GBF1 (Gartenzwerg)-dependent secretion is required for Drosophila tubulogenesis

Tube formation in multicellular organisms depends on the ability of epithelial cells to polarize and to form basal and apical membrane compartments de novo. During tube maturation, the initially narrow lumen, formed by the apical domains of one or several cells, is considerably expanded before becoming fully functional. Although tube formation and tube branching differ mechanistically between tubular organs (Andrew and Ewald, 2010; Baer et al., 2009; Bryant and Mostov, 2008; Chung and Andrew, 2008; Strilic et al., 2010), epithelial polarization is common to all systems with the orchestrated activity of the secretory pathway representing a prerequisite for polarization. In addition to the delivery of proteins to distinct membrane compartments, sorting machineries at the membrane are required that define, for example, adhesive domains to mediate cell–cell contact between lumen-forming cells and luminal membrane regions established at the luminal surface. Correspondingly, traffic from the Golgi complex to the cell surface was shown to be crucial for lumen formation. In the Drosophila melanogaster tracheal system, which has become an important model to study the molecular and cellular details of tubulogenesis, lack of e.g. Sec24 (also known as Stenosis and Haunted), a cargo-binding subunit of the COPII complex, causes a reduction in luminal diameter, cell flattening and incomplete cuticle assembly (Förster et al., 2010). The observed defects are cell-autonomous and the apical–basal cell polarity remains unaffected in sec24 mutant animals. One major function of COPII is the anterograde trafficking of cargo from the ER to the Golgi, from which surface proteins are further delivered towards the cell membrane through exocytic post-Golgi compartments (Lee et al., 2004; Spang, 2009). COPI, by contrast, plays a major role in targeting and shuttling vesicles from the Golgi back to the ER for recycling purposes. Nevertheless, components of the COPI complex are also essential for the formation of epithelial tubes. Mutations in the Drosophila γ-COP, a central component of the tetrameric subcomplex that forms the inner layer of a COPI vesicle, result in a reduction of the luminal diameter of the trachea (Grieder et al., 2008) and salivary glands (Jayaram et al., 2008), which is another well-defined model system for studying tubulogenesis. Indeed, virtually all mutants identified so far that affect assembly or function of the COPI or COPII complexes interfere with proper tubulogenesis in multicellular organisms. ADP-ribosylation factor guanine nucleotide exchange factors (ArfGEFs) are a major class of proteins that are key regulators of intracellular vesicle trafficking. As indicated by their names, they function as guanine nucleotide exchange factors (GEFs) for Arfs and play a central role in all eukaryotic cells by controlling the activation of small monomeric G proteins (Gillingham and Munro, 2007).

Arf proteins were first identified as cofactors for cholera-toxin-A-dependent ADP ribosylation of an adenylate cyclase subunit (Kahn and Gilman, 1986). Activation of Arf1 by GTPbinding is mediated by a GBF-type (Golgi-specific Brefeldin A resistance factor) large ArfGEF. Activated Arf1 triggers the assembly of the COPI complex and other downstream effectors. Depletion, drug inhibition or RNAi-mediated knockdown of the GBF-type ArfGEF Gea1 and Gea2 in budding yeast, of GNL1 in Arabidopsis and of GBF1 in mammalian cell cultures, respectively, all lead to COPI dispersal and malformation of the Golgi complex. This is accompanied by a cargo-selective transport defect with a partial block of membrane protein transport from or to the ER or early Golgi. In addition, selective endocytosis is also impaired (Peyroche et al., 2001; Peyroche and Jackson, 2001; Richter et al., 2007; Sáenz et al., 2009; Szul et al., 2005; Szul et al., 2007; Teh and Moore, 2007). These findings indicate central roles of GBF-type ArfGEFs in coordinating bidirectional protein and lipid transport and in maintaining the structural integrity of the Golgi.

Systematic genome-wide searches for ArfGEFs in several genomes have led to the conclusion that GBF-type and BIG-type ArfGEF subfamilies are common to all eukaryotes (Mouratou et al., 2005). Contrary to plants and budding yeast, however, metazoan genomes possess only a single member of the GBF-type family, indicating that these proteins are not functionally redundant. The single Drosophila orthologue of the GBF-type ArfGEF, encoded by the gene gartenzwerg (garz; CG8487) was identified in several genome-wide gain-of-function screens. Overexpression of garz in neuronal tissue causes pathfinding defects of motoneurons and the formation of ectopic synaptic branches in nerves of first instar larvae (Kraut et al., 2001). Furthermore, garz was identified in a gain-of-function screen for genes involved in salivary gland morphogenesis. Overexpression of garz induces hooks in salivary glands during embryogenesis, albeit with a low level of penetrance (Maybeck and Röper, 2009). Essential functions exerted by Garz were also uncovered by RNAi-mediated genome-wide knockdown screens in Drosophila S2 cells for genes involved in secretion and Golgi morphogenesis. Thereby, Garz was identified as being crucial for both processes (Bard et al., 2006). A similar approach led to the confirmation of Garz as being essential for the regulation of lipid homeostasis in S2 cells (Beller et al., 2008; Guo et al., 2008). Depletion of GBF1 in mammalian HeLa cells and Garz in Drosophila third instar larvae salivary glands provides further evidence for a crucial role of Garz in protein trafficking and secretion (Szul et al., 2011). Recently, garz was identified as a regulator of the non-canonical but highly conserved GEEC (GPI–AP enriched early endosomal compartments) clathrin- and dynamin-independent endocytic pathway (Gupta et al., 2009).

