Analysis of GTPase-activating proteins: Rab1 and Rab43 are key Rabs required to maintain a functional Golgi complex in human cells Free

Vectorial protein transport between the ER and Golgi complex requires the coordinated action of many different small GTP-binding proteins of the Ras superfamily, each controlling distinct aspects of the cargo sorting, vesicle formation and vesicle fusion processes (Bonifacino and Glick, 2004; Zerial and McBride, 2001). These GTP-binding proteins share a common mode of action involving the recruitment of cytosolic protein complexes to localized domains on membrane surfaces (Munro, 2002; Short et al., 2005). Sar1 and ARF1 recruit COPII and COPI coat complexes to ER and Golgi membranes, respectively, thus promoting vesicle formation (Bonifacino and Glick, 2004). Rabs recruit protein complexes important for directed vesicle movement and target recognition (Pfeffer, 1999; Waters and Hughson, 2000). To do this, these proteins cycle between a GDP bound inactive state, and a GTP bound membrane-associated active state (Pfeffer and Aivazian, 2004; Vetter and Wittinghofer, 2001). It is this GTP bound state that makes specific interactions with so-called effector complexes, and thus promotes their recruitment to membranes. This cycle of activation and inactivation is under the control of GDP-GTP exchange factors (GEFs) and GAPs (Pfeffer and Aivazian, 2004; Zerial and McBride, 2001).

In the case of ER to Golgi trafficking, Rab1 and Rab2 have been implicated in the tethering and fusion reactions, and are thought to act sequentially (Allan et al., 2000; Alvarez et al., 1999; Segev, 1991; Tisdale and Balch, 1996; Tisdale et al., 1992). First, COPII vesicles, formed from specialized exit sites on the endoplasmic reticulum (ER exit sites: ERES), are tethered together by the action of Rab1 together with its effector p115 to give rise to vesicular-tubular clusters (VTCs) adjacent to the Golgi (Allan et al., 2000; Alvarez et al., 1999; Bannykh et al., 1996; Cao et al., 1998). TRAPP, the GEF for Rab1, promotes tethering of COPII vesicles by activating Rab1 and also serves as a structural component of the tether (Barrowman et al., 2000; Jones et al., 2000; Sacher et al., 2001; Wang et al., 2000). Recent evidence shows that TRAPP is assembled onto forming COPII vesicles by direct interaction with the coat, thus providing a mechanism to ensure all vesicles can recruit active Rab1 (Cai et al., 2007). Second, Rab2 and its effectors promote the maturation of these VTC structures and their incorporation into the Golgi (Short et al., 2001; Tisdale and Balch, 1996). Finally, in addition to its role in COPII vesicle tethering, Rab1 may also have a function in regulating the exit of secretory cargo from the ER. This idea arose from experiments showing that extraction of Rabs using the Rab chaperone GDI (guanine nucleotide dissociation inhibitor) prevented the exit of the vesicular stomatitus virus G-protein (VSV G) from the ER and that this could be complemented by a complex of Rab1 and GDI (Peter et al., 1994). Later, Ypt1 and Uso1, the yeast homologues of Rab1 and p115, respectively, were found to play a role in coupling the sorting of secretory cargo to vesicle formation (Morsomme and Riezman, 2002), providing further support for this idea.

To determine which specific Rabs are of general importance for ER to Golgi trafficking and Golgi organization in mammalian cells, we have screened 38 human Rab GAPs for their ability to disrupt the organization of the Golgi complex and protein transport. Rab GAPs are characterized by a conserved catalytic domain, the TBC (Tre2/Bub2/Cdc16) domain (Albert and Gallwitz, 1999; Albert et al., 1999; Strom et al., 1993), that promotes GTP hydrolysis by a dual arginine-glutamine-finger catalytic mechanism related to the arginine-finger mechanism of Ras GAPs (Ahmadian et al., 1997; Albert et al., 1999; Pan et al., 2006; Rak et al., 2000). As we have shown previously, in the case of Rab5 (Haas et al., 2005), Rab GAPs can be used to specifically inactivate the endogenous pool of a Rab and thus interfere with the process that this Rab is involved in. By mutating the catalytic arginine residue of the TBC domain to alanine, the non-specific effects of GAP expression can easily be discriminated from the specific effects of Rab inactivation (Haas et al., 2005).

