Bchs, a BEACH domain protein, antagonizes Rab11 in synapse morphogenesis and other developmental events

BEACH proteins are conserved throughout eukaryotes. These large proteins,many of which exceed 400 kDa in size, have been implicated in cellular processes ranging from cytokinesis to synaptic transmission(Kwak et al., 1999; Su et al., 2004). Furthermore,mutations in several BEACH genes cause disease in humans. LYST(lysosomal trafficking regulator), the first BEACH gene to be discovered, is disrupted in Chédiak-Higashi syndrome - an often-fatal disease characterized by severe immunodeficiency, albinism, poor blood coagulation and neurologic involvement (Introne et al.,1999). Another BEACH family member, neurobeachin (NBEA), has been implicated as a candidate gene for autism(Castermans et al., 2003). Furthermore, upregulation of LRBA (LPS-responsive and beige-like anchor) is seen in several types of cancer and appears to facilitate cancer growth (Wang et al.,2004).

Localization of several BEACH proteins to subcellular membranes, as well as loss-of-function phenotypes that affect organelle morphology and function,suggest that these proteins may play roles in membrane trafficking. For instance, the hallmark of cells mutant in the Lyst gene is the presence of large intracellular granules of lysosomal origin, which probably result from increased lysosome fusion(Harris et al., 2002; Nagle et al., 1996). However,the mechanisms by which Lyst and other BEACH proteins regulate vesicle trafficking are not understood.

In the present study we have found a genetic interaction between bchs (blue cheese), a BEACH family member recently described in Drosophila (Finley et al.,2003), and rab11. The Rab family of small GTPases has a well-established involvement in membrane traffic; Rabs can regulate vesicle formation, motility, docking and fusion(Zerial and McBride, 2001). Rabs, like other GTPases, act as molecular switches: active in the GTP-bound form and inactive in the GDP-bound form. Moreover, each Rab must be associated with its particular subset of cellular membranes in order to carry out its function (Pfeffer and Aivazian,2004; Seabra and Wasmeier,2004). Thus, Rabs impart specificity to membrane trafficking events.

Rab11 localizes to the pericentriolar recycling endosome, the trans-Golgi network and post-Golgi vesicles(Chen et al., 1998; Ullrich et al., 1996), and plays a role in both the exocytic biosynthetic pathway and the recycling pathway (Chen et al., 1998; Ren et al., 1998; Satoh et al., 2005; Ullrich et al., 1996). In Drosophila, recent work has established a role for rab11 in multiple developmental events, including polarization of the oocyte,cellularization of the embryo and morphogenesis of the rhabdomere, the photosensing organelle of photoreceptor neurons(Dollar et al., 2002; Pelissier et al., 2003; Riggs et al., 2003; Satoh et al., 2005). In addition, subcellular localization of Rab11 may contribute to asymmetric cell division (Jafar-Nejad et al.,2005).

In this paper we describe the characterization of Drosophila Bchs and its interaction with Rab11. We find that Bchs is highly expressed in the nervous system, where it is associated with vesicles and concentrated in synaptic regions. We show that reductions in bchs function suppress the effects of loss-of-function rab11 mutations in multiple developmental contexts. In particular, bchs mutations suppress a newly described anatomical phenotype of rab11 at the neuromuscular synapse. Our data identify a role for these regulators of vesicle trafficking in developmental events, such as synaptic morphogenesis, and provide insight into how BEACH proteins could be involved in vesicle trafficking.

EP lines (Rorth, 1996) were obtained from Rebay lab (Whitehead Institute, Cambridge, MA). rab11^P2148, rab11^93Bi, Df(2L)cl7, pr[1]cn[1]/CyO, and the Drosophila DK1, DK2 and DK3 deficiency kits(2001) were obtained from the Bloomington Stock Center, rab11^ex1/TM3, Sb and rab11^ex2/TM3, Sb from R. Cohen(University of Kansas). w, Canton S flies were use as control, unless otherwise indicated.

Both sense and antisense RNA probes were made using DIG RNA labeling kit(Roche); the template was made by PCR from genomic fly DNA using the following primers:5′-GAATTAATACGACTCACTATAGGGAGAGCACACAAAGTTCGATCTTGAC-3′and 5′-AATTAACCCTCACTAAAGGGAGAGTTCGCCTACAAGCACATCG-3′. Embryo in situ hybridization was done as in Tautz and Pfeifle(Tautz and Pfeifle, 1989).