We identified the ArfGEF Garz in a genetic screen for heart malformations. garz mutant embryos exhibit a severe defect in secretion of the type IV collagen Pericardin in pericardial cells of the circulatory system. To gain further insight into the in vivo function of Garz we generated and identified several garz mutant alleles. All investigated alleles cause secretion defects accompanied by malformations of the ER and the Golgi complex. In addition we found that tubulogenesis in salivary glands and trachea is severely impaired in the absence of Garz. Herein we present the first phenotypic analysis of a metazoan GBF-type ArfGEF loss-of-function mutant in an organismic context.

Sequence analysis revealed that the Drosophila genome contains a single homologue of the human ArfGEF GBF1, called Gartenzwerg (Garz). The deduced amino acid sequence of Garz shares 44% identical and 61% similar residues with its human homologue, and has all conserved domains typical for ArfGEFs of the GBF-type protein subfamily (Fig. 1A,B). We initially analyzed the temporal and spatial expression of garz during embryogenesis and later stages of development by in situ hybridization and northern blot analysis. A garz cDNA sequence present in all putative splice variants predicted by the latest Drosophila genome annotation was used as a template to generate specific probes. Northern blot analysis revealed that two transcripts, presumably corresponding to the predicted isoforms garz-RA and garz-RB, are expressed during development, with transcript garz-RB being highly abundant in all stages tested and transcript garz-RA being present mainly in pupae and adults, but in much lower amounts (Fig. 1C). The third predicted isoform, garz-RC, was not detected. Whole-mount in situ hybridization revealed that in the embryo garz is most prominently expressed in invaginating ectodermal derivatives during gastrulation, i.e. in the primordia of the salivary glands, the proventriculus part of the stomodeum, the hindgut part of the proctodeum and in the tracheal placodes of embryos from stage 10 to 11 (Fig. 1D–R). Expression in these tissues is maintained later in development. In addition to the strong expression of garz in tissues that form tubular structures of epithelial character, a weak maternal contribution and a weak ubiquitous zygotic expression is apparent.

In order to analyze the subcellular localisation of Garz, we raised polyclonal antibodies against the protein (see Materials and Methods for details). The monospecificity of the affinity purified antibodies was verified by western blot analysis using protein extracts isolated from S2 cells and from whole flies (Fig. 2A). In protein preparations from S2 cells, a single band was detected at ~220 kDa, which corresponds well to the longer Garz isoform B (predicted molecular mass: 220.64 kDa). In protein extracts isolated from flies, a protein of similar size and abundance was labelled. However, in addition to this strong signal, a weakly expressed smaller protein with an apparent molecular mass of slightly less than 200 kDa was also detected, which indicates that not only isoform B but also the putative isoform A (predicted molecular mass: 194.9 kDa) is expressed during Drosophila development. Western blot analysis with protein extracts from wild-type and heterozygous garz^Δ137, heterozygous garz^EMS667 and heterozygous garz^S4-50 animals (garz mutants used in this study are described in the following section) revealed a reduction in full-length Garz abundance of ~50% in the heterozygotes compared with the wild type (Fig. 2B). In agreement with these data, immunodetection showed no Garz in salivary glands of homozygous mutant garz embryos (Fig. 2F,H,J), which further confirms antibody specificity as well as the absence of a full-length Garz protein in these mutant alleles. To obtain initial information on the cellular compartments Garz is associated with, subcellular fractions from untransfected S2 cells were obtained by sucrose density gradient centrifugation and probed with the Garz antiserum and established markers for the cis-Golgi network (GM130) and early endosomes (Rab5). Rab5 was chosen because a recent publication provided evidence of a role of Garz in the endocytic pathway (Gupta et al., 2009). As shown in Fig. 2C, Garz co-fractionated with cis-Golgi membranes as revealed by an identical distribution of Garz and GM130 with a peak accumulation in fraction 7. This result clearly indicates a function for Garz at this subcellular compartment. Rab5, however, accumulated predominately in fractions 3, 4, 5, 8 and 9, which is distinct from the distribution pattern of Garz. However, because of a certain overlap between Garz- and Rab5-positive fractions, e.g. fraction 6, an association of Garz with Rab5 endosomes cannot be excluded at this stage. To analyze Garz protein distribution in more detail, we performed colocalization studies in Drosophila S2 cells and in the budding yeast, Saccharomyces cerevisiae (Fig. 3). Simultaneous immunohistological stainings of Drosophila tissue with anti-Garz antibodies and for Golgi–EYFP (LaJeunesse et al., 2004) revealed that Garz is highly enriched in the ultra-proximal region of the Golgi complex, which corresponds to the cis-Golgi compartment (Fig. 3A–I). Clearly, Garz did not colocalize with markers for the trans-Golgi compartment as shown by simultaneous staining of salivary gland cells and S2 cells with anti-Garz, anti-AP1 (Benhra et al., 2011) and anti-GCC185 (Sinka et al., 2008) antibodies (Fig. 3D–F,J–O). This finding is further supported by comparing the subcellular distribution of Garz expressed in yeast cells with the localization of its yeast homologue Gea1. We found that Drosophila Garz and yeast Gea1 both colocalize with the yeast cis-Golgi marker Mnn9 (Fig. 3P–W). In summary, our results established Drosophila Garz as a protein of the cis-Golgi compartment.

