Movie [264]8). In contrast, eggs maternally expressing
   the rubicon shRNA transgene display a delayed appearance of
   mCherry-Atg8a on FVS (40–50 min AEL), which coated only a few short
   flagellar segments (Supplementary Fig. [265]7b and Supplementary
   Movie [266]9). These findings indicate that Rubicon is required for
   both fast and efficient accumulation of Atg8a on FVS.

FVS-derived vesicles with MD fragments fuse with lysosomes

   In the final stage of LAP, Atg8/LC3 is suggested to promote fusion of
   the LAPosome with lysosomes to accelerate degradation of the
   phagolysosomal cargo (Fig. [267]1d)^[268]49,[269]52,[270]102. To
   examine whether during PME lysosomes fuse with the FVS to promote
   efficient degradation of the MD, we monitored live eggs ubiquitously
   expressing the transgenic lysosomal marker GFP-LAMP1^[271]103,[272]104,
   and fertilized by the red-MD sperm. As shown in Supplementary
   Movie [273]10 and in Fig. [274]8a, multiple lysosomes are detected at
   the vicinity of the degrading MD, some of which appear to specifically
   co-localize with MD fragments. To examine whether the MD fragments
   correspond to the observed degrading MD fragments within the string of
   large Atg8a positive vesicles derived from the FVS (Fig. [275]6a), we
   imaged GFP-LAMP1-expressing eggs fertilized by WT sperm and stained to
   visualize LAMP1 (anti-GFP), Atg8a, and the axoneme. The lysosomal
   marker, GFP-LAMP1, was readily detected within the Atg8a positive large
   vesicles, indicating that degradation of the MD fragments within these
   vesicles is mediated by lysosomal hydrolases (Fig. [276]8b, c).

Fig. 8. Atg8 positive strings of vesicles derived from the FVS and contain MD
fragments fuse with lysosomes.

   [277]Fig. 8
   [278]Open in a new tab

   a Live confocal imaging of a sperm flagellar region in an egg
   ubiquitously expressing a transgenic construct that marks the lysosomes
   (magenta; tub-GFP-LAMP1), and fertilized by a red-MD sperm cell
   (green). Scale bar, 2 μm. b A sperm flagellar region in an egg
   ubiquitously expressing the lysosomal transgenic marker, GFP-LAMP1
   (magenta), fertilized by a WT (non-fluorescent) sperm cell, and
   immunostained to visualize Atg8a (cyan) and the axoneme (yellow). Note
   the budding of the string of Atg8 positive large vesicles from the FVS
   (i.e., Atg8a positive regions that are not associated with the
   axoneme), and the association of GFP-LAMP1 with these large vesicles,
   which correspond to the vesicles that contain degrading MD fragments (a
   and Fig. [279]6a). Scale bar, 3 μm. c Enlargement of the area confined
   by a dashed rectangle in (b). Scale bar, 2 μm. d An integrated model of
   PME by a LAP-like pathway in Drosophila. An illustration of a sperm
   flagellum inside the anterior part of an early fertilized egg. Relevant
   flagellar and LAP pathway components are indicated by different signs
   and colors as defined on the right-hand side. Seven PME pathway steps
   are indicated by Roman numerals. Egg-derived Rubicon positive MVBs
   engage the sperm flagellum plasma membrane in an asynchronous manner
   immediately after fertilization (I). The MVBs densely coat the sperm
   flagellum, forming extended vesicular sheaths, within which the sperm
   plasma membrane rapidly breaks down (II). The MD is exposed to the
   degradative luminal area of the FVS that express the LAP-specific
   Rubicon PI3KC3 (III). Active Rubicon PI3KC3 generates PtdIns(3)P (IV),
   which presumably with the help of ROS (produced by the egg and the MD),
   promotes the recruitment of Atg8/LC3 to the FVS (V). The Atg8/LC3
   positive FVS bud off strings of large vesicles that contain degrading
   MD fragments (VI). Lysosomes fuse with the strings of large vesicles,
   promoting efficient degradation of the MD fragments (VII).

The Atg5-Atg12-Atg16 complex is not required for PME

   In conventional autophagy, Atg7 serves as an E1 ligase-like enzyme that
   activates two different ubiquitin-like proteins, Atg12 and
   Atg8/LC3^[280]105,[281]106. Atg12 then undergoes conjugation to Atg5,
   and the Atg12-Atg5 complex interacts with Atg16L1 to form a large
   multimeric complex that facilitates efficient Atg8/LC3
   lipidation^[282]107. Likewise, it has been shown that at least in some
   LAP paradigms, the Atg12-Atg5-Atg16L1 complex is required for Atg8/LC3
   conjugation to the LAPosome membrane^[283]49,[284]108–[285]110.
   However, whereas maternal mutant and knockdown of atg7 significantly
   attenuates PME (Supplementary Fig. [286]6f, g)^[287]42, maternal
   knockdowns of atg12, atg5, and atg16 had almost no effect on PME
   kinetics, indicating that this complex is dispensable for the LAP-like
   pathway underlying PME (Supplementary Fig. [288]4l–n and Supplementary
   Fig. [289]6h–m). It is noteworthy that similar results were also
   obtained with Atg5 knockout early mouse embryos^[290]21. Given that
   Atg8/LC3 also decorates the mouse sperm mitochondria immediately after
   fertilization^[291]21, it is plausible that a similar non-conventional
   autophagy pathway might also mediate PME in mammals.

Discussion

   In this work, we uncovered the pathway by which the sperm MD is
   eliminated after fertilization in Drosophila. We show that special
   MVBs, which are abundant in the early fertilized egg, engage the sperm
   flagellum immediately after fertilization, while triggering a LAP-like
   pathway to specifically target the MD for degradation. This highly
   efficient process takes about 20–40 min to complete, as revealed by the
   disappearance of transgenic fluorescent signals that label the MD, with
   an average MD degradation rate of 100 nm/min. According to the sequence
   of events we uncovered, numerous egg MVBs that are loaded with the
   Rubicon-containing PI3K complex (PI3KC3) on their limiting membrane
   approach and densely coat large segments of the penetrating sperm
   flagellum in an asynchronous manner (Fig. [292]8d, I). These MVBs wrap
   around the sperm flagellum plasma membrane, forming FVS within which
   the sperm plasma membrane breaks down and the MD is degraded
   (Fig. [293]8d, II–VI). At the molecular level, activated LAP-specific
   PI3KC3 generates PtdIns(3)P and recruits Atg8 to the FVS, presumably
   with the assistance of ROS generated by Duox in the egg and/or the MD
   itself (Fig. [294]8d, IV, V). Consequently, lysosomes fuse with strings
   of large Atg8 positive vesicles derived from the FVS and contain
   degraded MD pieces and debris, leaving behind the axoneme
   (Fig. [295]8d, VI, VII). Ultimately, this process assures inheritance
   of the mitochondria from the maternal lineage only.

