Parcas is the predominant Rab11-GEF for rhodopsin transport in Drosophila photoreceptors Free

Sensory neurons exploit a common mechanism of polarized epithelial cell differentiation, amplification and specialization in apical plasma membranes to build sensory organelles. An example is Drosophila photoreceptor morphogenesis, where a late-pupal surge of secretory membrane traffic expands the apical photosensory membrane, the rhabdomere.

Rab proteins are small GTPases that control membrane traffic and maintain distinct organelle identities. More than 60 mammalian and 31 Drosophila Rab proteins regulate specific transport steps and pathways (Stenmark, 2009; Welz et al., 2014; Zhang et al., 2007). In Drosophila photoreceptors, three Rab proteins sequentially regulate the transport of Rh1 (also known as NinaE), the rhodopsin expressed in R1–R6 outer retinal photoreceptor cells. Rab1 and Rab6 regulate Rh1 transport from the ER to the cis-Golgi and from the trans-Golgi network (TGN) to the recycling endosome (RE), respectively (Iwanami et al., 2016; Satoh et al., 1997, 2005). Rab11 and its effectors, Rip11 and MyoV (also known as Didum), allow the Rh1-bearing post-Golgi vesicles to invade the exclusive retinal terminal web, the bundle of actin filaments with plus-ends that are anchored to the microvilli base (Li et al., 2007; Satoh et al., 2005).

The activities of Rab proteins are regulated by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs) (Barr and Lambright, 2010; Ishida et al., 2016). GEFs activate their specific Rab proteins by exchanging a bound GDP for a GTP. GAPs inactivate Rab proteins by facilitating the GTPase activity of the Rab proteins to complete the Rab cycle. Two GEF proteins are reported to be involved in Rh1 transport, specifically, Rich (a Rab6-GEF) and Crag (a Rab11-GEF). Rich-null mutant (Rich¹) photoreceptors exhibit phenotypes similar to those exhibited by Rab6-null mutant photoreceptors (Iwanami et al., 2016); however, the phenotypes of Crag-null mutant (Crag^CJ101) photoreceptors are quite different from those of Rab11-deficient photoreceptors, consistent with the idea that Crag facilitates light-dependent Rh1 transport in adult flies, but not during pupal development (Xiong et al., 2012).

Transport protein particle (TRAPP) complexes were originally identified in yeast as Rab-GEFs, and these proteins share a core set of subunits (Jones et al., 2000; Pinar et al., 2015; Sacher et al., 2008; Thomas and Fromme, 2016; Thomas et al., 2018). A recent study indicated that Drosophila possesses two TRAPP complexes (II and III), and both activate Rab1. One of these complexes, TRAPPII, also activates Rab11 (Riedel et al., 2018). Despite this, specific subunits of the TRAPPII complex (i.e. those present in TRAPPII but not TRAPPIII) are not essential for viability, and one of these subunits has been shown to localize on the cis-side of the Golgi rather than the trans-side of the Golgi or REs, where Rab11 is thought to be activated (Riedel et al., 2018).

REI-1 is another Rab11-GEF that was recently identified in Caenorhabditis elegans. This protein does not contain any known Rab-GEF domain, yet exhibits a strong Rab11-GEF activity in vitro (Sakaguchi et al., 2015). C. elegans models of double-mutant REI-1 and its close paralog REI-2 exhibit a loss of Rab11 localization to the late-Golgi compartments, RE and cleavage furrows, as well as delayed cytokinesis (Sato et al., 2008). Both Drosophila Rab11 and the REI-1 ortholog Parcas (Pcs) are necessary for the specific localization of oskar mRNA and Oskar protein to the oocyte posterior pole, although the mechanism underlying this localization remains unknown (Jankovics et al., 2001; Sinka et al., 2002). In the Pcs-deficient fly line, a hobo insertion into the first intron of the pcs gene, pcs^gs, is not lethal, and this mutant exhibits a grandchildless phenotype possibly caused by the lack of Rab11 activation in the oocyte. Pcs is also involved in muscle morphogenesis and Wnt signaling (Beckett and Baylies, 2006). Further, a recent study indicated that in addition to the TRAPPII complex, Pcs possesses GEF activity toward Rab11 in vitro, and pcs^gs causes synthetic lethality in combination with the null mutation of TRAPPC9, a specific subunit of TRAPPII (Riedel et al., 2018). Here, we investigated the function of Pcs and the TRAPPII complex in Drosophila photoreceptors.

Rab11 is essential for the post-Golgi trafficking of Rh1 to the rhabdomeres, and loss of Rab11 causes Rh1 accumulation within the photoreceptor cytoplasm (Li et al., 2007; Satoh et al., 2005). If Pcs functions as a Rab11-GEF in the developing photoreceptors, lack of Pcs should cause defects in Rh1 transport similar to those caused by the loss of Rab11. Thus, we investigated Rh1 localization in two hypomorphic alleles with transposon insertions, pcs^GS7166 in the first intron, or pcs^NP2623 in the 5′ UTR. Additionally, we used a CRISPR-Cas9 system to generate a deletion mutant allele, pcs^Δ1, which is thought to act as a null allele (Fig. 1A). Although these three alleles are viable, we observed mosaic retinas with both the wild-type and mutant photoreceptors, which allowed us to assess the phenotype relative to the wild type. The yellow lines in images in Fig. 1 demarcate the border between mutant and normal cells in genetic mosaics. We detected cytoplasmic accumulation of Rh1 in the late pupal retinas of pcs^GS7166, pcs^NP2623 and pcs^Δ1 homozygous photoreceptors (Fig. 1B,D; Fig. S1A), indicating that Pcs functions as a possible Rab11-GEF in the developing photoreceptors. The cytoplasmic accumulation of Rh1 disappeared completely in pcs^{GS7166 ex1} homozygous photoreceptors, where the P-element insertion is precisely excised from pcs^GS7166 by the Delta2-3 transposase (Fig. 1C). Thus, P-element insertion into the first intron of the pcs gene is the sole cause of Rh1 accumulation in the pcs^GS7166 photoreceptor cytoplasm. Quantification of the ratio of Rh1 staining in the cytoplasm against that in the whole photoreceptor (including both rhabdomeres and cytoplasm) in wild-type, pcs^GS7166, pcs^{GS7166 ex1} and pcs^Δ1 alleles (Fig. 1H) verified more than 80% of Rh1 staining in the cytoplasm of pcs^GS7166 and pcs^Δ1 homozygous photoreceptors. In pcs^gs, a piggyBac insertion within the first intron that has been reported as a null mutation (Sinka et al., 2002), Rh1 localized to the rhabdomeres (Fig. S1B), suggesting that this piggyBac insertion does not effectively reduce the expression of Pcs protein in the eyes.

