The apical protein Apnoia interacts with Crumbs to regulate tracheal growth and inflation

Most organs of multicellular organisms are built from epithelial tubes. To exert their functions, tubes rely on apico-basal polarity, on junctions, which form a barrier to separate the inside from the outside, and on a proper lumen, required for gas or liquid transport. Here we identify apnoia (apn), a novel Drosophila gene required for tracheal tube elongation and lumen stability at larval stages. Larvae lacking Apn show abnormal tracheal inflation and twisted airway tubes, but no obvious defects in early steps of tracheal maturation. apn encodes a transmembrane protein, primarily expressed in the tracheae, which exerts its function by controlling the localization of Crumbs (Crb), an evolutionarily conserved apical determinant. Apn physically interacts with Crb to control its localization and maintenance at the apical membrane of developing airways. In apn mutant tracheal cells, Crb fails to localize apically and is trapped in retromer-positive vesicles. Consistent with the role of Crb in apical membrane growth, RNAi-mediated knockdown of Crb results in decreased apical surface growth of tracheal cells and impaired axial elongation of the dorsal trunk. We conclude that Apn is a novel regulator of tracheal tube expansion in larval tracheae, the function of which is mediated by Crb.

Introduction Animal organs consist of epithelial tissues, which form the boundaries between internal and external environment [1][2][3]. During development, epithelia are instrumental to shape the various organs. Many epithelial tissues form tubular organs, such as the gut, the kidney or the respiratory system. A fundamental feature of epithelial tubes and sheets is to keep the balance between the maintenance of structural integrity and tissue rigidity during organ growth and morphogenesis. To understand how this balance is achieved during rapid, temporally regulated developmental transitions from juvenile to adult body shapes, several studies in various animal models have focused on elucidating how cell proliferation, cell polarity, cell shape changes and trafficking contribute to the formation of the tubular lumen length and diameter [4][5][6][7]. The correct coordination of these processes is crucial for normal organ function. This is reflected in the fact that several human diseases are linked to defects in epithelial tube formation and maintenance, such as polycystic kidney disease or cystic fibrosis [8][9][10][11].
The developing tracheae of Drosophila melanogaster, a network of branched epithelial tubes that ensure oxygen supply to the cells of the body, has emerged as an ideal system to study cell fate determination and morphogenesis of epithelial tubes. The available genetic tools, as well as the ease to image the tracheal system in the fly embryo, has provided detailed insights into the developmental processes required to form tubular structures with defined functional lumens and have contributed to elucidating the interplay between tissue growth, differentiation and cell polarity [12][13][14][15].
The stereotypically branched tracheal system of Drosophila is set up at mid-embryogenesis. Once a continuous tubular network has formed, the tube expands to accommodate an increased oxygen supply to all tissues during animal growth. Tube expansion occurs by growth along the diameter and along the anterior-posterior axis. Growth is accompanied by the formation of a transient cable, comprised of a chitinous apical extracellular matrix (aECM), which fills the lumen of the tube. The generation of this cable requires the secretion of chitin and chitin-modifying enzymes. Mutations in genes affecting secretion or organization of the chitin cable result in excessively elongated tracheal tubes or tubes with irregular diameter (with constricted and swollen areas along the tube) [15][16][17]. Axial growth, on the other hand, depends on the proper elongation of tracheal cells along the anterior-posterior axis. At later stages of embryogenesis, the lumen becomes cleared and filled with air.
After hatching, the larvae undergo two molts, a process during which animals rapidly shed and replace their exoskeleton with a new one, bigger in size. For this, a new chitinous aECM is secreted apically, thus surrounding the old tube. Remodeling of this aECM permits tissue growth between larval molts. The molting process is initiated by the separation of the old aECM from the apical surface of the epithelial cells and the secretion of chitinases and proteinases, which partially degrade the old cuticle. The remnants of the old cuticle in each metamer are shed through the spiracular branches. This process, called ecdysis, is followed immediately by clearance of the molting fluid and air filling [18][19][20]. Interestingly, while the diameter of the dorsal trunk only increases at each molt, tube length increases continuously throughout larval life, particularly during intermolt periods [14]. Despite the importance of tube expansion and elongation for larval development [18], the underlying mechanisms that control tracheal growth at this stage remain poorly understood.
A well-established regulator of apical domain size in developing epithelia is Crumbs (Crb). Crb is a type I transmembrane protein, which acts as an apical determinant of epithelial tissues [21]. It has a large extracellular, a single transmembrane and a short cytoplasmic domain. Loss-and gain-of-function experiments have shown that apical levels of Crb are important for proper cell polarity, tissue integrity and growth. For instance, absence of Crb in embryonic epithelia results in loss of apical identity and disruption of epithelial organization [22][23][24]. In contrast, overexpression of Crb triggers apical membrane expansion, which leads to a disordered epithelium, abnormal expansion of tracheal tubes and/or tissue overgrowth [21,[25][26][27][28][29][30][31][32][33]. These results underscore the importance of Crb levels for epithelial development and homeostasis.
To gain further insight into the molecular mechanisms that regulate Crb and its activity during epithelial growth, we set out to identify novel interacting partners of Crb using the yeast two-hybrid system. One of the candidates identified, CG15887, encodes a transmembrane protein, which localizes to the apical surface of tracheal tubes. We found that the CG15887 protein physically interacts with Crb. Based on the phenotype of mutations in CG15887, which is characterized by defects in tracheal growth and inflation during larval stages, we named this gene apnoia (apn). apn mutant animals die as second instar larvae with dorsal trunks displaying reduced axial growth and impaired apical surface area expansion, resulting in shorter tubes. This phenotype is correlated with the absence of Crb from the apical surface. RNAi knock-down of crb phenocopies the apn mutant phenotype of impaired longitudinal growth. These results identify Apn as a novel regulator of tracheal tube growth in the larvae, which acts through Crb to control axial tube expansion.