   An interesting question is how the MVBs specifically recognize and
   mediate degradation of the sperm MD, but not the axoneme or the
   maternal mitochondria. Since we show that the egg MVBs first engage the
   sperm flagellum plasma membrane, it is likely that they recognize
   specific lipid/s or protein/s placed on the exterior side of the sperm
   plasma membrane. Breakdown of the sperm plasma membrane exposes the two
   main flagellar organelles, the MD and the axoneme, to the luminal
   contents of the FVS. This leads to the degradation of the MD, but the
   axoneme is spared from a similar fate. In this study we also observed
   that the two organelles reside in separate compartments within the FVS,
   suggesting the existence of active mechanisms that direct the MD, but
   not the axoneme, to a degradative compartment within the FVS.
   Interestingly, cargo ubiquitination is a major MVB sorting
   mechanism^[296]111–[297]113, and we previously demonstrated that the
   MD, but not the axoneme, is specifically decorated with lysine
   63-linked ubiquitin chains after fertilization^[298]42, which may
   explain the different fates of these two organelles.

   Whereas LAP is known to target extracellular pathogenic microbes by
   phagocytosis, targeting of the MD occurs within the egg ooplasm,
   following sperm entry. Nevertheless, these two processes are not only
   molecularly reminiscent of each other but also anatomically similar. In
   LAP, a single membrane LAPosomal vesicle encapsulates the cargo by
   inward invagination of the plasma membrane, while during PME, single
   membrane MVBs form extended FVS that encapsulate large flagellar
   segments. Interestingly, in addition to LAP, other pathways, such as
   LC3-associated endocytosis (LANDO) and LC3-dependent extracellular
   vesicle loading and secretion (LDELS), were reported to involve
   conjugation of Atg8/LC3 to the limiting membrane of endosomes and MVBs,
   respectively, indicating that functional convergence of the endocytosis
   and autophagy pathways may be more common than has been previously
   recognized. Furthermore, similar to Drosophila PME, induction of LDELS
   does not require surface receptor activation, and LANDO requires
   Rubicon^[299]63,[300]110,[301]114–[302]116.

   A generally held belief is that elimination of the paternal
   mitochondria is important to prevent mtDNA heteroplasmy (i.e., a state
   of having more than one type of mtDNA in a cell), partially because
   sperm mtDNA was reported to have a higher mutation rate than egg
   mtDNA^[303]117. Furthermore, deliberate major mtDNA heteroplasmy in
   mice was reported to lead to reduced animal activity, increased
   anxiety, reduced feeding, and impaired spatial learning^[304]118.
   However, prevention of mtDNA heteroplasmy may not be the entire story,
   as the heteroplasmic state has been shown to be very unstable, and
   documented negative effects of paternal mtDNA leakage on the well-being
   of the organism are rare^[305]8,[306]17,[307]118–[308]129. Furthermore,
   several reports suggest that in some cases, including in human and
   Drosophila, the sperm mtDNA is degraded much before the observed
   degradation of the vacuolar paternal
   mitochondria^[309]18–[310]22,[311]24, implying that the latter
   degradation is a second line of defense against paternal mtDNA
   transmission, or that other paternal mitochondrial factors are
   deleterious to the organism. An alternative theory is that, similar to
   the role of macroautophagy in providing nutrients during starvation,
   recycling of the paternal mitochondria may support proper embryonic
   development, by providing developmental cues or important substances to
   the fertilized egg, such as iron, copper, folates,
   etc^[312]130–[313]132. Our findings that the egg triggers a conserved
   cellular defense pathway against invading microbes to target the MD for
   degradation, may support the idea that the paternal mitochondria
   might deliver deleterious factors to the embryos, as they imply that
   the egg regards the paternal mitochondria as a potentially dangerous
   trespasser, reminiscent of the manner by which other cells react
   against intrusive bacteria.

   How Atg8 is attached to the FVS is unclear. Although the
   Atg5-Atg12-Atg16 complex is required for conventional
   autophagy^[314]107, and is also reported to participate in some LAP
   paradigms^[315]49,[316]108–[317]110, maternal knockdown of these genes
   had no effect on PME. In contrast, Atg7 and Rubicon are both involved
   in recruitment of Atg8 to the FVS. Interestingly, double knockdown of
   atg8a and atg8b significantly attenuates PME. These findings suggest
   that the Atg5-Atg12-Atg16 complex is likely dispensable for Atg8
   recruitment to the FVS. It is noteworthy that other mechanisms of Atg8
   recruitment and roles, in a manner independent of the Atg5-Atg12-Atg16
   complex, have also been documented^[318]100,[319]133–[320]138.

   Paternal mitochondrial elimination and maternal mitochondrial
   inheritance are highly ubiquitous across the animal kingdom. Scattered
   evidence from different research groups suggests that at least some
   elements of the autophagy pathway are expressed on the flagellum of
   mammalian sperm, either before and/or immediately after
   fertilization^[321]21,[322]29–[323]31. Similar to the finding in the
   current study that the Atg12-Atg5-Atg16 complex is dispensable for PME
   in Drosophila, Atg5-deficient early mouse embryos also displayed normal
   PME^[324]21. Evidence from electron microscopy studies showed the
   presence of MVBs in the vicinity of the paternal mitochondria in some
   mammalian early embryos^[325]27,[326]65. Collectively, it is attractive
   to propose that a LAP-like pathway similar to the pathway underlying
   PME in Drosophila might also mediate the process in mammals and perhaps
   also in other organisms that produce flagellated sperm.

