Here, we describe that depletion of the Drosophila homologue of p115 (dp115) by RNA interference in Drosophila S2 cells led to important morphological changes in the Golgi stack morphology and the transitional ER (tER) organization. Using conventional and immunoelectron microscopy and confocal immunofluorescence microscopy, we show that Golgi stacks were converted into clusters of vesicles and tubules, and that the tERs (marked by Sec23p) lost their focused organization and were now dispersed throughout the cytoplasm. However, we found that this morphologically altered exocytic pathway was nevertheless largely competent in anterograde protein transport using two different assays. The effects were specific for dp115. Depletion of the Drosophila homologues of GM130 and syntaxin 5 (dSed5p) did not lead to an effect on the tER organization, though the Golgi stacks were greatly vesiculated in the cells depleted of dSed5p. Taken together, these studies suggest that dp115 could be implicated in the architecture of both the Golgi stacks and the tER sites.
The Golgi apparatus exhibits a unique membrane architecture comprising Golgi stacked cisternae. There is accumulating evidence, particularly from yeast studies, suggesting a relationship between the presence of the Golgi stacks and a focused organization of the transitional ER (tER)* sites (Glick, 2002). tER sites are defined as specialized ER subdomains at which proteins destined for the Golgi apparatus are packaged into transport vesicles. They are defined by the presence of COPII vesicles, which carry the secretory cargo out of the ER. Cells contain many tER sites, where the COPII components Sec23p, Sec13p, and Sar1p have been localized (Orci et al., 1991; Barlowe et al., 1994). tER sites often display an elaborate architecture of clustered pleiomorphic membranes comprising one cup-shaped ER cisternae where budding profiles can be observed, small 50–70-nm vesicles (sometimes coated), and short tubules (Bannykh et al., 1996). However, the molecular mechanism that generates them is still mysterious (Rossanese et al., 1999).
In yeast Pichia pastoris, the Golgi apparatus exhibits stacked cisternae with established polarity, and the pattern of Sec13–GFP representing the tERs comprised two to six spots surrounding the nucleus. In contrast, Saccharomyces cerevisiae exhibits a Golgi apparatus in a form of multiple single cisternal elements surrounded by vesicles and tubules, which represent either the cis or the trans side of the Golgi (Preuss et al., 1992). Sec13–GFP appears in multiple fluorescent dots, indicating that the tER sites are dispersed throughout the ER. The reasoning sustaining the relationship between the tER organization and the presence of polarized Golgi stacks is that the concentration of vesicles budding from focused tER sites would remain high enough, maybe due to the presence of a tER matrix, to allow cisternal formation and stacking (Rossanese et al., 1999; Bevis et al., 2002), whereas the membrane derived from dispersed tERs is too low in concentration for proper cisternal formation and further stacking. However, the nature of the tER matrix is far from being clear.
On the other hand, studies performed in mammalian cells have established a number of so-called “Golgi matrix proteins,” most of them Golgins such as p115, GM130, and giantin, that are involved in the building and maintenance of the Golgi stacks (Shorter and Warren, 2002). These proteins were characterized by the fact that they are not relocated to the ER upon brefeldin A treatment (Seemann et al., 2000a). These proteins have been shown and/or postulated to help the Golgi stacks adopt this typical architecture of stacked cisternae (see Discussion).
Golgi matrix proteins have been localized in mammalian cells to the Golgi area but also in earlier secretory compartments, rather away from the Golgi complex. p115, for instance, has been localized in the intermediate compartment and has been involved in ER to Golgi transport (Nelson et al., 1998; Alvarez et al., 1999, 2001). p115 has also been shown to be recruited on COPII vesicles by Rab1 and prime them for fusion with the Golgi membranes by recruiting a SNARE protein complex (Allan et al., 2000). Finally, Uso1p, the yeast functional orthologue of p115, has been implicated in early transport events such as ER-derived vesicle tethering and proper protein sorting (Sapperstein et al., 1995; Morsomme and Riezman, 2002), indicating that it could be localized in this part of the exocytic pathway. This opens the possibility that, in addition to its role in the different transport steps, p115 could also have a role in the organization of the tER sites.
In fly, the organization of the exocytic pathway is slightly different from the mammalian one. Golgi stacks are not interconnected to form a single copy organelle capping the nucleus. Instead, they remain dispersed in the cytoplasm (Rabouille et al., 1999), as is the case in plants and Pichia. The simplified organization observed in Drosophila provides us with the possibility to examine the molecular mechanisms underlying both structures in relation to each other. This could allow us to investigate whether the architecture of the Golgi apparatus in Drosophila S2 cells is a direct consequence of the tER organization or whether their structural organization is regulated independently of each other.
In the present study, we show that depletion of the Drosophila homologue of p115, dp115, which is localized both in the Golgi stacks and in dSec23p-positive tERs, led to a quantitative breakdown of Golgi stacks that are converted into Golgi clusters of vesicles and tubules and strongly affected the general organization of the tER sites. In addition, we show that despite the presence of a morphologically modified exocytic pathway, the intracellular transport was largely unaffected, suggesting that the disorganized tER and the Golgi clusters still form a functional exocytic pathway.
