Tango1 coordinates the formation of endoplasmic reticulum/Golgi docking sites to mediate secretory granule formation

Regulated secretion is a conserved process occurring across diverse cells and tissues. Current models suggest that the conserved cargo receptor Tango1 mediates the packaging of collagen into large coat protein complex II (COPII) vesicles that move from the endoplasmic reticulum (ER) to the Golgi apparatus. However, how Tango1 regulates the formation of COPII carriers and influences the secretion of other cargo remains unknown. Here, through high-resolution imaging of Tango1, COPII, Golgi, and secretory cargo (mucins) in Drosophila larval salivary glands, we found that Tango1 forms ring-like structures that mediate the formation of COPII rings rather than vesicles. These COPII rings act as docking sites for the cis-Golgi. Moreover, we observed nascent secretory mucins emerging from the Golgi side of these Tango1–COPII–Golgi complexes, suggesting that these structures represent functional docking sites/fusion points between the ER exit sites and the Golgi. Loss of Tango1 disrupted the formation of COPII rings, the association of COPII with the cis-Golgi, mucin O-glycosylation, and secretory granule biosynthesis. Additionally, we identified a Tango1 self-association domain that is essential for formation of this structure. Our results provide evidence that Tango1 organizes an interaction site where secretory cargo is efficiently transferred from the ER to Golgi and then to secretory vesicles. These findings may explain how the loss of Tango1 can influence Golgi/ER morphology and affect the secretion of diverse proteins across many tissues.

synthesized and transported to the cis-region of the Golgi apparatus in a COPII-dependent process (1)(2)(3)(4)(5). Appropriately modified and folded proteins are packaged into secretory vesicles emanating from the trans-Golgi network that then await appropriate signals before fusing with the plasma membrane to release their contents. However, the mechanisms whereby bulky cargo is efficiently packaged into small vesicular transport vehicles and moved between compartments of the secretory apparatus are largely unknown.
Recently studies have identified an essential cargo receptor (Tango1 or MIA Src homology 3 (SH3) domain ER export factor 3; MIA3) responsible for the efficient packaging and secretion of high-molecular-weight collagen (6,7). Tango1, a type I transmembrane protein located at the ER exit sites (ERES) was first identified in a screen for genes that affect general secretion and Golgi morphology in Drosophila cells (8). Subsequent studies demonstrated a role for Tango1 in collagen secretion, whereby it is thought to mediate the formation of large COPII megacarriers capable of transporting the large procollagen rods from the ER to the Golgi apparatus (6,7). Current models suggest that the luminal SH3 domain of Tango1 binds to the procollagen chaperone HSP47 (9), directing procollagen to sites of COPII vesicle formation. Tango1 is thought to modulate the size of the COPII vesicles by recruiting factors that limit Sar1GTPase activity, thus allowing the vesicles to grow in size to accommodate this large cargo (1,6,10,11).
Many recent studies have suggested that Tango1 is important for the secretion of additional molecules other than collagen. In mammalian cells, Tango1 affects the export of bulky lipid particles such as pre-chylomicrons/very low-density lipoproteins (12). In Drosophila, loss of Tango1 results in defects in the secretion of mucins, laminins, perlecan, and other extracellular matrix proteins (13)(14)(15)(16)(17). Additional genetic studies suggest that Tango1 influences general secretion rather than the specific secretion of certain proteins (18). These studies also identified many additional Golgi proteins that interact with Tango1, either directly or indirectly, such as Grasp65 and GM130, and suggest that there may exist more direct contacts between the ER and Golgi that are mediated by Tango1 (14). Work by Ríos-Barrera et al. (16) suggests that although Tango1 is important for the secretion of bulky cargo, it has an additional role in ER-Golgi morphology. However, high-resolution visualization of Tango1 dynamics and COPII vesicle formation relative to endogenous cargo biosynthesis and packaging has been challenging given the small size of these structures and the res-olution limits of light microscopy. Thus, the exact roles Tango1 plays in the packaging and secretion of diverse cargos, as well as ER-Golgi morphology, remain unclear.
Here, we use the Drosophila larval salivary gland (SG) to image the relationship between Tango1 and the synthesis and packaging of secretory cargo (mucins) in real time, taking advantage of the increased spatial resolution unique to this gland. The SG undergoes hormonally regulated secretory granule formation that results in secretory granules of 3-8 microns in diameter (ϳ10 -100ϫ larger than those seen in mammalian systems) that are filled with highly O-glycosylated mucin proteins (19 -23). Drosophila mucins are similar in structure to mammalian mucins (having serine/threonine-rich O-glycosylated regions) but are typically smaller in size (24). Fly lines expressing GFP-tagged versions of one secretory mucin (Sgs3-GFP) (19) have allowed high-resolution, real-time imaging of secretory granule formation and secretion (21-23). The genetic tractability of Drosophila has also allowed the identification of factors that control secretory vesicle formation, morphology, and extrusion of bulky cargo, such as mucins (21)(22)(23)(25)(26)(27). Through real-time imaging using this system, we find that Tango1 undergoes regulated self-association and dynamic shape changes during hormonally induced secretion to form ring structures that mediate the formation of COPII rings rather than vesicles. These Tango1-COPII rings act as docking sites for the cis-Golgi. Moreover, we image nascent secretory mucins emerging from the Golgi side of these Tango1-COPII-Golgi complexes, suggesting that these structures represent functional docking sites/fusion points between the ER exit sites and the Golgi. Taken together, our data suggest that Tango1 acts as a scaffold for the formation of functional ER-Golgi junctions that allow the efficient synthesis, intraorganellar transport, and packaging of diverse secretory cargo.