Similar to Rab11-deficient photoreceptors, both TRP, a cation channel essential for photo-transduction (Montell and Rubin, 1989), and Chaoptin (Chp), an adhesion molecule bundling microvilli into the rhabdomeres (Reinke et al., 1988), accumulated in the cytoplasm of pcs^GS7166 and pcs^Δ1 photoreceptors (Fig. 1E,G), but not in the cytoplasm of pcs^{GS7166 ex1} photoreceptors (Fig. 1F). Thus, the transport of rhabdomeric proteins was inhibited by the loss of Pcs. Na⁺K⁺ATPase, however, localized normally in basolateral membranes in pcs^GS7166, pcs^Δ1 and pcs^NP2623 homozygous photoreceptors (Fig. 1B,D; Fig. S1A).

A recent study indicated that in Drosophila photoreceptors, PIP4K regulates clathrin-mediated endocytosis (CME) from the plasma membrane, and its absence causes increased CME uptake, leading to an increased number of Rh1-loaded early endocytic vesicles (Kamalesh et al., 2017). This cytoplasmic accumulation of Rh1 in PIP4K-deficient photoreceptors is light-dependent, likely because of light-stimulated endocytosis, and this phenotype is rescued by the inhibition of endocytosis by Rab5 RNAi. To investigate whether Rh1 accumulation in Pcs-deficient cells is caused by increased endocytosis, we investigated Rh1 localization in dark-reared homozygous pcs^GS7166 and pcs^Δ1 flies, and also in pcs^GS7166 mosaic retinas expressing Rab5 RNAi. We found high levels of Rh1 accumulation in the cytoplasm of dark-reared pcs^GS7166 and pcs^Δ1 homozygous photoreceptors (Fig. 1I,J) and also in the cytoplasm of pcs^GS7166 photoreceptors expressing Rab5 RNAi (Fig. 1K). Quantification of Rh1 accumulation indicated that the degree of Rh1 accumulation is not reduced under dark conditions compared to that observed under light conditions, and accumulation is also not affected by the expression of Rab5 RNAi (Fig. 1H). These results indicate that Pcs is likely to regulate biosynthetic pathways and not the endocytosis of Rh1.

To more precisely compare the phenotypes between Pcs deficiency and the loss of the Rab11, we observed thin sections of wild-type w¹¹¹⁸, pcs^NP2623, pcs^GS7166, pcs^{GS7166 ex1} and pcs^Δ1 pupal photoreceptors using electron microscopy (Fig. 2A–H; Fig. S1C,D). We found smaller rhabdomeres and cytoplasmic vesicle accumulations in pcs^GS7166, pcs^NP2623 and pcs^Δ1 photoreceptors, while the wild-type w¹¹¹⁸ and pcs^{GS7166 ex1} photoreceptors remained unchanged. The appearance of these vesicles resembled the vesicles observed previously to accumulate upon deficiency of Rab11 (Li et al., 2007; Satoh et al., 2005). These vesicles look unusual but are different from autophagosomes, which are enclosed by a double membrane, although Rab11 is reported to be involved in autophagy (Longatti et al., 2012; Puri et al., 2018). We measured the area of cytoplasmic vesicles in wild-type, pcs^GS7166, pcs^{GS7166 ex1} and pcs^Δ1 pupal photoreceptors, and found that the areas occupied by the cytoplasmic vesicles were significantly increased in pcs^GS7166 and pcs^Δ1 pupal photoreceptors (Fig. 2I). The cisternae of Golgi units appeared slightly swollen, but they still stacked well (Fig. 2G,H, arrows). Mitochondria and ER were normal in pcs^GS7166 and pcs^Δ1 photoreceptors (Fig. 2G,H, asterisks and arrowheads).

We investigated Rh1 localization by means of electron microscopy using the post-embedding method with LR White resin. Although the detail of membrane structures is not visualized clearly in this methodology, Rh1 seems to localize on the cytoplasmic vesicles accumulated in pcs^GS7166 photoreceptors (Fig. 3A,C,D). By contrast, Rh1 predominantly localizes on the rhabdomeres in pcs^{GS7166 ex1} photoreceptors (Fig. 3B).

Rab11 localizes to the trans-side of Golgi units and the post-Golgi vesicles at the rhabdomere base in fly photoreceptors (Iwanami et al., 2016; Satoh et al., 2005). As Rab11 is widely accepted as a RE marker, the RE might associate with the Golgi in fly photoreceptors. We postulated that the Rab11-positive membrane on the trans-side of Golgi units is the RE in fly photoreceptors, and refer to this as the ‘Golgi-associated RE’.

Next, we sought to compare the localization of Pcs and Rab11. We failed to detect endogenous Pcs proteins using anti-Pcs antibodies, which were kindly gifted by Dr Mary Baylies (Sloan Kettering Institute, New York, USA) and Dr Miklos Erdelyi (Hungarian Academy Sciences, Szeged, Hungary) (Beckett and Baylies, 2006; Sinka et al., 2002). Therefore, we created transgenic flies expressing UAS-V5::Pcs, and expressed V5::Pcs in late pupal photoreceptors using Rh1-Gal4. V5::Pcs co-localized with Rab11 both in the cytoplasmic puncta (Fig. 4A, arrowheads; Fig. S2), which are presumably Golgi-associated RE, and in the puncta at the base of the rhabdomeres (Fig. 4A, arrows; Fig. S2), which are the post-Golgi vesicles bearing newly synthesized rhabdomere proteins. Line plots of fluorescent intensities of anti-V5 and anti-Rab11 antibodies (Fig. 4C; Fig. S2C,F,I) through the perinuclear region from a Golgi unit to the rhabdomere (arrow in Fig. 4B; Fig. S2B,E,H) indicate that the V5::Pcs signal at the Golgi-associated RE was greater than the V5::Pcs signal at the base of the rhabdomere, although the Rab11 signal detected at the base of the rhabdomere was greater than that observed at the Golgi-associated RE.