Template for mouse RNA probe corresponding to 4737-5413 bp of Wdfy3 cDNA was made by PCR from mouse cDNA (Gertler Laboratory, MIT),primers used 5′-T3CCTAAGCCTGTCGCCACTACTTTAC-3′and 5′-T7CCAAACTTCTTCTTCTGCTCCCG-3′. Probe was synthesized using Strip-EZ RNA (Ambion) and a-P³²UTP 800 ci/mM (Amersham). Hybridization was done according to the manufacturer's instructions(Ambion).

Ethyl methanesulfonate (EMS) mutagenesis was performed as described(Lewis and Bacher, 1968). bchs alleles were PCR amplified from genomic DNA and sequenced by the MGH Core Facility. The following primers were used.

Pair 1: 5′-CAAACCCCACGGACATGC-3′and 5′-GCTGGTGTGGACTGACGCC-3′.

Pair 2: 5′-GCACGCTCCCTCCGTTCG-3′and 5′-CAAACTTGGAGCACTGCCTGAG-3′.

Pair 3: 5′-CAACCAGTTACAGGGTCGGAATC-3′and 5′-GCGCTGACCACTTTTGTAGTCTG-3′.

Full-length bchs cDNA was assembled by combining a partial cDNA(clone LD02084) with cDNA produced via RT-PCR from S2 cell RNA (provided by Pardue Laboratory, MIT). RT (RETROscript, Ambion) and PCR (Expand High Fidelity PCR System, Roche) were done according to the manufacturer's instructions. The following primers were used:5′-CGGGATCCATGAATGTAATGCGTAAGCTGCG-3′,5′-CGGAATTCGCCACCAAGGACTTGATGATTTCG-3′,5′-CGGAATTCTGCTTCGCACCACGCAGGTC-3′,5′-CGGGATCCCGAGCGGACAACAAAAGCATTG-3′,5′-ACGCGTCGACCAGATTCCGACCCTGTAACTGG-3′,5′-GCAACCACGAGTTGGAATTCATTGGC-3′and 5′-ATTTGCGGCCGCCCTAATTGTCCAACGAGTTCGTGC-3′.

Fragments were sequenced before assembly into full-length cDNA, which was modified with 5′HA tag in pcDNA6/V5-His (Invitrogen) and inserted into pUASt (Brand and Perrimon,1993).

Bchs aa 2237-2590 were expressed as 6XHis fusion protein in bacteria and purified according to the manufacturer's instructions (Amersham). Polyclonal antisera were produced in rats (Covance).

Each lane of a 6% SDS-polyacrylamide gel contained nine adult heads homogenized in 1×Laemmli buffer in PBS (130 mmol/l NaCl, 175 mmol/l Na₂HPO₄, 60 mmol/l NaH₂PO₄). After electrophoresis, proteins were transferred to Hybond-P membrane (Amersham Pharmacia) and membranes were blocked in 5% nonfat milk and probed with anti-Bchs (1:1000) and rat anti-Elav (1:1000), followed by HRP-conjugated goat anti-rat antibody (1:5000) (Jackson).

Immunohistochemistry of larval and adult brains was performed as previously described (Garrity et al.,1999). Anti-Bchs was preabsorbed using bchs¹⁷animals and used at 1:500. Primary antibodies: mouse MAb 24B10 anti-Chaoptin(Fujita et al., 1982) (1:200),rabbit anti-Syt1 (T. Littleton, MIT) (1:500) and mouse anti-HA (Covance)(1:1000) were used. Secondary antibodies were goat anti-mouse HRP (1:200),goat anti-rat Cy3 (1:500), goat anti-rabbit FITC (1:200) and goat anti-mouse Cy3 (1:1000) (Jackson). Fluorescent samples were visualized using a Nikon PCM2000 confocal microscope.

Larval body wall dissections were done in PBS and fixed in 4% PFA in PBS. Anti-Bchs was preabsorbed using bchs null animals and used at 1:500. Mouse anti-Rab11 antibody (BD Biosciences) was used at 1:200. Subtracted goat anti-rat Cy3 at 1:100 (Jackson), goat anti-mouse Alexa-488 at 1:100. Cy5- and FITC-conjugated anti-HRP antibodies at 1:100 (Jackson) were used. Confocal data was acquired as single images or image stacks of multi-tracked, separate channels with a Zeiss LSM 510 microscope.

Each stock from the Drosophila Deficiency Kits (195 stocks that together delete over 85% of the genome) was crossed to GMR-GAL4;EP-bchs flies to determine whether heterozygosity for particular genomic region modified the adult eye phenotype caused by Bchs overexpression. Smaller deficiency chromosomes and mutations within genomic regions of interest were obtained (Bloomington Stock Center) and examined for interactions with GMR-GAL4;EP-bchs.