We identified the S4-50 allele by screening a collection of EMS-induced mutants for heart formation defects (Albrecht et al., 2006; Hummel et al., 1999). At late stage of embryogenesis, garz mutant embryos show a secretion defect of the collagen Pericardin at the heart tube (Fig. 2G,I,K). The mutated gene in S4-50 was mapped by complementation analysis to chromosomal region 48F–49A on the second chromosome. S4-50 failed to complement the previously described transposon-induced gartenzwerg allele garz^EP(2)2028 (Rørth, 1996), in which the EP element is inserted into the putative promoter region of garz. A second EMS-induced garz allele, garz^EMS667, was isolated in the Technau laboratory (Vef et al., 2006). Transheterozygous animals with the genotype S4-50/EMS667, S4-50/EP(2)2028 or any other mutant combination are 100% lethal and show the same phenotypes as described in the following sections. To identify the EMS-induced point mutations in garz^S4-50 and garz^EMS667 mutant chromosomes we sequenced the garz locus from mutant animals (supplementary material Fig. S1). The garz^S4-50 chromosome carries a mutation that consists of a 2 bp deletion and the substitution of three nucleotides found in the third exon of the garz gene. This results in a frame shift and a premature stop codon, leading to a protein of 138 amino acids that lacks all conserved domains (supplementary material Fig. S1). The mutation in garz^EMS667 is a single G to A transition causing a premature stop codon leading to a truncated protein of 1529 amino acids. This protein lacks the C-terminus including the HUS3 domain (supplementary material Fig. S1). In addition, we generated a series of new garz deletion lines by remobilizing the transposable P-element EPgy2^EY07592, obtained from the Berkeley Drosophila gene disruption project (Bellen et al., 2004), in order to induce imprecise excision events. EPgy2^EY07592 harbours an EP element located in the 5′-garz region, as we verified by sequencing. EPgy2^EY07592 individuals are viable both as homozygotes and in trans over the deficiency Df(2R)Exel6061 that includes garz and several additional genes. Therefore, we considered EPgy2^EY07592 suitable for jump-out mutagenesis and screened for lethality caused by genomic deletions as a result of imprecise excision events. The 181 lethal mutant lines that were recovered belong to two complementation groups. Further genetic and molecular characterization revealed that all alleles from one complementation group are allelic to S4-50, EMS667, EP(2)2028 and the garz deficiency Df(2R)Exel6061. Sequencing showed the presence of deletions ranging from 556 bp in garz^Δ221 up to 1825 bp in garz^Δ137 (Fig. 1B). In all cases the deletion breakpoints map precisely to the previous insertion site of the P-element and sequences downstream within the garz locus. We focused on the largest deletion present in the garz^Δ137 allele that removes the putative garz promoter region as well as the first exon (Fig. 1). garz^Δ137 homozygous animals develop until the end of embryogenesis but fail to hatch and subsequently die. Moreover, mutant individuals exhibit asynchronous development and considerable retardation in comparison with wild-type controls (data not shown). Similar results were obtained for garz^S4-50/Df (2R)Exel6061 and garz^Δ137/Df(2R)Exel6061 mutant animals, indicating that these lines represent strong garz alleles. Lethal phase analysis of garz^EMS667/garz^EMS667, garz^EMS667/Df(2R) Exel6061 and garz^EP(2)2028/garz^EP(2)2028 mutants revealed early first instar larval lethality with a very low number of individuals surviving until the end of larval development, which indicates that these mutants represent weaker alleles.