Apnoia is an apical transmembrane protein expressed in Drosophila tracheae
To identify novel interactors of Crb, we searched for binding partners using a modified yeast two-hybrid screen (MBmate Y2H) [42,43], allowing bait and prey to interact at the yeast plasma membrane. The bait consisted of the C-terminal-most extracellular EGF (epidermal growth factor)-like repeat, the transmembrane domain and the cytoplasmic tail of Drosophila Crb. One of the Crb interacting clones contained a 414bp cDNA insert encoded by the fulllength CG15887 gene. Based on the tracheal inflation phenotype described below we named the gene apnoia (apn) (άπνοια, Greek for: lack of air).
The apn mRNA encodes a single protein isoform of 137 amino acids. Apn is predicted to contain a signal peptide at the amino terminus (1-23 aa) and two transmembrane domains (amino acids 50-72 and 79-101), based on the TMHMM transmembrane algorithm prediction [44]. Both the amino and carboxy terminus are located extracellularly, separated by a small intracellular loop (S1 Fig). The PFAM algorithm (PFAM domains database 27.0) predicts that Apn contains two LPAM domains (47-56 aa and 78-90 aa), known as prokaryotic membrane lipoprotein lipid attachment sites. Apn is highly conserved within the insect order (S1 Fig) but does not appear to have a true orthologue in vertebrates.
To determine the tissue distribution and subcellular localization of apn mRNA and Apn protein we performed in situ hybridizations and immunostainings of wild-type or transgenic animals, which either carried the fosapn sfGFP , a fosmid encoding the Apn protein C-terminally tagged with superfolded (sf) GFP [45], or a UAS-transgene encoding fluorescently-tagged Apn (UAS-apn mCitrine ). In addition, anti-Apn antibodies were raised in rabbits against a peptide of the N-terminal extracellular domain (aa [24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40]. Expression of both apn mRNA (Fig 1A-1C) and Apn protein (Fig 1D-1F and S2A and S2B Fig) were first detected in embryos at stage 13 in tracheal fusion cells. During embryonic stages 15 and 16, expression could also be detected in the dorsal and lateral trunks, in the visceral and dorsal branches and in the transverse connective branches. In the larvae, Apn is continuously expressed in the entire tracheal system (S2C and S2D Fig). As shown by antibody staining or Apn mCitrine fluorescence, Apn is restricted to the apical plasma membrane, where it co-localizes with the apical markers Stranded at second (Sas) (Fig 1G-1G" and cross section in 1G‴) and Uninflatable (Uif) (Fig 1H-1H" and cross section in 1H‴). Apn co-localizes with Crb in the subapical region, a small region of the apical membrane apical to the adherens junctions (AJ) (Fig 1I-1I" and cross section in 1I‴).
This co-localization and the interaction in the yeast 2-hybrid system (S2E Fig) prompted us to further analyze the interaction between Apn and Crb in co-immunoprecipitation experiments. Full-length Apn (Apn FL ) expressed in S2R + cells co-immunoprecipitated the full-length Crb (Crb FL ) ( Fig 1J). In situ interactions between Crb and Apn were corroborated by Proximity Ligation Assays (PLA) [46] using fosapn sfGFP . We found that Crb and Apn-sfGFP interact in the larval tracheae (Fig 2A-2A", 2D-2D" and 2C), whereas no interaction between Apn and the Drosophila E-cadherin, tagged to GFP (DEcad-GFP) (negative control), was detected ( Fig 2B-2B", 2E-2E" and 2C), indicating that the observed signal was specific for the Crb-Apn interaction. To exclude any random interactions between Crb and Apn in the apical domain we have tested a different apical protein (SAS-Venus) [47] for its interaction with Apn and found no increased PLA signal as compared to the signal between Crb and Apn (  The uninflated tubes observed in apn 1 deficient animals suggested defects in tracheal maturation. In wild-type embryos as well as in each molting step of larval development, tracheal maturation is characterized by distinct sequential processes: i) secretion of a chitinous apical extracellular matrix (aECM) into the lumen, which confers rigidity to the tube and is responsible for tube expansion; ii) a pulse of endocytosis, resulting in the removal of luminal proteins, and iii) liquid clearance and air filling [16,48]. Electron micrographs of apn 1 mutant larvae revealed a disorganized lumen with "tongues" of cellular protrusions into the lumen (S3G and S3H Fig). This phenotype is probably a consequence of the irregularly twisted tubes (compare  However, the last maturation steps, involving the liquid clearance and gas filling, were strongly affected in apn 1 mutant tracheae (compare Fig 3K, 3K', 3L and 3L'). We could exclude leakage of the septate junctions (SJ) and hence loss of paracellular barrier from being a cause of this phenotype, since Contactin (Con) and Discs Large (Dlg), two SJ components [53,54] were properly localized in the tracheae of apn 1 mutants (S3M-S3N' Fig).
Taken together, our data demonstrate that loss of apn affects the late steps of tracheal tube maturation, including liquid clearance and gas filling, and impairs the growth and morphology of the dorsal trunk at the second larval stage.