Tango1 is required for secretory granule formation in diverse tissues
tango1 is broadly expressed across diverse tissues in Drosophila that form secretory granules and undergo regulated secretion (www.flybase.org, identifier FBgn0286898), 3 and loss of tango1 (via RNAi or conventional mutations) results in altered secretory apparatus structure and disruption of secretion (14,16,17). Although Tango1 has well-documented roles in the packaging and secretion of collagen, we set out to examine its effects in tissues that form secretory granules containing diverse cargo proteins (Fig. 1). For example, the male accessory gland secretes a variety of seminal peptides and proteins that are transferred to the female, affecting postmating behavior (28); the female spermatheca is known to secrete proteases, lectins, and enzymes that play a role in sperm storage (reviewed in Refs. 29 and 30); the larval proventriculus packages and secretes a highly O-glycosylated mucin (Muc26B) that is a component of the protective peritrophic membrane of the digestive tract (31,32); and the SG secretes a variety of O-glycosylated mucins and small proteins that form the glue that mediates prepupal adhesion to a substrate prior to metamorphosis (20). RNAi directed against tango1 (tango1 RNAi ) in each of these tissues results in the disruption of secretory granule formation (Fig. 1). Secretory granules in the female spermatheca and male accessory gland (detected by the lectin Helix pomatia, HPA, which recognizes GalNAc linked to serine or threonine) are readily visible in a WT background but are no longer present in tango1 RNAi animals ( Fig. 1, a and b). tango1 RNAi in the larval proventriculus resulted in the loss of Muc26B containing secretory granules and the accumulation of Muc26B in the basal region of the cells (Fig. 1c and Ref. 17). tango1 RNAi in the larval salivary glands likewise resulted in the loss of the mucin-containing secretory granules (Sgs3-GFP) and abnormal accumulation of Sgs3 within the secretory cells of the gland ( Fig. 1d and Fig. S1). Taken together, these results highlight a role for Tango1 in the proper packaging of diverse cargo into secretory granules across many tissues.