   The importance of understanding the mechanisms underlying PME is not
   only of academic interest, as the use of medically assisted
   reproduction (MAR) treatments is constantly increasing, leading to
   diversion of research efforts from molecular studies to improvement of
   MAR technologies. Consequently, much less attention has been given to
   possible negative impacts of some of these treatments, which may
   involve paternal mitochondrial leakage, on embryo development,
   implantation, and clinical and obstetric outcomes^[327]139–[328]141.
   Furthermore, the novel reproductive technologies known as mitochondrial
   donation or mitochondrial replacement therapy (MRT), raise challenging
   ethical and safety issues, as only a few successful MRT case reports
   are available. In addition, since we still lack long-term studies of
   MRT outcomes, the potential risks of both major mtDNA heteroplasmy and
   the mixing of genetic materials (nuclear and mitochondrial genomes)
   from three parents are under much debate^[329]142–[330]146. Therefore,
   delineating the molecular pathway underlying PME and the (possible
   negative) effects of paternal mitochondrial persistence on the
   developing embryo, will allow for a deeper understanding of this highly
   conserved phenomenon for embryo development, and may ultimately lead to
   the development and/or modification of MAR technologies that guard
   against these effects.

Methods

Fly strains used in this study

   Fly stocks were maintained on a standard yeast/molasses medium at
   25 °C. Unless indicated otherwise, yellow white (yw) flies were used as
   WT control. Fly strains, including mutant and transgenic flies used in
   this study are listed in Supplementary Table [331]1.

Generation of the rubicon mutant alleles

   The rubicon^Δ8, rubicon^Δ10, rubicon^Δ13 and rubicon^Δ196 mutant
   alleles were generated using CRISPR/Cas9-mediated mutagenesis. Briefly,
   nos-Cas9 expressing flies were crossed to flies ubiquitously expressing
   a single guide RNA (gRNA) against rubicon (BDSC #81781). F1 females
   carrying both the nos-Cas9 and the rubicon gRNA, and hence subjected to
   potential germline mutagenesis in the rubicon gene, were crossed with
   males carrying an X chromosome balancer (FM7), as the rubicon gene is
   located on the X chromosome. F2 females were then individually crossed
   again with FM7 males to establish stable fly lines with potential
   mutations in rubicon. Genomic DNA was extracted from each line, the
   rubicon gene was PCR amplified using forward (F) primer:
   CCGAGACAGTGCGTCAATG and reverse (R) primer: CTTCCTCGCTGGGATAGTACA, and
   sequenced to identify rubicon mutant lines.

Generation of DNA constructs and transgenic lines

dj-(MTS)tdTomato (red-MD)

   The tdTomato coding sequence (CDS) was PCR amplified from the
   pRSETB-tdTomato vector^[332]66 using F primer:
   CATTCGTTGGGGGATCCACCGGTCGCCACCATGGTGAGCAAGGGCGAGGAGG and R primer:
   ATCTTGTTGTTTCGCAGGCGATTGCGGCCGTTACTTGTACAGCTCGTCCATGC. Applying the
   restriction free technique (RF)^[333]147, the DsRed CDS in the
   dj-(MTS)DsRed plasmid^[334]42 was replaced with the tdTomato CDS. Note
   that for expression during late-stage spermatogenesis, the
   (MTS)tdTomato is placed under the regulatory elements of the dj gene
   (i.e., promoter and 5′ UTR)^[335]80,[336]148, as well as the 3′ UTR of
   the male germ cell-specific gene cyt-c-d^[337]149 [this vector is
   termed dj(P + 5′ UTR)//cyt-c-d(3′ UTR)]. Embryo injection to introduce
   the transgene into the attP2 site was carried out by BestGene Inc.

dj-CD8-PARP^(DEVG)-Venus (sperm plasma membrane reporter)

   This line was generated using restriction-ligation cloning. The
   CD8::parp^DEVG::Venus (UAS-CP^GV) transgene^[338]150 was PCR amplified
   from the pUASt-CD8:PARP:Venus plasmid^[339]151 using F primer:
   CGTAGATCTATGGCCTCACCGTTGACCCGCTTT and R primer:
   ATCCGCGGCCGCCTATTTGTACAATTCATCCATAC, and subcloned into the dj(P + 5′
   UTR)//cyt-c-d(3′ UTR) vector. Embryo injection to introduce the
   transgene into the 3^nd-chromosome through P element-mediated
   transformation was carried out by Genetic Services, Inc.

UASp-eGFP-hCD63

   The eGFP-hCD63 CDS was PCR amplified from a pUASt-eGFP-hCD63 plasmid
   (kindly provided by Suzanne Eaton, the Max Plank Institute of Molecular
   Cell Biology and Genetics, Germany) using F primer:
   CCGCGCGGCCGCATGGTGAGCAAGGGCGAGGAGCTG and R primer:
   GCGGACTAGTCTACATCACCTCGTAGCCAC. The eGFP-hCD63 CDS was subcloned into
   the pUASp2 vector^[340]68 using the restriction enzymes NotI and SpeI
   for the insert, and NotI and XbaI for the vector. Embryo injection to
   introduce the transgene into the 2^nd-chromosome through P
   element-mediated transformation was carried out by BestGene Inc.

UASz-rubicon-eGFP

   This line was generated using the In-fusion cloning technique (Takara)
   in two steps as follows: First, the eGFP DNA sequence was PCR amplified
   using F primer: GTACCGCCTCTCTAGGTGAGCAAGGGCGAGGAG and R primer:
   GAATTCACACTCTAGTTACTTGTACAGCTCGTCCATGCC. This amplicon was inserted
   into the pUASz1.1 plasmid^[341]69, obtained from the Drosophila
   Genomics Resource Center (DGRC; stock #1433), following digestion of
   the plasmid with XbaI to obtain a pUASz-eGFP vector. Next, the
   full-length rubicon CDS was PCR amplified from the cDNA clone RH61467
   (DGRC, Stock #11152) using F primer: CAAAGGATCCCTCGAATGACCACGCCCCCG and
   R primer: AGGCGGTACCCTCGAGCTGGCACGGCTTTGA. The amplicon was inserted
   into the pUASz-eGFP vector following digestion with XhoI to generate
   the pUASz-Rubicon-eGFP vector. Embryo injections to introduce the
   transgene into the attP2 and attP40 sites were carried out by FlyORF.

UASz-hCD63-tdTomato

   The tdTomato gene was first isolated from a plasmid by digestion with
   BamHI and ligated into BamHI linearized pUASz1.1 plasmid to obtain the
   pUASz-tdTomato vector. hCD63 CDS was PCR amplified from the
   UASp-eGFP-hCD63 vector using F primer:
   CAAAGGATCCACCGGATGGCGGTGGAAGGAGGAAT and R primer:
   CATGGTGGCGACCGGCATCACCTCGTAGCCACTTCTGA. The amplicon was then inserted
   into the pUASz-tdTomato vector using the In-fusion cloning technique
   (Takara) following digestion with AgeI. Embryo injection to introduce
   the transgene into the attP2 site was carried out by FlyORF.