The new Drosophila garz alleles enabled us to study the consequences of its loss of function in a multicellular organism. This is of particular interest because until now none of the known Garz homologues from higher eukaryotes was studied in an organismic context because of the absence of mutants. Nevertheless, a crucial role of Garz in the secretorypathway was demonstrated by biochemical analyses and studies in yeast and mammalian cell cultures (Chantalat et al., 2003; Claude et al., 1999; Niu et al., 2005; Peyroche et al., 2001; Peyroche and Jackson, 2001; Ramaen et al., 2007; Sáenz et al., 2009; Spang et al., 2001; Szul et al., 2007) and in Drosophila by an RNA interference (RNAi) approach (Szul et al., 2011). Recently, Gupta and colleagues also provided evidence of a role of Garz in the endocytic pathway in flies (Gupta et al., 2009). The assumed function of Garz in protein secretion suggests defects in extracellular protein deposition in garz mutant animals. Therefore, we stained wild-type and mutant embryos for secreted extracellular matrix (ECM) components and membrane-associated factors and focused on tissues that are known to be highly secretory. We utilized the embryonic salivary glands as a model organ to address the question of whether ECM formation at luminal membrane compartments depends, at least partially, on functional secretion and the underlying Garz activity. The second issue we were interested in was the extent to which garz-dependent secretion contributes to lumen diameter control in salivary glands. Staining of salivary glands with antibodies against GM130, a cytoplasmic protein tightly bound to membranes of the cis-Golgi network (CGN) in all higher eukaryotes (Nakamura et al., 1995), revealed that the CGN compartments of salivary gland cells were severely reduced in number in garz mutant embryos (Fig. 4A,B,E,F). Consistent with this observation, we found that vesicles secreting electron-dense ECM material, which are normally numerous close to the apical membrane of salivary gland cells at stage 16, were entirely absent in garz mutant embryos (Fig. 4Q,R). At the same time, the highly enriched rough ER, which is characteristic of active secretory cells, vanished in garz mutants. Instead, the ER compartment appeared as ‘bloated balloons’. It is noteworthy that mutations in genes crucial for retrograde vesicle trafficking such as that for γ-COP, as well as anterograde vesicle trafficking such as sec23 (also known as ghost) sec24 and sar1, result in a very similar ER phenotype (Förster et al., 2010; Jayaram et al., 2008; Norum et al., 2010; Tsarouhas et al., 2007). Failure of luminal diameter expansion was also clearly seen by stainings for the apical membrane determinants Crumbs and Bazooka (Tepass et al., 1990; Wodarz et al., 1999) (Fig. 4C–F and insets) and in cross sections analyzed by transmission electron microscopy (TEM; Fig. 4Q,R). The formation of zonulae adherentes and septate junctions is not affected in garz mutant embryos, as most clearly seen in TEM sections (insets in Fig. 4Q′,R′) and in stainings for the septate junction markers Fasciclin III (FasIII) and Discs large (Dlg; Fig. 4G–J).

Taken together, our observations provide clear evidence for a specific role of Garz in luminal diameter expansion in salivary glands. Interestingly, the luminal length remains unaffected in garz mutant embryos (Fig. 6), indicating that Garz is not involved in invagination or in the elongation of salivary gland cells, but is involved in apical membrane expansion.

Recent studies revealed an essential role of cell–matrix interactions in tube morphogenesis, including the generation, shaping and maintenance of epithelial tubes (Affolter and Caussinus, 2008; Schottenfeld et al., 2010). Several components of the nascent tracheal lumen, e.g. the zona pellucida (ZP) protein Piopio (Pio) or the chitin deacetylases Vermiform (Verm) and Serpentine (Serp) have been shown to play essential roles in length and size control of trachea (Jazwinska et al., 2003; Luschnig et al., 2006). Furthermore, the convoluted gene, coding for the Drosophila acid-labile subunit (ALS), which forms a ternary complex with Drosophila insulin-like peptide 2 (Dilp2) and the binding protein IMP-L2 (Arquier et al., 2008), is required for the dynamic organization of the transient luminal matrix and for the establishment of cuticular structures that line the tracheal lumen (Swanson et al., 2009). These components are secreted from the apical membrane facing the lumen and their deposition is dependent on the secretory machinery of the cells. On the basis of our data, we assumed that Garz plays a key role in the delivery of tracheal ECM proteins to the apical membrane. Using the monoclonal antibody mAb2A12, which exclusively labels a yet unknown luminal antigen, we observed a large reduction of the tracheal luminal diameter in garz mutants (Fig. 5A,A′). Retention of the mAb2A12 antigen within the tracheal cells implies that loss of Garz prevents the antigen from being secreted into the lumen. We furthermore examined the distribution of the ZP protein Piopio (Bökel et al., 2005; Jazwinska et al., 2003). Piopio, which is normally secreted across the apical membrane into the luminal space and tightly linked to the ECM lining the surface of the lumen, was retained inside the cells in garz mutant embryos (Fig. 5B,B′). Similar results were obtained for the chitin deacetylase Vermiform (Luschnig et al., 2006) (Fig. 5C,C′). Similar to Piopio and mAb2A12, Vermiform secretion and Crumbs delivery were severely impaired in garz mutants. The reduction of cell matrix components at the luminal surface of trachea might also affect airway clearance. At approximately stage 17 trachea are filled with gas that has replaced the liquid present in all tracheal branches at earlier stages (Behr et al., 2007; Tsarouhas et al., 2007). In order to prove an effect of garz on the clearance process, we used video bright-field microscopy of living specimens and found that airways clearance was inhibited in garz mutant animals (Fig. 5D,D′ and supplementary material Movies 1 and 2). From this we conclude that Garz plays an important role in airway morphogenesis and maturation. Similar to our observations on salivary glands, delivery of apical membrane proteins such as Crumbs or septate junction proteins such as FasIII was not severely disturbed in tracheal cells of garz mutant embryos (Fig. 5E–M). Notably, before the death of homozygous garz mutant individuals, the tracheal lumen collapses (own observation). This might be caused by improper formation of cuticle structures that mechanically support tube stability, and such defects inside tracheal lumina should be visible in cross sections analyzed by transmission electron microscopy. As expected, besides a reduction in luminal diameter we found an aberrant organization of chitin cables inside the tracheal lumen in garz mutants (Fig. 5N,O). On the basis of these data, we addressed the question of whether cuticle development in general is affected in garz. Indeed, inspection of stage 17 embryos revealed that also epidermal secretion of the cuticle is disturbed to various degrees in garz mutants (supplementary material Fig. S2), which is consistent with a requirement for Garz in secretion processes mainly, but not exclusively, in tubular structures.