Tango1 undergoes dynamic conformational changes to form rings prior to secretory granule formation
To further investigate the role of Tango1 in secretory granule formation in vivo, we took advantage of the larval SG, which offers increased spatial resolution because of the size of the secretory granules (3-8 microns in diameter) (21-23). Additionally, the SG allows one to image secretory vesicle formation and secretion in real time (21-23). Previous work from our laboratory (22) and the Shilo laboratory (21) employed a Drosophila line carrying a fluorescently labeled mucin cargo (Sgs3-GFP or Sgs3-RFP) (19) to detail the steps and factors involved in secretory granule fusion with the apical plasma membrane and secretion of vesicular cargo. Here, we use this same approach to image early stages of secretory granule biogenesis and Tango1 dynamics. As shown in Fig. 2a, granules are first visible when they are ϳ0.5 m in diameter (Fig. 2a). They then undergo a series of homotypic fusion events ( Fig. 2a and Movie S1) (23) until they reach the size of mature granules (3-8 m in diameter), at which point homotypic fusion stops. To image Tango1 dynamics during the process of secretory granule formation, we constructed a Drosophila line carrying a fluorescently labeled version of Tango1 (Tango1-GFP). Using this line, we performed super resolution real-time lattice structured illumination microscopy (SIM) imaging of Tango1 in SGs. This novel imaging platform, which provides extended imaging time while negating bleaching effects, demonstrated that Tango1 undergoes dynamic movements and self-associates to form a circular structure ( Fig. 2b and Movie S2). These circular structures were also seen when using an antibody to detect endogenous Tango1 in a WT background, indicating that these structures are not due to overexpressed recombinant Tango1-GFP (Fig. 2c).
To investigate the temporal relationship between these Tango1 structures and the formation of secretory granules, we next stained SGs expressing Sgs3-GFP (a cargo protein) with antibodies to endogenous Tango1. The formation of secretory vesicles is not synchronous within the cells of a SG, with secretory vesicle formation and secretion beginning in 3 Please note that the JBC is not responsible for the long-term archiving and maintenance of this site or any other third party hosted site.

Tango1 creates ER-Golgi docking sites
the most distal cells of the gland. One can typically observe cells at various stages of secretory granule formation within an individual SG, allowing one to see snapshots in time of the cellular changes that occur as secretory granules form. Fig.  2d shows three juxtaposed cells of a SG (separated by white dashed lines) at various stages of secretory granule biogenesis based on the presence of Sgs3-GFP. In Fig. 2d, the cell in the top right corner does not yet have visible Sgs3-GFP granules, but Tango1 is present in circular structures, suggesting that Tango1 adopts this structure early in the process before secretory granules are visible by light microscopy. The cell on the left side of Fig. 2d has visible small (ϳ1 m in diameter) secretory granules present, and Tango1 is still in a circular structure. Finally, the cell in the bottom portion of Fig.  2d has larger, more mature secretory granules, and Tango1 remains in circular structures, similar to Tango1 structures seen previously (33). Higher magnification images from each cell (boxes 1, 2, and 3) rotated in three dimensions reveal that the Tango1 structure is a circular ring (not a sphere or disc) and that this ring exists before secretory granules are visible (Fig. 2e, panels 1, 2, and 3, respectively). These results indicate that Tango1 undergoes self-association and dynamic structural changes to form a ring structure prior to secretory granule biosynthesis and maintains this structure throughout the duration of secretory granule formation.

Tango1 mediates structured COPII/Golgi docking sites
Previous studies have suggested models in which Tango1 initiates the formation of large COPII vesicles containing cargo (megacarriers) that bud from the ERES and eventually fuse with the cis-Golgi (6,(33)(34)(35), whereas other studies suggest that Tango1 may coordinate interactions between ERES, COPII vesicles, and the Golgi to enhance cargo transit (14). To further investigate these two models, we examined the spatial relationship of each of these components at high resolution using the SG system. We first stained SGs with COPII to examine its structure in three dimensions. Surprisingly, we found that, like Tango1, COPII was also present as a circular ring rather than disc or sphere ( Fig. 3a and Fig. S2). High-magnification images rotated in three dimensions and reconstructed show COPII in a ring structure, suggesting that it is not forming a separate vesicular carrier (Fig. 3a). We next stained WT SGs with antibodies to Tango1, COPII, and the cis-Golgi marker GMAP (Fig. 3b). Similar to what has been reported in previous studies, Tango1 and COPII are closely associated and colocalize throughout all of the cells of the SGs in ring-like structures (Fig. 3b). Interest-