Apn supports apical membrane growth in larval tracheae
A striking defect observed in apn 1 mutant larvae was the reduction in the length of the dorsal trunk ( Fig 3E-3H). To determine the cellular basis of this phenotype we stained for DEcad to visualize the cell outline. We could not detect significant differences in cell number within different metameres. This led us to hypothesize that shortening of tracheal tubes is caused by defective apical cell surface expansion. Therefore, we measured the long and the short axes of cells (referred to as axial and circumferential length, respectively) (see Fig 4A) as well as their cell surface area. While the circumferential cell length was not significantly different, the axial cell length of apn 1 mutants was reduced in comparison to that of wild type cells (Fig 4A, 4B, 4E and 4F). This difference was also reflected by a reduced aspect ratio of the two axes (axial to circumferential length) ( Fig 4G) and the overall reduction of the apical surface area (Fig 4C,  4D and 4H). From these results we conclude that Apn is required for anisotropic apical surface expansion and hence tracheal tube elongation.

Apnoia is required for maintenance of Crumbs on the apical membrane of tracheal cells
Regulation of apical cell surface area during axial growth of tracheal tubes has been shown to require junctional and polarity proteins as well as the apical protein Uif [55][56][57][58]. Therefore, to Rigid and gas filled tracheal tubes are present in wild type (WT) larvae (A), whereas absence of apn causes twisted and gas deficient tubes (B). apn down-regulation (apn RNAi) recapitulates apn 1 mutant tracheal defects (C). Tube morphology defects are rescued by one genomic copy of apn (fosapn mCherry.NLS ) (D). Anterior is to the left. Scale bar: 500μm. (E-G) The 9 th metamer of the dorsal trunk (yellow dotted line) of apn 1 mutant second instar larvae (F) is shorter than that of WT larvae (E). Expression of one genomic copy of apn (fosapn mCherry.NLS ) significantly rescues the metamer elongation defects of apn mutant larvae (G). Anterior is to the left. Scale bar: 200μm. (H) Quantification of the length of the 9 th metamer of second instar tracheal tubes of WT, apn 1 mutants and apn 1 mutants rescued with one extra apn copy (fosapn mCherry.NLS ;apn 1 ). Measurements were pooled from 6 larvae. (I-J) Chitin binding probe (CBP) allows the visualization of the tracheal tube structure. Note the wrinkled and twisted tube of apn 1 mutants (J) as compared to that of WT (I) larvae. Scale bar: 50μm. (K-L') Brightfield images to show dorsal trunk diameter expansion (K, L) and gas filling (K', L') of the newly formed tracheal tube of WT (K, K') and apn 1 mutant (L, L') second larval instar. Yellow indicates the tracheal cells that line the newly formed lumen (green). Note the bubble filling the newly formed lumen in WT (white arrow) (K'), which is absent in the tube of apn 1 mutants (L'), which fails to fill with gas. Scale bar: 50μm.