We next investigated the detailed localization of V5::Pcs to Golgi units using young pupal retina, a tissue that possesses well-developed Golgi units. V5::Pcs signals were localized to the trans-side of the Golgi units, and separated from both the cis-Golgi marker GM130 and medial-Golgi marker p120 (Fig. 4D,E). V5::Pcs, however, co-localized with Rab11 at the RE and extended toward the TGN, where Rab6 localizes (Fig. 4F,G). Thus, Pcs co-localized with Rab11, but Pcs localization was shifted slightly away from Rab11 and toward the upstream TGN in the polarized transport pathway to the rhabdomere. These localization studies suggested that Pcs works upstream of Rab11, and this is in agreement with the idea that Pcs works as a Rab11-GEF.

To investigate Pcs localization using electron micrography (EM), we employed a recently developed genetic EM tag, APEX2, that catalyzes the polymerization and local deposition of DAB, which in turn provides EM contrast after treatment with OsO₄ (Lam et al., 2015; Martell et al., 2012). First we confirmed that myc::APEX2::Pcs associates with Golgi units in stably transformed Drosophila S2 cells by immunostaining using anti-myc antibody (Fig. 4H). In the electron micrographs of S2 cells expressing myc::APEX2::Pcs, electron-dense DAB staining was found on the trans-side of Golgi units (Fig. 4I,J) or at some distance from the trans-side of Golgi units (Fig. 4K,L). DAB staining of myc::APEX2::Pcs primarily associated with 150–300 nm distorted vesicles or cisternae (Fig. 4J–L). These vesicles or cisternae are often clustered. These observations suggest that these membrane structures are likely to represent REs in Drosophila.

We next investigated whether Pcs is required for Rab11 localization to the Golgi-associated RE and post-Golgi vesicles, by means of immunostaining of pcs-deficient mosaic retina. The strong punctate staining of Rab11 seen in the wild-type photoreceptors was absent in pcs^GS7166 and pcs^Δ1 homozygous photoreceptors (Fig. 5A,C,D). There was no significant difference in the staining of the medial-Golgi marker MPPE between the wild-type and homozygous pcs^GS7166 and pcs^Δ1 photoreceptors, indicating that the Golgi units themselves were not eliminated in pcs-deficient photoreceptors, consistent with our observations by electron microscopy (Fig. 2G,H, arrows). The defect in Rab11 localization to Golgi-associated RE and post-Golgi vesicles in pcs^GS7166 was also rescued in pcs^{GS7166 ex1} photoreceptors, indicating that Pcs function is essential for Rab11 localization to these compartments (Fig. 5A,D; Fig. S1E). We also investigated Rab1 localization in pcs^GS7166 mosaic retina (Fig. 5B). In contrast to Rab11, Rab1 localized normally to Golgi units with GM130, indicating that Pcs is not required for the Rab1 localization. Thus, Rab11 localization to Golgi-associated RE and post-Golgi vesicles is dependent upon Pcs, suggesting that Pcs works as a Rab11-GEF in fly photoreceptors.

As Rab-GEFs regulate the recruitment of their target Rab proteins to the specific organelle membrane, we compared the relative amount of Rab11 in membrane fractions in w¹¹¹⁸, pcs^GS7166, pcs^{GS7166 ex1}, w¹¹¹⁸/Df(ED2423) and pcs^NP2623/deletion fly heads. In every genotype test, 52–60% of Rab11 was found in the membrane fraction with no significant differences between the wild-type (w¹¹¹⁸, pcs^{GS7166 ex1}, w¹¹¹⁸/Df(ED2423)) and pcs-deficient (pcs^GS7166, pcs^NP2623/Df(ED2423)) heads (Fig. 5E,H), similar to observations from a previous report (Sakaguchi et al., 2015). We confirmed that membrane and cytoplasm were well fractionated in our procedure (Fig. 5F), and the total amount of Rab11 was unchanged in these flies (Fig. 5G,I). Thus, Pcs is necessary for the specific localization of Rab11 to the trans-side of Golgi units and post-Golgi vesicles (Fig. 5A,C), but the protein is not essential for membrane recruitment (Fig. 5E,H).

To identify the nature of the membrane to which Rab11 localizes under Pcs deficiency, we investigated whether Rab11 co-localizes with an ER marker, Calnexin (Cnx), or Rh1 in cytoplasmic vesicles. In pcs^Δ1 homozygous photoreceptors, Rab11 was partially co-localized with Rh1, but not with Cnx (Fig. 5J,K). Pearson's correlation coefficient between Rab11 and Rh1 was 0.49±0.12 (mean±s.d.), but between Rab11 and Cnx it was 0.03±0.17. These results indicate that some, if not all, of Rab11 was localized on the cytoplasmic vesicles accumulated in Pcs-deficient photoreceptors. Thus, Rab11 is recruited on the membrane even in Pcs-deficient cells, but is not concentrated on the trans-side of Golgi units.