Tango1 creates ER-Golgi docking sites
ingly, cis-Golgi (GMAP) staining was closely associated with both Tango1 and COPII and was clearly present in the inner portion of the Tango1-COPII rings, as seen in magnified en face and side views (Fig. 3b). We repeated these stainings using antibodies to a trans-Golgi marker (Golgin-245) along with Tango1 and the cis-Golgi marker GM130 and obtained similar results (Fig. 3c). Magnified en face and side views of these structures reveal that Tango1 forms a ring that appears to encircle the cis-Golgi marker, with the trans-Golgi marker (Golgin-245) in close proximity to the cis-Golgi marker (Fig. 3c). These results suggest that Tango1 may form a ring structure that is lined with COPII and encircles the cis-Golgi, indicative of a docking site for the Golgi apparatus.
To address the role of Tango1 in forming these structures, we next performed RNAi to tango1 in SGs and stained for COPII and GMAP. Interestingly, in the absence of Tango1, COPII fails to form the ring structures and remains in dispersed puncta throughout the cell (Fig. 4a). GMAP also has a more dispersed appearance throughout the cell and fails to colocalize with COPII to the same extent seen in WT SGs (Fig. 4a). Two-dimensional line scans illustrate the degree of spatial association between COPII and GMAP in WT and the dispersed pattern of both COPII and GMAP throughout the cell in the absence of Tango1 (Fig. 4, b and c, and Fig. S3).
We further examined the ability of Tango1 to initiate the formation of COPII rings and Golgi docking sites by overexpressing full-length and deletion constructs of Tango1 in Drosophila cells (Fig. 5a). As shown in Fig. 5b, full-length Tango1-V5 in Drosophila S2Rϩ cells forms ring structures that are lined by COPII and appear to encircle the Golgi marker GM130. Previous studies on Drosophila Tango1 defined the cytoplasmic coiled-coil domain (CCD) as being required for localization of Tango1 to the ERES (14). However, studies on mammalian Tango1 identified a region of the cytoplasmic CCD as being responsible for Tango1-Tango1 self-association (35). Upon further analysis, we identified three separate CCDs within the cytoplasmic region of Drosophila Tango1 (CCD1, CCD2, and CCD3) and tested the effects of each on Tango1 self-association and the ability to form ring structures (Fig. 5a). As shown in Fig. 5 (c and d), full-length Tango1-FLAG and Tango1-V5 colocalize with one another in ring structures and are able to immunoprecipitate one another, indicating self-association. However, Tango1 lacking only the CCD1 domain (⌬CCD1) failed to form rings. All other deletion constructs were able to form rings (Fig. 5c). Additionally, ⌬CCD1 no longer colocalized with full-length Tango1 and was no longer able to immunoprecipitate full-length Tango1 (Fig. 5d).

Tango1 creates ER-Golgi docking sites
We then further tested whether each deletion construct still associated with COPII and GM130, as seen for the full-length Tango1-V5. As shown in Fig. 5 (e and f), all deletion constructs showed spatial proximity to COPII and GM130, including ⌬CCD1. However, regions of colocalization of COPII or Golgi with ⌬CCD1 were confined to small puncta scattered through- Figure 3. The cis-Golgi lies within Tango1-COPII rings. a, WT third instar larval SGs were stained with COPII (yellow), and the image in the box was magnified and rotated to show that COPII is present as a ring rather than a sphere or vesicle. Three-dimensional reconstructions below each image were generated using three-dimensional rendering software in Imaris. b, third instar larval SGs were stained for Tango1 (red), COPII (green), and the cis-Golgi marker GMAP (blue). Tango1 and COPII are closely associated (regions of colocalization appear yellow), whereas the cis-Golgi marker lies within the Tango1-COPII rings. Lower panels show magnified en face views (row 1) or side views (row 2) of the Tango1-COPII-cis-Golgi structure within the boxed regions from the top panels. c, SGs stained for Tango1 (red), the cis-Golgi marker GM130 (green), and the trans-Golgi marker Golgin-245 (blue). GM130 lies within the Tango1 rings (regions of colocalization appear yellow) and Golgin-245 lies in close proximity to GM130. Lower panels are magnified views of the boxed regions in c showing en face (row 1) or side views (row 2) of the Tango1-GM130 -Golgin-245 structures. Representative images from at least five biological replicates are shown. Scale bars, 1 m for all panels.