Tagging the endogenous rubicon gene with tdTomato

   A single gRNA was designed to target Cas9-mediated excision around the
   stop codon of the rubicon gene (TTCTTTGATCGGATGTTAGC in the pCFD5
   plasmid). For homology-directed repair, we generated a donor plasmid in
   several steps as follows: We first inserted the tdTomato gene into the
   pHD-DsRed-attP vector, which contains a loxP flanked 3XP3-DsRed
   cassette that produces eye-specific expression of the DsRed^[342]152.
   Then, using In-fusion cloning (Takara), we subcloned into the
   pHD-tdTomato-DsRed-attP vector the two homology arms (HAs) which
   contain sequences flanking the Cas9 cleavage site. The HAs were PCR
   amplified from yw Drosophila genomic DNA using F primer:
   CTGGGCCTTTCGCCCCCAGTCTGACCAGAGGCACC and R primer:
   CCCCATAATTGGCCCTTGCTGGCACGGCTTTGAAGG, for the 5′ HA, and F primer:
   ATAGAAGAGCACTAGCATCCGATCAAAGAAAATCGAAGGG and R primer:
   GGAGATCTTTACTAGACGCGTTCGGCAAAATACC, for the 3′ HA. The
   pHD-tdTomato-DsRed-attP vector was digested with SmaI to first insert
   the 5′ HA and subsequently with SpeI to insert the 3′ HA, obtaining the
   pHD-5′HA-tdTomato-3′HA-DsRed-attP donor plasmid. Embryo injections to
   obtain transgenic lines were carried out by FlyORF. F1 progeny was
   screened for insertion events using the eye DsRed expression. Positive
   lines were validated and then crossed to nos-Cre flies to remove the
   3XP3-DsRed cassette.

MVB purification from Drosophila eggs

   To isolate the egg-derived MVBs, we devised a procedure that is based
   on two protocols, one for mitochondria isolation from Drosophila early
   fertilized eggs and the second for isolation of exosomes from cultured
   cells^[343]153,[344]154. To adjust the procedure, MVB isolation trials
   were first performed on small-scale eGFP-hCD63 expressing early
   fertilized eggs (i.e., early fertilized eggs laid by females carrying
   the maternal driver and the UASp-eGFP-hCD63 transgene and fertilized by
   WT males). Once adjusted, the procedure was then scaled up to produce a
   large amount of protein that is needed for MS analysis. To avoid slow
   fertilized egg collection due to repeated crossings, the large-scale
   MVB isolation was performed with WT (yw) early fertilized eggs.
   Collection, preparation, and MVB extraction from the eGFP-hCD63
   expressing early fertilized eggs was performed as follows: Flies were
   placed in a population cage, allowing to lay eggs on juice agar plates
   with a dollop of thick yeast paste for 1 h. Early fertilized eggs were
   then collected in mesh baskets, dechorionated in 6% bleach, washed with
   water, shock frozen in liquid nitrogen, and stored at −80 °C until the
   day of procedure. All further steps were performed on ice and using a
   cooled centrifuge. 60 mg of early fertilized eggs were thawed on ice
   and transferred to a Dounce homogenizer containing 6-times volume of
   homogenization buffer (HB; 0.25 M sucrose, 1 mM EDTA, 0.03 M Tris pH
   7.4, supplemented with protease inhibitor cocktail (Sigma-Aldrich,
   P8340)). The homogenate was centrifuged three times at 500 x g for
   15 min to pellet cell debris. A fraction of the supernatant was removed
   and served as cell lysate control, while the remaining supernatant was
   centrifuged at 12,000 x g for 20 min. The pellet was then resuspended
   in HB and layered onto a 5-50% OptiPrep™ Density Gradient Medium
   (Sigma-Aldrich, D1556), and ultracentrifuged at 50,000 x g for 3 h in a
   SW41Ti rotor (Beckman). 12 fractions of 1 ml each were manually
   separated, diluted 1:4 in OptiPrep™ buffer (OB, 0.07 M EDTA, 0.03 M
   Tris pH 7.4) and centrifuged at 30,000 x g for 15 min. Fractions pellet
   was resuspended in SDS sample buffer and analyzed by Western blotting
   (WB) using anti-GFP antibody (Abcam, ab 290, 1:1,000). GFP-positive
   fractions (5 and 6) were subjected to TEM analysis to assess the level
   of MVB purification. The scaled-up procedure was applied to 2 gm of yw
   early fertilized eggs, and the homogenization buffer volume was
   accordingly adjusted (all other steps were not modified). Fractions 5
   and 6 were combined and subjected to proteomic analysis.

Transmission electron microscopy

   Electron micrographs of early fertilized eggs shown in Supplementary
   Fig. [345]1a–c, were generated essentially as described in ref.
   ^[346]42.

   Electron micrographs of isolated MVBs were prepared as follows. Samples
   were fixed in a solution containing 4% paraformaldehyde (Electron
   Microscopy Sciences, EMS) and 2% glutaraldehyde (EMS) in 0.1 M
   cacodylate buffer containing 5 mM CaCl2 (pH 7.4). Fixed samples were
   drawn into cellulose capillary tubes with an inner diameter of 200
   µm^[347]155. Samples were postfixed in 1% osmium tetroxide (EMS)
   supplemented with 0.5% potassium hexacyanoferrate tryhidrate (BDH
   chemicals) and potassium dichromate (BDH chemicals) in 0.1 M cacodylate
   for 1 h, stained with 2% uranyl acetate (EMS) in double distilled water
   for one hour, dehydrated in graded ethanol solutions and embedded in
   epoxy resin (Agar scientific Ltd.). Ultrathin sections (70 nm) were
   obtained with a Leica EMUC7 ultramicrotome. Sections were transferred
   to 200 mesh copper transmission electron microscopy grids (SPI). Grids
   were stained with lead citrate (Merk) and examined with a Tecnai T12
   transmission electron microscope (Thermo Fisher Scientific). Digital
   electron micrographs were acquired with a bottom-mounted TVIPS
   TemCam-XF416 4k × 4k CMOS camera.