https://doi.org/10.1371/journal.pgen.1007852.g003 better understand the mechanism by which apn ensures apical membrane growth, we examined the subcellular distribution of junctional and polarity proteins in the tracheae of apn 1 mutants. The AJ markers Armadillo (Arm), the Drosophila β-catenin [59] (Fig 5A and 5B), Polychaetoid (Pyd), the single Drosophila ZO-1 orthologue [60] (Fig 5C and 5D) and DEcad (Fig 5E and 5F) localized similar to the wild type tracheae. apn 1 mutant tracheal cells also showed normal distribution of Uif (Fig 5G and 5H). These results indicate no major defects in apico-basal polarity and epithelial integrity of the tracheal tube in apn 1 mutant larvae. The physical interaction between Apn and Crb motivated us to analyze the expression of Crb in apn 1 mutants. In wild type tracheal cells of second instar larvae, Crb is localized in the subapical region, outlining the cell (Fig 5I). In contrast, Crb strongly accumulated in cytoplasmic vesicles of multicellular, autocellular and seamless tubes in apn 1 mutant tracheae and upon knock-down of apn (Fig 5J and 5R and S4A-S4D' Fig). Consistent with these results, not only multicellular, but also autocellular and seamless tubes were twisted and uninflated (S4E-S4F' Fig). However, the total protein levels of Crb were unchanged as revealed by western blotting (S2F Fig). To investigate whether apn is required for Crb apical localization only in the trachea, we analyzed another epithelial tube, the salivary gland. A uniform apical localization of Crb was observed in both wild type and apn 1 mutant salivary glands, indicating a tracheae-specific role of apn (S4H-S4I' Fig).
Similar to Crb, Stardust (Sdt) (Fig 5K and 5L) and Moesin (Moe) (Fig 5M and 5N), whose subapical localization depend on Crb in many epithelia [61][62][63][64], are found in the same vesicular compartments as Crb. The introduction of one copy of the apn genomic locus (fosapn m-Cherry.NLS ) into the apn 1 mutant background restored Crb membrane localization and suppressed the accumulation of Crumbs loaded vesicles (CLVs) (Fig 5O-5Q). These results indicate that Apn is required for Crb trafficking to or maintenance at the plasma membrane of tracheal cells.
In order to distinguish between these two possibilities, we blocked endocytosis in apn 1 mutant tracheae chemically and genetically. After 2 hours incubation with dynasore, an inhibitor of Dynamin [65], Crb was mostly localized at the plasma membrane in apn 1 mutant tracheal cells (Fig 6A). In contrast, apn 1 mutant cells incubated with dynasore-free medium showed only punctate staining of Crb (Fig 6B). To corroborate this result, we blocked endocytosis by using shibire ts1 (shi ts1 ), a temperature-sensitive allele of shi, which encodes Dynamin. When incubated at the restrictive temperature (34˚C) shi ts1 ;apn 1 double mutant tracheae retained Crb at the apical plasma membrane (Fig 6C), as compared to shi ts1 ;apn 1 mutant tracheae, incubated at the permissive temperature (25˚C) (Fig 6D). From these results we concluded that Apn is required for Crb maintenance at the apical membrane.
We noticed that the majority of Vps35-positive vesicles were significantly larger in apn 1 mutants, measuring around 0.7μm (n = 194 vesicles) in diameter, as compared to 0.27μm (n = 152 vesicles) in control larval tracheal cells (Fig 8B). No significant size differences in two other trafficking compartments, such as the Arl8-(lysosomal) and the Golgin245-(trans-Golgi) [68] positive vesicles, were observed between the two genotypes (Fig 8C and 8D). This result suggests that the size increase specifically in the Vps35-positive compartment is an aspect of the apn 1 mutant phenotype.
Taken together, these results suggest that Apn maintains apical Crb by preventing its clathrin-dependent endocytosis. Loss of apn results in Crb accumulation in Vps35/retromer-positive vesicles of increased size.