Tango1 creates ER-Golgi docking sites
out the cytoplasm rather than in ring structures, further highlighting the specific role of this region in Tango1 self-association. These results identify the CCD1 domain of Drosophila Tango1 as being responsible for Tango1 self-association to form ring structures. These data also support the existence of conserved functional regions within Tango1 across species (35).  To address the relationship between the Tango1-COPII-Golgi structure and the formation of secretory vesicles, we next imaged secretory vesicle formation in the presence of these three markers (Fig. 6). Interestingly, secretory vesicles (containing Sgs3-GFP; green) were seen forming directly at the Golgi side of these structures (Fig. 6, a and b). High-magnification views and reconstructions in three dimensions clearly show the Tango1 ring lined by COPII, with Golgi markers in the center of the ring and secretory vesicles emanating from the Golgi side (Fig. 6, a and b, and Fig. S4). This suggests that Tango1 and COPII form a docking site for the Golgi that allows passage of cargo directly from the ERES to the Golgi, with secretory granules then forming at the trans-Golgi face.
To further interrogate this model, we examined the effect of the loss of Tango1 on the Golgi-specific O-glycosylation that is normally found on the Sgs3 mucin (27). An antibody to Sgs3 was made, and specificity was verified (Fig. S5) to be able to follow the glycosylation of endogenous SG cargo. As shown in Fig. 6c, endogenous Sgs3 is highly O-glycosylated with the disaccharide Gal␤1,3GalNAc-S/T (as detected by the lectin peanut agglutinin (PNA)) and runs as a high-molecular-weight aggregate. However, in the absence of Tango1, Sgs3 is no longer heavily glycosylated, as evidenced by the dramatic reduction in PNA reactivity and significant decrease in molecular weight (Fig. 6c). These results further support a model where loss of Tango1 interferes with the movement of cargo proteins from the ER to Golgi, thus disrupting Golgi-specific protein modifications. Taken together, our study supports a model where Tango1 self-association to generate rings mediates the formation of COPII rings that form functional interaction/docking sites between the ER and Golgi, allowing the efficient transfer of cargo between these compartments (Fig. 7). This mechanism may be operative in cells required to synthesize and package large amounts of cargo over short periods of time. This model may also explain how the loss of Tango1 can result in the disruption of secretory vesicle formation across tissues that synthesize many types of diverse secretory cargo.

Discussion
Here, taking advantage of the high spatial resolution of secretory structures in the Drosophila larval SG, we demonstrate that Tango1 coordinates the formation of ER-Golgi interaction sites to mediate secretory vesicle formation. Our high resolution imaging of Tango1 relative to COPII, Golgi, and secretory cargo demonstrates that Tango1 self-associates to form rings that appear to orchestrate the formation of COPII rings. Moreover, cis-Golgi markers localize within these rings, forming a distinct Tango1-COPII-Golgi structure. In support of this Tango1-COPII-Golgi structure being a functional site of interaction between the ER and Golgi, we found secretory granules emanating from the trans-Golgi face of this structure. Our results support a model where Tango1 mediates a functional interaction point between the ERES and Golgi to allow efficient transfer of cargo and the subsequent formation of secretory granules.
This unique structure and the formation of secretory vesicles are dependent on Tango1. We found that loss of Tango1 resulted in the loss of the COPII rings, loss of the organized association of the cis-Golgi with COPII, and loss of secretory vesicles. Additionally, Tango1 overexpression in Drosophila cells was sufficient to drive COPII ring formation and Golgi association. Previous studies have demonstrated direct binding of Tango1 and COPII components (6, 10), which likely orchestrates the overlapping ring formation. Likewise, COPII components are known to interact with various cis-Golgi proteins, which likely drives their association in this structure. The formation of this entire structure depends on Tango1 self-association via the CCD1 domain in the cytoplasmic region. This is similar to the domain responsible for Tango1 selfassociation in mammals (35), suggesting conserved aspects of Tango1 action between these species.
Interestingly, the COPII structures present in this system exist as rings rather than spherical vesicular structures. Whether the COPII ring represents a structure unique to Drosophila or whether these structures might also be present in mammals awaits further investigation. However, evidence exists for diverse COPII structures in different systems, including tubules and protruding saccules from the ER membrane (36,37). The flexibility of the COPII coat may allow unique adaptations, depending on the biological context (38). Indeed, one recent study in mammalian cells suggests that procollagen transport occurs via a "short-loop pathway" from the ER to the Golgi in the absence of large COPII vesicular carriers (39). This study offers support for the possibility that similar COPII-dependent ER-Golgi interaction sites may exist in mammals.
Previous work in Drosophila also supports a model where Tango1 mediates a connection to the Golgi (14). In this study, the authors demonstrated that overexpression of the Tango1 cytoplasmic domain can increase the size and density of ERES and increase the number of Golgi units, strongly suggesting a role for Tango1 in organizing both ERES and the Golgi (14). Indeed, the authors propose that large COPII carriers may begin fusion with the Golgi before separating from the ERES. Our imaging clearly demonstrates that Tango1, COPII, and the Golgi lie in close proximity and that their spatial separation likely precludes the formation of a separate, large COPII vesic- with COPII (green) is shown. Note that regions of colocalization of ⌬CCD1 with GM130 or COPII are confined to puncta scattered throughout the cytoplasm as ⌬CCD1 is unable to form rings. Representative images from at least three independent experiments are shown.