A recent study indicated that both Drosophila TRAPPIII and TRAPPII complexes activate Rab1, and TRAPPII also activates Rab11 in a biochemical GEF-assay (Riedel et al., 2018). Given these findings, we investigated the functions of TRAPPIII and TRAPPII on Rh1 transport. TRAPPIII and TRAPPII share seven subunits, TRAPPC1 (Bet5), TRAPPC2 (Trs20), TRAPPC2L, TRAPPC3 (Bet3), TRAPPC4 (Trs23), TRAPPC5 (Trs31) and TRAPPC6 (Trs33). TRAPPC8 (Trs85), TRAPPC11 (Gryzun), TRAPPC12 (CG11396) and TRAPPC13 (CG4953) are TRAPPIII-specific and TRAPPC9 (Brun) and TRAPPC10 (SIDL) are TRAPPII-specific subunits (Riedel et al., 2018). We first investigated the functions of the shared core subunits TRAPPC2 and TRAPPC6, and the TRAPPIII-specific subunits TRAPPC8 and TRAPPC11, as their null insertional mutations, TRAPP2C^c00766, TRAPPC6^f00985, TRAPPC8^L3809 and TRAPPC11^MB06920, are available publicly. As all of the insertional mutations are lethal, we combined them with corresponding FRTs and investigated the resulting mosaic retinas containing both the wild-type and null homozygous photoreceptors. We found that Rh1 accumulation to the rhabdomeres was completely inhibited in TRAPP2C^c00766, TRAPPC6^f00985, TRAPPC8^L3809 and TRAPPC11^MB06920 homozygous photoreceptors (Fig. 6A–D). In contrast to conditions of Rab11 or Pcs deficiency, Rh1 was not accumulated in the cytoplasm, but instead became undetectable. The immunostaining of Na⁺K⁺ATPase at the basolateral membrane was also reduced. These phenotypes were similar to those caused by the loss of Syx5, which regulates ER-to-Golgi transport, and these findings are in agreement with those of a recent study that showed that fly TRAPPIII complex is a Rab1-GEF (Riedel et al., 2018; Satoh et al., 2016). We next investigated the influence of TRAPPIII deficiency on Rab1 and Rab11 localization. In both TRAPPC6^f00985 and TRAPPC8^L3809 homozygous photoreceptors, the punctate staining of Rab1 on the cis-Golgi was undetectable, while Rab11 foci were still clearly detected. Punctate staining of the Golgi markers GM130 and MPPE was also significantly reduced (Fig. 6E–H). Thus, TRAPPIII is likely to function as a Rab1-GEF, and this complex is necessary for Rh1 transport, presumably from ER to cis-Golgi in fly photoreceptors.

We made deletion mutants for the TRAPPII-specific subunits TRAPPC9 and TRAPPC10 by imprecise excisions of P-elements. Both deletion mutants, TRAPPC9^3D12 and TRAPPC10^10F4, are thought to be null; however, both were homozygous-viable, as reported previously (Riedel et al., 2018), and no phenotype was observed on their retinas (data not shown). Even TRAPPC9^3D12; TRAPPC10^10F4 double-mutant flies were viable, and Rh1 and Na⁺K⁺ATPase normally localized on rhabdomeres and the basolateral membranes of their photoreceptors (Fig. 7A). Additionally, Rab1 and Rab11 also localized normally to the Golgi units (Fig. 7D,F). In contrast, the TRAPP9^3D12, pcs^GS7166 double-homozygous mutant exhibited synthetic lethality, and this is in agreement with the results of a previous study (Riedel et al., 2018). The lethal stage of TRAPP9^3D12, pcs^GS7166 double-homozygous mutant flies occurred immediately prior to emersion. Interestingly, the retinas of TRAPP9^3D12, pcs^GS7166 double-homozygous mutant pupae were small and exhibited severe roughness (Fig. 7C). We dissected these small eyes and investigated the localization of Rh1 and Na⁺K⁺ATPase. As expected from the rough eye phenotype, ommatidia were not well aligned and vast spaces were found between the ommatidia in TRAPP9^3D12, pcs^GS7166 retinas (Fig. 7B). Thus, the developmental processes of retinas were severely affected by simultaneous loss of TRAPPII and Pcs. Conversely, Rh1 transport defects of TRAPP9^3D12, pcs^GS7166 double-homozygous mutant photoreceptors were not significantly worse than those of pcs^GS7166 single-homozygous mutant photoreceptors, as the majority of Rh1 was accumulated in the cytoplasm, but some Rh1 is still visible in the rhabdomeres (Fig. 7B), similar to that in pcs^GS7166 single-mutant photoreceptors. The quantification of the ratio of Rh1 staining in the cytoplasm against that in the whole photoreceptor in pcs^GS7166, TRAPPC9^3D12 double-homozygous photoreceptors shows 81.1±0.063% Rh1 staining in the cytoplasm (Fig. 7H). This value is not significantly higher than that of pcs^GS7166 or pcs^Δ1 homozygous photoreceptors (80.1±0.044%, 83.4±0.062%) (Fig. 1H). Na⁺K⁺ATPase localized normally at the basolateral membrane, and Rab1 localized normally on Golgi units with GM130, but Rab11 was largely diffused in TRAPPC9^3D12, pcs^GS7166 photoreceptors (Fig. 7E,G,I).

Collectively, eye development seems to require both TRAPPII and Parcas; however, Rh1-transport depends on Pcs, but not on TRAPPII. Although the effect of TRAPPII deficiency is not apparent in Rh1 transport, the TRAPPII-specific subunits, TRAPPC9 and TRAPPC10, are expressed constitutively in late pupal eyes (Fig. 7J). These results do not necessarily exclude the role of TRAPPII in Rh1 transport, nor redundancy of TRAPPII and Pcs. One possible explanation is that Pcs deficiency alone might reduce Rab11-GEF activity severely enough to stall Rh1 transport, because robust Rh1 transport in the late pupal stages demands high Rab11-GEF activity.

We have previously demonstrated that Rab11 and its effectors, Rip11 and MyoV, regulate the polarized transport of rhodopsin to the Drosophila rhabdomeres (Li et al., 2007; Satoh et al., 2005). Although Crag functions as a Rab11-GEF in Drosophila photoreceptors, it is required only for light-dependent Rh1 transport in adult flies (Xiong et al., 2012). Thus, a Rab11-GEF that regulates Rh1 transport at the pupal stage had yet to be identified. Here, we show that Pcs and Rab11 strongly co-localize on the Golgi-associated RE and also on post-Golgi vesicles at the base of rhabdomeres in pupal stages. Furthermore, loss of Pcs diminishes Rab11 concentration on both the trans-side of Golgi units and post-Golgi vesicles, resulting in impaired Rh1 transport to the rhabdomere and accumulation of cytoplasmic vesicles containing Rh1. These results strongly indicate that Pcs is the predominant Rab11-GEF for post-Golgi transport to the photosensitive apical membrane, rhabdomeres, in pupal flies.

