A trimeric metazoan Rab7 GEF complex is crucial for endocytosis and scavenger function

Garland and pericardial nephrocytes, which are present in the larval and adult fly, are excellent model cells for investigating potential deficits in endosomal pathways in mutants (Aggarwal and King, 1967; Koenig and Ikeda, 1990; Kosaka and Ikeda, 1983; Lőrincz et al., 2016, 2017; Psathaki et al., 2018; Rizki, 1978; Tutor et al., 2014; Weavers et al., 2009). Mutants that lack functional nephrocytes are viable, but suffer from a diminished toxic tolerance (Ivy et al., 2015). Garland and pericardial nephrocytes have been characterised at the ultrastructural level in great detail (Bowers, 1964; Brockhouse et al., 1999; Crossley, 1972, 1985; Hochapfel et al., 2018; Kawasaki et al., 2019; Lehmacher et al., 2012; Mills and King, 1965; Psathaki et al., 2018; Sanger and McCann, 1968), and they display all signs of highly active endocytic cells. Endocytic activity is one of the major functions of nephrocytes, which take up surplus material from the circulating haemolymph (Das et al., 2008; Helmstädter and Simons, 2017; Ivy et al., 2015; Weavers et al., 2009). Endosomes then accumulate waste material, such as circulating signalling peptides, or by-products from various metabolic pathways, and target it for storage in specific endosomes or degradation in lysosomes. In all multicellular organisms, specialised cells or tissues, such as pericardial and garland nephrocytes in Drosophila or kidney tissue in vertebrates, take on this function (Helmstädter et al., 2012; Helmstädter and Simons, 2017; Ivy et al., 2015; Psathaki et al., 2018; Weavers et al., 2009); thereby, Drosophila nephrocytes have been used as a versatile model for human nephropathies (Carrasco-Rando et al., 2019; Dlugos et al., 2019).

A second important function of endocytosis is the removal or recycling of membrane proteins, including transmembrane receptors. At early endosomes, the subsequent cellular fate of such proteins – either lysosomal degradation or recycling – is determined by specific modifications such as ubiquitination and their recognition by endosomal sorting proteins. Rab GTPases play a key role in coordinating trafficking routes and dynamics in the endomembrane system of the cell (Huotari and Helenius, 2011). Rabs are switch-like proteins, which are inactive in their GDP-bound form and are kept soluble in the cytosol by GDP-dissociation inhibitors (GDIs). To function on membranes, Rabs require a specific guanine-nucleotide-exchange factor (GEF), which triggers nucleotide exchange. When bound to GTP, Rabs can interact with their effectors, such as tethering complexes or signalling proteins. GTPase-activating proteins (GAPs) then promote GTP hydrolysis to inactivate the Rab, which is then extracted from membranes by GDI.

Early endosomes contain Rab5, whereas Rab7 is present on late endosomes. The biogenesis of early Rab5-positive endosomes requires their homotypic fusion to form larger vesicular compartments, which is accompanied by their acidification (Huotari and Helenius, 2011; Huss et al., 2011; Zeigerer et al., 2012). Endosomal complexes required for transport (ESCRTs) function on the endosomal surface to channel membrane proteins into intraluminal vesicles, which accumulate in the endosomal lumen (Henne et al., 2011; Hurley, 2015). Such endosomes eventually form structures called multivesicular bodies (MVBs), and have been identified in all eukaryotes. During their formation, Rab5 is replaced by Rab7.

Maturation of early endosomes into late endosomes and their final fusion with lysosomes relies on two heterohexameric tethering complexes, CORVET and HOPS, and their associated small GTPases. The CORVET complex is an effector of Rab5 on early endosomes, whereas HOPS functions together with Rab7-GTP and mediates the fusion of late endosomes with lysosomes (Balderhaar et al., 2013). The yeast HOPS complex directly binds the Rab7-like Ypt7 to tether membranes (Hickey et al., 2009; Ho and Stroupe, 2015; Lürick et al., 2017), whereas metazoan HOPS interacts with Rab2 and Arl8, and possibly Rab7, to trigger tethering and fusion (Fujita et al., 2017; Gillingham et al., 2014; Hofmann and Munro, 2006; Khatter et al., 2015; Lőrincz et al., 2017; Marwaha et al., 2017). Moreover, Rab7 directly interacts with the retromer complex on endosomes and controls not only endosomal maturation but also endosomal positioning, and thus function, within cells (Balderhaar et al., 2010; Bonifacino and Neefjes, 2017; Jimenez-Orgaz et al., 2018; Liu et al., 2012; Purushothaman et al., 2017; Rojas et al., 2008). Thus, both in yeast and metazoan cells, Rab7 function is critical for endosomal maturation and fusion with lysosomes.

The Rab7 GEF was initially characterised in yeast as the heterodimeric Mon1–Ccz1 complex (Nordmann et al., 2010). Its metazoan orthologues are likewise required for Rab7 function, as shown in mammalian cells (Gerondopoulos et al., 2012; Kinchen and Ravichandran, 2010; Poteryaev et al., 2010) and Drosophila (Dhiman et al., 2019; Hegedűs et al., 2016; Yousefian et al., 2013), and a similar role has been described for Mon1–Ccz1 orthologues in plants (Cui et al., 2014; Singh et al., 2014). Several studies have further suggested direct interaction of the GEF complex with Rab5 and the endosomal lipid phosphatidylinositol 3-phosphate (PI3P) (Cui et al., 2014; Hegedűs et al., 2016; Kinchen and Ravichandran, 2010; Lawrence et al., 2014; Singh et al., 2014). A previously reported crystal structure revealed that a conserved interface between the longin domains of Mon1 and Ccz1 and the Rab Ypt7 is required for nucleotide exchange (Kiontke et al., 2017). The activated Ypt7 then recruits the fusion machinery to allow fusion of the late endosome with the vacuole (Kiontke et al., 2017; Langemeyer et al., 2018a,b). Metazoan Rab7 does not seem to directly recruit the fusion machinery, but rather functions in retromer-mediated recycling, endosome and lysosome positioning, and mediating contacts between lysosomes and mitochondria, among other roles (Baba et al., 2019; Jimenez-Orgaz et al., 2018; Marwaha et al., 2017; Wong et al., 2018). However, Mon1–Ccz1 and Rab7 are nevertheless required not only for endosome maturation and lysosomal biogenesis, but also for autophagosome fusion with lysosomes, as in yeast (Gao et al., 2018; Hegedűs et al., 2016; Kinchen and Ravichandran, 2010; Yasuda et al., 2016; Yousefian et al., 2013). This suggests that Mon1–Ccz1 must be specifically recruited to multiple locations to provide sufficient active Rab7. Recently, we provided direct evidence that Mon1–Ccz1 is activated by membrane-bound Rab5 to facilitate nucleotide exchange of Rab7 on membranes (Langemeyer et al., 2020).

In this study, we now report the identification and characterisation of a third subunit of the Mon1–Ccz1 complex in metazoan cells, the Drosophila protein Bulli (encoded by CG8270). Bulli is homologous to the RMC1 protein in human cells (Vaites et al., 2018; van der Boomen et al., 2019 preprint), yet its physiological role and its contribution to Rab7 GEF activity have remained unexplored. Taking advantage of the genetics of Drosophila, we demonstrate that Bulli is required for endosome maturation in nephrocytes and for stress tolerance in Drosophila. Our data suggest that Bulli extends the ability of Mon1–Ccz1 to coordinate the Rab5 to Rab7 transition at endosomes.