Tango1 creates ER-Golgi docking sites
ular carrier. Moreover, our finding that mucin cargo emerges from the trans-Golgi face of this structure strongly supports a model where Tango1 serves to organize ordered ERES/Golgi interactions sites through which cargo passes. Our results and model would also explain previous Drosophila studies where loss of Tango1 affects Golgi structure as well as the secretion of diverse proteins. Tango1 was originally discovered in an RNAi screen in Drosophila cells for genes that affect both secretion

Tango1 creates ER-Golgi docking sites
and Golgi structure (Tango ϭ transport and Golgi organization) (8). Likewise, more recent studies have suggested that Tango1 plays a role in the interaction of the Golgi and ER that is independent of its role in trafficking bulky cargo proteins (16). Many of these studies also present evidence that loss of Tango1 affects constitutive secretion of all proteins, including small reporter proteins (8). If Tango1 functions to mediate docking sites between ERES and Golgi, then the loss of Tango1 would be expected to result in changes in Golgi structure. Indeed, we see evidence of Golgi structural changes when we deplete Tango1 from this system. Likewise, if the rate of constitutive secretion also benefits from these contact sites, one would expect the loss of Tango1 to affect this as well. Our results and model are therefore consistent with previous studies that suggest a role for Tango1 in Golgi structure and constitutive secretion.
Studies investigating the role of Tango1 in other systems have proposed diverse models for how Tango1 coordinates the secretion of specific cargo. In mammals, it is proposed that the SH3 region of Tango1 interacts with the HSP47 chaperone, which then binds collagen to mediate its entry into the nascent COPII vesicle. However, this model does not explain how diverse cargo and constitutive secretion can be affected by the loss of Tango1. Additionally, this model necessitates a second packaging event for bulky cargo on the trans-side of the Golgi that must take place. The data presented here suggest that tissues under a high secretory burden may use Tango1 to reduce the number of independent packaging steps required for bulky, highly glycosylated cargo (such as mucins) by forming direct connection points between the ER and Golgi. This may ensure the efficient production and packaging of large amounts of cargo into secretory vesicles over a short period of time.
The size of the secretory structures present in this genetically tractable system and its amenability to real-time imaging during the secretory process have led to key insights with regard to secretory granule biogenesis and secretion (19). Using this system, it was previously shown that clathrin and AP-1, which localize to the trans-Golgi network, are required for proper secretory granule formation (26). Subsequently, the activity of the phosphatidylinositol kinase PI4KII was shown to be essential for secretory granules to reach mature size, likely because of influences on homotypic fusion events (25). Our previous work has identified a role for O-glycosylation in secretory granule morphology during granule maturation (27). Real-time imaging has outlined the steps involved in secretory granule fusion with the apical plasma membrane and identified factors required for proper secretion of the mucinous contents (21, 22). Our current results are consistent with these prior studies and shed light on how cargo moves efficiently from the ER to the Golgi through a unique secretory structure whose organization depends on Tango1. This structure may explain how Tango1 has diverse effects on both regulated and constitutive secretion across many cell and tissue types. Moving forward, this tractable genetic imaging system will be amenable to identifying additional factors responsible for the highly organized and incredibly robust secretory program of the SG. Moreover, understanding the mechanisms by which biological systems maximize secretory capacity and efficiency may provide insights into novel strategies to restore defective secretion in disease states.