Loss of the TRAPPII complex, another known Rab11-GEF, does not impact on eye development or transport of Rh1 to the rhabdomere (Fig. 7). Simultaneous loss of TRAPPII and Pcs gives larval lethality, but a small population of larvae survive until just before eclosion. These surviving double-mutant pupae have small eyes with much fewer ommatidia; however, photoreceptors still alive in these small retinas showed similar Rh1 accumulation to that seen in Pcs-deficient single mutants. In contrast, loss of Rab11 leads to strong cell lethality. Rab11-deficient flies die as embryos (Dollar et al., 2002; Jankovics et al., 2001), and Rab11^EP3017 mosaic retinas contain only a small number of mutant photoreceptors that retain trace Rab11 immunoreactivity (Satoh et al., 2005). The expression of dominant-negative Rab11^N124I from the early stage of eye development gives rise to flies lacking eyes. Thus, loss of Rab11 causes a more severe phenotype than that caused by simultaneous loss of Pcs and TRAPPII. These results suggest the presence of additional Rab11-GEFs in flies other than TRAPPII and Pcs. As the Rab11-GEF activity of Crag has been shown to be necessary for light-dependent transport in adult photoreceptors (Xiong et al., 2012), Crag might also have a function in light-independent transport. Alternatively, GEF-independent activation of Rab11 might be enough to keep photoreceptors alive and cause a milder phenotype than Rab11 deficiency.

Membrane-localized Rab-GEFs are sufficient to recruit specific Rabs, as shown by the mislocalized Rab-GEFs on the mitochondrial outer membrane (Barr and Lambright, 2010; Blümer et al., 2013; Gerondopoulos et al., 2012). Consistent with this idea, in Pcs-deficient photoreceptors, accumulation of Rab11 was lost both on the trans-side of Golgi units and in post-Golgi vesicles at the base of rhabdomeres. In spite of this, we also found that membrane recruitment of Rab11 is not dependent on Pcs function; moreover, Rab11 partially localized on Rh1-bearing vesicles accumulated in the cytoplasm of Pcs-deficient photoreceptors. This kind of GEF-independent membrane binding of Rab GTPases has also been reported for REI1 and REI2 in C. elegans (Sakaguchi et al., 2015) and several Rab-GEFs in yeast (Cabrera and Ungermann, 2013). The Pcs-independent membrane localization of Rab11 might be explained by the action of a GTP-dissociation inhibitor (GDI), as biophysical measurements suggest that GDIs can rapidly transfer Rab GTPases back and forth between the cytoplasm and lipid bilayer without the help of other factors (Wu et al., 2010). Alternatively, membrane recruitment of Rab11 might be mediated by an unidentified factor, such as the previously suggested GDI displacement factor (GDF), which recruits Rab-GDP from the cytoplasm to the membrane (Sivars et al., 2003). Our results fit to either of these models, and future work will resolve these outstanding questions further.

Flies were grown at 20–25°C on standard cornmeal-glucose-agar-yeast food unless indicated otherwise. Carotenoid-deficient food was prepared from 1% agarose, 10% dry yeast, 10% sucrose, 0.02% cholesterol, 0.5% propionate and 0.05% methyl 4-hydroxybenzoate.

The fly stocks obtained from the Bloomington Drosophila Stock Center (BDSC), the Kyoto Drosophila Genomics and Genetic Resource (DGGR) and Exelixis at Harvard Medical School are referred to by the prefixes BL, KY and HV, respectively, followed by the stock numbers. Two pcs gene insertion lines, y¹, w^67c23; P{GSV2}pcs^GS7166/SM1 (KY201057) and w*; P{GawB}pcs^NP2623/CyO (KY104263) were crossed to y, w, eyFLP; FRT42D flies and combined FRT42D with pcs^NP2623 or pcs^GS7166. The flies with FRT42D, pcs^NP2623 or pcs^GS7166 chromosomes were crossed to y, w, eyFLP; FRT42D, P3RFP to obtain mosaic eyes. To excise the pcs^NP2623 or pcs^GS7166 insertions, the fly lines with the FRT42D combined with pcs^NP2623 or pcs^GS7166 chromosomes were crossed to w*; wg^Sp−1/CyO; ry⁵⁰⁶, Sb¹, P{ry[+t7.2]=Delta2-3}99B/TM6B, Tb⁺ (KY107139). In the flies with white eyes, excisions were confirmed by sequencing. A pUAS-V5::Pcs construct made in our labs was crossed to Rh1–Gal4 and GMR–Gal4 to investigate the localization of Pcs. Rab11^EYFP was a gift from Dr Marko Brankatschk (Biotechnology Center of TU Dresden, Dresden, Germany) (Dunst et al., 2015).

To target the pcs gene, two CRISPR targets (5′-ACACCCGGCTCGCGGGGTCCTGG-3′ and 5′-GTGCGACTATCCCTCCATAGCGG-3′) on exon 4 of the pcs gene were designed using CRISPR Optimal Target Finder (http://tools.flycrispr.molbio.wisc.edu/targetFinder/) (Gratz et al., 2014). A BbsI-digested fragment containing a gRNA core and dU6:3 promoter was PCR-amplified from pCFD4-U6:1_U6:3-tandemgRNAs (Addgene plasmid #49411, deposited by Simon Bullock) (Port et al., 2014), with primers pcs-g3-F1 (5′-GTGCGACTATCCCTCCATAGGTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGG-3′) and pcs-g2-R1 (5′-GGACCCCGCGAGCCGGGTGTCGACGTTAAATTGAAAATAGGTCTATATATACGAAC-3′) and then pcs-g3-F2 (5′-GGAGGGAGGGAAGACCCTTCGTGCGACTATCCCTCCATAGGT-3′) and pcs-g2-R2 (5′-CCACCCACCGAAGACCCAAACGGACCCCGCGAGCCGGGTGTCG-3′), digested with BbsI, and cloned between BbsI sites of pCFD4-U6:1_U6:3-tandemgRNAs, resulting in a phi31-transformation vector Pcs-CR1, which encodes the two guide RNAs targeting the pcs gene. The plasmid pCFD4-Pcs was injected into embryos of PBac (Savidis et al., 2016) VK00033 (BL9750) by BestGene Inc. (Chino Hills, CA, USA) to generate transgenic lines carrying gRNA on the third chromosome.