In a computational search for proteins sharing similarities to HOPS/CORVET complex proteins in Drosophila, we noticed a single gene, CG8270, located on the third chromosome at position 65B3 (Fig. 1A,B). CG8270 encodes a 642 amino acid-long protein, for which a DIOPT-based search (Hu et al., 2011) via FlyBase (Flybase, 1999) retrieves orthologues across all model organisms, including mammals (311002H16Rik in mouse and WDR98 in rat, both also known as RMC1), amphibians (Rmc1 in Xenopus tropicalis), fish (Rmc1 in Danio rerio), nematodes (Y50D4A.4 in Caenorhabditis elegans) and plants (AT3G12010 in Arabidopsis thaliana), but not in budding or fission yeasts. The deduced amino acid sequence encoded by CG8270 shares approximately 60% similarity and 37% identity with its human homologue (Fig. S1). Web-based prediction algorithms indicate the presence of two large and highly conserved structural motifs, a predicted N-terminal β-propeller and a C-terminal α-solenoid motif (Fig. 1C). CG8270, which we named Bulli (Bulli is the nickname of a classical German bus from the 1970s that facilitated cargo trafficking), was previously observed in a large-scale proteomic screen (Aradska et al., 2015), as a genome-wide candidate among Rab5-interacting proteins (Gillingham et al., 2014), and in a microarray-based genome-wide transcriptional analysis (Gelbart and Emmert, 2013). The latter study showed that CG8270/bulli has low to moderate expression in almost all tissues and at all developmental stages. CG8270/bulli thus represents a ubiquitously expressed gene. Owing to its predicted secondary structure with a β-propeller followed by an α-solenoid domain, and thus similarity to several HOPS and CORVET subunits, we began to determine its function using Drosophila nephrocytes as a model.

Antibodies that recognise endogenous Bulli protein in protein extracts or within intact tissues are not available. To identify possible interaction partners, we established transgenic Drosophila lines carrying a bulli–GFP fusion gene. In animals, Bulli–GFP expression was driven with daughterless-Gal4, and thus overproduced. We purified Bulli–GFP via GFP-trap beads using protein extracts isolated from whole flies and determined its interaction partners by mass spectrometry. Only two proteins were highly enriched in the co-purification – the Rab7 GEF complex proteins Mon1 and Ccz1 (Fig. 1D). Importantly, Mon1, Ccz1 and Bulli–GFP had similar peptide counts, indicating that the cellular expression levels of the daughterless-Gal4 driven Bulli–GFP fusion protein are comparable to those of Mon1–Ccz1.

The human Bulli orthologue RMC1 interacts with Mon1–Ccz1 to form a trimeric complex (Vaites et al., 2018; van der Boomen et al., 2019 preprint). Similarly, our data thus far revealed that the Drosophila Bulli protein interacts with the corresponding Drosophila dimer Mon1–Ccz1 to form a trimeric complex. However, neither activity nor stoichiometry of the trimeric complex had been analysed. To clarify the role of Bulli within the complex, we purified Mon1–Ccz1 as a GST-tagged dimer from insect cells and tested its interaction with full-length and FLAG-tagged Bulli. Bulli was specifically retained on beads carrying GST–Mon1–Ccz1, but not GST (Fig. 1E). We also attempted to generate N- and C-terminal truncations of Bulli to narrow down the interaction interface, but did not obtain a folded protein. Encouraged by the reconstitution of the trimeric complex, we co-expressed Mon1–Ccz1 either alone or together with Bulli. Both trimeric and dimeric Mon1–Ccz1 complexes were efficiently purified and resolved by gel filtration as a single peak (Fig. 1F). Not unexpectedly, some dimer of Mon1–Ccz1 was also observed in the trimer purification (Fig. 1F). We conclude that Mon1 and Ccz1 form a stable complex alone or together with Bulli.

To determine the effect of Bulli on GEF activity, we loaded Drosophila Rab7, carrying a C-terminal His₆-tag, with fluorescently-labelled MANT-GDP, and determined nucleotide exchange in the presence of unlabelled GTP, which results in a strong decrease in fluorescence (Nordmann et al., 2010). When Rab7 was incubated with either the dimeric or trimeric Mon1–Ccz1 complex in solution, efficient nucleotide exchange was observed for both GEF complexes (Fig. 1G–I). The trimeric complex was under these conditions slightly more efficient than the dimer. Because GEFs act on membranes, we repeated this assay in the presence of multi-lamellar vesicles (MLVs), which mimic the surface of endosomes. In previous studies, we recruited His-tagged Ypt7 to DGS-NTA-containing MLVs and observed that the Mon1–Ccz1 dimer had far higher activity than in solution (Cabrera et al., 2014b). Employing a similar setup, we obtained slightly higher activity of both complexes in the presence of membranes than in solution, and some 40% increase in activity of the trimer in comparison to the activity of the dimer. We thus conclude that Bulli has a positive effect on the GEF activity of the Mon1–Ccz1 complex, possibly by improving its membrane association. Importantly, the main GEF activity for Rab7 requires just the Mon1–Ccz1 dimer, consistent with our previous structural analyses (Kiontke et al., 2017).

Given the previously described cellular function of Mon1–Ccz1 in recruiting Rab7 to the endosome, and thereby converting Rab5-positive early endosomes into late endosomes, we next asked whether Bulli colocalises with Rab5 or Rab7 in cells. We thus expressed the Bulli–GFP fusion protein in pericardial nephrocytes using handC-Gal4_8.0 as driver (Sellin et al., 2006) (Fig. 2). We observed the protein in puncta distributed throughout the cell with enrichment at vesicular compartments, presumably endosomes at different stages of maturation (Fig. 2A–D, green channel, arrows and circles). Co-staining of Bulli–GFP and Rab5 (Fig. 2B–B″) revealed that both proteins display largely independent localisations within the cell with only minor overlap. The Pearson colocalisation coefficient was R=0.28 (range of −1 to +1, with +1 indicating complete overlap; Fig. 2B‴). In contrast, co-staining of Bulli–GFP and Rab7 (Fig. 2C–C″) indicated that a substantial portion of both proteins colocalise to membranes (Pearson coefficient of about 0.8; Fig. 2C‴). As further control, we co-stained for Bulli–GFP and Vps8 (Fig. 2D–D″), a core component of the hexameric CORVET complex and marker of early endosomes. Vps8 interacts directly with Rab5 (Balderhaar and Ungermann, 2013; Lőrincz et al., 2016). However, Bulli–GFP and Vps8–mCherry showed only poor colocalisation (Pearson coefficient=0.26; Fig. 2D‴). Our results indicate that Bulli is most likely localised to maturing (late) endosomes, which are typically characterised by association with Rab7, although a small fraction of Bulli might remain cytoplasmic or be associated with other, as yet unknown, cellular compartments.

To reveal the physiological importance of Bulli in vivo in a complex animal system, we established bulli (CG8270) mutants using CRISPR-Cas9-mediated targeted mutagenesis. In total, six guide RNA (gRNAs) were generated by in vitro transcription and injected into vasa-Cas9 transgenic Drosophila (Fig. 3A). The new bulli alleles (Fig. 3C,E) all contain small deletions causing frame shifts and premature stop codons, thus resulting in truncated Bulli proteins. All bulli alleles were homozygous viable and fertile. During our investigations, transposon insertion lines became available as part of larger genome-wide mutagenesis approaches (Fig. 3B). Of note, the bulli mutant CG8270^f03412, a PiggyBac insertion line and the bulli mutant CG8270^Mi03821, a Minos element insertion line, were described as being homozygous lethal, and stocks were maintained using balancer chromosomes. For clarification of this discrepancy regarding the observed homozygous viability of our lines, we performed a complementation analysis (Fig. 3D). Using our new bulli alleles as well as the molecularly defined deficiency line, Df(3L)ED211, which removes the single bulli gene together with ∼18–20 additional genes, showed that lethality in homozygous CG8270^f03412 and CG8270^Mi03821 flies is caused by secondary hits elsewhere in the genome (Fig. 3D). Additional PiggyBac insertion lines listed in Flybase (f07172, f00597 and f07847) have not been characterised in detail thus far. However, some of our established CRISPR-Cas9-induced lines (18 out of 349) were also homozygous lethal, caused by secondary hits, as verified by complementation tests.