To measure viability, 60 larvae (second instar) of each genotype were placed in identical vials at 25°C and monitored daily for survival to adulthood. Bristle loss was quantified by counting the numbers of bristle-filled and empty sockets in the last row of abdominal tergites(segments) 2, 3 and 4 in adults (≤1-day old). Fraction of bristle-filled sockets was calculated. At least ten animals per genotype were counted. All P-values were determined using two-tailed unpaired Student's t-tests.

Sucrose gradient was prepared by Bill Adolfsen (Littleton Laboratory, MIT)as described in Adolfsen et al.,2004 (Adolfsen et al.,2004). Western blots were as described above using: rat anti-Bchs 1:1000, mouse anti-Rab11 (BD Biosciences) 1:1000, mouse anti-Rop (4F8)(Harrison et al., 1994) (H. Bellen, Baylor) 1:1000, and rabbit anti-Synaptotagmin (rabbit) (T. Littleton,MIT) 1:500. HRP-conjugated goat anti-rat, mouse, rabbit and guinea pig secondary antibodies (Jackson) were used at 1:5000.

Confocal stacks through third instar NMJ 6/7 were flattened into projections using the LSM 510 software. Each bouton was marked on the image by a colored dot using Adobe Photoshop and the resulting dots were counted using ImageJ software. The number of branching boutons was counted in the same manner. Muscle surface area was estimated as described(Schuster et al., 1996).

We identified bchs in a genetic screen as a modifier of photoreceptor growth cone morphology when expressed at a high level in the retina. In this screen, we used a retina-specific promoter fused to GAL4(GMR-GAL4) to overexpress genes adjacent to GAL4-responsive UAS elements. These genes were represented in a collection of about 2000 Drosophila lines that contained insertions of the UAS-bearing EP transposon (Rorth, 1996). One of these lines, EP(2)2299 (henceforth termed EP-bchs), contained an EP transposon insertion adjacent to bchs. When bchs was highly expressed in the developing retina by the GMR-GAL4 driver,photoreceptor growth cones had an unusually bulbous central region and were less expanded than controls, although they appeared to grow normally toward their targets (Fig. 1A′A″,B′,B″). While patterning of the third instar eye disc appeared normal in these GMR-GAL4;EP-bchsanimals, suggesting that early eye development was not significantly affected(R.K., unpublished), overexpression of bchs in the retina resulted in a smaller, glazed adult eye (Fig. 1A,B).

Bchs has a predicted mass of 390 kDa and has been evolutionarily conserved from the slime mold Dictyostelium to humans(De Lozanne, 2003). In all BEACH proteins, including Bchs, the BEACH domain is preceded by a pleckstrin-homology (PH) domain and followed by four to six WD40 repeats (five in Bchs) (Fig. 2A). The large size of most BEACH proteins combined with the presence of multiple WD40 domains, which serve as protein-protein interaction motifs, suggest that BEACH proteins could be involved in the assembly of large protein complexes. BEACH proteins have been subdivided into five classes(De Lozanne, 2003; Wang, N. et al., 2002). Whereas the PH-BEACH-WD40 module is conserved across the family, each BEACH protein has additional regions of homology particular to members of its class. There are five predicted BEACH proteins in Drosophila, and Bchs falls in a class that includes LvsA in Dictyostelium, several uncharacterized proteins in Arabadopsis and Caenorhabditis elegans, and two human members: Alfy (autophagy-linked FYVE protein;WDFY3 - Human Genome Nomenclature Database) and the uncharacterized molecule KIAA1607. Bchs has extensive homology to Alfy throughout the entire protein length: 51-84% amino acid identity in a carboxyl portion that includes the PH-BEACH-WD40 regions and 41% identity in the remainder of the protein,hereafter called CRAB (conserved region in Alfy and Bchs)(Fig. 2A). In addition to the PH-BEACH-WD40 module, Bchs and Alfy also have a C-terminal FYVE (Fab1p, YOTB,Vac1p and EEA1) domain (Fig. 2A), a motif that can mediate interactions with phosphatidylinositol 3-phosphate (Misra et al., 2001).

Both bchs and genes expressing the Alfy protein are expressed in the nervous system. RNA in situ hybridization indicated that bchsmRNA was enriched in, but not restricted to, the embryonic brain and ventral nerve cord (Fig. 2B) (see also Kraut et al., 2001). Similarly, transcripts for murine Alfy (Wdfy3 - Mouse Genome Informatics) were expressed highly, but not exclusively, in the brain(Fig. 2C). Thus, Bchs and its homologs are neuronal, but are likely to also function in other cell types.