The localization of dp115 in Drosophila S2 cells
Drosophila p115 (dp115) exhibits 60% similarity to its rat counterpart (Fig. 1 A). dp115 does not possess the acidic stretch of the 50 COOH-terminal amino acids that has been shown in rat p115 to be involved in binding GM130 and giantin (Dirac-Svejstrup et al., 2000). In dp115, there are only a few acidic amino acids scattered along the last 100 COOH-terminal portion of the protein. Besides Neurospora, the tail of Caenorhabditis elegans and Arabidopsis p115 homologues is not acidic either, except for the last 5 and 15 amino acids, respectively (our search). However, the domain in GM130 that binds p115 is conserved in dGM130 (Kondylis et al., 2001; Fig. 1 A), suggesting that both Drosophila proteins could bind each other (albeit possibly through a different domain in dp115).
Despite the high level of homology between mammalian and Drosophila p115, we were interested to verify the localization of the Drosophila homologue. dp115 has been shown to localize to the Golgi apparatus and Golgi clusters in Drosophila imaginal discs by an antibody raised against rat p115 that cross reacted with dp115 (Kondylis et al., 2001). We raised an antibody against a dp115-specific peptide (dp115/584) that recognizes three bands by Western blotting (Fig. 1 B), one at 40 kD (resulting from the secondary antibody, because it is present in the absence of primary antibody; unpublished data), one at 65 kD, and one at 85 kD (Fig. 1 B, lane 1), which has the predicted molecular mass for dp115. We fractionated the S2 cell extract into cytosolic and membrane fractions. After blotting, the 85-kD band seemed only associated with the membrane fraction (lane 3), although its absence from the cytoplasmic fraction (lane 2) might be due to insufficient loading. In larval extract, this band was the major one, and the 65-kD band was much weaker (lane 5). The 65-kD band was considered nonspecific because Western blotting of embryo extracts did not show any bands at 85-kD but revealed a 65-kD band with the same intensity as the larval extract (lane 4). Furthermore, the intensity of this band did not vary according to the amount of protein loaded (unpublished data), and it was present in other Western blots using different primary antibodies.
When used in immunoelectron microscopy (IEM), the dp115/584 antibody gave a specific signal (Fig. 1, C–F). In S2 cells, 30 ± 5% of the gold particles were on the cytoplasm. The Golgi area was labeled by 25 ± 8% of the membrane-associated gold particles. The ER cisternae were decorated by 21 ± 6%. Pleiomorphic membranes (Fig. 1, C and D, asterisk) comprising tubules and vesicles (50–70 nm in diameter, but also larger vessels) close to the Golgi stacks or in their neighborhood contained 54 ± 7% of the membrane-associated labeling. A double labeling with an antibody recognizing the Drosophila homologue of Sec23p (dSec23p; Fig. 1, E and F) suggested that these pleiomorphic membranes represented tERs (see Fig. 5). The linear density of gold particles over the membrane of these three compartments was estimated to be 0.57 gold/μm on the Golgi membrane, 0.37 on tER membrane, and 0.15 on ER cisternae. A similar pattern of distribution was observed when salivary glands of third instar larvae were labeled with the same antibody (Fig. 1 D).
dGM130 has been described in Kondylis et al. (2001) and localizes to the Golgi apparatus in Drosophila imaginal discs. When MLO7 (a polyclonal antibody directed against the first 73 amino acids of GM130; a gift from Martin Lowe, University of Manchester, Manchester, UK) was used in S2 cells for indirect immunofluorescence, it labeled dGM130 in a Golgi-specific manner (unpublished data).
Effect of depleting dp115 and dGM130 on Golgi stack morphology
We depleted S2 cells of dGM130 using a dsRNA corresponding to the second exon of dGM130 (ds dGM130). When samples were analyzed by Western blotting, two bands were detected, one very faint, which was neglected (Fig. 2, top, lane C), and a strong one. The strong band migrated at the predicted position for a protein of the mass of dGM130 (arrow) and was reduced below detectable level after 48 h (Fig. 2, top) and remained so up to 120 h of incubation. This depletion, however, did not lead to any effect on Golgi stack morphology, as assessed by EM (Fig. 3 B).
The percentage of cell sections exhibiting at least one Golgi stack per cell profile was scored. 100% of mock-treated (incubated without dsRNA) and mock-depleted cells (incubated with ds EGFP) exhibited at least one Golgi stack per cell profile (Fig. 3 A). This percentage decreased slightly to 85% for incubations up to 96 and 120 h. On average, these Golgi stacks had a cross-sectional diameter of 0.368 ± 0.047 μm and 3.7 ± 0.8 cisternae per stack.
In cells depleted of dGM130, the percentage of cells exhibiting Golgi stacks (Fig. 4 A), the number of cisternae per stack, and the mean diameter of the stacks were comparable to the figures obtained for mock-treated cells (diameter of 0.347 ± 0.078 μm and 3.5 ± 1.6 cisternae per stack).
We depleted S2 cells of dp115. dsRNA corresponding to the NH2-terminal portion of dp115 (ds dp115) was added to S2 cells for up to 120 h. We first assessed the depletion of dp115 mRNA by RT-PCR using primers corresponding to the 5′ end of the dp115 gene (Fig. 2 B, RT PCR). After 48 h, dp115 mRNA could not be detected and remained so up to 120 h. Western blotting with the dp115/584 antibody showed that the 85-kD band disappeared after 96 h of incubation with ds dp115 (Fig. 2 B, WB).