We therefore decided to focus on the molecularly defined CRISPR-induced mutants, in particular the CG8270^6-61 allele (hereafter referred to as bulli^6-61). In this mutant, a truncated Bulli protein is expressed, which lacks the complete α-solenoid region of the protein. Transheterozygous animals, in which the loss of Bulli (CG8270^f03412) is combined with bulli^6-61, displayed identical phenotypes in nephrocytes compared to homozygous CRISPR-Cas9 alleles (see below), indicating that bulli^6-61 represents a loss-of-function allele without carrying secondary unrelated genomic mutations.

To clarify the overall alterations within the endolysosomal system potential caused by loss of bulli, we analysed the ultrastructure of nephrocytes from third instar larvae (Fig. 4). Late endosomes (called α-vesicles in nephrocytes) in wild-type nephrocytes contain one or two electron-dense deposits (Fig. 4A,A′) (Mills and King, 1965). The pericardial nephrocytes of all three tested mutant alleles, bulli^1-12, bulli^3-40 and bulli^6-61, and the transheterozygous bulli^6-61/CG8270^f03412, (Fig. 4B–E) accumulated a high number of endosomes with a significantly enlarged diameter (Fig. 4B′–E′, F). All of the enlarged endosomes contained up to ten electron-dense deposits in an ultrathin cross section, compared to one or two in wild type.

Insect nephrocytes have been established as a highly useful model for vertebrate kidney podocytes (Helmstädter and Simons, 2017; Hermle et al., 2017; Hochapfel et al., 2017; Psathaki et al., 2018), because they form a molecularly similar filtration apparatus, the slit diaphragm. Slit diaphragms serve as a filtration barrier for components larger than ∼70 kDa. Material that successfully passed the slit diaphragm is potentially endocytosed at membranes of the labyrinth channel system. We thus analysed the integrity of the slit diaphragm connections present in wild type and bulli mutant nephrocytes. These appeared normal in all bulli mutant alleles (Fig. 4A″–E″, arrows). Likewise, clathrin-coated vesicles that are present at the tips of the labyrinth channels appeared similar in wild type and bulli mutants (Fig. 4A″–E″, boxed arrows). This indicates that endocytosis and the formation of early endosomes is not affected in mutant cells.

We additionally determined possible alterations of other organelles in bulli mutants (Fig. 4G). We found no alterations in the nuclear envelope, Golgi apparatus, or the endoplasmic reticulum (ER) of nephrocytes. However, the labyrinth channel system, which is the primary site of endocytosis in nephrocytes, was slightly enlarged, disorganised, and extended throughout the entire cell of bulli mutants.

The endolysosomal deficits seen on the ultrastructural level (Fig. 4, Fig. S2) and upon immunohistochemical or physiological characterisation (Figs 5–8, Fig. S2) occurred in bulli mutant animals with a penetrance of nearly 100%. However, the observed phenotypes could vary within a single animal, due to fact that nephrocytes act as scavenger cells with an individual fate, which depends on how much material they take up during their lifetime (Psathaki et al., 2018).

To determine the role of Bulli in endosomal maturation, we analysed Rab5 and Rab7 distribution and localisation in nephrocytes by immunofluorescence (Fig. 5). In the wild-type cells, Rab5 showed a punctate staining pattern throughout the cell with the highest abundance in the cortical region (Fig. 5A,A′). However, Rab5 distribution and localisation was dramatically altered in bulli mutants (Fig. 5D,D′). Enlarged vesicular compartments, already visible in our transmission electron microscopy (TEM) analysis (Fig. 4), became prominent, and Rab5 accumulated at high densities in large clusters and aggregates (arrows in Fig. 5D′). The size of the vesicles in bulli mutants was significantly enlarged in diameter (Fig. 4F) compared to those in wild type, and immunofluorescence revealed that such enlarged vesicles contained multiple large Rab5-positive aggregates. To identify the nature of the enlarged vesicles, we stained for Rab7 (Fig. 5). In wild-type nephrocytes, Rab7-positive endosomes of different sizes were clearly visible (Fig. 5B–C′), reflecting the different states of endosomal maturation. In bulli mutant nephrocytes (Fig. 5E,F) we also found Rab7-positive endosomes of a different size, which were preferentially located in the cortical region of cells (Fig. 5E–F′). However, some of the enlarged vesicles characteristic for bulli mutant nephrocytes were Rab7 negative (Fig. 5E–F′, grey circles), but both types of large vesicle, the Rab7 positive (Fig. 5F′, grey circle) and the Rab7 negative (Fig. 5F′, red circle), contained Rab5 accumulations. Moreover, Rab7 appeared largely diffuse and less localised in mutant cells compared to in wild-type cells. This indicates that a certain amount of Rab7 remains cytoplasmic in bulli mutants, compared to the localisation seen in wild type. Quantification of the size of Rab7-positive and -negative vesicles is shown in Fig. 5L.

To verify the accumulation of Rab5 signal within the large vesicles of bulli mutant nephrocytes, we applied intensity profiling of single vesicles from wild type (Fig. 5G) and bulli mutant nephrocytes (Fig. 5H). We observed Rab5 within the endosomal compartment and not at the membrane surface of the endosome. Interestingly, it has been previously shown that Rab5 enriches at the surface of endosomes in mon1 mutant imaginal disc cells (Yousefian et al., 2013), which is a clearly different phenotype compared to our observations in bulli mutants.

A characteristic feature of bulli mutant nephrocytes is the presence of giant Rab7-negative endosomal vesicles with strong Rab5 accumulations. Rab7 is required for the maturation of the late endosome by regulating their fusion with acidic lysosomes, thus forming endolysosomes. We therefore asked next whether the Rab7-negative vesicles of bulli mutants are acidified and contain cargo. Nephrocytes were thus co-stained for Rab7 and LysoTracker as a marker of acidic lysosomal compartments (Fig. 5I,J). In wild-type cells, Rab7-positive endosomes were LysoTracker positive (Fig. 5I, white arrows). In bulli mutant nephrocytes, however, we observed three types of vesicles: (i) acidified and Rab7-decorated normally-sized late endosomes (Fig. 5J, upper panels, type 1), (ii) enlarged Rab7-positive vesicles, which were positive for LysoTracker (Fig. 5J, middle panels, type 2) and (iii) vesicles that were highly enlarged and Rab7 negative (Fig. 5J, lower panels, type 3). These type 3 vesicles failed to become acidified. Rab5 accumulation was observed in type 2, and even more in type 3 vesicles. Our results indicate that bulli mutants have a disturbed endolysosomal system, with defects in endosomal maturation and lysosomal acidification.

To verify that Rab5 accumulates within enlarged vesicles of bulli mutants, we correlated images of immunostained nephrocytes with TEM images in a two-step approach. Nephrocytes were stained for Rab5 and Rab7 with fluorescently labelled antibodies, embedded in Technovit and sectioned (see Materials and Methods). Technovit embedding preserves fluorescent staining and allowed us to visualise Rab5 and Rab7 localisation by fluorescence microscopy (Fig. 6A,B). Afterwards, the same section was stained with Toluidine Blue, which is an acidophilic metachromatic dye that stains electron-dense structures in cells (Sridharan and Shankar, 2012). The sections were imaged under bright-field illumination (Fig. 6C,D). Overlay of the images showed that the Rab5 accumulation and clustering matched the Toluidine-positive deposits in vesicles in bulli mutant nephrocytes (Fig. 6B,D,F), but not in wild type (Fig. 6A,C,E). To correlate TEM images with Toluidine-stained images, we used subsequent sections, either processed for TEM analysis (Fig. 6G,H) or Toluidine Blue staining (Fig. 6G′,H′). This showed that Toluidine-positive aggregates (which are Rab5 positive in bulli mutants; Fig. 6B,D,F) correspond to the electron-dense deposits in late endosomes of bulli mutants.