Mass spectrometry

   MVB-containing fractions and cell lysate control were resuspended in
   lysis buffer (5% SDS, 50 mM Tris pH 7.4) and centrifuged at 14,000 x g
   for 5 min. Pellets were discarded, and the supernatant was collected.
   Samples were reduced in 6 mM dithiothreitol and alkylated with 12 mM
   iodoacetamide in the dark. Each sample was loaded onto S-Trap
   minicolumns (Protifi, USA) according to the manufacturer’s
   instructions. In brief, after loading, samples were washed with a
   solution of 90% methanol and 10% 50 mM Triethylammonium bicarbonate.
   Samples were then digested with trypsin (Promega) at 1:50
   trypsin/protein ratio for 90 min at 47 °C. The digested peptides were
   eluted using 50 mM ammonium bicarbonate; trypsin was added to this
   fraction and incubated overnight at 37 °C. Two more elutions were made
   using 0.2% formic acid and 0.2% formic acid in 50% acetonitrile. The
   three elutions were pooled together and vacuum-centrifuged to dry.

   UPLC/MS grade solvents were used for all chromatographic steps. Each
   sample was loaded using splitless nanoUltra Performance Liquid
   Chromatography (10k psi nanoAcquity; Waters, Milford, MA, USA). The
   mobile phase was: A) H2O + 0.1% formic acid and B) acetonitrile + 0.1%
   formic acid. The samples were desalted online using a reversed-phase
   Symmetry C18 trapping column (180 µm internal diameter, 20 mm length,
   5 µm particle size; Waters). The peptides were then separated using a
   T3 HSS nano-column (75 µm internal diameter, 250 mm length, 1.8 µm
   particle size; Waters) at 0.35 µL/min. Peptides were eluted from the
   column into the mass spectrometer using the following buffer B (99.9%
   Acetonitrile, 0.1% formic acid) gradient in three steps: (1) 4% to 25%
   of buffer B for 150 min, (2) 25% to 90% of buffer B for 5 min, (III)
   5 min incubation in 90% buffer B, and then back to initial conditions.
   Samples were run in random order.

   A nanoESI emitter (20 µm tip, Fossilion Tech, Madrid, Spain) was used
   on a FlexIon source (Thermo Fisher), mounted on a Tribrid orbitrap mass
   spectrometer (Fusion Lumos, Thermo Fisher). Data were acquired in
   data-dependent acquisition (DDA) mode, using a Top Speed method for
   3 sec. MS1 resolution was set to 120,000@200 m/z, mass range of
   380–1650 m/z, AGC at Standard and maximum injection time was set to 50
   msec. Precursors selected for MS2 were limited to charge states of 2–8
   with a minimum intensity of 50000. MS2 resolution was set to
   15,000@200 m/z, quadrupole isolation 1 m/z, first mass at 130 m/z, AGC
   set at Standard. Dynamic exclusion of 30 sec with +/−10 ppm window and
   maximum injection time set to 100 ms. Ions were fragmented using HCD at
   30 NCE.

   The data analysis was performed using Byonic (v3.3.11) search engine
   (Protein Metrics)^[348]156 against the Drosophila melanogaster proteome
   database (UniProt Nov 2018). Data were searched against specific
   C-terminal cleavage at K and R, allowing one missed cleavage at a
   tolerance of 10 ppm in MS1 and 20 ppm in MS2, using HCD settings for
   precursors up to 5500 Da. Modifications allowed were fixed
   Carbamidomethylation on C and variable modifications on the following
   amino acids: oxidation on M, deamidation on NQ and protein N-terminal
   acetylation. Data was filtered for 1% FDR. Quantification was performed
   by FlashLFQ^[349]157 without match between runs or normalization since
   the samples were very different from one another. Protein annotations
   obtained from MS analysis were scored by the number of peptide
   sequences that are unique to a protein group (unique peptides value),
   and the number of peptide spectrum matches (PSM value), which is the
   total number of identified peptide spectra matched for the protein. Low
   abundance proteins (PSM < 10) were disregarded, and Log[2] fold change
   (log[2]FC) ratios were calculated to obtain a list of MVB-enriched
   proteins (log[2]FC > 1). Pathway enrichment analysis was performed
   using Flymine^[350]158.

Western blotting

   Samples were run in SDS-PAGE and transferred to nitroglycerin membrane.
   After blocking in 5% milk solution (dry milk in PBTw), the membrane was
   incubated with anti-GFP (Abcam, ab290 1:1000), followed by incubation
   with anti-rabbit HRP secondary antibody (1:10,000) for 1 h at room
   temperature (Jackson Immuno-Research). After washes in PBTw, images
   were obtained using the ImageQuant LAS 4000 mini.

Live imaging

MD elimination kinetic assay

   Female flies of the appropriate genotype were crossed with males
   producing red-MD sperm in population cages, and were allowed to lay
   eggs on juice agar plates with a dollop of thick yeast paste for
   15 min. Early fertilized eggs were collected in mesh baskets,
   dechorionated in bleach and washed with water. Early fertilized eggs
   were then aligned on an agar pad, picked up on an embryo glue
   (Permanent double-sided tape, Scotch®, 665)-coated coverslip and
   covered with halocarbon oil 27 and 700 (1:1 mixture, Sigma-Aldrich
   H8773 and H8898, respectively). Early fertilized eggs were then imaged
   under a Nikon Eclipse fluorescence microscope. 2D images were acquired
   for each early fertilized egg at 1 h intervals for the duration of 3 h
   AEL. Live imaging assays of atg14, atg8b and double knockdown of atg8a
   and atg8b were imaged under a Zeiss Axio Observer Z1 inverted
   brightfield/fluorescence microscope and a Hamamatsu ORCA-Flash4.0
   digital CMOS camera.

High-resolution 3D live imaging

   Early fertilized eggs at 0–15 min AEL of the appropriate genotype were
   collected and mounted for imaging as described above. Imaging was
   performed using a Dragonfly Spinning disc confocal system (Andor
   Technology PLC) mounted on an inverted Leica Dmi8 microscope (Leica
   GMBH). Images shown in Fig. [351]3a–d and Fig. [352]5a, b, d, e were
   deconvolved with the internal Andor Fusion deconvolution application.
   MIP (Max-Intensity-Projection) processing was performed using Imaris
   9.5.0 (Bitplane, [353]http://www.bitplane.com/, RRID:SCR_007370).