Cells depleted of dp115 up to 72 h did not show significant changes in their Golgi stack morphology when compared with controls (Fig. 3 C). 76.3 ± 4.0% exhibited at least one Golgi stack per cell section (Fig. 4 A). However, this percentage decreased sharply in the cells incubated for 96 h. In only 18 ± 6.3% of the cell sections, one or more Golgi stacks were visible (Fig. 4 A). These stacks were also smaller than in mock-depleted cells with a mean diameter of 0.268 ± 0.050 μm and 3.2 ± 0.4 cisternae/stack. The remaining 82% of the cell sections exhibited a Golgi area under the form of clusters of vesicles and tubules (Fig. 3 D). The percentage of membrane in cisternae per Golgi area decreased by 12% at 72 h, and by 87% at 96 h, when compared with control (Fig. 4 B). This latter decrease was paralleled by a decrease in stacking (Fig. 4 B) and was mirrored by an increase of 32% in vesicular profiles and an increase of 50% of small tubules.
The surface density of the Golgi area (SDgo) and total cisternae (SDcis) (stacked and single) was estimated in mock- and dp115-depleted cells. A reduction of 21% of SDgo was observed after dp115 depletion together with a reduction of 88% of the SDcis (Fig. 4 C), suggesting that cisternal profiles were lost or not merely diluted, for instance, by incoming vesicles that would accumulate around them, unable to fuse due to lack of dp115.
Effect of dp115 depletion on the tER organization
Given the impact of dp115 depletion on the structure of the Golgi stacks, the proposed relationship between the tER organization and the presence of Golgi stacks, and the presence of dp115 at the tERs, we were prompted to look at the organization of the tER sites in cells depleted of dp115.
For this purpose, two antibodies were used. First, a polyclonal antibody raised against a rat Sec23p peptide (conserved in Drosophila) was used for IEM on S2 cell cryosections. This antibody decorated pleiomorphic membrane populated with 50–70-nm vesicular profiles (some coated; Fig. 5, D and E), ER coated buds, as well as larger vesicular and tubular membrane structures, often contained in the concavity of an ER cisterna (Fig. 5, A and C–E). These structures are reminiscent of those described by Bannykh et al. (1996), referred to as vesicular tubular clusters (VTCs). Since then, the VTC terminology has been used to describe other structures in the intermediate compartment. To avoid a possible misunderstanding, we refer to our membrane structures as tER sites on the basis of their dSec23p labeling and their colocalization with a transport cargo, such as the plasma membrane protein Delta (Fig. 5 F; see below). The gold labeling density on the tER sites was six to seven times higher than in the surrounded cytoplasm. Furthermore, 10 ± 3% of the gold particles decorated the ER cisternae.
Second, a monoclonal antibody recognizing a 120-kD Drosophila Golgi membrane antigen (d120kd; Stanley et al., 1997) was tested by IEM. The gold labeling was specific to the Golgi apparatus. 55% labeled the Golgi stacks, and 35% labeled vesicles and tubules abutting the Golgi stack (Fig. 5, B and C), a result in agreement with immunofluorescence data (Stanley et al., 1997; Munro and Freeman, 2000).
In immunofluorescence, these two antibodies gave similar patterns. The dSec23p pattern corresponded to 20 ± 8 large fluorescent objects dispersed in the cytoplasm (Fig. 6 A). The d120kd pattern corresponded to 18 ± 7 similar large fluorescent objects (Fig. 6 B). These two patterns overlapped partially (Fig. 6 C). By IEM, the region of overlap corresponded to the interface between the Golgi stack and the tER where the two antigens are in close proximity (Fig. 5 C).
In cells depleted of dp115 for 96 h, however, both patterns were differentially affected. In 90% of the cells, the immunofluorescence pattern of dSec23p appeared as numerous scattered small dots all over the cytoplasm (84.5 ± 20/cell; Fig. 6 D). The size of these fluorescent objects was approximately three to four times smaller when compared with mock-depleted cells. The d120kd pattern of fluorescence also changed after dp115 depletion (Fig. 6 E). The number of fluorescent objects corresponding to this antigen was slightly higher than in control cells (26 ± 7 vs. 18 ± 6 in control cells), but their size was smaller, and the intensity of fluorescence reduced, as if this antigen was dispersed. d120kd and dSec23p immunofluorescence patterns did not overlap as much as in mock-treated cells. Numerous dSec23p dots seemed to be free of d120kd, and ∼50% of the d120kd-positive dots were also observed free of dSec23p (Fig. 6 F). Similar changes in patterns were also observed when the cells were fixed with stronger fixatives and for a longer period of time, suggesting that the observed effects were not due to mild fixation (see Fig. S1). Furthermore, dSec23p and d120kd were not degraded upon dp115 depletion (Fig. 2 B), suggesting that this was not the cause of the change in patterns.
In dp115-depleted cells observed by IEM, the dSec23p-positive small and scattered dots observed in immunofluorescence represented pleiomorphic membranes containing vesicles and tubules, reminiscent of those observed in control cells, but with a smaller size (compare Fig. 5 A with Fig. 7, A and B; see Fig. S2, A and B). They appeared more dispersed throughout the cytoplasm than in control cells and sometimes exhibited a reduced labeling density (Fig. 7, B and E). The number of these dSec23p-positive sites per cell section was 6.7 ± 2.3, and only 20% of them were positive for d120kd. This is to be compared with 2.7 ± 1.5 dSec23p-positive sites per section of mock-depleted cells, 90% of them also positive for d120kd (Fig. 5 C), suggesting that small tER sites have been generated that lack the spatial relationship with Golgi areas.