To exclude the possibility that the electron-dense deposits were fixation artefacts, we compared the ultrastructure of late endosomes in chemically fixed nephrocytes and in nephrocytes that had undergone high-pressure freezing and freeze substitution (HPFFS), a method known to preserve the ultrastructure of membranes (Fig. S2). Both methods showed that electron-dense deposits accumulated in nephrocytes of wild-type (Fig. S2A,C) and bulli mutant specimens (Fig. S2B,D). Intravesicular compartments were observed with both methods (Fig. S2A″,B″,C′,D″); however, the double lipid layer of membranes was detectable only in HPFFS-processed specimens. We postulate that the α-vesicles present in nephrocytes represent MVBs and thus late endosomes. Moreover, we speculate that the electron-dense deposits, which contain Rab5 in bulli mutants, represent clusters of intraluminal vesicles (Fig. S2D).

Next, we asked whether deficits in the endolysosomal pathway of nephrocytes cause any physiological defects. To test bulli mutants for endosomal uptake capacity, larval and adult flies were dissected and treated either with silver nitrate (Fig. 7A–D) or with FITC-labelled albumin (Fig. 7I–N). Pericardial nephrocytes from control flies as well as from homozygous bulli mutant larvae and adults efficiently took up silver nitrate and albumin by endocytosis and accumulated both compounds in endosomes within minutes, indicating that early steps of endocytosis are not inhibited in bulli mutants.

The nephrocytes of adult flies undergo structural changes upon ageing (Psathaki et al., 2018) and show an age-dependent decrease in endocytic uptake efficiency. Nephrocyte ageing includes degeneration of the endolysosomal compartment and a loss of slit diaphragms. We thus measured FITC-albumin uptake efficiency of nephrocytes from wild-type (Fig. 7I–K) and bulli mutant flies (Fig. 7L–N) at 1, 4 and 6 weeks after hatching. To calculate uptake efficiency at different ages, the mean pixel intensity was measured and uptake efficiency determined (Fig. 7O). In the wild type, young flies showed an initially high uptake efficiency, which decreased upon ageing. In contrast, young bulli^6-61 mutants were already impaired in endocytosis rates, with levels comparable to those of old wild-type flies. These data suggest that endocytosis, and therefore nephrocyte function, is severely impaired in bulli mutant flies under physiological stress conditions such as ageing.

Under our laboratory conditions, bulli mutants were homozygous viable and fertile. To test for semi-lethality or reduced survival rate we first counted the number of hatched first larvae from egg-laying plates (Fig. 7P), and found no difference between wild-type and bulli^6-61 mutant specimens. Furthermore, homozygous bulli^6-61 larvae developed equally into pupae and adult flies (Fig. 7Q), indicating no developmental delay or semi-lethality. However, when we tested for longevity of adult flies we recognised a significantly reduced lifespan in bulli^6-61 mutant specimens (Fig. 7R). This indicates that Bulli plays a critical role in long-term adaptation and stress resistance of animals, which require a functional and robust endolysosomal system. Indeed, animals lacking functional nephrocytes have a diminished lifespan due to impaired scavenger function and waste uptake (Das et al., 2008; Weavers et al., 2009). Thus, the cellular deficits of bulli mutant nephrocytes might contribute to the reduced lifespan. However, lack of Bulli does not cause earlier developmental deficits, suggesting that it can be compensated by the cell to some extent.

Our data show that the metazoan Rab7 GEF complex functions in cells with Bulli (known as RMC1 in vertebrates) as an additional subunit. Unlike in the yeast, the metazoan GEF complex is thus a trimer in vivo. Using Drosophila as a physiological model system, we have shown that Bulli is required for efficient Rab7 localisation in tissues, and for the release of Rab5 from early endosomes during endosomal maturation. Thus, Bulli is essential for proper biogenesis of endosomes in endocytically active tissues. Consequently, loss of Bulli causes deficits in endocytic uptake efficiency and impairs longevity of adult specimens. However, Drosophila Mon1–Ccz1 does not depend on Bulli for its GEF activity, in line with the recently solved mechanism (Kiontke et al., 2017). Bulli is thus more likely a cofactor of the Mon1–Ccz1 complex involved in coordinated release of Rab5 and recruitment of Rab7.

Studies in diverse organisms have clearly demonstrated that GEF activity of Mon1–Ccz1 depends on these two subunits (Cabrera et al., 2014a,b; Hegedűs et al., 2016; Kiontke et al., 2017; Nordmann et al., 2010; Yousefian et al., 2013). Rab7 is required not only on endosomes, but also on autophagosomes and lysosomes. Consequently, localisation of Mon1–Ccz1 to these membranes likely follows specific cues that are present on each of these membranes. We recently showed that Mon1–Ccz1 is a direct effector of Rab5. At the early endosome this interaction is crucial to Rab7 activation by Mon1–Ccz1, and thereby the fusion of late endosomes with lysosomes (Langemeyer et al., 2020). Interaction studies in plants, Drosophila and mammalian cells suggest that targeting of Mon1–Ccz1 to endosomes depends on PI3P and/or Rab5 (Cui et al., 2014; Hegedűs et al., 2016; Kinchen and Ravichandran, 2010; Singh et al., 2014). Likewise, PI3P is required for Mon1–Ccz1 targeting to autophagosomes (Hegedűs et al., 2016). We recently showed that with yeast Mon1–Ccz1, the Ccz1 subunit binds specifically to the LC-3 homologue Atg8, and requires this interaction for its specific targeting to autophagosomes in vivo (Gao et al., 2018).

The identification and analysis of human RMC1 (Vaites et al., 2018; van der Boomen et al., 2019 preprint) and Drosophila Bulli (this study) as a third subunit of the Mon1–Ccz1 complex suggests that, at least in metazoan cells, additional aspects have to be considered. Both cells with depleted RMC1 and bulli mutants have deficient Rab7 localisation, which results in impaired endocytosis (Vaites et al., 2018; van der Boomen et al., 2019 preprint; this study). Given that Drosophila Mon1 and Ccz1 also form a very stable complex, which has almost the same GEF activity as the heterotrimeric complex, Bulli most likely contributes to the specific targeting of Mon1–Ccz1 to membranes or might control the transition from Rab5 to Rab7 (Poteryaev et al., 2010). In line with this, we observe that Rab5 remains on endosomal membranes and becomes internalised into enlarged giant endosomal compartments (MVBs) in bulli mutants. As recently shown, Mon1–Ccz1 is a direct effector of Rab5, both in yeast and Drosophila, and this interaction strongly promotes Rab7 recruitment and activation on membranes (Langemeyer et al., 2020). It is possible that Bulli supports Mon1–Ccz1 membrane binding in vivo and thus Rab7 activation. Bulli has also been identified as a Rab5-interacting protein in Drosophila (Gillingham et al., 2014). In light of our findings we suspect that Bulli was more likely co-purified with Mon1–Ccz1. Based on our biochemical analysis, we find no evidence that Bulli directly regulates Mon1–Ccz1 activity as previously suggested (Vaites et al., 2018; van der Boomen et al., 2019 preprint). It was shown that Mon1 (specifically the nematode orthologue SAND-1) has an effect on the negative feedback-loop resulting in the inactivation of Rab5 on late endosomal membranes (Poteryaev et al., 2010). It is possible that membrane-bound Rab5 becomes ubiquitinated in the absence of Bulli and mis-sorted into MVBs afterwards. Therefore, Bulli might have an additional role in the inactivation of Rab5 on the early endosome, a subject for future studies.