ROS staining

Early fertilized eggs

   WT early fertilized eggs at 0–15 min AEL were collected and
   dechorionated as described above. The early fertilized eggs were then
   permeabilized by 1 min incubation in EPS solution (90% D[+] Limonene
   [Thermo Fisher Scientific, FL/1860/07], 5% ethoxylated alcohol
   [Bio-Soft N1-7], 5% cocamide DEA [kind gift from Sano International;
   [354]https://www.sano-international.com/], diluted 1:10 in pre-warmed
   [37 °C] PBS)^[355]159. Early fertilized eggs were washed four times in
   PBS and incubated in 10 μM 2,7-Dichlorodihydrofluorescein diacetate
   solution (H2DCFDA; Cayman chemical, 85155) for 3 min. Early fertilized
   eggs were then washed twice in PBS and mounted and imaged as described
   above.

Sperm

   Seminal vesicles (SVs) were dissected from males producing the red-MD
   sperm. The SVs were transferred to a drop of Schneider’s Drosophila
   Medium with L-Glutamine (Biological industries, Beit Haemek, Israel,
   01-150-1 A) on a Poly-L-Lysine coated coverslip. SVs were punctured
   using forceps and incubated in 1 μM H2DCFDA for 10 min. H2DCFDA
   solution was then replaced with fresh Schneider’s medium and imaged as
   described above.

Early fertilized egg fixation and immunostaining

   Early fertilized eggs at 0–1-hour AEL of the appropriate genotype were
   collected and dechorionated as described above. Early fertilized eggs
   were placed in boiling Triton/Salt solution (0.7% NaCl, 0.04%
   TritonX-100) for 5 sec, and immediately moved to cool Triton/Salt
   solution on ice for 15 min. Vitelline membranes were removed by manual
   shaking in 50% methanol: 50% heptane for 1 min. Early fertilized eggs
   were rehydrated by passing through a series of solutions with
   decreasing methanol concentrations, washed with 1× PBS + 0.1%
   TritonX-100 (PBTx), and blocked in 5% normal goat serum in PBTx for
   1 h. Then, the early fertilized eggs were incubated with the relevant
   primary antibody overnight at 4 °C, washed in PBTx, and incubated with
   the secondary antibody at room temperature for 2 h. Early fertilized
   eggs were washed in PBTx and then mounted onto slides in Fluoromount
   medium (SouthernBiotech, Birmingham, AL, USA).

   For expansion microscopy, early fertilized eggs were fixed as described
   in ref. ^[356]160. Briefly, dechorionated early fertilized eggs were
   placed in a mix of 4% formaldehyde (diluted in PBS) and heptane (1:1
   ratio) and vigorously shaken for 20 min. After removal of the aqueous
   layer, the early fertilized eggs were placed in ice-cold 50% methanol:
   50% heptane and vigorously shaken for 1 min. All further steps were
   conducted as described above.

   Primary antibodies used in this study are biotin Anti-GFP antibody
   (1:100, Abcam, ab6658), anti-Drosophila melanogaster Atg8a polyclonal
   antibody (1:100, Creative-diagnostics,CABT-L1690), anti-pan
   polyglycylated Tubulin antibody, clone AXO 49 (1:5000, Sigma-Aldrich,
   MABS276). Anti-RFP (1:100, ROCKLAND, 600-401-379). All secondary
   antibodies (Jackson Immuno-Research) were used at a dilution of 1:250.
   For expansion microscopy, antibody concentrations were doubled.
   Anti-Mouse-IgG-Atto-647N (Sigma-Aldrich, 50185) was used at a dilution
   of 1:10. All images were acquired using the Andor Dragonfly Spinning
   disc confocal imaging system (Andor Technology PLC) mounted on an
   inverted Leica DMi8 microscope (Leica GMBH).

Expansion microscopy

   Early fertilized eggs expansion protocol was adapted from^[357]82.
   Early fertilized eggs were incubated with 1 mM MA-NHS (methacrylic acid
   N-hydroxy succinimidyl ester, Sigma-Aldrich, 730300) in PBS for 1 h at
   room temperature, followed by 3 × 20 min washes in PBS. Samples were
   then incubated in monomer solution [2 M NaCl, 8.625% sodium acrylate
   (Sigma-Aldrich, 408220), 2.5% acrylamide (Bio-Rad, 1610140), 0.15%
   bisacrylamide (Bio-Rad, 1610142) in PBS] for 1 h at 4 °C. The solution
   was then replaced with gelation solution {monomer solution + 0.01%
   4-hydroxy-2,2,6,6-tetramenthylpiperidin-1-oxyl [4-hydroxy-TEMPO
   (Sigma-Aldrich, 176141)], 0.2% tetramethylethylenediamine (Bio-Rad,
   1610800) and 0.2% ammonium persulfate (Bio-Rad, 1610700)}. The samples
   were mounted in bridged glass slides (Superfrost microscope slide with
   two glued 24 mm coverslips with a gap in between) and incubated in a
   humid chamber at 37 °C for 2 h. At this stage, the gelated samples were
   taken to the microscope for pre-expansion imaging. Next, the gels were
   cut around each early fertilized egg and digested in 8 U/ml proteinase
   K solution in digestion buffer (40 mM Tris, pH 8, 1 mM EDTA, 0.5%
   Triton, 0.8 M guanidine HCl) for 1 h at 37 °C. The samples were washed
   3 X in PBS and placed in double distilled water for expansion. Water
   was exchanged three time before a final overnight incubation. The
   samples were transferred into Poly-L-Lysine coated 8 well chamber slide
   (µ-Slide 8 Well high, Ibidi, 80806), covered with water and imaged as
   described above.

   Expansion factors were calculated by dividing post-expansion average
   early fertilized egg width by its pre-expansion measurement. Scale bars
   have been divided by their respective measured expansion factors of
   approximately four times and therefore correspond to pre-expansion
   dimensions.