The d120kd-positive dots that were observed in immunofluorescence represented clusters of vesicles and tubules labeled by 73 ± 3.6% of the gold particles associated with d120kd (Fig. 7, C–F). In these clusters, we could observe profiles reminiscent of short cisternal remnants (Fig. 7, D–F, arrows), indicating that these clusters could derive from Golgi stack breakdown. d120kd also labeled the ER cisternae (27 ± 3.6%; Fig. 7 E, arrowhead), suggesting perhaps a retrograde movement of Golgi membrane to the ER.
Pleiomorphic membranes of vesicles and tubules were also labeled by both dSec23p and d120kd, showing that some tERs have remained in close proximity to the Golgi membrane, as they were in nondepleted cells. In some cases, these two markers did not label the pleiomorphic membrane homogeneously and seemed to retain their original polarity (Fig. 7, D–F).
Overall, this experiment shows that the depletion of dp115 leads to the disorganization of both the tER sites and the Golgi stacks, suggesting that dp115 could be involved in the organization of both. Alternatively, the disorganization of the tERs could lead to the destabilization on the Golgi stacks. dp115 could thus have only one role in tER organization. To test this hypothesis, we assessed the immunofluorescence pattern of dSec23p and d120kd in dp115-depleted S2 cells between 72 and 96 h after ds dp115 addition (see Fig. S3). After a 72-h incubation with ds dp115, both dSec23p and d120kd patterns were almost indistinguishable from mock-treated cells (∼20 large fluorescent objects partially overlapping, which we will refer to as “control patterns”). However, after 84, 88, and 92 h incubation, a mixture of patterns was observed. The number of cells exhibiting control patterns decreased gradually over time to reach 12.6% after 96 h incubation. Conversely, the cells exhibiting a pattern where both dSec23p and d120kd were affected increased gradually from 17.6 to 83.8%. The percentage of cells where only the dSec23p pattern was affected was roughly the same (2.3%) as that of cells where only the d120kd pattern was affected (0.9%). This result suggests that depletion of dp115 leads to a concomitant loss of Golgi stacks and tER organization, indicating that dp115 could have a distinct role in the organization of both structures.
These effects were specific for dp115 depletion. dGM130 depletion did not have any effect either on the structure of the Golgi complex (Fig. 3 B) or on tER organization (Fig. 8 E). On the other hand, depletion of dSed5p, the Drosophila homologue of mammalian syntaxin 5 (Banfield et al., 1994), had a very strong effect on the Golgi stack morphology (complete and quantitative vesiculation; Fig. 8, A and B), but the organization of the tERs was kept intact (Fig. 8 D) when compared with mock-depleted cells (Fig. 8 C). The fragmentation of the Golgi stacks therefore did not cause the redistribution of the tER sites.
Effect of the depletion of dp115 and dGM130 on anterograde protein transport
Depletion of dp115 led to quantitative breakdown of the Golgi stacks and dispersal of the tER sites. That prompted us to test whether this morphologically modified exocytic pathway was still functional.
We used an S2 cell line stably transfected with a construct expressing the full-length plasma membrane protein Delta (Klueg et al., 1998), referred to as Delta S2 cells. Delta is a glycosylated Drosophila ligand of Notch (Panin et al., 2002) that uses the exocytic pathway for its plasma membrane deposition (Fig. 5 F; see Fig. S4). The expression of Delta in this cell line is under the control of a metallothionein promoter, and addition of CuSO4 drives its expression. The morphological effects of the various depletions were strictly similar in Delta S2 cells and in wild-type S2 cells, and all our stereological analyses have been obtained with both cell lines.
We verified that the induction and transport of Delta for as long as 2.5 h did not induce the rebuilding of Golgi stacks. In cells depleted of dp115 for 96 h, the percentage of cells with at least one Golgi stack was the same before and after Delta induction (18.0 ± 6.3 vs. 18.9 ± 1.2%, respectively).
We first induced Delta expression for 1 h followed by a 90-min chase as a way of measuring steady-state transport. In mock-treated and mock-depleted cells, a large majority of the staining was found at the plasma membrane (Fig. 9 A). Only 3.3 ± 2.5% of the cells exhibited exclusively intracellular (Fig. 9 B), with no plasma membrane, labeling.
The intensity of surface labeling at the plasma membrane varied considerably (Fig. 9 A). The estimation of the percentage of cells exhibiting the different labeling intensities allowed the calculation of the total intensity of the mock-treated cells (Fig. 9 D; Table I). This was set as 100% and was 90% inhibited by brefeldin A (Fig. 9, B and C; Table I), a known inhibitor of transport in the early exocytic pathway. Furthermore, an 89% transport inhibition was also obtained when cells were depleted of dSed5p, a protein largely documented for its role in intracellular transport (Fig. 9 C; Table I), suggesting that our transport assay was sufficiently sensitive to monitor an inhibition in transport and that Delta was a suitable marker for anterograde transport.
Depletion of dGM130 had only a negligible effect on the steady-state transport of Delta (Fig. 9 C; Table I). Surprisingly, depletion of dp115 reduced steady-state transport to a very small extent (∼12%; Fig. 9 C; Table I).
The initial rate of transport in cells depleted for dp115, measured by inducing Delta expression for 25 min (minimal induction period to detect expression of the protein by Western blotting; unpublished data) and chasing for 45–90 min, was found to be reduced by 30.5 ± 3% when compared with mock-depleted cells (Fig. 9 D). The distribution of plasma membrane intensity was essentially the same in mock- and dp115-depleted cells where the exocytic pathway was largely modified (Fig. 9, compare E with F). This suggests that the large modifications to the morphology of the Golgi stacks and the tER organization caused by dp115 depletion did not make the cells incompetent for transport.