Our ultrastructural and functional analyses show that Bulli is not essential during early stages of endocytosis, but specifically affects the biogenesis of endosomes and lysosomes in pericardial nephrocytes. We find that the maturation of α-vesicles and thus endosomal maturation is strongly affected in bulli mutants (Fig. 8). Interestingly, Drosophila mutants lacking CORVET, Mon1 or HOPS result in similar accumulations of large α-vesicles (Hegedűs et al., 2016; Lőrincz et al., 2016, 2017), yet do not cause a similar Rab5 accumulation as observed here.

Surprisingly, the loss of Bulli resulted in a relatively mild phenotype of impaired longevity of adult Drosophila but not embryonic or larval lethality (Fig. 7), even though Bulli is ubiquitously expressed and present in all tissues. We thus do not know whether other tissues that do not depend so strongly on endocytosis can adjust to the loss of Bulli and use just the Mon1–Ccz1 dimer for endolysosomal biogenesis. We know from many studies in yeast that cells can strongly adapt to the loss of genes through suppressor mutations or selective upregulation of partner proteins (van Leeuwen et al., 2016). For Mon1–Ccz1 we observed strong GEF activity, also in the absence of Bulli, even though Rab7 was poorly localised in bulli mutants (Figs 5, 6). In agreement with partial targeting of Mon1–Ccz1 to endosomes, Rab7 localised to a subset of endosomal vesicles in bulli mutant nephrocytes (Fig. 5). In contrast, the mon1 null allele is homozygous semi-lethal with occasional escapers (Hegedűs et al., 2016), indicating that GEF activity of the Mon1–Ccz1 complex per se is crucial for animal survival. How Bulli thus affects Mon1–Ccz1 function in vivo and possibly under different stress conditions requires further studies.

Our findings agree with recent investigations in human cells that were published while our study was underway (Vaites et al., 2018; van der Boomen et al., 2019 preprint). RMC1 interacts with Mon1–Ccz1, and localises to LAMP2-positive endosomes and autophagosomes. RNAi-mediated knockout of RMC1 impairs autophagy and lysosomal cholesterol export and results in enlarged endosomes, very similar to our observations in nephrocytes. RMC1 and Bulli thus seem to have similar functions. Our studies strongly extend these findings through the detailed analysis of the endolysosomal system and the physiology of bulli mutant flies.

Apart from endolysosomal transport, Mon1–Ccz1 and Rab7 are required for autophagy, likely at the autophagosome–lysosome fusion stage, in yeast (Gao et al., 2018), flies (Hegedűs et al., 2016) and mammals (Ravikumar et al., 2008), and RMC1 was initially identified as an autophagy factor (Vaites et al., 2018). Although not analysed in detail so far, we also observed aberrant autophagosomes in fly adipocytes from bulli mutants (M.J. and A.P., unpublished). These fat cells remobilise lipids from lipid droplets during starvation by autophagy, and are thus an attractive model to unravel this aspect of Bulli function.

RMC1/Bulli appears to be specific for metazoans and plants, whereas yeast has only the Mon1–Ccz1 dimer (Nordmann et al., 2010). Intriguingly, Bulli shows homology with proteins that have a similar architecture of an N-terminal β-propeller and a C-terminal α-solenoid, such as HOPS and CORVET subunits, proteins of the nuclear pore complex, COPI and COPII vesicle coats, and subunits of the SEA complex involved in lysosomal signalling (Balderhaar and Ungermann, 2013; Beck and Hurt, 2017; Field et al., 2011). It is thus possible that Bulli evolved by duplication of one of these genes, and took over a function previously encoded within the non-metazoan Mon1–Ccz1 dimer. Future studies will need to clarify how Bulli cooperates with Mon1–Ccz1 in endolysosomal biogenesis.

The following fly stocks were obtained from the Bloomington Drosophila Stock Center (BDSC, Bloomington, USA): BL36975 (RRID:BDSC_36975; CG8270^MI03821/TM3, Sb, Ser), BL18651 (RRID:BDSC_18651; CG8270^f03412/TM6B, Tb), BL8063 (RRID:BDSC_8063; Df(3L)ED211/TM6C, cu, Sb) and UAS-eGFP (BL 5428; RRID:BDSC_5428). As a control strain, white¹¹¹⁸ was used. The daughterless-Gal4 line (previously named BL55849) was from Andreas Wodarz (University of Göttingen, Göttingen, Germany). UAS-Vps8–mCherry expression plasmids were generated based on an Escherichia coli/Saccharomyces cerevisiae/Drosophila melanogaster triple-shuttle derivative of the pUAST vector (Brand and Perimon, 1993) adapted for cloning by homologous recombination in vivo. The vector was assembled as described in Paululat and Heinisch (2012). Resulting constructs were injected at BestGene (Chino Hills, CA, USA) to generate transgenic lines. Fly husbandry was carried out as described previously (Wang et al., 2012).

Monoclonal antibodies were used to detect Rab5 (1:250, Abcam 31261, Cambridge, UK), RFP (1:500, Abcam 62341, Cambridge, UK) and Rab7 (1:10; Developmental Studies Hybridoma Bank, University of Iowa, USA; Riedel et al., 2016). Monoclonal rabbit anti-GFP (1:2000) was from Abcam (Ab6556), monoclonal mouse anti-GFP (3E6, 1:500) was from Invitrogen (A-11120; Thermo Fisher Scientific, USA). FITC-albumin A9771 was purchased from Sigma-Aldrich (Munich, Germany). Antibody staining of Drosophila tissue was performed as described previously (Drechsler et al., 2018, 2013). LysoTracker Red DND-99 was from Thermo Fisher Scientific, USA.

Specimens were first stained for Rab5 and Rab7. Afterwards, fluorescently-labelled secondary antibodies were used. Probes were then dehydrated through a series of ethanol steps (30%, 50%, 70%, 80%, 95%, 100%; each 10 min on ice). Afterwards, specimens were embedded in Technovit 8100, following the protocol provided by the manufacturer (Kulzer, Hanau, Germany). 3 µm sections were cut on a Leica EM UC6 microtome (Leica Mikrosysteme, Vienna, Austria) with a glass knife.

Third instar wandering larvae were dissected following a standard protocol (Lehmacher et al., 2012, 2009). Animals were pinned on their dorsal side on a SYLGARD plate (Sylgard 184, Dow Corning, Wiesbaden, Germany), covered with BBT (1× PBS, 0.1% Tween-20 and 0.1% BSA) and opened ventrally with a longitudinal cut along the posterior–anterior axis. The cuticle was pinned aside to allow fixation and staining of the heart. Viscera were carefully removed with fine forceps. Specimen were fixed for 1 h at room temperature in 4% methanol-free paraformaldehyde under constant shaking. Fixed larvae were then transferred to reaction cups for better handling. Three subsequent washing steps in BBT followed the fixation process, each taking 15 min. Samples were then permeabilised with 1% Triton in PBS for 1 h. After three washing steps in BBT for 15 min each, saturation to block unspecific epitopes was performed in saturation buffer (1× PBS, 0.1% Tween-20 and 3% BSA) for 1 h at room temperature. Primary antibodies were diluted in BBT buffer containing 1% Tween and were subsequently added to the tissue to incubate overnight at 4°C under constant shaking. The following day, the primary antibody solution was removed and replaced by BBT buffer containing 1% Tween for thorough washing to remove unbound antibodies, which was repeated three times for 15 min each. This procedure was followed by another 1 h blocking step with saturation buffer. Secondary antibodies with coupled fluorophores were diluted in BBT buffer containing 1% Tween and incubated at room temperature in the dark for a maximum of 2 h. All unbound antibodies were finally removed by three washing steps in BBT buffer containing 1% Tween. Probes were either stored in the fridge or in Fluoromount-G™ (Thermo Fisher Scientific, Waltham, USA) with DAPI or Roti^®-Mount FluorCare with DAPI (Roth, Karlsruhe, Germany) for imaging with a laser scanning microscope (Zeiss, Jena, Germany). Images were analysed using Fiji ImageJ software. For evaluating colocalisation, the plugin Colocalisation threshold was used (Schindelin et al., 2012).