We reasoned that mutations disrupting bchs function would decrease or eliminate the ability of Bchs to alter the structure of the adult eye when overexpressed. Therefore, we mutagenized EP-bchs animals using EMS and isolated 17 independent EP-bchs chromosomes that no longer caused strong eye defects when combined with GMR-GAL4(Fig. 1 and Table 1). DNA sequencing identified mutations in the bchs gene in eight of these strains(Table 1). Examination of the larval photoreceptor axons in these animals showed that those mutations that strongly suppressed the adult eye phenotype also strongly suppressed the growth cone phenotype (Fig. 1C). The identification of these mutations within bchsnot only provided loss-of-function bchs alleles for further analysis,but also confirmed that the growth cone and eye phenotypes resulted from bchs overexpression.

To characterize bchs alleles, we analyzed protein levels using antisera raised against amino acids 2237-2590 of Bchs, a region between the positions of the lesions in bchs¹³ and bchs¹⁷. These antisera recognized a protein species migrating well above the 250 kDa marker on western blots, consistent with the predicted molecular weight for Bchs of 390 kDa(Fig. 2D). This protein species was increased in flies overexpressing bchs(GMR-GAL4;EP-bchs) and absent in flies homozygous for the bchs alleles bchs¹², bchs¹⁷, and bchs⁵⁸ (these alleles retain the EP transposon insertion present in EP-bchs). These data confirmed that the antibody detects Bchs protein and showed that three of our loss-of-function bchsalleles do not produce detectable full-length Bchs protein.

We examined multiple bchs alleles for mutant phenotypes, including the putative protein null alleles bchs¹²,bchs¹⁷ and bchs⁵⁸ in combination with a deficiency chromosome, Df(2L)cl7, missing the bchs locus. Mutants in bchs were viable, fertile and had no overt defects. Furthermore, no defects in axon guidance or growth cone morphology could be detected in the larval visual system or the embryonic central nervous system(CNS) and motor axons (R.K., unpublished).

To gain insight into the function of Bchs, we searched for genes that could alter the adult eye phenotype caused by Bchs overexpression. We examined a collection of 195 chromosomal deficiencies that deleted defined portions of the Drosophila genome (in total, covering ≈85% of the genome). Five chromosomal deficiencies dominantly suppressed the GMR-GAL4;EP-bchs eye phenotype, while eight dominantly enhanced it. In particular, deficiency Df(3R)e-R1 acted as a strong enhancer. Genes within the interval deleted by Df(3R)e-R1 were examined, and mutations in rab11 were found to strongly enhance the Bchs overexpression phenotype (Fig. 3). Multiple loss-of-function rab11 alleles were tested,including rab11^ex1, rab11^ex2,rab11^E(To)3, rab11^E(To)11 and rab11^J2D1(also known as rab11^P2148)(Dollar et al., 2002; Jankovics et al., 2001). Heterozygocity for rab11 did not cause a disruption in eye development in otherwise wild-type animals, but all the examined rab11 mutants showed a strong dominant enhancement of the bchs overexpression phenotype(Fig. 3). Thus, Bchs overexpression makes the eye sensitive to partial reductions in rab11gene function, raising the possibility that Bchs and Rab11 functionally antagonize one another.

As decreased dosage of rab11 enhanced the Bchs overexpression phenotype, we wanted to test whether the loss of bchs would, in turn,suppress rab11 loss-of-function phenotypes. Null rab11mutants die in mid-embryogenesis; however, rab11 hypomorphs exhibit different degrees of lethality, sterility and bristle defects(Dollar et al., 2002; Jankovics et al., 2001). The combination of a hypomorphic allele rab11^93Bi and a putative null allele rab11^ex1(Dollar et al., 2002) is mostly lethal (Jankovics et al.,2001), but a small number of rab11^ex1/rab11^93Bi flies survive to adulthood. To test whether the loss of bchs had an effect on the viability of rab11^ex1/rab11^93Bimutants, we introduced into this genotype the putative protein null alleles bchs¹², bchs¹⁷ and Df(2L)cl7. Loss of one or both copies of bchs strongly enhanced the viability of rab11^ex1/rab11^93Bi animals(Fig. 4A,B). For example, in one experiment, survival to adulthood was improved from 17±4% in rab11^ex1/rab11^93Bi mutants to 77±3% in bchs¹²/bchs¹⁷;rab11^ex1/rab11^93Bidouble mutants (all values are mean±s.e.m., n=3 independent vials, P<0.001) (Fig. 4A). In another experiment, survival was improved from 5±1%for rab11^ex1/rab11^93Bi mutants to 48±6% for bchs¹⁷/+;rab11^ex1/rab11^93Bianimals (n=8 independent vials, P<0.001)(Fig. 4B). Thus, reduction in,or loss of, bchs function increases the survival of rab11mutants. This suggests that Bchs antagonizes essential functions of Rab11.