Last, using a second S2 cell line where the expression of a secreted marker could be induced and monitored, we showed that the depletion of dp115 and dGM130 did not affect the secretion of this marker (unpublished data). This result strengthens the notion that the disorganization of the exocytic pathway does not lead to a significant decrease in transport efficiency.
We have used Drosophila S2 cells to perform depletion of dp115 and dGM130 by RNA interference (RNAi). Despite small differences with mammalian cells, Drosophila cells are a novel, but adequate, biological system to investigate the issues related to membrane traffic and organelle structure.
dp115 was localized on dSec23p-positive pleiomorphic membrane structures representing tERs, on the Golgi area, and on the ER cisternae. This distribution is slightly different from that in mammalian cells, where p115 has been localized in the cis Golgi, the Golgi stack, and the intermediate compartment (Nelson et al., 1998; Alvarez et al., 1999, 2001). Its localization on the tER sites was inferred only by the role it plays in priming COPII vesicles for fusion (Allan et al., 2000). This difference could be explained by the fact that the region between the ER and the Golgi stack comprised more layers in mammalian cells, including one positive for ERGIC53 that does not overlap with Sec13p (Hammond and Glick, 2000). In contrast, in S2 cells, the tER sites were found abutting the Golgi stack itself, as if the intermediate compartment was missing (a situation that seems similar to yeast where sec12/sec13 labeling seems to overlap with the Golgi apparatus; Rossanese et al., 1999; Bevis et al., 2002) or compressed into a single layer.
Golgi stacks are breaking down upon dp115 depletion
Depletion of dp115 in S2 cells led to the fragmentation of their Golgi stacks into clusters of vesicles and tubules. These clusters were labeled by a fraction of d120kd that, in mock-treated cells, preferentially marked the Golgi stacks. Some of them also seemed to contain remnants of Golgi cisternae. Furthermore, the Golgi stacks, in the small percentage of cells where they could still be observed, were 30% shorter than control. This suggests that upon dp115 depletion, Golgi stacks are fragmented into clusters retaining Golgi identity and function (see below). Similar results have been reported in mammalian cells that exhibited a fragmented Golgi ribbon upon depletion of p115 by antibody injection (Alvarez et al., 1999; Puthenveedu and Linstedt, 2001).
What could be the mechanism? Intracellular transport could be inhibited in dp115-depleted cells, and transport intermediates could accumulate and create the membrane clusters. In mammalian cells, p115 has been shown to have a role in intracellular transport by at least three mechanisms. First, it forms a complex with GM130 on the Golgi cisternae, and giantin on the COPI vesicles, that helps the vesicles to be in close proximity to their target membrane, enhancing docking efficiency and fusion (Nakamura et al., 1997; Sönnichsen et al., 1998). Second, p115 is recruited by activated rab1 onto the COPII budding vesicles, priming them for fusion (Allan et al., 2000). Third, p115 has recently been shown to catalytically promote the syntaxin5/Gos28 SNAREpin formation, suggesting a direct role in vesicle docking and fusion (Shorter et al., 2002). Given the significant homology between dp115 and its mammalian counterpart, dp115 could have an equivalent role in intracellular transport. However, we found that anterograde intracellular transport was largely unaffected (see below), suggesting that as a sole mechanism, the reduction of docking/fusion of ER-derived transport vesicles is unlikely to be sufficient to explain the observed Golgi breakdown.
Mammalian p115 has also been involved in the building and maintenance of the Golgi stacks, independent of its role in transport (Puthenveedu and Linstedt, 2001; Shorter and Warren, 2002). First, in vitro NSF-mediated rebuilding of Golgi cisternae was shown to require p115 (Rabouille et al., 1995) (though that could due to its role in SNARE pairing). p115 was also involved in the in vitro stacking of the p97-mediated Golgi cisternae (Shorter and Warren, 1999). Last, it has been identified as a binding partner of GM130 in the GM130/GRASP65 complex (Barr et al., 1998). This tripartite complex is proposed to be involved in the stacking of Golgi cisternae (Barr et al., 1997; Linstedt, 1999).
dp115 could be involved in a similar complex. Our unpublished results show that dp115 and dGM130 interact genetically and could therefore form a complex, perhaps with dGRASP, the single homologue of GRASP65 and 55. If the stacking mechanism were impaired by the removal of dp115, this would affect the morphology of the Golgi area. However, we did not observe single cisternae, suggesting that dp115 might have an additional role in the maintenance of the cisternae themselves, maybe in preventing their fission into smaller fragments. Furthermore, we have shown that depletion of dGM130 does not affect Golgi stack architecture, in agreement with Puthenveedu and Linstedt (2001) and Vasile et al. (2003). That dGM130, and its mammalian counterpart, has a direct role in the structure of the Golgi stack remains to be shown.