During the last few years, much insight has been gained into the activities and roles of ArfGEFs using cultured mammalian cells, plants and budding yeast in conjunction with in vitro approaches. However, up to now little has been known about the physiological and developmental consequences of mutations affecting ArfGEFs of the Garz family in multicellular animals. Sheen and colleagues provided evidence that point mutations in the human ArfGef2 (BIG2), identified in two patients suffering from microcephaly caused by disruption of neural precursor proliferation, inhibit protein transport from the Golgi to the cell surface (Sheen et al., 2004). A third patient with heterozygosity for two ArfGEF2 (BIG2) mutations was recently reported to suffer from movement disorder, bilateral periventricular nodular heterotopia (BPNH) and secondary microcephaly (de Wit et al., 2009). However, in this case molecular data explaining these symptoms were not reported. In human HeLa cells, small-interfering RNA based suppression of ArfGEF1 (BIG1) and ArfGEF2 (BIG2), which constitute, together with GBF1, the three human large ArfGEFs, revealed that BIG1 is required to maintain the normal morphology of the Golgi complex and BIG2 is essential for endosomal compartment integrity (Boal and Stephens, 2010).

Homologues of Garz were identified in all eukaryotes in the course of genome projects; however, their in vivo role was analyzed only in Arabidopsis thaliana and in Saccharomyces cerevisiae. Mutations in the Arabidopsis paralogues GNOM and GNOM-like (GNL1) lead to impaired auxin transport, caused by the requirement of GNOM for correct polar localization of PIN1, a crucial efflux facilitator for auxin (Bonifacino and Jackson, 2003; Geldner et al., 2003; Geldner et al., 2004; Richter et al., 2007; Teh and Moore, 2007). In budding yeast, two paralogous genes for Gartenzwerg are present, namely Gea1 and Gea2. Double mutants are lethal, causing failure to bud, and the dying cells show severe defects in the organization of the actin cytoskeleton (Peyroche et al., 2001; Spang et al., 2001; Zakrzewska et al., 2003).

As reported in this study, in Drosophila the morphologies of both the ER and Golgi complex clearly depend on Garz. Our TEM analysis of highly active secretory cells in salivary glands revealed a considerable expansion of the ER lumen and many fewer Golgi-derived vesicles in garz mutant animals. Secretory vesicles that normally accumulate close to the apical membrane and subsequently release their cargo into the luminal space are virtually absent in garz mutants. Instead, an accumulation of electron-dense material on the basal side of salivary gland cells is visible, which is accompanied by a large number of apparently empty vesicles. Therefore, we postulate that Garz interferes with the secretory pathway of the cells. This interpretation of our results is in agreement with previous studies that have established Garz as a major component of the Arf1–COPI secretory pathway. Blocking Arf1–COPI either in budding yeast, cultured mammalian cells or in Drosophila mutants causes a progressive disassembly of the Golgi with its components being redistributed back to the ER. In addition, mutations in Arf1–COPI hinder the transport of secretory vesicles towards the plasma membrane. Similarly, inhibition of the Sar1–COPII machinery (Förster et al., 2010; Norum et al., 2010; Tsarouhas et al., 2007) also causes disassembly of the Golgi complex and cargo fails to reach its correct membrane destination (Ward et al., 2001). In the absence of functional Garz, the activation of Arf1, which is involved in the formation and function of COPI complexes, fails, eventually leading to the disruption of one or several trafficking routes. This idea is corroborated by our findings that several proteins known to be transported either to distinct membrane compartments or secreted into the lumen or to cuticle structures, are retained inside the cell and accumulate in vesicular compartments. We conclude that the phenotypes seen in garz mutants on the cellular and ultrastructural level are consistent with the function of the protein as a guanine nucleotide exchange factor for Arf1 in the Arf1–COPI secretory pathway machinery.