Artificial haemolymph (Vogler and Ocorr, 2009) was removed after dissection of the animal and replaced by LysoTracker solution [0.5 µM LysoTracker™ Red DND-99 (Thermo Fisher Scientific, Waltham, USA) in artificial haemolymph]. Cells were incubated for 15 min at room temperature under lightly shaking conditions. Afterwards, the LysoTracker solution was replaced by artificial haemolymph for a 1-min chasing step. Subsequently, the tissue was fixed with 4% methanol-free paraformaldehyde in PBS for 30 min at room temperature with light shaking. The fixation solution was exchanged up to three times to ensure to wash-off of excess LysoTracker. After fixation, cells were washed three times for 15 min each in BBT to remove the fixation solution. Permeabilisation of the tissue was enhanced with 10 µg/ml digitonin (Merck Millipore, Darmstadt, Germany) (Diaz and Stahl, 1989). After permeabilisation, three further washing steps with BBT were performed, before a 1 h step of saturation was carried out. The primary antibody, diluted in BBT solution, was applied afterwards and incubated overnight at 8°C under shaking conditions. The next day, primary antibody was replaced by BBT and samples were washed three times for 15 min each. From here, the standard protocol for antibody staining was followed and probes were embedded for analysis using Roti^®-Mount FluorCare with DAPI (Roth, Karlsruhe, Germany).

Specimens were processed as described previously (Beyenbach et al., in press; Lehmacher et al., 2012, 2009; Psathaki et al., 2018; Rotstein et al., 2018). Briefly, specimens were prepared in artificial haemolymph and subsequently fixed for 4 h at room temperature in fixative [2% glutaraldehyde (Sigma-Aldrich, Germany)/4% paraformaldehyde (Merck, Germany) in 0.05 M cacodylate buffer pH 7.4]. Specimens were post-fixed for 2 h at room temperature in 1% osmium tetroxide in 0.05 M cacodylate buffer pH 7.4 (Sciences Services, Germany), dehydrated stepwise in a graded ethanol series followed by 100% acetone. Specimens were embedded in Epon 812 (Merck, Darmstadt, Germany) and polymerised for 48 h at 60°C. Ultrathin sections (70 nm), were cut on an ultramicrotome (UC6 and UC7, Leica, Wetzlar, Germany), mounted on formvar-coated copper slot grids. Sections were stained for 30 min in 2% uranyl acetate (Sciences Services, Germany) and 20 min in 3% lead citrate (Roth, Germany). All samples were analysed at 80 kV with a Zeiss 902, Zeiss LEO912 and at 200 kV with a Jeol JEM2100-Plus transmission electron microscope (Zeiss, Jena, Germany; Jeol, Tokyo, Japan). TEM images (pericardial nuclear cross-section level) were used for measuring vesicle size in wild-type and mutant nephrocytes utilising the area tool implemented within the Fiji software package (Schindelin et al., 2012).

Drosophila specimens were prepared in artificial haemolymph and shortly dropped in 20% BSA before placing in an aluminum planchette (diameter 3 mm, cavity 150 µm), which was then covered with another flat aluminum planchette. The cavities between the specimen and the aluminum planchettes were filled with hexadecane. The planchette doublet was thereafter placed in an HPF-holder and immediately frozen using a Wohlwend HPF Compact 03 high-pressure freezer (Engineering Office M. Wohlwend GmbH, Sennwald, Switzerland). The frozen samples were stored in liquid nitrogen until freeze substitution (FS). For FS the aluminum planchettes were opened in liquid nitrogen, and hexadecane, which enrobes the specimen, was largely removed. The specimen was immersed in substitution medium containing 1% osmium tetroxide, 0.2% uranyl acetate and 5% water in water-free acetone, pre-cooled to −90°C. The FS was performed in a Leica AFS2, following the protocol of −90°C, 24 h; −90°C to −60°C, 10 h; −60°C, 7 h; −60°C to 30°C, 7 h; −30°C, 6 h; −30°C to 0°C, 9 h; 0°C, 5 h, and samples were then washed with water-free acetone, stepwise embedded in Epon 812 mixed with acetone (30% Epon, 60% Epon, 100% Epon) and finally polymerised for 48 h at 60°C. Ultrathin sections of 70 nm were cut with a Leica UC7 ultramicrotome using diamond knives (Diatome, Switzerland). Sections were collected on formvar-coated grids and post-stained for 30 min with 2% uranyl acetate and 20 min in 3% lead citrate (Roth, Germany) and analysed with a 200 kV TEM (JEM 2100-Plus, Jeol, Japan).

Albumin uptake: adult flies (1-, 4- or 6-weeks old) were anaesthetised with carbon dioxide and fixed, ventral side upwards, on SYLGARD plates. Preparation was performed in artificial haemolymph buffer (Vogler and Ocorr, 2009). The ventral body side and head were removed with a razor blade, and viscera were carefully removed. Preparation buffer was replaced by fresh artificial haemolymph buffer containing 0.2 mg/ml FITC-albumin (A9771; albumin–fluorescein isothiocyanate conjugate; MW 66 kDa; Sigma-Aldrich, Munich, Germany) for fluorescent uptake assays; specimens were incubated in the dark for exactly 5 min. Uptake was stopped by fixation in 3.7% paraformaldehyde (PFA) in PBS for 3×5 min and 1×45 min. After washing with PBS, tissues were embedded in Fluoromount-G mounting medium (Thermo Fisher Scientific, Waltham, USA) for microscopic analysis. Uptake efficiency was quantified by imaging respective nephrocytes (LSM5 Pascal, Zeiss, Jena, Germany). Imaging settings were identical for wild-type and bulli mutant animals, and the mean pixel-intensity measurement function provided by the Fiji ImageJ software package was used to quantify uptake efficiency. Pixel intensity was measured in relation to the perimeter of the cell.

Silver nitrate uptake: third-instar larvae or adult flies employed for silver nitrate-staining experiments on pericardial nephrocytes were collected from the vials as described before, prepared in PBS media (Lammers et al., 2017), and afterwards fixed by incubation in PBS buffer containing 3.7% PFA. Preparations were then transferred to a microscope slide for imaging. To enhance the contrast of the internalised silver nitrate, samples were darkened using the Zeiss LED Colibri system at 365 nm for exactly 6 min. Subsequently, samples were embedded in Fluoromount-G. Bright-field pictures were captured using a Zeiss AxioScope.A1 (Zeiss, Jena, Germany).

Individuals raised for the lifespan assay were kept in breeding bottles at a constant temperature of 22°C from the embryonic stage, and afterwards fed with yeast. Corresponding larvae were provided with fresh yeast every day. First-instar larvae, pupae and freshly hatched adults were counted daily for statistical analysis. For the lifespan assay, ten flies of a gender ratio of 1:1 were collected on the day of eclosion and transferred twice per week to fresh breeding vials for survival analysis. Vials were kept constantly at 22°C, and dead animals were counted after vial changes and immediately removed. Test animals were subjected to a daily 12 h/12 h light/dark cycle (Linford et al., 2013). We analysed at least 100 animals per genotype.