Fly stocks and genetics
All fly stocks and crosses were kept on MM media (corn meal, yeast, sucrose base) (KD Medical, Inc.) at 25°C. The c135-Gal4-driver line (Bloomington catalog no. 6978) was used to generate RNAi in the salivary gland and to express Tango1-GFP in the salivary gland. To visualize glue granules, the sgs3-GFP line (Bloomington catalog no. 5885) (19) was recombined with the c135-Gal4-driver line to generate sgs3-GFP, c135-Gal4 (22). Bloomington Drosophila Stock Center catalog no. 6978 (c135-Gal4, the proventriculus and salivary gland-Gal4 driver line), catalog no. 6983 (c729, the adult male accessory gland-Gal4 driver line), catalog no. 6314 (the adult female spermatheca-Gal4 driver line), catalog no. 5884 (sgs3-GFP line), catalog no. 62944 (sgs3 RNAi line), and catalog no. 5885 (sgs3-GFP line) (19) were used. Vienna Drosophila RNAi Center (VDRC) catalog no. 21594 (tango1 RNAi line) and catalog no. 60000 (w 1118 ) were used. For WT controls, w 1118 (VDRC catalog no. 60000) stock was crossed with the various Gal4 driver lines specific for the tissue of interest. For WT controls in the salivary glands, w 1118  This model suggests that Tango1 forms docking sites for COPII aggregation into rings, which then allows organized association between COPII and the Golgi, allowing efficient cargo movement from the ER to the Golgi; the cargo is then packaged into secretory granules on the trans-Golgi side. Implied in this model is the formation of a fusion pore within the Tango1-COPII-Golgi structure to allow cargo to move from the ER to the Golgi.

Transgenic fly lines
cDNA encoding Tango1 (Drosophila Genomics Resource Center) was cloned into the pUAST plasmid with GFP sequence conjugated at the 3Ј-terminal end. The plasmid pUAST-tango1-sGFP was used to create transgenic flies. Transformants were produced by Genetic Services Inc. (Cambridge, MA) using methodology based on the procedure described previously (40).

Staining of Drosophila tissues
The male accessory glands and female spermathecae were dissected from 3-5-day-old adult flies. The proventriculus and salivary glands were dissected from third instar larva. After dis-section, the tissues were incubated with 4% paraformaldehyde in PBS, then washed with 0.3% Triton X-100 in PBS, and incubated with 2% BSA, 0.3% TritonX-100 in PBS. For staining of proventriculus, the tissues were incubated with anti-Muc26B antibody (1:200) overnight at 4°C and then followed by incubating with anti-rabbit IgG antibody (dilution 1:200; Jackson ImmunoResearch Laboratories). For staining of male accessory glands and female spermathecae, tissues were incubated with TRITC-HPA (Thermo Fisher Scientific; 1:1000) and Alexa Fluor 647-phalloidin (Thermo Fisher Scientific; 1:500). After washing with 0.3% Triton X-100 in PBS, the samples were mounted in Vectashield mounting medium with 4Ј,6Ј-diamino-2-phenylindole (Vector Laboratories) on a slide with a spacer and imaged on Nikon A1Rϩ confocal microscope with a 60ϫ 1.4NA oil objective. Images were processed using ImageJ.
Salivary glands were dissected from early third instar larva and fixed with 4% paraformaldehyde in PBS for 20 min on ice. The glands were then washed (three times for 15 min) with 0.1% Triton X-100 in PBS and blocked in 2% BSA, 0.1% Triton X-100 in PBS for 1 h. The samples were incubated with primary antibodies overnight at 4°C, washed (three times for 15 min), and incubated with secondary antibodies for 2-3 h before a final wash (three times for 15 min) with 0.1% Triton X-100 in PBS. The samples were mounted onto a slide with ProLong glass antifade mountant (Thermo Fisher Scientific). Stained samples were imaged using Leica SP8 confocal using a 100ϫ 1.4NA oil objective and white laser excitation with 16-bit color depth and pixel resolution of 1024 ϫ 1024. Image acquisition was performed in sequential mode with the following excitation lines and emission windows: sequence 1: 488-nm excitation with 498 -540-nm emission and 647-nm excitation with 657-800-nm emission; sequence 2: 545-nm excitation with 555-590-nm emission; and sequence 3: 594-nm excitation with 604 -635-nm emission. Z-stacks were imaged in 100-nm steps over a volume of 1 m.