Furthermore, we found that animals lacking garz suffer from morphological defects in lumen formation in salivary glands and trachea. Although the formation of biological tubes is mechanistically different in these tissues (Baer et al., 2009), apical membrane and luminal matrix expansion probably play a general role in tube formation. Garz might be directly involved in these processes through its known ability to convert Arf1 from an inactive into an active state (Kawamoto et al., 2002; Szul et al., 2005; Zhao et al., 2002). On the one hand, activated Arf1 regulates COPI coat assembly and vesicle budding at the Golgi membrane. Disruption of this process impairs the proper delivery of secreted components or membrane regulators to their physiological destinations, e.g. the expanding lumen of salivary glands. On the other hand, Arf1 modulates Golgi structure by stimulating the assembly of Spectrin and Actin cytoskeletal elements at Golgi membranes (Fucini et al., 2002; Fucini et al., 2000; Godi et al., 1998). In garz mutants ARF1 presumably remains inactive, which ultimately leads to the disassembly of Golgi architecture. Such an idea is in agreement with studies demonstrating that mutations affecting COPI coatomers γ-COP and Δ-COP (Grieder et al., 2008; Jayaram et al., 2008), COPII coatomers Sec23 and Sec24 (Förster et al., 2010; Norum et al., 2010) and Sar1 (Tsarouhas et al., 2007) compromise ER and Golgi morphology and additionally cause severe defects in tube expansion in salivary glands and trachea. Our finding that garz mutants affect the delivery of luminal matrix components and thereby cause luminal diameter reduction is also in agreement with previous studies that have shown that, for example, α-subunits of resident ER enzymes PH4αSG1 and PH4βSG1, which are hydroxylate proline residues in procollagen and other secreted proteins, are responsible for altered salivary gland secretion causing regional tube dilation and constriction, with intermittent tube closure (Abrams et al., 2006). Although tracheal lumen formation is mechanistically distinct from lumen formation in salivary glands, garz mutations lead to perturbed lumen dilation also in this tissue. As a direct consequence of loss of Garz activity, the transport vesicles carrying lipids for apical membrane growth and transient luminal proteins are not able to be targeted to the luminal side. Thus, protein secretion into the lumen and incorporation into the apical membrane is severely disturbed.

Recently it was shown that Garz has a function in the endosomal system (Gupta et al., 2009). The authors provided evidence, that Garz is required for the Arf1-dependent fluid uptake mediated by the clathrin- and dynamin-independent GEEC pathway. After internalization, GEEC endosomes fuse with Rab5-containing endosomes. Initially, GEEC endosomes are mainly devoid of early endosomal markers such as Rab5 and EEA1 (Kalia et al., 2006). This is in accordance with our observation that Garz and Rab5 cofractionate only to a very limited amount (Fig. 2C), indicating, that Garz might play a role in early steps of the GEEC pathway rather than being associated with other endosomal pathways. Although speculative, it might be that Garz and the GEEC pathway are required for fluid uptake upon tracheal liquid clearance.

Our analysis of garz loss-of-function mutants led us to conclude that Garz is essential to initiate assembly of the Arf1–COPI machinery and to maintain the integrity of the Golgi complex. In the secretory pathway, lack of intact Golgi and COPI-vesicle formation leads to failure in directed cargo export and accumulation of cargo within and at the ER, which is in agreement with the previously proposed models (Ward et al., 2001). During organogenesis, lack of Garz becomes manifested primarily in tubular organs and epithelial structures, which are strongly impaired.

The wild-type strain used was white¹¹¹⁸. The EMS allele S4-50 was generated in the lab of C. Klaämbt (Hummel et al., 1999) and identified by us in a genetic screen for heart mutants (Albrecht et al., 2006). The EMS-induced mutant garz^EMS667 was generated in the laboratory of G. Technau and identified in a screen for nervous system mutants and verified as a garz allele in this study and by O. Vef and B. Altenhein (Vef et al., 2006). The garz^EP(2)2028 allele was generated in P. Rørth's lab and obtained from the Szeged Stock Center (Rørth, 1996). The EPgy2^EY07592 P-element line was generated by the Drosophila genome project (Bellen et al., 2004) and provided by the Bloomington Stock Center. All other lines were obtained from the Bloomington Stock Center. We used the Bloomington deficiency kit (released March 2005) and the DrosDel isogenic core deletion kit (Ryder et al., 2004) for mapping of the EMS-induced alleles.