The bulli open-reading frame region, including the ATG start codon but lacking the stop codon, was amplified by PCR [primers: forward (fw), 5′-TACTGAAATCTGCCAAGAAGTAATTATTGAATACAAGAAGAGAACTCTGAGGTACCATGGATAATTCCAATGGAATC-3′; reverse (rv), 5′-CTTGCTCACCATGCTTACAGGCTGTCTGGTAGGATTATCAGC-3′]. As a template, the BO28916 clone obtained from the Drosophila Genomics Resource Center (DGRC, Bloomington, IN, USA) was used. The eGFP reading frame was amplified from pH-Stinger vector (Barolo et al., 2000) via PCR (fw, 5′-AGCCTGTAAGCATGGTGAGCAAGGGCGAG-3′; rv, 5′-ATTATGTCACACCACAGAAGTAAGGTTCCTTCACAAAGATCCTCTAGAGGCTCGAGTTACTTGTACAGCTCGTCCAT-3′). The two amplicons were inserted into the UAS-YED-vector pJJH1784 via homologous recombination in yeast (Paululat and Heinisch, 2012), specifically taking advantage of the 48 bp overlapping regions of the amplicons to generate a Bulli–GFP C-terminal fusion construct. Subsequently, plasmid DNA was isolated from yeast to transform E. coli cultures, which were then used to generate highly purified plasmid DNA suitable for injection into BL24483 embryos for 2nd chromosome insertion and BL24749 for 3rd chromosome insertion. The BL24483 and BL24749 strains harbour landing sites for site-specific integration of the injected constructs. Transgenic fly lines were generated by Best Gene, Chino Hills, USA. Transgenic UAS-CG8270–eGFPII.2 (insertion on the 2nd chromosome) line no. 2 was used for GFP pull-down experiments. The same line, then named Bulli–GFP, was used for immunostaining.

The following sequences were employed for sequence alignment (Fig. S1): RMC1 (NCBI reference number NP_037458.3) and CG8270 (Bulli; NCBI reference number AAF50661.3). We used the Needle algorithm provided as part of the EMBL-EBI sequence analysis software package (Madeira et al., 2019).

Suitable CRISPR targets were selected using the ‘CRISPR Optimal Target Finder’ (Gratz et al., 2014). Only sites that were predicted to lack any off-targets within the genome were chosen. sgRNAs were generated via in vitro transcription of double-stranded DNA templates, which were designed following established protocols (Böttcher et al., 2014). Among the six tested sgRNAs (g1–g6; positions are indicated in Fig. 3), we obtained mutant fly lines for three of them: g1, g3 and g6. Oligonucleotides for g1, g3 and g6 DNA templates are listed below. The gRNAs g1 and g2 targeted the first exon of bulli, g3 the second exon and g4–g6 the fourth exon. Guide RNAs g1, g2 and g3 potentially induce mutations in the N-terminal region of the gene, whereas guide RNA g4, g5 and g6 were chosen to potentially induce mutations in the central region of the gene to generate truncated forms of Bulli lacking the C-terminal half of the protein.

Oligonucleotides for g1: fw, 5′-TAATACGACTCACTATAGCTCAATGTAGTGGATTCCATGTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTCGGTGCTTTT-3′; rv, 5′-AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAACATGGAATCCACTACATTGAGCTATAGTGAGTCGTATTA-3′. Oligonucleotides for g3: fw, 5′-TAATACGACTCACTATAGGTCTGGAGGAGCTACTGGTGGTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTCGGTGCTTTT-3′; rv, 5′-AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAACCACCAGTAGCTCCTCCAGACCTATAGTGAGTCGTATTA-3′. Oligonucleotides for g6: fw, 5′-TAATACGACTCACTATAGATTAAAAGTCATTGGACAATGTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTCGGTGCTTTT-3′; rv, 5′-AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAACATTGTCCAATGACTTTTAATCTATAGTGAGTCGTATTA-3′. Corresponding oligonucleotides were annealed prior to in-vitro transcription with T7 RNA polymerase.

Resulting sgRNAs were individually injected into flies carrying vasa-Cas9 [y M (vas-Cas9) ZH-2A w1118/FM7c; BL51323; RRID:BDSC_51323]. Injections were done by Best Gene, Chino Hills, USA. Genomic DNA was prepared from resulting lines, and PCR amplification of the bulli locus (primers for g1 and g3: fw, 5′-TGAGTTGGAAGGGGTTAAATA-3′ and rv, 5′-TCTCCCTGAAAACATATGAAC-3′; primers for g6: fw, 5′-CCCTCACTTTCGCCTGATGGA-3′, and rv, 5′-TTGCAGCTCCAACTGGACCCA-3′) was used to identify potential bulli mutants via sequencing.

Drosophila Rab7 was amplified from cDNA (clone GH03685, DGRC) and subsequently cloned into pET24b (Novagen, Madison, USA) to express and purify recombinant protein as C-terminal His₆-tag fusion from E. coli. Heterologous expression of Mon1, Ccz1 and CG8270 was performed in Sf21 cells (RRID:CVCL_0518) using the Bac-to-Bac Baculovirus Expression System (Life Technologies, Carlsbad, USA). Generation of double and triple constructs was carried out using the biGBac system (Weissmann et al., 2016). Initially, pLIB clones of GST-pre-Mon1, Ccz1-3×FLAG and Bulli were generated. Subsequently, all subunits were amplified with linker sequences and assembled into the pBig1a vector using Gibson assembly. Clones were tested by digesting the plasmid with SwaI to test for incorporation of all subunits. Subsequently, DH10EMBacY electrocompetent cells were transformed with respective pBig1a plasmids, and bacmids were isolated through the isopropanol/ethanol extraction method (Sambrook et al., 1989). Viruses were generated by transfecting Sf21 cells with the constructed bacmids using Fugene6 transfection reagent (Promega, Mannheim, Germany). Cells were incubated for 3 days, and the culture medium (Insect Xpress, Biozym, Hessisch Oldendorf, Germany) containing the recombinant virus was collected. For virus amplification, the corresponding medium was employed to re-infect Sf21 cells. To track infection efficiency, expression of eGFP, inserted into the virus backbone, was monitored. Infected and non-infected cells were cultured in 75 cm² flasks and harvested by centrifugation (300 g, 5 min).

Drosophila Rab7–His was purified from E. coli BL21 (DE3) Rosetta cells. Cells were grown to an OD₆₀₀ of 0.5 at 37°C and induced with 0.2 mM IPTG overnight at 16°C. Cells were lysed in lysis buffer [20 mM HEPES-NaOH, pH 7.4, 150 mM NaCl, 10% glycerol, 1 mM PMSF and 1× protease inhibitor cocktail (0.1 mg/ml of leupeptin, 1 mM o-phenanthroline, 0.5 mg/ml of pepstatin A and 0.1 mM Pefabloc)] using a Microfluidiser (model M-110L, Microfluidics, Newton, USA). Lysates were centrifuged for 20 min at 30,000 g, and the cleared supernatant was incubated with Ni-NTA beads for 2 h at 4°C. Beads were washed with 20 ml cold lysis buffer containing 20 mM imidazole. Bound protein was eluted with buffer containing 300 mM imidazole. Buffer was exchanged to remove imidazole through a NAP-10 column (GE Healthcare, Penzberg, Germany).