Tracheal defects caused by apn depletion are mediated by crb
Since loss of apical Crb is often associated with reduced apical membrane [28,69,70] and Crb is depleted from the apical membrane in apn 1 mutant tracheal cells, we asked whether the impaired apical surface growth observed in apn 1 mutant tracheae is due to its effect on apical Crb. Since homozygous crb mutant embryos die with severe defects in many epithelia, including the tracheae [24,71], we knocked-down crb in tracheal tubes by expressing crb RNAi ubiquitously (using da-Gal4) or specifically in the tracheae (using btl-Gal4). This resulted in a strong depletion of Crb and its binding partner Sdt (Fig 9A, 9B, 9A' and 9B'), but had no effect on Apn expression and localization (Fig 9C, 9D, 9C' and 9'D). RNAi-mediated downregulation of crb reproduced several aspects of the apn 1 mutant phenotypes, such as twisted tracheal tubes, lack of gas filling (Fig 9E-9H) and reduced apical surfaces of tracheal tube cells (Fig 9I-9K). No defect in apico-basal polarity was observed upon knockdown of Crb (Fig 9I, 9J, 9I" and 9J"). In addition, most animals died at L2 (larval stage 2) with some surviving until L3 (larval stage 3).
To assess whether the increased size of Vps35 positive vesicles in apn 1 mutants are due to Crb accumulation in these vesicles, we knocked-down crb in apn 1 tracheae using btl-Gal4. We found a small, yet significant, reduction in the size of Vps35 positive vesicles in apn 1 tracheal cells upon crb RNAi expression, compared to that of apn 1 single mutants (Fig 10A, 10B and  10D and S6A-S6C Fig). In contrast, Vps35 positive vesicles in tracheal cells expressing crb RNAi in otherwise wild-type animals are comparable in size to those of wild type Vps35 vesicles (compare Fig 10C and 10D and Fig 8B).
These results are the first to show that loss of crb results in a reduction of the apical surface area of larval tracheal cells, which in turn prevents proper tube elongation. In addition, they identify Apn as a novel regulator of apical Crb in the developing tracheae, which controls dorsal trunk maturation and expansion. Absence of apn leads to accumulation of Crb in Vps35 positive vesicles, which may contribute to the increase in vesicular size.