The homozygous viable P-element insertion line EPgy2^EY07592 was used to generate garz mutations by imprecise excision. EPgy2^EY07592 (w⁺ mc, light orange eyes) was mobilized as follows: 100 y w¹¹¹⁸; EPgy2^EY07592 virgin females were crossed to jump starter males y w¹¹¹⁸; PBac^Δ2–3 (w⁺ red eyes). Resulting y w¹¹¹⁸; EPgy2^EY07592/PBac^Δ2–3 males (100) were individually crossed to y w¹¹¹⁸; If/CyO, kr–GFP virgin females. Individual white eyed males from the F2 generation (100) were selected and crossed back to y w¹¹¹⁸; If/CyO, kr–GFP females to establish stocks. We tested ~800 lethal jump out lines for complementation with Df(2R)Exel6061 in which garz is absent. Lines that failed to complement Df(2R)Exel6061 were kept for further analysis. The newly induced stocks fall into two complementation groups. One group turned out to be allelic to the EMS-induced mutant alleles S4-50 and EMS667, and to the EP(2)2028. We sequenced genomic DNA isolated from theses alleles to determine the deletion breakpoint (see Result section). The second complementation group is allelic to the larger deficiency Df(2R)Exel6061 but not to garz alleles. Sequencing of genomic DNA isolated from alleles belonging to this second complementation group revealed that they carry deletions with breakpoints mapping to the original EP-element insertion site and within the gene CG8841. These alleles were not further characterized. For lethal phase determination we balanced garz alleles over CyO, kr–GFP. Homozygous mutant individuals were identified by the absence of GFP expression. A knock-in line carrying a Crumbs-GFP fusion protein was obtained from Y. Hong (University of Pittsburgh) (Huang et al., 2009).

A cDNA containing the whole open reading frame of garz was amplified from isolated total RNA from Oregon-R flies using the SuperScript One-Step RT-PCR system and subsequently cloned into pCR2.1-TOPO vector (Invitrogen, Carlsbad, CA). Primers used were: forward: 5′-ATGGCGCTTCCAGGCAACG-3′ and reverse: 5′-TCACTGCTGGCCGTAGAGCA-3′. To determine the EMS-induced point mutations in garz^S4-50 and garz^EMS667, we isolated genomic DNA from heterozygous adult flies or selected homozygous mutant embryos. PCR amplicons were cloned and sequenced twice in both directions. To distinguish between polymorphisms and EMS-induced point mutations, we amplified and sequenced corresponding gene regions from EMS alleles from the same mutant collection with isogenic background in parallel (kuz^J2-11, mam^S2-29) (Albrecht et al. 2006).

For probe synthesis we cloned a 1937 bp fragment (nucleotides 302–2268) of the full-length garz cDNA into the pGEM-T easy cloning vector (Promega, Madison, WI) using the following primers: forward 5′-ATCCCACGTCTCCAAATCTGG-3′, and reverse 5′-ATGCCAATGATCAGAAAAGTG-3′. Sense and anti-sense RNA probes were synthesized using the DIG RNA Labelling Kit (Roche, Basel, Switzerland). In situ hybridization and double labelling with antibodies were performed as described previously (Duan et al., 2001).

A 939 bp cDNA fragment encoding amino acids 1368–1680 of the predicted isoform A was cloned into the pET16b vector and transformed into Escherichia coli Rosetta (DE3) cells (Novagen, EMD Biosciences, Madison, WI). The sequences of the respective primers were: forward 5′-TACTCACATAT-GCGCTGCATCCGCATCTTT-3′ GTG, and reverse 5′-TACTCGGGATCC-TCTAATTTGCGTATAGTCAAG-3′. Expression was done essentially as previously described (Meyer et al., 2009) and the refolded protein (fused to a N-terminal 10xHis tag) was purified using Ni-NTA Agarose (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Finally the antigen was used for immunization of two rabbits (Pineda Antibody Service, Berlin, Germany). To improve specificity, the resulting sera were purified by antigen affinity, and the monospecificity was confirmed by western blotting.

S2 cells were grown in Schneider's Drosophila medium supplemented with L-glutamine (Invitrogen) in 175 cm² flasks to 95% confluency and harvested by centrifugation (100 g, 10 minutes). Cell were disrupted in PBS supplemented with Protease-Inhibitor Mix M (Serva, Heidelberg, Germany), by the freeze–thaw method and the coarse protein extract was centrifuged at 2000 g for 10 minutes to remove cell debris and nuclei. The resulting postnuclear supernatant was then layered onto a 0.3–2.0 M linear sucrose gradient and centrifuged at 36,000 r.p.m. (160,000 g) for 18 hours using a SW 41 Ti rotor (Beckman Coulter, Brea, CA). A total of 12 fractions of 1 ml each were collected from the bottom of the tube and analyzed by SDS-polyacrylamide gel electrophoresis and western blot analysis using anti-Garz antibody and the following antibodies for the indicated membrane compartments: anti-GM130 (cis-Golgi) and anti-Rab5 (early endosomes).