Image analysis

MD elimination

   To quantify MD intensity, two-channel 2D images were acquired for each
   early fertilized egg: A bright field (BF) channel and a fluorescent
   channel of the (MTS)tdTomato (red-MD). Early fertilized eggs that
   failed to complete cellularization were omitted from the analysis. Each
   image includes one or two early fertilized eggs. We used the BF channel
   to segment the whole early fertilized egg using Ilastik AutoContext
   pixel classifier^[358]161. We trained the Ilastik classifier on
   multiple images from multiple different conditions. To avoid artifacts
   on the edge of the early fertilized egg, we eroded the early fertilized
   egg segments by 15 pixels. Small non-early fertilized egg segments were
   discarded from further analysis. We then identified positive MD regions
   by applying background subtraction with Rolling ball (sigma = 35
   pixels) to the fluorescent channel and selecting all the pixels above a
   fixed selected value (1200). For each early fertilized egg, we measured
   the total and average signal in the whole early fertilized egg, MD
   positive regions and MD-negative regions, and calculated the difference
   between the average MD positive signal to the average MD-negative
   signal. We saved segmented regions on top of the BF channel and the
   fluorescent channel for quality control. The regions-of-interest of all
   early fertilized eggs were saved into a file, to enable manual
   correction in case of incorrect early fertilized egg segmentation.
   Normalized MD intensity values were calculated for each early
   fertilized egg, by subtracting the mean intensity value of the
   background (MD-negative) from the mean MD intensity value (MD
   positive), and multiplying by the fraction area of intensity above
   threshold, to take into account the MD size. The above workflow was
   implemented as a Fiji 125 macro^[359]162, which also enables the above
   measurement from manually modified regions instead of automatic
   regions, to overcome segmentation problems. Experiments imaged on the
   Zeiss Axio Observer Z1 inverted brightfield/fluorescence microscope
   were analyzed by the same code with new Ilastik classifier to segment
   the early fertilized eggs. The intensity threshold value was adjusted
   to 250. All other parameters were the same.

Atg8a recruitment

   To determine Atg8a recruitment levels to the sperm flagellum, we used
   an Arivis Vision4D pipeline to analyze the relative volume of Atg8a as
   a function of the distance from the axoneme, in 0–1-hour AEL early
   fertilized eggs immunostained to reveal the axoneme and Atg8a. First,
   the axoneme was identified using a random forest pixel classifier.
   False positive axoneme objects were manually removed. Then, the Atg8a
   signal was identified using a random forest pixel classifier. To save
   computation time and avoid background signal, the smallest 40% Atg8a
   objects were filtered out from each file, such that only 60% of the
   total Atg8a volume was further analyzed. We verified on a subset of the
   data that we get the same spatial distribution of the Atg8a objects
   when analyzing all the objects or just the bigger ones as described
   herein. Next, total Atg8a object volume was measured in 10 envelopes of
   1 µm thick around the identified axoneme (the first envelope captures
   Atg8a objects with less than 1 µm distance from the axoneme, the second
   envelope captures Atg8a objects residing between 1 and 2 µm distance
   and so on). Then, the normalized Atg8a volume was obtained by dividing
   the total Atg8a volume within an envelope by the corresponding envelope
   volume.

Sperm plasma membrane analysis

   To determine sperm plasma membrane volume, we used an Arivis Vision4D
   pipeline to analyze the sperm plasma membrane signal in proximity to
   (touching) the axoneme. 0–1-hour AEL early fertilized eggs (further
   staged to 0–4 nuclei) were immunostained to reveal the axoneme and the
   sperm plasma membrane. First, the axoneme was identified using a random
   forest pixel classifier. False-positive axoneme objects were manually
   removed. Then, the sperm plasma membrane signal was identified using a
   random forest pixel classifier. The normalized sperm plasma membrane
   volume was obtained by dividing the total sperm plasma membrane volume
   with the corresponding axoneme volume.

Rubicon-eGFP vesicles and hCD63-tdTomato analysis

   To determine the correlation between Rubicon and hCD63 signal in
   Rubicon positive vesicles, we used an Arivis Vision4D pipeline, applied
   on live images of 0–15 min AEL early fertilized eggs, dually expressing
   Rubicon-eGFP and hCD63-tdTomato. First, Rubicon vesicles were
   identified using a random forest pixel classifier. Objects with a
   volume smaller than 15 µm^3 were filtered out. hCD63-tdTomato mean
   intensity value was determined as follows: First, hCD63-tdTomato mean
   intensity signal was measured within each Rubicon-eGFP vesicle. To
   evaluate hCD63-tdTomato background intensity values, two surrounding
   envelopes were generated around each vesicle. A first 200 nm thick
   hCD63-tdTomato signal envelope was used to separate between each
   vesicle and its background. A second 300 µm thick hCD63-tdTomato signal
   envelope was used to determine hCD63-tdTomato background signal, by
   averaging the total mean intensity of all second envelopes in each
   early fertilized egg. To obtain the normalized hCD63-tdTomato mean
   intensity value of each vesicle, the total mean intensity value of the
   background was extracted from each vesicle. Vesicles in which the
   fluorescent signal was lower compared to the engulfing envelopes
   signal, were omitted from further analysis. Rubicon vesicles that
   exhibit the hCD63-tdTomato signal above background were considered
   positive.

eGFP-hCD63 signal analysis

   To determine the eGFP-hCD63 signal intensity on the MD, we used an
   Arivis Vision4D pipeline, applied on live images of 0–15 min AEL early
   fertilized eggs, expressing eGFP-hCD63 and fertilized with red-MD
   sperm. First, the red-MD was identified using a random forest pixel
   classifier. Then, eGFP-hCD63 mean intensity signal within the red-MD
   was measured. To separate between the red-MD and its background, a
   first 0.5 µm thick eGFP-hCD63 envelope signal was not measured. Then,
   an additional 1 µm thick eGFP-hCD63 envelope was generated, and its
   mean intensity was considered as background. The eGFP-hCD63 mean
   intensity of the second envelope was calculated (background) and
   extracted from the corresponding red-MD, to obtain the normalized
   eGFP-hCD63 mean intensity value.

3D reconstruction and movie of ExM images

   Images generated by expansion microscopy were computed using the Imaris
   surfaces model. To generate the movie presented in Supplementary
   Movie [360]3, we used 3D maximum intensity projection of confocal
   optical section, followed by surface reconstruction computationally
   rendered with the Imaris software. The sequence of z-slices was
   generated using the Ortho slicer.