Discussion
This work identifies Apn as an essential protein for airway maturation in Drosophila larval stages. Apn is localized apically in tracheal epithelial cells, where it co-localizes and physically interacts with Crb. apn 1 mutant larvae exhibit loss of tracheal tissue structure, manifested by tube size defects and impaired gas filling, resulting in body size reduction and lethality at second instar. At the cellular level, exclusion of Crb from the apical membrane in apn 1 mutant larval tracheae goes along with apical cell surface reduction and an overall tracheal tube shortening. Absence of apn leads to Crb inhibition and accumulation in enlarged, Vps35/retromerpositive vesicles.
Elongation of the tracheal tube has been extensively studied in embryos where it has been shown to rely on different mechanisms, such as the organization of the aECM and cell shape changes [33,[72][73][74][75][76]. Anisotropic growth of the apical plasma membrane is an additional mechanism to achieve proper longitudinal tube expansion. However, only few proteins have been described so far to regulate this process. One of these, the protein kinase Src42A, is required for the expansion of the cells in the axial direction, and loss of Src42A function results in tube length shortening, which is associated with an increased tube diameter [72,75,77]. Src42A has been suggested to exert its function, at least in part, by controlling DEcad recycling and hence adherens junctions remodeling [72] and/or by its interaction with the Diaphanous-related formin dDAAM (Drosophila Dishevelled-Associated Activator of Morphogenesis), loss of which results in reduced apical levels of activated pSrc42A [75]. More recently, Src42A has been suggested to control axial expansion by inducing anisotropic localization of Crb preferentially along the longitudinal junctions [78]. However, we never observed any anisotropic distribution of Crb in wild type larval tracheal cells, making it unlikely that, at this developmental stage, axial expansion is regulated by a Src42A-dependent mechanism. This assumption is corroborated by the observation that, unlike in Src42A mutants, the lack of longitudinal expansion in apn 1 mutant larval tubes is not associated with circumferential expansion. Another protein regulating tube elongation in the embryo is the epidermal growth factor receptor, EGFR. Expressing a constitutively active EGFR results in shortened tracheal tubes with smaller apical cell surfaces, but with increased diametrical growth. In this condition, Crb shows altered apical distribution [78,79]. This phenotype differs from the apn 1 phenotype, where apical localization of Crb is almost completely lost and only longitudinal tube growth is affected. This suggests that Apn executes tube length expansion by a different mechanism.
How does decrease in tubular growth lead to loss of tracheal structure? During development, the larval body, including the tracheal tissue, elongates about 8-fold [18]. Impaired axial tracheal cell growth in apn 1 mutants thus may affect the balance between the forces exerted by apical membrane growth on the one hand and the resistance provided by the luminal aECM on the other, an important mechanism described previously to control tube shape in the embryo [33]. This could lead to physical rupture of tubes mutant for apn 1 , allowing fluid entry. The presence of fluid would, in turn, disrupt proper gas filling, resulting in hypoxia and, consequently, in impaired body growth.
Several studies have shown that, in some tissues, Crb accumulation on the apical membrane is mediated by the retromer complex, which controls either the retrograde transport of Crb to the trans-Golgi [80] or the direct trafficking from the endosomes to the plasma membrane [39,41,79]. The physical interaction of Apn and Crb, the functional requirement of Apn for Crb apical localization and the fact that in apn 1 mutants Crb is trapped in Vps35-positive/retromer vesicles all suggest that Apn is required for trafficking and/or maintenance of Crb at the apical membrane (Fig 11).
However, the increase in the size of Vps35-positive vesicles in apn 1 mutant cells, which is, to some extent, due to the accumulation of Crb, suggests defects in retromer function, which may prevent Crb lysosomal degradation. Further studies will help to elucidate at which level Apn controls Crb trafficking in larval tracheae.

Generation of apn 1 mutant allele by CRISPR/Cas9
Target sites were designed using the settings of the flyCRISPR Optimal Target Finder (http:// tools.flycrispr.molbio.wisc.edu/targetFinder/), to guide Cas9 to two target sites, one at the In the wild type (WT) dorsal trunk, the apical membrane grows along the axial axis (thick red arrows) and pulls the apical extracellular matrix (aECM), until the aECM resistance (indicated by green arrows) balances the forces provoked by tube elongation [33]. Bottom: In WT tracheal cells, Crb (blue) is enriched at the apical membrane where it controls apical surface growth. Apn (green) in the apical membrane is responsible for Crb maintenance and therefore ensures tube elongation. Crb trafficking involves recycling by the retromer to either the trans-Golgi network (TGN; blue vesicles) or to the plasma membrane (red vesicles), or by Rab11, or its degradation in the lysosome. (B) Top: In the apn 1 mutant dorsal trunk, the pulling forces of the apical membrane expansion are decreased due to decreased growth of the apical membrane (thin red arrows), whereas the forces mediated by the aECM are likely to remain unchanged, causing breakage of the tube. Bottom: In the absence of Apn, Crb is depleted from the apical surface due to increased endocytosis. Crb is trapped in enlarged, Vps35 (retromer)-positive vesicles. It fails to be recycled (as shown by the lack of colocalization with Rab11), but is also not degraded, pointing to a functional defect of the retromer.
Dynasore treatment of larval tracheae apn 1 mutant tracheae from second instar larvae were dissected in Grace's medium supplemented with Pen/Strep. Tissues were incubated in 60μM dynasore (Enzo Life Sciences) in Grace's medium containing Pen/Strep and 2.5% FCS at room temperature for 2hr. The dynasore was washed out and tracheae were fixed in 4% FA in Grace's medium for 30min.