Sf21 insect cells were infected with recombinant viruses encoding Mon1–Ccz1, Mon1–Ccz1–Bulli or Bulli for 72 h. Cells were harvested by centrifugation (300 g, 5 min) and lysed in buffer (20 mM HEPES-NaOH; pH 7.4, 150 mM NaCl, and 10% Glycerol with 1× protease inhibitor cocktail mix) with a glass-Teflon homogeniser followed by sonication. Further lysis was conducted by addition of 0.1% CHAPS and incubation for 30 min at 4°C. Lysates were centrifuged at high speed (30,000 g) for 30 min, and cleared lysate was incubated with prewashed Glutathione Sepharose 4B (GE Healthcare, Dach, Germany) or anti-FLAG M2 affinity gel (Sigma-Aldrich, Darmstadt, Germany) for Bulli for 2 h at 4°C. Subsequently, beads were transferred to a Mobicol column (MoBiTec GmbH, Memmingen, Germany) and washed with 15 ml buffer. To elute the complexes, the sample was incubated with PreScission protease (20 µg/ml) at 4°C for 16 h. The eluted complex was subsequently subjected to gel-filtration chromatography using a Superdex^® 200 Increase 10/300 GL column (GE Healthcare, Dach, Germany). Peak fractions were collected and analysed by SDS–PAGE. For Bulli, bound protein was eluted with FLAG peptide (200 µg/ml). The SF21 expression protocol has been established previously (Hallier et al., 2016).

To perform GST pulldown binding assays, GST or GST-fused Mon1–Ccz1 were used as bait, and Bulli-FLAG as prey. GST or GST-tagged proteins (20 µg) were incubated with Glutathione Sepharose 4B (GE Healthcare, Dach, Germany) for 1 h at 4°C on a rotating wheel. Beads were washed three times with buffer (20 mM HEPES-NaOH pH 7.4, 150 mM NaCl and 0.1% NP-40) and immobilised proteins were then incubated with Bulli (5 µg) for 2 h at 4°C on a rotating wheel. Beads were washed three times with buffer. Bound proteins were eluted by boiling at 95°C for 10 min in SDS-sample buffer, resolved by SDS–PAGE, and interaction was analysed by immunoblotting with monoclonal mouse anti-FLAG M2 antibody (stock solution 1 mg/ml, working dilution 1:1000; Sigma-Aldrich, Darmstadt, Germany).

GEF activity of trimeric and dimeric complexes was determined in solution and on membrane surfaces as described previously (Cabrera et al., 2014b; Langemeyer et al., 2014; Nordmann et al., 2010). MLVs consisting of 83 mol% palmitoyl-oleoyl phosphatidylcholine, 10 mol% palmitoyl-oleoyl phosphatidylserine, 5 mol% 1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)succinyl] (DOGS-NTA), 2 mol% phosphatidylinositol 3-phosphate (PI3P), were prepared, and 50 nmoles of MLVs were used in the assay. A total of 50 pmoles of Rab7–His were preloaded with MANT-GDP and incubated with MLVs for 15 min at 26°C before addition of the GEF (dimeric and trimeric complexes). In solution GEF activity was measured by omitting MLVs from the assay. MANT-GDP fluorescence was detected in a SpectraMax M3 Multi-Mode microplate reader (Molecular Devices, Biberach, Germany). Samples were excited at 355 nm, and emission was detected at 448 nm. After allowing for baseline stabilisation for 15 min, 0.1 mM GTP was added to trigger the exchange reaction. The decrease of MANT-GDP fluorescence indicated nucleotide exchange. K_obs was obtained for different concentrations of GEF using a nonlinear exponential, first-order decay equation via GraphPad Prism8 software (GraphPad Software, San Diego, USA). K_obs was plotted against the GEF concentration and slope was determined to calculate k_cat/K_m.

da>CG8270–eGFP flies as well as their control flies, da>eGFP, were raised at 27°C and collected one week after hatching at the latest. Adult flies (40 animals per experiment) were anaesthetised with carbon dioxide, collected and instantly frozen in liquid nitrogen. A total of 500 μl lysis buffer [150 mM NaCl, 50 mM HEPES-NaOH, pH 7.4, 1 mM MgCl₂, 0.15% IGEPAL, 10% glycerol and protease inhibitors (Mix G, Serva, Heidelberg, Germany)] and 500 μl acid-washed glass beads were added to the specimen. Probes were lysed twice with the S. cerevisiae program on a FastPrep^®-24 homogenizer (MP Biomedicals, Santa Ana, USA) with a resting time of 2 min on ice. Lysates were separated from debris via centrifugation (4000 g, 4°C) and further cleared by centrifugation at 14,000 g at 4°C for 5 min. A volume of 1 ml of supernatant was incubated with 15 µl GFP-Trap Beads (GFP-Trap^®_A, Chromotek, Martinsried, Germany) in a new test tube for 30 min at 4°C under constant movement. Prior to use, beads were washed twice with lysis buffer. After incubation, beads were separated from the supernatant via centrifugation (1000 g, 2 min, 4°C) and washed twice with 1 ml of lysis buffer and four times with 1 ml lysis buffer lacking IGEPAL.

Elution of bound proteins was done by adding 50 µl of denaturing buffer (6 M urea, 100 mM Tris-HCl, pH 8.5, 100 mM DTT). Samples were incubated at room temperature for 30 min. Afterwards, 55 mM iodoacetamide was added for alkylation (20 min in the dark at room temperature). Subsequently, 1 µl of trypsin (1 µg/µl, Promega, Mannheim, Germany) was added, and proteolytic digestion was conducted at 37°C overnight. Resulting peptide samples were centrifuged twice (2 min, 10,000 g), and the supernatant was subjected to mass spectrometry (MS) analysis.

Analysis of the peptide mixture was performed as described previously (Eising et al., 2019; Fröhlich et al., 2013; González Montoro et al., 2018). Briefly, peptides were separated on a 50 cm PepMap^® C18 easy-spray column with a 75 µm inner diameter (Thermo Fisher Scientific, Dreieich, Germany) on a Thermo Ultimate 3000 RSLCnano HPLC (Thermo Fisher Scientific, Dreieich, Germany). Peptides were eluted at a constant flow rate of 300 nl for 112 min with a linear acetonitrile gradient from 10 to 35% and directly sprayed into an online coupled Q ExactivePLUS mass spectrometer (Thermo Fisher Scientific, Dreieich, Germany). Mass spectra were acquired in a data-dependent mode to automatically switch between full-scan MS and up to ten data-dependent MS/MS scans. The maximum injection time for full scans was 50 ms with a target value of 3,000,000 at a resolution of 70,000 at m/z=200. The ten most intense multiply charged ions (z≥2) from the survey scan were selected with an isolation width of 1.6 Th and fragmented with higher energy collision dissociation (Olsen et al., 2007) with normalised collision energies of 27. Target values for MS/MS were set at 100,000 with a maximum injection time of 80 ms at a resolution of 17,500 at m/z=200. The resulting MS and MS/MS spectra were analysed via MaxQuant (version 1.6.0.13, www.maxquant.org) with its integrated ANDROMEDA search algorithm (Cox and Mann, 2008; Cox et al., 2011; Fröhlich et al., 2013). Carbamidomethylation of cysteine was set as fixed, and methionine oxidation as well as N-terminal acetylation set as variable modifications. Maximum mass deviation was 6 ppm for MS peaks and 20 ppm for MS/MS peaks with a maximum of two missed cleavages allowed and a minimum peptide length of six amino acids. Label-free quantitation was performed utilising the QUBIC software package as described previously (Hubner et al., 2010). All calculations and plots were carried out with the R software package (https://www.r-project.org/). A list of all proteins identified in the label-free pulldown experiments with Bulli are shown in Table S1.

We thank Kerstin Etzold, Birgit Hemmis, Mechthild Krabusch, Martina Biedermann and Angela Perz for expert technical assistance, Eva Cordes for sequencing bulli alleles, Melanie Benetze for her help during the CRISPR mutagenesis process and Maria Schischkin for help with uptake assays and lifespan analyses. We thank Thomas Klein, Allison Gillingham and Sean Munro for discussing results prior to publication.