The posterior abdomens of viable rab11 mutants are missing many small mechanosensory bristles (or microchaetae)(Fig. 5A)(Jankovics et al., 2001). In addition, the posterior scutellar macrochaetae, large mechanosensory bristles,are shortened in rab11^ex1/rab11^93Bimutants (Fig. 5B). Mutations in bchs suppressed both the bristle shortening and the bristle loss defects of rab11 mutants (Fig. 5A,B). To quantify this suppression we calculated the fraction of empty sockets in the last row of abdominal tergites 2, 3 and 4 for two different allelic combinations of rab11 (the viable rab11^93Bi homozygote and the semi-lethal rab11^ex1/rab11^93Bi heterozygote) alone and in combination with bchs alleles. Reductions in bchssuppressed bristle loss in both rab11 allelic combinations in a dosage-dependent manner. The removal of one copy of bchs strongly suppressed bristle loss, and the removal of both copies returned the number of bristles to control levels (Fig. 5C). As expected from the western blot above, bchs¹⁷ and bchs¹² were indistinguishable in this suppression assay from Df(2L)cl7,suggesting that they are null alleles. These data further support the conclusion that Bchs and Rab11 function antagonistically during Drosophila development.

Interestingly, bchs⁸, which produces full-length Bchs protein, suppressed bristle loss to the same extent as the protein null bchs⁵⁸ allele (Fig. 5C, bottom panel). This suggests that the missense mutation in bchs⁸, which alters a conserved threonine in the interface between the PH and the BEACH domains of the protein(Fig. 2A and Table 1)(Jogl et al., 2002), strongly disrupts protein function.

We examined the localization of Bchs protein in the larval brain and found that it was enriched in synaptic regions within the CNS(Fig. 6A). The specificity of the Bchs immunoreactivity was demonstrated by its absence from the neuropil of bchs¹².

To examine the distribution of Bchs protein within a neuron at greater resolution, we expressed an HA-tagged version of full-length Bchs in ellipsoid body neurons using EB1-GAL4(Wang, J. et al., 2002). EB neurons extend a single neurite that splits into distinct dendritic and axonal processes, with the dendrite arborizing in the lateral triangle, and the axon in the ellipsoid body ring (Hanesch et al., 1989). Whereas a membrane-associated CD8-GFP fusion(Chang et al., 2002) was evenly distributed throughout these neurons, HA-Bchs preferentially accumulated near the axon terminals (Fig. 6B),suggesting that Bchs can preferentially localize to the presynaptic regions of a neuron.

The larval neuromuscular junction (NMJ) provides the opportunity to examine individual synaptic terminals isolated from other neuronal elements. In these preparations, some Bchs immunoreactivity was present in the muscle fibers, but it was most prominent within the nerve terminals(Fig. 6C,D). Bright puncta of Bchs immunoreactivity were observed presynaptically, as was established by using anti-HRP to label the surrounding neuronal plasma membrane. In double-labeling experiments, the Bchs-immunoreactive puncta failed to correlate with other subcellular compartments that have been characterized at the NMJ, including early endosomes marked with 2XFYVE-GFP and Rab5-GFP(Wucherpfennig et al., 2003),dense-core vesicles marked with ANF-GFP(Rao et al., 2001) and recycling vesicles marked with Clathrin-GFP(Chang et al., 2002)(Fig. 7A,B,C,D,respectively).

Rab11 immunoreactivity had not been previously characterized at the NMJ;therefore, we examined its distribution to determine whether it, like Bchs,was enriched at the synapse. Rab11-positive puncta were present in nerves,synaptic boutons and muscles (Fig. 8A). The observed staining was specific for Rab11, as it was greatly reduced in rab11^ex1/rab11^93Bi mutants,which also showed reduced protein levels (compared with controls) by western blot analysis (Fig. 8A,B). While both Rab11 and Bchs labeled puncta at the NMJ, Bchs was more enriched in the boutons than Rab11, while Rab11 was more abundant in the muscles than Bchs. Double labeling these preparations for Rab11 and Bchs demonstrated that their subcellular localization substantially overlapped both in neurons and muscle cells (Fig. 8C),although some puncta were immunoreactive for only one protein or the other. Within puncta that contained both proteins, Bchs and Rab11 staining was not always perfectly congruent: Bchs immunoreactivity at times predominated to one side of the stained structure and Rab11 to the other(Fig. 8C, arrows). These localization data, showing that Bchs and Rab11 are present in overlapping locations at the NMJ, together with the genetic interaction data, suggest that Bchs and Rab11 function in the same or closely related processes.