Live salivary gland sample preparation and imaging
Salivary glands from early third instar larvae were prepared for imaging as described by Tran et al. (22). Imaging of secretory granule homotypic fusion was performed on a Nikon A1Rϩ confocal microscope using a Plan Apochromat Lambda S 100ϫC 1.35NA Sil objective. Tango1 live imaging was performed on a Zeiss Elyra 7 with Lattice SIM using a C-Apochromat 40 ϫ 1.2NA W objective. Imaging volumes of 1.5-2 m were acquired in Apotome mode using optimal volume param-Tango1 creates ER-Golgi docking sites eters. SIM processing was performed in Zen Black using sharpness values between 3 and 5.

Image processing and analysis
Image deconvolution was performed using Huygens Professional with standard confocal presets. Thresholding and threedimensional surface reconstructions were performed using three-dimensional rendering software in Imaris 9.3.0, with perspective set at 45°. Videos and image rotations were generated using Imaris key frame animation. Two-dimensional line scans were generated with the ImageJ plot profile tool, and the graphs were generated in Microsoft Excel. The figures were constructed using Adobe Illustrator.

Western blotting
S2Rϩ cells were transfected with plasmids using Effectene transfection reagent (Qiagen) according to manufacturer's instructions. After 3-4 days, the cells were collected and lysed with radioimmune precipitation assay buffer (Sigma). The lysates were incubated with anti-V5-agarose (Bethyl) to purify the proteins with V5 tag. The cell lysates or the purified proteins were electrophoresed in the 4 -12% SDS-PAGE gel, and proteins were transferred to nitrocellulose membranes. After blocking, the membranes were incubated with anti-V5 antibody (dilution 1:1000, Invitrogen) or anti-FLAG antibody (dilution 1:1000, mouse antibody from Sigma, rabbit antibody from Invitrogen) to detect interacting proteins in co-immunoprecipitation experiments. After washing, the membranes were incubated with IRDye 680LT or 800CW-conjugated goat antimouse antibody (dilution 1:5000, LI-COR) or goat anti-rabbit antibody (dilution 1:5000, LI-COR). The membranes were washed with PBST (0.1% Tween 20), rinsed in PBS, and scanned using a Li-COR Odyssey IR imaging system.
Protein extracts were prepared from wandering third instar larvae salivary gland lysates from WT and tango1 RNAi crosses. Tissues were dissected in PBS and transferred to radioimmune precipitation assay buffer containing proteinase inhibitors (Thermo Fisher) and then homogenized on ice. Samples were separated on 4 -12% SDS-PAGE gradient gel (Invitrogen) under reducing conditions and then transferred onto a 0.45-m pore size nitrocellulose membrane (Invitrogen). Membranes were blocked with Odyssey blocking buffer (LI-COR) and then incubated with rabbit anti-Sgs3 primary antibody (1:2000) overnight. The membrane was then washed with PBS ϩ 0.1% Tween three times and then probed with IRDye 680CW-conjugated goat anti-rabbit (1:5000) and with IRDye 800LT-conjugated PNA (1:5000). The membranes were analyzed and processed using LI-COR Odyssey CLx and Image Studio software.

Antibody preparation
Polyclonal antibodies to Sgs3 were raised in rabbits using the peptide CQDLNGVLRNLERKIRQ (GenScript) and were affinity-purified. Antibody specificity was tested on Western blots of SG extracts from WT and animals expressing RNAi directed against sgs3 (using the c135-Gal4 driver along with the sgs3 RNAi line).

Statistical analyses
Number of replicates used for each analysis is specified in the figure legends. The p values were calculated using the twotailed Student's t test. No statistical method was used to predetermine sample size, no randomization methods were used for these studies, and no blinding studies were performed. No data sets were generated or analyzed during the current study.