w; FRT42D; Pcs guide RNA/Vasa-Cas9 or Nos-Cas9 were crossed with y, w, eyFLP; Sp, heat-shock Hid/CyOGFP and the male progeny, y, w, eyFLP; * 42D/CyOGFP was crossed with y, w, eyFLP; Sp, heat-shock Hid/CyOGFP to produce stock lines. Males of each stock line were crossed with y, w, eyFLP; RFP42D; Arrestin2GFP and expression of Arrestin2–GFP was observed to judge the localization of Rh1. We observed 42 stock lines and obtained three independent stocks lines, Pcsⁿ¹, Pcsⁿ² and Pcs^Δ1 with four, two and 121 nucleotide deletions in the Pcs coding region. As the result of the frameshift, in the Pcsⁿ¹ allele, S273 to stop476 was changed to 38 amino acids (LRAARCLWAQKLPKQLPKLKMRKMPATTMKREPGSSEV*). In the Pcsⁿ² allele, I274 to stop476 was changed to 18 amino acids (SGQPDVFGRKNSPSSCRN*). In the Pcs^Δ1 allele, P237 to stop476 was changed to 35 amino acids (ARCLWAQKLPKQLPKLKMRKMPATTMKREPGSSEV*).

Insertions in two TRAPP genes, PBac{PB}TRAPP2C^c00766” (HVc00766) and Mi{ET1}TRAPPC11^MB06920 (BL25338) were combined with FRT80B. TRAPPC6 insertion, TRAPPC6^f00985 (BL18399) was combined with FRT82B. TRAPPC8 insertion line, y^d2 w¹¹¹⁸ ey-FLP; P{lacW}l(3)76BDm(TRAPPC8)^L3809, FRT80B/TM6B was obtained from the Kyoto Drosophila Genomics and Genetic Resource (DGGR) (KY111022). To obtain mosaic eyes, the flies with the TRAPP2C^c00766, FRT80B or TRAPPC8^L38090, FRT80B chromosome were crossed to y, w, eyFLP; P3RFP, FRT80B, and the flies with the TRAPPC11^MB06920, FRT80B chromosome were crossed to y, w, eyFLP, FRT80B as TRAPPC11^MB06920 hovers GFP. Flies with the FRT82B, TRAPPC6^f00985 chromosome were crossed to y, w, eyFLP; FRT82B, P3RFP to obtain mosaic eyes.

We obtained the deletion alleles, TRAPPC9^3D12 and TRAPPC10^10F4 by imprecise excisions of the insertions in the 5′ UTR of TRAPPC9^KG04460 (BL13600) and in the promoter of TRAPPC10^EY00704 (BL15036). TRAPPC9^3D12 and TRAPPC10^10F4 have a deletion from 2L:20654257–20656299(R6) and 3R:15217094–15219239(R6), respectively.

Synthesized DNA fragment K-miniSOG-Ap (Table S1) was digested and cloned into the KpnI-ApaI site of pMT-puro (Addgene plasmid #17923, deposited by David Sabatini) to generate pMT-mSOG4m. A DNA fragment encoding APEX2 was amplified from pcDNA3-APEX2-NES (Addgene plasmid #49386, deposited by Alice Ting) using primers GL3N2-APEX2 (5′-ggaggttctggtggtggtGCGGCCGCcGGAAAGTCTTACCCAACTGTGA-3′) and APEX2-AGL4 (5′-ACCAGAACCTCCACCACCaGGCGCGCCGGCATCAGCAAACCCAAGCT-3′), then with primers Sp-GL3 (5′-CTGACTAGTggaggaggaggttctggtggtggt-3′) and GL4-Xh (5′-CTCACTCGAGCCaccgccACCAGAACCTCCACCACC-3′), digested and cloned into SpeI and XhoI sites of pMT-mSOG4m to generate pMT-mAPEX2m.

A gene fragment of pcs was amplified from cDNA of w¹¹¹⁸ third instar larvae, using pcs-GF2 (5′-GGATGTGTCTGTGTAGCAACGAG-3′) and pcs-GR2 (5′-TGATATGGGGCTGGCTGAAGAAGTC-3′) as primers. To construct pMK-V5-pcs, the coding region of pcs amplified with pcs-MK-Sp (5′-gatcttcatggtcgactagaATTATTCCAGAGAGCGCCTACGCAG-3′) and GL3-pcs-F (5′-ggaggaggttctggtggtggtTCGAGTGCAGAAGACGGCGAG-3′) primers, together with an N-terminal V5-epitope, were cloned into the Kpn-ApaI site of pMK33-CFH-BD using Gibson assembly.

The coding region was amplified from pMK-V5-pcs, using msXh-pcs-F (5′-gaggttctggtggcggtGGCTCGAGTGCAGAAGACGGCGAG-3′) and msAp-pcs-R (5′-AGGCTTACCttcgaaGGGCCTTATTCCAGAGAGCGCCTACGCAGC-3′) primers and cloned into the XhoI-ApaI site of pMT-mAPEX2m to obtain pMT-mAPEX2-pcs.

To produce stable transformants, 0.5 ml of Drosophila S2 cells (a gift from Dr Gota Goshima at Nagoya University, Japan) were transfected with 1 μg of pMT-mAPEX2-pcs DNA using 3 μl of FuGENE HD (Promega, Medison, WI, USA) and selected with 2 μg/ml puromycin for 2 weeks.