The results of biochemical fractionation experiments were also consistent with the partially overlapping distribution of Bchs and Rab11 proteins detected by immunohistochemistry at the NMJ. We probed fractions [previously characterized in Adolfsen et al. (Adolfsen et al., 2004)] from a 10-30% sucrose velocity gradient of head extract to determine how Bchs migrated with respect to known membrane fractions (Fig. 9). Bchs migrated with an intermediate density characteristic of a membrane-associated protein, but did not co-migrate with either the plasma membrane marker Syntaxin 1A (Schulze et al.,1995) or the synaptic vesicle markers neuronal-Synaptobrevin and Synaptotagmin1 (DiAntonio et al.,1993; Littleton et al.,1993). Rab11 showed a different distribution across the gradient than Bchs (Fig. 9). Nonetheless, a significant amount of Rab11 was recovered in fraction 17, the fraction in which Bchs was most abundant. Such partial co-migration of Bchs and Rab11 is consistent with localization of these proteins to partially overlapping membrane compartments.

Given the genetic interactions between bchs and rab11observed elsewhere in the animal, and the overlap of Bchs and Rab11 protein distribution at the NMJ, we investigated whether rab11 and bchs interacted in NMJ development. We compared the phenotype of the hypomorph rab11^ex1/rab11^93Bi to controls, with and without bchs, using the null genotype bchs¹⁷/bchs¹². Although overexpression of Bchs causes bulges at the junctions between the axons and the synaptic branches(Kraut et al., 2001), bchs loss-of-function mutants had no detectable defects at the NMJ. rab11 hypomorphs, however, showed significant alterations in synaptic bouton patterning at all NMJs. The defect was quantified at the muscle 6/7 synapse in abdominal segment A3, where the total number of synaptic boutons increased by ≈60% in rab11 mutants compared with controls(166±11 versus 98±5, n=14 for each genotype;mean±s.e.m.; P<0.0001). Because synapses at the NMJ of wild-type larvae grow in proportion to muscle size(Schuster et al., 1996), we also calculated the number of boutons per unit of estimated muscle area. When the decreased size of rab11 muscles was taken into account(50±2×10³μm² in rab11 versus 67±1×10³ μm² in control), the density of synaptic boutons/μm² muscle area in the mutants was more than double that of controls (3.3±0.2×10^-3 versus 1.5±0.1×10^-3 boutons/μm²; n=14; P<0.0001) (Fig. 10A,B).

In addition to the increase in number and density, the boutons in rab11 mutants were often arrayed in tight clusters resembling bunches of grapes, rather than the normal beads-on-a-string morphology. Among the previously described phenotypes at the NMJ, rab11 most closely resembled that of nervous wreck, a mutation influencing the WASP signaling pathway (Coyle et al.,2004). The unusual clustering of boutons in rab11 mutants was related to an increase in the number of branching boutons - those connected to more than two neighboring boutons. Normally such branch points are uncommon, but in rab11 mutants the fraction of boutons that branch increased 2.6-fold (22.2±0.8% of all boutons versus 8.4±0.4% in control; n=14; P<0.0001)(Fig. 10A,B). Thus, rab11 mutants are abnormal in synaptic growth and morphogenesis.

The NMJ phenotypes observed in rab11 mutants were partially suppressed by loss-of-function mutations in bchs. Muscle size was restored to near control in bchs,rab11 double mutant animals(64±3×10³μm² in bchs;rab11 vs. 67±1×10³ μm² in control, n=14). The bchs,rab11 double mutant and rab11 single mutant had similar numbers of synaptic boutons. However, the density of boutons per muscle area in bchs,rab11 animals was significantly less than in rab11 mutants alone (2.5±0.2×10^-3boutons/μm²; 169% of control in bchs,rab11 versus 3.3±0.2×10^-3 boutons/μm²; 225% of control in rab11; n=14, P<0.01). This represents a 44% suppression of the rab11 phenotype(Fig. 10A,B). The percentage of branching boutons in the bchs,rab11 double mutant was also significantly less than in rab11 alone (15.6±1.1% versus 22.2±0.8%; n=14; P<0.0001). This represents a 48%suppression of the rab11 phenotype(Fig. 10A,B). Thus, loss of bchs partially suppressed the increases both in bouton density and in bouton branching exhibited by rab11 mutants. Consequently, Bchs probably has a modulatory role in synaptic development, acting as an antagonist of Rab11.