Depletion of dp115 affects tER organization
The alteration in the tER organization that we demonstrated in S2 cells depleted of dp115 could contribute to the breakdown of the Golgi stack architecture (see Introduction; Rossanese et al., 1999; Ward et al., 2001). Instead of the focused organization observed in control cells, the tERs were scattered into a multitude of smaller sites in dp115-depleted S2 cells. The organization of the tER/Golgi stacks in our mock-treated cells was reminiscent of that in Pichia pastoris (see Fig. 6 B in Rossanese et al., 1999), whereas in dp115-depleted S2 cells, it resembles S. cerevisiae (see Fig. 5 in Rossanese et al., 1999). Given our results, a differential localization of Uso1p or a binding partner between Pichia and S. cerevisiae could perhaps explain the difference in the organization of their exocytic pathways.
dp115 could therefore only have a role in tER organization. If that were the case, the dispersion of the tER sites would precede the change in the Golgi stack. Despite intensive searching for such profiles, only a very small percentage of cells fit these criteria, and to the same extent as cells in which the Golgi stack organization was lost while still exhibiting a normal tER pattern. Of course, the effect of the tER disorganization on the Golgi stack morphology could be very rapid. Only video microscopy could lead to a definitive answer on that issue. Nevertheless, this result suggests that dp115 is involved in Golgi stack morphology independently of its role in tER organization.
We propose that in Drosophila, dp115 has a role in both tER and Golgi stack organization. dp115 could be involved in at least two different complexes, one in the tERs (with unknown partners; Nelson et al., 1998) and one in the Golgi stacks (with perhaps dGM130). These two complexes could either form part of a common matrix underlying both the tERs and the Golgi stacks, as is the case in Pichia (Mogelsvang et al., 2003). The fact that in cells depleted of dSed5p, the Golgi stacks are completely vesiculated while the tER organization is intact leads us to propose that there are two independent matrices that have dp115 as a common component. Furthermore, it has been proposed that tERs and Golgi stacks would operate as a positive feedback loop on each other's organization by exchanging membrane and molecules (Hammond and Glick, 2000). dp115 could be an important part of this exchange program.
The dispersion of the tERs could also contribute to Golgi stack breakdown, by allowing the dispersion of budded vesicles that would not participate in the homeostasis/building of Golgi stacked cisternae. Whether Golgi stacks can form in cells comprising dispersed tER sites remains to be shown.
Intracellular transport is only marginally reduced in cells depleted of dp115
The anterograde protein transport in cells depleted of dp115, in which both the Golgi stacks and the tER sites are disorganized, was only marginally affected, shown by two independent assays. Supporting this result, cell proliferation was not affected in dp115 depletion (unpublished data), suggesting that, as for the protein markers used in our experiments, endogenous proteins were likely to be transported as efficiently in depleted cells.
This result is different from those recently published where the injection of a p115 antibody in mammalian cells disrupts the Golgi ribbon and strongly inhibits the ER to Golgi transport (Alvarez et al., 1999, 2001). Furthermore, the injection of a GM130 cDNA encoding a form of GM130 lacking its p115 binding domain also inhibited anterograde transport by 65% (Seemann et al., 2000b). The same holds true for the depletion of dGM130 that did not lead to an inhibition of intracellular transport in our assay, contrary to mammalian cell studies (Alvarez et al., 2001). These differences could be explained by the methodology used for the depletion or perhaps suggest that dp115 and dGM130, unlike their mammalian counterparts, do not have a role in intracellular transport.
However, we find it quite surprising. A strong structural similarity between dp115 and mammalian p115 exists. The SNARE motif identified in mammalian SNARES (Weimbs et al., 1997) is also present in dp115 (amino acids 664–729; unpublished data), suggesting that like its mammalian counterpart, it could also catalyze SNAREpin assembly between dSed5p and dGos28, which also contains a SNARE motif (Weimbs et al., 1997; unpublished data). Given that the role of p115 is catalytic at least in vitro, a trace amount of dp115 could serve its function in transport. Although we show that the dp115 mRNA and protein were successfully depleted, we think that trace amounts of protein remain, a limitation of RNAi techniques. Therefore, we instead conclude that the role of dp115 in contributing to the Golgi stack structure and tER organization can be separated from its role in intracellular transport.
We show here, as is the case in yeast, that Drosophila intracellular transport can be sustained by more than one type of exocytic pathway, including dispersed tERs and fragmented Golgi stacks. Clearly, smaller and multiple tER sites and Golgi clusters of vesicles and tubules can efficiently perform this transport function. This could explain the existence of developmental stages in Drosophila in which cells do not exhibit Golgi stacks but exhibit clusters of vesicles and tubules instead, as in the case of the third instar larval imaginal discs (Kondylis et al., 2001). Perhaps this flexibility is not offered to mammalian cells, with their unique organization of the Golgi complex involving the formation of the single copy Golgi ribbon.
Thus, what is the role of Golgi stacks if efficient anterograde transport can take place in their absence? Golgi stacks (but not Golgi clusters) could have a role in retrograde transport, in the proper and complete maturation of protein-born O- and N-linked oligosaccharide moieties and the addition of sorting signals. This is for further investigation.
Materials And Methods
dp115 cDNA was cloned by screening an embryonic cDNA library with a DNA probe made using the EST LD41079. The dp115 cDNA (1P2C1) was used to PCR a 682-bp fragment with flanking T7 RNA polymerase binding sites using the 5′ primer TAATACGACTCACTATAGGGAGA(T7)-acccagaatagac and the 3′ primer T7-tcaaaaaggcggtca. The 774-bp T7-dGM130 template was obtained by PCR from the clone p5.6KK (provided by Terry Orr-Weaver, Whitehead Institute, Cambridge, MA) containing the MeiS332 gene (Kerrebrock et al., 1995) and the coding sequence of dGM130 using the 5′ primer T7-cgccagcaacaacaa and the 3′ primer T7-tgctccttgtcctgcgtt. The 635-bp T7/T3-EGFP template (gift from Thomas Vaccari, European Molecular Biology Laboratory [EMBL], Heidelberg, Germany) was obtained using the 5′ primer T7-taaacggccacaagttcag and the 3′ primer AATTAACCCTCACTAAAGGGAGA(T3)-gtgatcgcgcttctcgttg. Finally, the 733-bp T7/T3-dsed5 template (gift from Thomas Vaccari) was obtained using the 5′ primer T7-gcttatttgatgacagac and the 3′ primer T3-aatatgagaacgccgaag.