Yeast-two-Hybrid screen
Part of the coding sequence of a Drosophila melanogaster crb cDNA (encoding aa: 2034-2189) (GenBank accession number NM_001043286.1) was PCR-amplified and cloned in pB102, in frame with the STE2 leader sequence at the N-terminus and ubiquitin (Cub) at the C-terminus of the bait which is coupled to the artificial transcription factor LexA-VP16 (STE2-Crb-Cub-LexA-VP16). The construct was verified by sequencing. Prey fragments were isolated from a MBmate screen with Drosophila melanogaster Crb as bait against a Drosophila Embryo NubGx (D3DE_dT) library (NubG stands for the N-terminal domain of mutated ubiquitin and x for the prey fragment). Interaction pairs were tested in duplicate as two independent clones. For each interaction, several dilutions (undiluted, 10 −1 , 10 −2 , 10 −3 ) of the diploid yeast cells (culture normalized at 5x10 7 cells) and expressing both bait and prey constructs were spotted in selective media. The DO-2 selective medium lacking tryptophan and leucine was used to control for growth and to verify the presence of both the bait and prey plasmids. The different dilutions were also spotted on a selective medium without tryptophan, leucine and histidine (DO-3). Six different concentrations of 3-AT, an inhibitor of the HIS3 gene product, were added to the DO-3 plates to increase stringency and reduce possible auto-activation. The following 3-AT concentrations were tested: 1, 5, 10, 50, 100 and 200mM. The 1-by-1 Yeast two-hybrid assays were performed by Hybrigenics Services, S.A.S., Paris, France (http://www. hybrigenics.com).

Cell culture and transfection
Drosophila S2R + cells were cultured at 25˚C in Schneider's Drosophila medium (Sigma) supplemented with 10% fetal bovine serum. pAct5-Gal4 together with UAS-apn FL and/or UAScrb FL [21] (encoding full-length Apn and Crb, respectively) was transfected into S2R + cells using FuGENE HD (Promega) according to the manufacturer's protocol.

Immunoprecipitation
Transfected cells were harvested after 48h, washed with ice-cold PBS (120mM NaCl in phosphate buffer at pH 6.7), resuspended in lysis buffer (containing 10% glycerol; 1% Triton X-100; 1.5mM MgCl 2 ; 120mM NaCl; 100mM PIPES, pH 6.8; 3mM CaCl 2 ; 1 mM PMSF and Complete). Cells were lysed on ice for 20min and lysates were spinned at 14.000rpm for 20min at 4˚C. The supernatant was incubated with the antibody for 2. In the meantime 50μl of Protein G were washed 3 times with blocking solution and incubated with the antibody solution overnight at 4˚C. Protein G beads were collected by centrifugation for 2min at 3.000rpm and washed 4 times with lysis buffer. Beads were resuspended in 1.5x SDS sample buffer and heated for 5min at 95˚C.

Western blot
Wild-type and apn 1 mutant embryos and dissected larval tracheae were homogenized on ice using a Dounce tissue grinder in 1mL of lysis buffer containing 130mM NaCl, 50mM Tris-HCl pH = 8, 0,5% Triton-X and protease inhibitor (Roche). After 30min at 4˚C under rotation the homogenate was centrifuged for 20min at 14.000rpm. Sample buffer 3x SDS was added to the supernatant and boiled for 5min at 95˚C.