While several BEACH-family proteins have been implicated in vesicle trafficking, the mechanisms through which they may regulate this process are unknown. We have shown that Bchs, the Drosophila relative of the human Alfy protein, is a functional antagonist of the vesicle-trafficking regulator Rab11. In particular, reduction in bchs strongly suppressed the defects in viability, synaptic morphogenesis and bristle extension exhibited by rab11 loss-of-function mutants. Additionally, reduction in rab11 activity strongly enhanced the eye phenotype caused by bchs gain of function. The reciprocity of genetic interactions between these two genes provides compelling evidence that Bchs participates in many of the same processes as Rab11 and, thus, suggests that Bchs, like Rab11,is a regulator of vesicle trafficking.

The subcellular localization of Bchs also supports the hypothesis that this protein functions in membrane traffic. Bchs was present exclusively in membrane fractions and exhibited punctate staining in the presynaptic motoneuron terminals and in the muscles at the NMJ. This pattern is consistent with the localization of Bchs to a membrane-bound organelle. Furthermore, in line with a functional relationship between Bchs and Rab11, we observed significant subcellular co-localization of Bchs and Rab11 at the NMJ, as well as partial overlap in the distribution of Bchs and Rab11 within membrane fractions. These data further support our hypothesis that Bchs regulates vesicle trafficking and that it may do so via an interaction with the Rab11 GTPase.

A prominent bchs phenotype was the suppression of the rab11 sensory bristle defects, which entailed both shortened and missing bristles. These defects probably arose from alterations in membrane traffic. The extension of sensory bristles involves several vesicle trafficking steps, including membrane addition at the tip of the growing bristle and the secretion of cuticle to support the bristle cell. The complete loss of mechanosensory bristles could result from extremely impaired bristle growth. Alternatively, bristle loss could arise from a cell fate transformation that prevents the specification of the bristle-producing cell,as Rab11-mediated vesicle trafficking has also been implicated in the asymmetric cell divisions of the precursors that give rise to these cells(Jafar-Nejad et al., 2005). The ability of mutations in bchs to strongly suppress all rab11 bristle defects implicates Bchs in bristle morphogenesis and is consistent with it playing a role in Rab11-mediated vesicle trafficking.

We have also uncovered a crucial role for rab11 in the formation of the Drosophila NMJ: rab11 mutants exhibit an increase in the density and branching of synaptic boutons and a decrease in the size of the muscles. Vesicle trafficking is important in determining the number and morphology of boutons at the NMJ (Dickman et al., 2006). In sculpting the synapse, membrane traffic is needed not only for the addition of new membrane and active zone proteins, but also for the insertion, removal and signaling of regulatory molecules at the cell surface (Marie et al.,2004; Sweeney and Davis,2002). Furthermore, exocyst-dependent membrane addition is required for the expansion of NMJs (Murthy et al., 2003), and Rab11 is a known regulator of exocyst function(Beronja et al., 2005; Zhang et al., 2004). Thus,Rab11 is involved in synaptic morphogenesis at the NMJ, probably via regulation of vesicle trafficking.

By virtue of suppressing rab11 NMJ phenotypes, Bchs is also implicated in a membrane-trafficking aspect of synaptic morphogenesis. Consistent with such a model, both Bchs and Rab11 showed punctate localization and partial overlap at the NMJ. A functional role of Bchs in presynaptic development may explain its concentration in the axonal rather than dendritic compartment of ellipsoid body neurons (Fig. 7B).

The link between Bchs and Rab11 function provides initial mechanistic insights into the trafficking pathways that may involve Bchs. Rab11 is involved in both biosynthetic exocytic traffic and membrane traffic through the recycling endosome (Pelissier et al.,2003; Satoh et al.,2005). As the loss of bchs suppresses lethality of rab11 alleles, it is likely to be involved in all the essential functions of Rab11.