The various T7 and T3 templates were purified from the agarose gel using the GFX PCR DNA and gel band purification kit (Amersham Biosciences). The purified products were used as templates for dsRNA synthesis using the MEGASCRIPT T7 and T3 transcription kit (Ambion) according to manufacturer's standard protocol.
Cell cultures and dsRNA interference
S2 cells were grown in Schneider's insect medium supplemented with 10% heat-inactivated and insect-tested FBS at 27°C in a humidified atmosphere. Delta-WTNdeMYC S2 cells (Delta S2 cells) were a gift from Kris Klueg (Indiana University, Bloomington, IN). Delta S2 is a stable cell line expressing the full-length Delta protein at the plasma membrane upon induction with 1 mM CuSO4 (Klueg et al., 1998). The Delta S2 cell line was cultured in the presence of 2 × 10−6 M methotrexate (ICN Biomedicals).
RNAi was performed in both cell lines as described in Clemens et al. (2000). 1,000,000 cells/well in DES serum-free medium (Invitrogen) were incubated with 20–30 μg of the various dsRNAs for 30 min at room temperature followed by the addition of 2 ml of Schneider's medium containing FBS and methotrexate in the case of Delta S2 cells. Mock-treated and mock-depleted cells were treated in the same way except that no dsRNA or ds EGFP was added, respectively. They were considered as controls.
dGM130 was detected using a rabbit polyclonal antibody directed against the first 73 amino acids of human GM130 (MLO7; gift from Martin Lowe), as described in Kondylis et al. (2001). dp115 was detected with a rabbit polyclonal antibody raised against the peptide (G)CSKLAEVSRHEAYSRA, which corresponds to amino acids 584–599 from dp115 (Biosynthesis), and purified against the peptide coupled to EAH–Sepharose 4B (Amersham Biosciences) activated by Sulfo-MBS cross-linker (Pierce Chemical Co.) (dp115/584). dSec23p was detected using a rabbit polyclonal antibody raised against the first 18 amino acids of mammalian Sec23p (Affinity BioReagents, Inc.), sharing 90% identity with the equivalent peptide of dSec23p. d120kd was detected with a mouse monoclonal antibody recognizing a 120-kD integral Golgi membrane protein in Drosophila (Calbiochem). dSed5p was detected using a rabbit polyclonal antibody raised against mammalian syntaxin 5 (JSEE1; gift from Martin Lowe). Delta was detected using C594.9B, a mouse monoclonal antibody raised against its extracellular domain (DSHB; University of Iowa).
The same number of cells (typically from 1,000,000 to 2,500,000) incubated with or without dsRNA for each time point was harvested, spun, and homogenized in 25 μl buffer containing 20 mM Tris-HCl, 1 mM EDTA, 10 mM MgCl2, 10 mM KCl, 10 mM NaCl, 1 mM DTT, 0.23 M sucrose, and 1% Triton X-100 and protease inhibitors using a motorized pestle. In the case of dp115 detection, cells were homogenized in the absence of Triton and with protease inhibitor cocktail (Roche). The SDS sample buffer was added to a 1× final concentration. The membrane fraction was prepared as follows. 10,000,000 cells were cracked using a 30-gauge needle in the homogenization buffer and protease inhibitors. After a short spin to remove cell debris and nuclei, the extract was spun for 1 h at 100,000 g to separate the cytosol from the membrane. The membrane pellet was recovered in 40 μl 1× SDS sample buffer. Total extract of one third instar Drosophila larvae and 20 embryos was prepared by homogenization using a motorized pestle in 40 μl of the homogenization buffer. The total extract, the membrane pellets, and 4% of the cytosolic fraction were fractionated on a 10% acrylamide gel and detected by Western blots using MLO7, the affinity-purified dp115 antibody, JSEE1, Sec23p antibody, and d120kd antibody, followed by the appropriate secondary antibody coupled to HRP (Vector Laboratories). ECL (Amersham Biosciences) was used for the visualization of the bands.
Total RNA was extracted from 1,000,000 S2 cells using the Purescript RNA isolation kit (Flowgen). The RT reactions were performed with the Superscript II reverse transcriptase kit (Invitrogen) according to the company's protocol and using 1 μg of total RNA.
The RT products were diluted 1/20 and were used in the PCR reaction with the 5′ primer agttcctgaagagtggcatcaa and the 3′ primer gctatctggacgaatacgat for dp115, leading to a 414-bp fragment (corresponding to the 5′ end of the dp115 gene), and the 5′ primer gtggaaaaggtggcaaagtgaa and 3′ primer atgctggtgacaacaagaa for histone 2A (H2A), leading to a 213-bp fragment. The four primers were added in the same reaction.
Cells were fixed for 2 h in 1% glutaraldehyde in 0.2 M phosphate buffer (pH 7.4) and processed for conventional EM as described in Kondylis et al. (2001).