Fixation and staining were performed as described previously (Satoh and Ready, 2005), except that PLP (10 mM periodate, 75 mM lysine, 30 mM phosphate buffer, 4% paraformaldehyde) was used as fixative. Primary antisera were as follows: rabbit anti-Rh1 (1:1000; Satoh et al., 2005), mouse monoclonal anti-Na⁺K⁺ATPase α subunit [1:300 ascite; α5, Developmental Studies Hybridoma Bank (DSHB)], mouse monoclonal anti-Chp (1:15 supernatant; 24B10, DSHB), rabbit anti-TRP (1:2000; a gift from Dr Craig Montell, Johns Hopkins University, Baltimore, MD, USA), rat monoclonal anti DE-Cad (1:20 supernatant; DCAD2, DSHB), rabbit anti-GM130 (1:300; ab30637, Abcam), rabbit anti-MPPE (1:1000; a gift from Dr Junhai Han, Southeast University, Nanjing, China), rat monoclonal anti-p120 (1:12; a gift from Dr Satoshi Goto, Rikkyo University, Tokyo, Japan; Yamamoto-Hino et al., 2012), guinea pig anti-Rab6 (1:300; Iwanami et al., 2016), mouse anti-V5 monoclonal antibody: 6F5 (1:150; CTN3094, WAKO Chemical), rabbit anti-V5 (1:300; PM003, MBL) and rat anti-Rab11 (1:250; produced in-house for this study). Secondary antibodies were anti-mouse, anti-rabbit, anti-rat, and/or anti-guinea pig antibodies labeled with Alexa Fluor 488, 568 and 647 (1:300; Life Technologies). Images of samples (except Fig. 3P,Q) were recorded using an FV1000 confocal microscope (PlanApo N 60×1.42 NA objective lens; Olympus). For Fig. 3P,Q, a FV3000 confocal microscope (UPlanSApo 60×S2 1.30 NA objective lens; Olympus) was used. To minimize bleed-through, each signal in double- or triple-stained samples was imaged sequentially. Images were processed in accordance with the Guidelines for Proper Digital Image Handling (Rossner and Yamada, 2004) using Fiji (Schindelin et al., 2012), Affinity Photo (Serif), and/or Photoshop CS3 (Adobe). For the quantification of the intensity of Rh1 or Rab11 staining in photoreceptor cytoplasm, we used >3 mosaic retinas with >20 wild-type and >20 mutant photoreceptors in each retina. The area of the cytoplasm or whole cells and also their staining intensities were measured using Fiji. Pearson's correlation coefficient between Rab11 staining and Rh1 and/or Cnx staining was measured using cellSens software (Olympus). The region of R1 to R6 photoreceptor cytoplasm including the nucleus was enclosed to form a ROI. We used >5 retinas and set five ROIs for each retina. Thresholds were set to exclude 30% of pixels with the weakest signals for each channel, then Pearson's correlation coefficients were calculated for both Rab11 and Rh1 and/or Cnx.

For the plot of the immunostaining intensity across the Golgi units, lines were drawn through each Golgi unit, and a typical representative plot is presented here. For the plot of the immunostaining intensity across a Golgi unit and the rhabdomere in a photoreceptor, the line was drawn through a Golgi unit to the center of the rhabdomere, and a typical representative plot is presented here.

S2 cells expressing myc::APEX2::Pcs were fixed in 4% paraformaldehyde in 1× PBS for 1 h on ice. Cells were rinsed three times for 2 min each in 1× PBS and then treated for 5 min in 1× PBS containing 0.1% Triton X-100, followed by another three 2 min rinses in 1× PBS. Cells were incubated 2 h in mouse monoclonal anti-myc (1:15 supernatant; 9E10C, DSHB) and guinea pig anti-Rab6 (1:150; Iwanami et al., 2016) with 5% bovine serum in 1× PBS. After three rinses for 2 min each in 1× PBS, cells were incubated for 2 h in anti-mouse antibodies labeled with Alexa Fluor 488 and anti-guinea pig antibodies labeled with Alexa Fluor 568 (1:150; Life Technologies). Imaging and data processing were the same as for fly retina immunostaining.

Flies were reared in the dark and the retinal samples were fixed at the late pupal stage. To avoid light-dependent Rh1 endocytosis, fixation was performed within 3 min of transferring the pupae to light. Electron microscopy was performed as described previously (Satoh et al., 1997). Samples were observed on a JEM1400 electron microscope (JEOL), and montage images were taken with a CCD camera system (JEOL). For the quantification of the area occupied by the vesicles in the cytoplasm, we used five photoreceptors of >3 retinas. The areas of vesicles and cytoplasm were measured using Fiji.

Retinas were dissected from late pupal flies and fixed for 2 h in PLP with 0.1% glutaraldehyde at room temperature. After overnight wash in 1× PBS, retinas were serial dehydrated in alcohol and embedded in LR White resin (Electron Microscopy Sciences). For immunogold labeling of the sections, specimens were reacted overnight at 4°C with mouse anti-Rh1 (1:20 concentrated supernatant; 4C5, DSHB), and then reacted overnight at 4°C with anti-mouse IgG-18 nm gold conjugates (1:40; No. 115-215-075, Jackson ImmunoResearch Laboratories, West Grove, PA). After 2 min fixation using 2% glutaraldehyde, 2% paraformaldehyde in 1× PBS, specimens were stained by means of 15 min incubation in 2% uranyl acetate and 3 min in lead stain solution (Sigma-Aldrich). Samples were observed on a JEM1400 electron microscope (JEOL), and montage images were taken with a CCD camera system (JEOL).