Garz-RB and Gea1 cDNAs were tagged with a yeast EGFP (Sheff and Thorn, 2004) at the C-terminus using PCR-based homologous recombination, and stably integrated at the gea1 locus by homologous recombination using standard techniques (cloning details available upon request). For colocalization studies we used a strain that carries the cis-Golgi protein Mnn9 (Jungmann and Munro, 1998) fused to the td-Tomato cassette (Meiringer et al., 2008). Tetrad analyses of the diploid strains were followed by succeeding selections on appropriate media for haploid segregants producing either Gea1–EGFP or Garz–EGFP together with Mnn9–td-Tomato.

Whole-mount immunostainings were carried out as described previously (Albrecht et al., 2011). Embryos were collected from grape juice agar plates and dechorionated in 50% DanKlorix (Colgate Palmolive), then fixed in a 1:1 (v/v) mixture of PBS, containing 50 mM EGTA and freshly prepared methanol-free 9% formaldehyde (Polysciences Inc., Eppelheim, Germany) and n-heptane for 30 minutes at room temperature. Devitellinisation was performed by shaking for 60 seconds in 1:1 (v/v) mixture of methanol and n-heptane. Confocal images were captured either with a Zeiss LSM 5 Pascal confocal microscope or a Leica SP1 confocal microscope. Z-stacks are depicted as maximum projections if not denoted otherwise. Whole-mount in situ hybridizations were documented with a Zeiss AxioCam MRc5 camera mounted on a Zeiss Axioskop 2, whereas immunofluorescence and live cell imaging of yeast and S2 cells were undertaken with a cooled CCD CoolSNAP HQ camera on a Zeiss Axioplan 2 then processed with MetaMorph v.6.2 software. For TEM analysis, embryos were prepared as described previously (Lehmacher et al., 2009; Tepass and Hartenstein, 1994). After fixation, the specimens were embedded in Epon 812, and ultra-thin sections (70 nm) were cut for TEM with a diamond knife on a Leica Ultracut UCT ultramicrotome. Sections were mounted on single slot grids, contrasted with uranyl acetate (40 minutes; 20°C) and lead citrate (6 minutes; 20°C) using a Leica EMstain. Specimens were examined with a Zeiss 902 transmission electron microscope (60 kV).

Monoclonal antibodies for Crumbs (mAbCq4, 1:100); Dlg (4F3, supernatant 1:10); FasIII (7G10; supernatant 1:10) and Pericardin (EC11, 1:100) were obtained from the Hybridoma Bank, University of Iowa. Monoclonal mouse anti-galactosidase (Z378B; 1:5000) was from Promega; monoclonal mouse anti-GFP (3E6; 1:1000) was from Invitrogen (A11120); polyclonal antibodies for GM130 and Rab5 were from Abcam (ab30637, ab31261; 1:100). The polyclonal rabbit anti-Pio antibody (Jazwinska et al., 2003) was used at a dilution of 1:200, the polyclonal rabbit anti-Verm (Luschnig et al., 2006) at 1:100, polyclonal rabbit anti-Bazooka at 1:1000 (Wodarz et al., 1999). The polyclonal mouse anti-AP1 (Benhra et al., 2011) was used at a dilution of 1:300, the polyclonal rat anti-GCC185 (Sinka et al., 2008) at a dilution of 1:100 and the polyclonal mouse anti-MHC at a dilution of 1:500 (Kiehart, 1990). Monoclonal mouse IgM 2A12 was kindly provided by Nipam Patel (University of California at Berkeley) and used at 1:10 with the Vectastain Kit from Vector Laboratories. The anti-Garz, described in this study, was used at a dilution of 1:100. Secondary antibodies coupled with Cy2 or Cy3 were obtained from Dianova (Pinole, CA). Fluorescein-conjugated chitin-binding probe (CBP) from New England Biolabs (P5211S; Ipswich, MA) was kindly provided by R. Schuh (Max Planck Institute for Biophysical Chemistry, Göttingen) and used at a dilution of 1:500.

Cuticles were prepared as described previously (Alexandre, 2008), with minor modifications. Briefly, embryos between 0 and 6 hours AED (after egg deposition) were collected and aged for an additional 24–36 hours at 25°C. Then, mobile larvae were removed from the agar plate. the remaining specimens were fixed in 4% formaldehyde. For cuticle preparation, a mixture consisting of 50% Hoyer's mountant and 50% lactic acid was used. The specimens were incubated at 60°C for 72 hours before mounting on a microscope slide for examination using dark-field microscopy.

We thank C. Klaämbt, R. Le Borgne, S. Goto, Y. Hong, S. Luschnig, S. Munro, N. Patel, R. Schuh, G. Technau, C. Ungermann and A. Wodarz for sharing fly stocks or reagents, O. Vef for initial mapping of the garz^EMS667 allele, S. Tejedor Vaquero for assisting DNA sequence analysis, and S. Vos for generating the time-lapse movie. We also thank K. Etzold for her assistance with TEM analysis, and E. Haß-Cordes, M. Krabusch and M. Biedermann for excellent technical support.