Control or depleted S2 cells (96 h) were fixed in 2% PFA and 0.2% glutaraldehyde (GA) (2 h at room temperature) or 4% PFA alone (overnight at 4°C) in 0.2 M phosphate buffer, pH 7.4. Oregon R third instar salivary glands were dissected as described in Dunne et al. (2002), fixed in 2% PFA and 0.2% GA overnight at 4°C in the same buffer as described above. Both cells and tissues were processed for IEM as described previously (Liou et al., 1996). In brief, immunolabeling was performed using the described primary antibodies followed directly by protein A conjugated to gold particles (protein A gold), in the case of rabbit antibodies. In the case of a mouse antibody, a bridging step of rabbit anti–mouse IgG was used followed by protein A gold.
The Golgi area was defined by the Golgi stacked cisternae and immediate surrounding vesicles and tubules. When Golgi stacks were not observed, the Golgi area was defined as the clusters of vesicles and tubules occupying the position where Golgi stacks are normally found, nested near an ER cisterna in 80% of the cases. Vesicles, tubules, cisternae, and Golgi stacks are defined in Kondylis et al. (2001).
The percentage of cell profiles exhibiting at least one Golgi stack was estimated by randomly analyzing under the transmission electron microscope >100 cell profiles taken from at least two grids and two different experiments. The percentage of membrane in cisternae, tubules, or vesicles per Golgi area and the surface density of the Golgi area and total cisternae were estimated as described in Kondylis et al. (2001). The linear density was determined by the intersection method.
Delta transport assay
Delta S2 cells plated on glass coverslips were treated with dsRNA for given periods of time to deplete the protein of interest, typically 96 h. 1 mM CuSO4 was added to the culture media to induce the production of Delta. After 25 min at 27°C, the media were replaced with new Schneider's media without CuSO4, and the transport of Delta to the plasma membrane was chased for up to 90 min (initial transport of Delta). We also induced the synthesis of Delta for 60 min followed by a chase of 90 min (steady-state transport of Delta). Delta transport was also measured in cells treated with 50 μM brefeldin A for 30 min at 37°C, followed by induction with CuSO4 and chase at 27°C in the presence of brefeldin A.
S2 cells were fixed for 20 min in 3% PFA in PBS at room temperature and processed for immunofluorescence (Rabouille et al., 1999) after permeabilization with Triton. dSec23p and d120kd were detected using the antibodies described above followed by anti–rabbit IgG coupled to Alexa 488 and anti–mouse IgG coupled to Alexa 568 (Molecular Probes), respectively. Delta S2 cells were processed in the same way. Delta protein was labeled using C594.9B antibody followed by an anti–mouse IgG coupled to Texas red (Vector Laboratories). Cells were viewed under a Leica confocal microscope. A series of 30 sections through the cells were collected, and projections were reconstituted and presented unless otherwise stated.
Quantitation of Delta transport to the plasma membrane
Equatorial sections of random cells from each sample were captured with the Leica confocal microscope using a 63× water lens. Control cells were analyzed first so that the highest intensity of labeling was within the dynamic range of the laser. This set up was maintained for the analysis of the depleted cells.
The surface labeling intensity was estimated using the Leica software, which is able to measure and display the intensity of labeling along a line that is drawn perpendicular to the plasma membrane. In this way, the intensity peak as well as the thickness of the labeled plasma membrane were estimated. Typically, cells induced for 1 h followed by a 90-min chase displayed a large range of intensity and were ranked into three categories: 210 ± 45, 130 ± 35, and 75 ± 20 intensity units per plasma membrane cross section. The intensity was essentially constant around the perimeter of the cell.
About 300–500 cells for each condition were quantitated, and the results were expressed as a percentage of cells exhibiting each of the three categories of intensity. The total intensity in each condition was calculated using an arithmetic sum ΣI × N × p, where I is the intensity, N is the percentage of cells exhibiting intensity I, and p is the perimeter length of the plasma membrane. The perimeter length was not taken into account because both control and depleted cells had similar mean diameters and their almost spherical shape did not change. The total intensity was considered as 100% for the control cells, while the efficiency of intracellular transport in treated/depleted cells was expressed as a ratio of their total intensity versus that of the control cells. The same procedure was reproduced for cells induced for 25 min and chased for 45, 60, and 90 min. Again the intensity of labeling at the plasma membrane was measured as described above and ranked into three categories that vary slightly with the time of chase. The total intensity was calculated as an absolute value.
Online supplemental material
The supplemental material (Figs. S1–S4) is available. A concise description of the data presented in each supplementary figure is introduced upon citation in the text. Details on the experimental procedure and further comments on data reported can be found in the legends.
We wish to thank Martin Lowe for his gift of antibodies, Kris Klueg for her gift of the Delta cell line, Terry Orr-Weaver for the cDNA clone containing dGM130 sequence, Jonathan Dunne for critically reading and editing the manuscript, Elly van Donselaar for her help with IEM, Dagmar Zeuschner for her help with the affinity purification of the dp115 antiserum, and Thomas Vaccari and Anne Ephrussi (EMBL) for their gift of EGFP and dsed5 primers. We acknowledge the use of Flybase, the Berkeley Drosophila genome project, and Gadfly.
V. Kondylis was supported by the Darwin Trust of Edinburgh. C. Rabouille was supported by a fellowship from the Medical Research Council (UK) and by the Department of Cell Biology, Utrecht.
The online version of this article includes supplemental material.
Abbreviations used in this paper: GA, glutaraldehyde; IEM, immunoelectron microscopy; RNAi, RNA interference; tER, transitional ER.