Determination of relative RNA expression levels by RT-qPCR

   Maternal gene knockdowns were validated by comparing the levels of the
   examined genes in control early fertilized eggs (expressing the
   matα-GAL4 driver alone) and early fertilized eggs expressing
   corresponding UAS-ShR transgenes. Total RNA was isolated from 0–30 min
   AEL early fertilized eggs, using the Quick-RNA Microprep Kit (Zymo
   Research, R1051). cDNA was then synthesized using the High-Capacity
   cDNA Reverse Transcription Kit (Applied Biosystems™, 4368814). Gene
   expression levels were determined by RT-qPCR using the StepOnePlus™
   Real-Time PCR System (Applied Biosystems™, 4376600) with the KAPA SYBR®
   FAST qPCR Master Mix Kit (KAPA Biosystems, KR0389_S-v2.17). Expression
   values were normalized to the αTub84B gene (CG1913). Fold change in the
   expression levels in the knockdown validation experiments were analyzed
   using the ΔΔCt method^[361]163, while relative mRNA levels in
   Supplementary Fig. [362]5d are represented by ΔCt values. Primer sets
   were designed using the FlyPrimerBank online database^[363]164 and are
   listed in Supplementary Table [364]2.

rubicon mRNA levels determined by RT-PCR

   mRNA levels of rubicon were determined as follows. cDNA was synthesized
   from 0–30 min AEL early fertilized eggs as described above. Fragments
   from three different areas across the rubicon gene were PCR amplified
   using Phusion® High-Fidelity DNA Polymerase (New England Biolabs,
   M0530S). aTub84b was used as cDNA template loading control (primers are
   indicated above). The primers used to amplify rubicon are listed
   herein.
   Fragment 1 ATGACCACGCCCCCG
              CAGTCCATGCTGCAAAATTCTAGTGC
   Fragment 2 GGCCCAAAGCTGCACTAGAAT
              AACAAGAACTGGTAGATGGGGCA
   Fragment 3 TTCCTGCGACGATCCGG
              AACAAGAACTGGTAGATGGGGCA
   [365]Open in a new tab

Fertility tests

   Fertility was evaluated using either 3-day-old adult males or females
   that are rubicon mutant or knockdown [germline knockdown (males) and
   maternal knockdown (females)]. Each fly was crossed for 3 (males) or 5
   (females) days with 3 WT virgin females or males, respectively.
   Subsequently, the parents were removed, and adult progeny numbers were
   scored for a week after the eclosion of first progeny.

   To determine embryonic cellularization stage, we used brightfield
   microscopy combined with MD fluorescence, as described in the “Image
   analysis” paragraph of the Methods section. Fertilization of the eggs
   was assessed by the presence of the red-MD sperm. Of these fertilized
   eggs, we counted the number of eggs that developed to the
   cellularization stage by the third hour AEL.

Data visualization, statistics, and reproducibility

   Illustrations in Fig. [366]1d and Supplementary Fig. [367]1d were
   generated using BioRender.com. The pathway enrichment analysis graph
   was generated using R (R Core Team, 2020
   [368]https://www.r-project.org/, RRID:SCR_001905). Illustrations in
   Fig. [369]1a and Fig. [370]8d were generated using Adobe Illustrator
   (Adobe Inc. [2019]. Retrieved from
   [371]https://adobe.com/products/illustrator).

   All graphs and statistical analyses in this manuscript were generated
   using the GraphPad Prism software version 9.5.1 for Windows (GraphPad
   Software, San Diego, California USA, [372]www.graphpad.com), except for
   the evaluation of the co-localization of Rubicon vesicles with hCD63,
   which was done using the colocr R package^[373]165. The specific
   statistical tests that were used for determining the significance
   between experimental groups, as well as the experimental
   reproducibility, are all indicated in the relevant figure legends.
   Significance is indicated by asterisks as follows: *p  <   0.05,
   **p  <   0.01, ***p  <   0.001 and ****p  <   0.0001.

   For all MD elimination kinetic assay experiments, 2–3 biological
   replicates were performed. Specific n numbers for each genotype are
   indicated in the figure legends. Note that for rubicon, several
   different mutant alleles and an shRNA transgene displayed similar
   phenotypes.

   At least three biological replicates were performed in Figs. [374]1b,
   [375]3a, [376]5d, e, and [377]6a. For Figs. [378]3d, [379]5h,
   [380]6d–k, and [381]8a, two biological replicates were performed. Note
   that early fertilized eggs stained with H2DCFDA display variations in
   intensity levels of this indicator signal, likely due to penetration
   efficiency. A single biological replicate was performed in
   Figs. [382]4a, f, and [383]8b, each containing at least eight early
   fertilized eggs with identical results. The MS analysis in this paper
   is a result of a single biological replicate.

   No statistical method was used to predetermine the sample size. Sample
   size was chosen according to standard practices in the laboratory and
   general area of study. For the PME kinetics assay, more than 20 early
   fertilized eggs were analyzed per genotype.

   To reduce background noise in the proteomic analysis (Supplementary
   Data [384]1 and Supplementary Fig. [385]1), proteins with the lowest
   abundance (PSM < 10) were discarded from the analysis (both from the
   MVB fraction and control lysate). To save computation time and avoid
   background signal in the Atg8a recruitment experiment (Fig. [386]7e),
   40% of the smallest Atg8a objects were filtered out from each image,
   and only 60% of the total Atg8a volume was further analyzed. Using a
   subset of the data, we verified that the same spatial distribution of
   the Atg8a objects is obtained when analyzing all the objects or only
   the larger (60%) objects as described herein.

   The experiments were not randomized.

   The investigators were not blinded to allocation during experiments and
   outcome assessment.

Reagents and software

   Reagents and software used in this study are listed in Supplementary
   Tables [387]3 and [388]4, respectively.

Reporting summary

   Further information on research design is available in the [389]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [390]Supplementary Information^ (4.1MB, pdf)
   [391]Peer Review file^ (4.8MB, pdf)
   [392]41467_2024_50041_MOESM3_ESM.pdf^ (428.1KB, pdf)

   Description of Additional Supplementary Files
   [393]Supplementary Data 1^ (271.1KB, xlsx)
   [394]Supplementary Movie 1^ (5.3MB, mp4)
   [395]Supplementary Movie 2^ (18.5MB, mp4)
   [396]Supplementary Movie 3^ (128.2MB, mp4)
   [397]Supplementary Movie 4^ (29.4MB, mp4)
   [398]Supplementary Movie 5^ (13.9MB, mp4)
   [399]Supplementary Movie 6^ (18.1MB, mp4)
   [400]Supplementary Movie 7^ (16.4MB, mp4)
   [401]Supplementary Movie 8^ (35.1MB, mp4)
   [402]Supplementary Movie 9^ (78.3MB, mp4)
   [403]Supplementary Movie 10^ (21.5MB, mp4)
   [404]Reporting Summary^ (2.7MB, pdf)

Source data

   [405]Source Data^ (213.1KB, xlsx)

Acknowledgements