Expression of myc::APEX2::Pcs in S2 cells was induced by adding 0.5 mM CuSO₄ to culture medium. DAB staining was performed using the protocol described by Martell et al. (2017). S2 cells expressing myc::APEX2::Pcs were fixed in 0.1 M cacodylate buffer (pH 7.4) with 2% glutaraldehyde, 2% paraformaldehyde and 2 mM CaCl₂ for 1 h on ice. Cells were rinsed five times for 2 min each in chilled cacodylate buffer, then treated for 5 min in buffer containing 20 mM glycine to quench unreacted glutaraldehyde, and this was followed by another five 2-min rinses in chilled cacodylate buffer. A freshly diluted solution of 0.5 mg/ml (1.4 mM) DAB (Sigma-Aldrich) in chilled cacodylate buffer (DAB solution) was added to cells for 1 min, and then the solution was replaced by DAB solution with 10 mM H₂O₂ in chilled cacodylate buffer and kept at room temperature for 3 min. Cells were rinsed five times for 2 min with chilled cacodylate buffer and then centrifuged at 100 g for 1–2 min. Cell pellets were mixed with pre-heated/melted 10% agarose in 0.1 M cacodylate buffer without CaCl₂ (∼40–50°C) and cooled for solidification. Solidified agarose with cells was cut into 0.5- to 1-μm cubes. Post-fixation staining was performed with 2% (w/v) osmium tetroxide (Electron Microscopy Sciences) for 30 min in chilled cacodylate buffer. Agarose cubes with cells were rinsed five times for 2 min each in chilled distilled water and then placed in chilled 2% (w/v) uranyl acetate in ddH₂O overnight. Agarose cubes with cells were dehydrated in a graded ethanol series (50%, 70%, 90%, 99.5%) for 5 min each time, and they were treated two times with 100% ethanol for 10 min. After treatment with propylene oxide two times for 10 min, agarose cubes with cells were embedded in EPON-812 resin (Electron Microscopy Sciences) using 1:1 (v/v) resin and propylene oxide for 3 h and were then transferred into 100% resin and left to sit overnight. Agarose cubes with cells in EPON-812 were polymerized at 100°C for 20 h. Embedded agarose cubes with cells were cut with a diamond knife into 70-nm sections and imaged on a JEM1400 transmission electron microscope (JEOL) operated at 80 kV, and montage images were taken with a CCD camera system (JEOL).

We designed new rat anti-Rab1 and anti-Rab11 antisera, as the mouse anti-Rab1 and anti-Rab11 antisera used previously (Satoh et al., 2005) were depleted. 6×His-tagged Rab1 and Rab11 proteins were expressed in E. coli pG-KJE8/BL21 (TAKARA) at 23°C using the previously created pQE60-Rab11 vector (Li et al., 2007; Satoh et al., 2005) and purified in native conditions using Ni-NTA Agarose (QIAGEN). To obtain antisera, three rats for each Rab protein were immunized six times with 120 µg 6×His-Rab1 or 80 µg 6×His-Rab11 proteins. We designated the resultant antisera Rat1 to Rat3 anti-Rab1 and anti-Rab11. New rat anti-Rab1 and anti-Rab11 antisera recognized endogenous Rab1 or Rab11 in immunoblots and exhibited staining patterns similar to those exhibited by the previous mouse anti-Rab1 and anti-Rab11 antisera in immunocytochemistry (Fig. S3).

All experiments were performed in accordance with the guidelines and approvals from the Hiroshima University Animal Care Committee (University of Toronto Protocol #20012022). The wild-type male mice, C57/Bl6 strain (Charles River Laboratories), were used to produce antibodies. Animals were housed in the Faculty of Arts and Science Biosciences Facility (BSF) under a 12-h light:12-h dark cycle with 2–5 animals/cage.

V5 tag and pcs genes were cloned into pUAST (Drosophila Genomics Resource Center, Bloomington, USA) to construct pP{UAST-V5::Pcs} and injected into embryos by BestGene Inc. to generate transgenic fly lines.

Immunoblotting was performed as described previously (Satoh et al., 1997). Rat3 anti-Rab1 (1:2000 serum), Rat3 anti-Rab11 (1:2000 serum), mouse anti-α-tubulin (1:200 supernatant; AA4.3-s, DSHB) and mouse anti-Na⁺K⁺ATPase α (1:10,000 ascite; α5, DSHB) as primary antibodies and HRP-conjugated anti-rat and anti-mouse IgG antibodies (1:20,000 No. 112-035-003 and No. 115-035-003, Jackson ImmunoReserch Laboratories). as secondary antibodies. Signals were visualized using enhanced chemiluminescence (Clarity Western ECL Substrate; Bio-Rad) and imaged using ChemiDoc XRS+ (Bio-Rad). For quantification of blots, three independent samples were immunoblotted, and the intensities of the bands were measured using Fiji.

For each time point, four tubes of samples containing two dissected pupal retinae were prepared. These samples were dissolved, digested with DNaseI and reverse-transcribed to prepare cDNAs using SuperPrep II Cell Lysis & RT Kit (TOYOBO). Quantitative PCR was performed using CFX-Connect (Bio-Rad), with KOD SYBR qPCR Mix (TOYOBO) and primers (Rab6, 5′-GCTGCGGAAGTTCAAGCTC-3′ and 5′-CCTGGTACGTGTTGTCGAAG-3′; TRAPPC9, 5′-GAAGAACCTGGCAGATCTATCG-3′ and 5′-GAATCACCTACCGACCTCAAAG-3′; TRAPPC10, 5′-GTCAAGCCTATTTGCTGCTTAC-3′ and 5′-CTCCTGACACTCCAGCTTTATG-3′; pcs-qF1, 5′-GCTAATTCCACCTTTCGCATCC-3′ and 5′-TCGTAGTAGGGTCTAGCCTTCT-3′). Dilution series of pooled cDNA were used to generate standard curves. Rab6 was used as the standard of total RNA. Levels of TRAPPC9, TRAPPC10 and Pcs cDNA on day 4, relative to Rab6 cDNA, were normalized as the mean amount with a value of 1.

We thank Drs U. Tepass, C. Montell, C. S. Goto, M. Brankatschk, M. Baylies and M. Erdelyi for kindly providing fly stocks and reagents. We also thank the Bloomington Drosophila Stock Center and the Drosophila Genomics and Genetic Resource (Kyoto Institute of Technology) for fly stocks.