The partial colocalization of Bchs and Rab11 suggests candidate sites for the function of Bchs. In particular, Rab11 has been observed on the trans-Golgi network, post-Golgi vesicles, recycling endosomes and vesicles that travel from the recycling endosome to the plasma membrane(Chen et al., 1998; Ullrich et al., 1996; Ward et al., 2005). In regulating traffic to the plasma membrane, Rab11 has been shown to physically interact with members of the exocyst complex(Beronja et al., 2005; Zhang et al., 2004). We found the distribution of Bchs to be highly polarized in neurons and enriched at synaptic endings, not cell bodies or dendrites. This suggests that interactions between Bchs and Rab11 may occur in a compartment adjacent to the presynaptic plasma membrane, rather than near the trans-Golgi network, the perinuclear recycling endosome or dendritic endosomes.

At the synapse, the most prominent membrane trafficking pathway is the synaptic vesicle cycle. However, Bchs immunoreactivity did not colocalize with synaptic vesicles or co-migrate with them in cell fractionation. These findings, together with the fact that loss of bchs does not alter the viability of the mutants, suggests that Bchs does not play a major role in the release of neurotransmitter. In this regard, Bchs is clearly distinct from Neurobeachin, a Beach-domain protein that is essential for transmitter release at the mouse NMJ (Su et al.,2004). Bchs did not colocalize with early endosomes, marked with either 2XFYVE-GFP or Rab5-GFP, through which at least some synaptic vesicles are thought to cycle, nor did Bchs immunoreactivity resemble the distribution of endocytic vesicles identified by clathrin-GFP. The former was perhaps surprising, given that Bchs possesses a FYVE domain. These data suggest that Bchs is unlikely to be involved in early endocytic events. The BEACH domain protein Lyst has been implicated in lysosomal trafficking; however, we observed no overlap of Bchs immunoreactivity with lysotracker (R.K.,unpublished). Bchs is, therefore, not a lysosomal protein. Rather, Bchs appears to reside on a novel synaptic compartment adjacent to or overlapping a Rab11-containing organelle. We speculate that this may be a previously unappreciated presynaptic sorting endosome through which either recycled or Golgi-derived proteins are trafficked.

What cellular and molecular mechanisms underlie the extensive genetic interactions between bchs and rab11? In one scenario, Bchs could negatively regulate Rab11 activity, perhaps by promoting a Rab11-GAP that restricts Rab11 function. Bchs could modulate the efficacy of Rab11 function, but might be only one of several negative regulators of Rab11. Such a model would be consistent with our genetic studies, as loss of bchsmight not cause defects on its own, but bchs overexpression would shut down the Rab11 pathway. Alternatively, rab11 and bchscould be involved in competing intracellular pathways. For example, Rab11 might direct endosomal cargos toward the plasma membrane, while Bchs diverts these cargos elsewhere. This hypothesis receives support from the observation that Rab11 and Bchs appear to concentrate in partially distinct subcompartments of those organelles on which they both reside(Fig. 8C). This pattern of partially overlapping localizations is reminiscent of other pairs of regulators of membrane traffic, including Rab4 and Rab5, and Rab4 and Rab11 on two sequential, yet distinct, populations of endosomes. The distribution of these Rabs reflects their participation in linked steps of cargo transport along the recycling pathway (de Renzis et al., 2002; Sonnichsen et al.,2000), which may also explain the localization pattern of Bchs and Rab11.

In addition to Bchs, other BEACH family proteins, such as Lyst and Neurobeachin, have been implicated in vesicle trafficking. Mutations in Lyst result in the accumulation of giant lysosomes in the cells of both beige mutant mice and Chédiak-Higashi syndrome patients,suggesting that Lyst is involved in lysosome trafficking, fusion or formation(Introne et al., 1999). Mouse Neurobeachin mutants lack evoked synaptic transmission at the NMJ, implicating neurobeachin in neurotransmitter release(Su et al., 2004). Our finding that Bchs antagonizes Rab11 raises the intriguing possibility that other BEACH proteins might also interact with Rab family members. It will be interesting to determine whether the defects observed in Lyst and neurobeachin mutants reflect the altered activity of particular Rabs and whether alterations in Rab function could ameliorate defects caused by the absence of Lyst or Neurobeachin.

We thank I. Rebay, R. Cohen, M. Gonzalez-Gaitan, D. Ready, H. Bellen, T. Littleton and the Bloomington Stock Center for flies and reagents; B. Adolfsen and T. Littleton for sucrose gradient fractions; T. Mosca and J. Salogiannis for training and assistance; and the DDRC Imaging Core of Children's Hospital. This work was supported by grants from the NEI (EY013874, P.A.G.) and NINDS(NS041062, T.L.S). R.K. was supported by the Medical Scientist Training Program 5T32GM007753-28.