Proximity ligation assay (PLA)
Tracheae from fosapn sfGFP third instar larvae were dissected and fixed in ice cold 4% FA in PBS. Primary antibodies against GFP (rabbit anti-GFP 1:250; Invitrogen A11122) and Crb (rat anti-Crb 1:500 [88]) were added and incubated overnight at 4˚C. The Duolink PLA Kit (Sigma) was used to incubate the tissue with the PLA probes PLUS and MINUS at 37˚C for 1h. Ligation of the PLA oligonucleotides and amplification were performed at 37˚C for 30min and 100min, respectively. Samples were mounted in Duolink mounting media and imaged using Zeiss LSM880.

Generation of Apn antiserum
Polyclonal antibodies against CG15887 were raised in rabbits using the KLH-conjugated synthetic peptide QQAANSSDSDSDVAESC (from the N-terminal extracellular part) for immunization. Antibodies were subsequently affinity-purified using the same peptide immobilized on SulfoLink Coupling Gel (ThermoFisher #20401) and following recommendations by the manufacturer. The work was performed by the MPI-CBG Antibody Facility.

Immunohistochemistry
Immunostainings on embryos were done as follows: embryos were dechorionated in 50% bleach for 2min and fixed for 20min in formaldehyde/heptane mixture. After devitellinization in methanol, embryos were permeabilized in 0.1% Triton X-100/PBS except for rabbit anti-Apn staining, for which embryos were permeabilized in 0.2% Saponin/PBS. After washing, embryos were incubated for 1h at RT in blocking solution [(0.5%w/v BSA in PBST/S (0.1%v/v Triton X-100) or (0.2%w/v Saponin)]. Second instar larvae were opened in PBS and fixed in 4% formaldehyde for 20min. After washing in either 0.1% Triton X-100/PBS or 0.2% Saponin/ PBS (for anti-Apn antibody staining), tracheae were dissected and incubated in blocking solution for 1h at RT. Embryos and tracheae were incubated with primary antibodies overnight at 4˚C, washed and incubated with secondary antibodies for 2h at RT. Samples were mounted in Vectashield (Vector Laboratories) and imaged with LSM880 Laser Scanning Confocal Microscope (Carl Zeiss). Unless otherwise indicated, images shown are z-stack projections of sections. Images were processed with Fiji software [89]. Cell area measurements were obtained using the Fiji Freehand selection tool.

RNA in situ hybridization
DIG-labelled RNA probes were synthesized from PCR templates amplified from for a fulllength apn (RE53127) cDNA clone. Sequence specific primers for pFLC-I vector (BDGP resources) were: M13 (-21) 5'-TGTAAAACGACGGCCAGT-3' and M13 (REV) 5'-GGAAA-CAGCTATGACCATG-3'. PCR products were purified by PCR purification columns (Promega, PCR CleanUp system). In vitro transcription reactions were performed by mixing the PCR product with the polymerase mix, which includes T3 RNA polymerase. RNA was labelled with digoxigenin-UTP (Roche Applied Science, #11277073910). Eggs were collected on apple juice plates for 12h. Embryos were dechorionated in 50% bleach for 2min and fixed for 20min in formaldehyde/heptane mixture. After devitellinization in methanol embryos were processed for hybridization, as modified from [91].

Electron microscopy analysis
Larvae were fixed in 2% glutaraldehyde in 0.1M PB buffer pH 7.2 for 20min at room temperature. Larvae were transferred in microcentrifuge tubes and fixed in 1% OsO 4 /2% Glutaraldehyde and then 2% OsO 4 . Further procedures were done according to the protocol described [92]. Ultrathin sections of 0.1μm were prepared and analyzed with Tecnai 12 BioTWIN (FEI Company).

Image analysis
We developed a Fiji script to quantify the co-localization of proteins of the trafficking machinery (e.g. retromer, lysosome, Golgi) with Crb-positive vesicles. Two channel images showing fluorescent Crb signal and protein X signal were imported into a script for the freely available Fiji software [89] and characterized as to their overlap. The plugin was tested on Fiji current version: (Fiji is just ImageJ) ImageJ 2.0.0-rc-65/1.51w. The code of the scripts and its documentation are available on the project repository (https://git.mpi-cbg.de/bioimage-informatics/ Skouloudaki_et_al_Crumbs_overlap_analysis).