TBC-domain containing proteins, putative Rab GAPs, were screened for their ability to alter Golgi structure when overexpressed in either HeLa or telomerase-immortalized human retinal epithelial (hTERT-RPE1) cells (Fig. 1). Of the 38 TBC domain proteins tested, only three showed any significant effects on Golgi organization in both HeLa and hTERT-RPE1 cells (Fig. 1A). One of these, TBC1D14 was eliminated since the catalytically inactive form showed equivalent or increased effects on the Golgi when compared to the wild-type protein (Fig. 1A). This left two candidate regulators of Golgi-localized Rab GTPases, TBC1D20 and RN-tre, both capable of causing activity dependent fragmentation of both the cis-Golgi marker GM130 and the trans-Golgi marker TGN46 (Fig. 1B,C).

To further define the site of action of the different GAPs, the effects of their expression on the ER-Golgi intermediate compartment (ERGIC) were examined (Fig. 2). In this screen TBC1D20 showed catalytic activity dependent ERGIC fragmentation (Fig. 2A,B). Interestingly RN-tre caused only subtle changes to the ERGIC (Fig. 2A,B), suggesting that this GAP is more important for regulating trafficking at other Golgi compartments. This is consistent with our previous findings that RN-tre, together with its target Rab43, is required for retrograde trafficking to the trans-Golgi, but not anterograde cargo transport through the Golgi (Fuchs et al., 2007) (see supplementary material Figs S1 and S2). Interestingly, TBC1D22B, a potential GAP for Rab33 (Pan et al., 2006) was found to cause ERGIC disruption (Fig. 2A). Other GAPs showed only weak or no effect on the ERGIC. Because TBC1D20 had a general effect on Golgi and ERGIC organization, and is a highly conserved protein (see supplementary material Fig. S3), it was studied further.

To establish what happens to the Golgi upon TBC1D20 expression, markers for different Golgi sub-compartments were investigated (Fig. 3). TBC1D20 expression results in the loss of recognizable Golgi or punctate vesicular staining for p115, golgin84 and golgin97, but has no effect on the staining of the endosomal Rab5 effector EEA1 (Fig. 3A). TBC1D20, therefore, perturbs all compartments of the Golgi. None of these effects were seen when catalytically inactive forms of TBC1D20 were used (Fig. 3B), indicating that these effects are likely to be due to the catalytic inactivation of one or more Rab family GTPases.

These observations raised the question of where the Golgi proteins have redistributed to in TBC1D20-expressing cells. Peripheral membrane proteins such as GM130 and p115 can relocate to the cytosol, but integral membrane proteins such as golgin84 and Golgi enzymes must still be present in membrane structures. Alternatively, Golgi proteins could be degraded. To test these possibilities a series of experiments was performed. First, the fate of a Golgi enzyme was followed. This revealed that in TBC1D20-expressing cells, N-acetylglucosaminyl-transferase (NAGT-I) is redistributed to the ER, identified by the marker calnexin (Fig. 3C). When a catalytically inactive mutant of TBC1D20 was used, NAGT-I remained in the Golgi (Fig. 3C). Second, to test if Golgi proteins were degraded, western blots were performed on extracts of cells expressing wild-type or catalytically inactive TBC1D20. Despite the high transfection efficiency, which was over 50% as judged by fluorescence microscopy, there was no change in the levels of either GM130 or p115 in cells expressing either form of TBC1D20 (Fig. 3D). Finally, the redistribution of peripheral Golgi membrane proteins to the cytosol was tested. This showed that GM130 behaved like the integral membrane protein golgin84 and was associated with the membrane pellet to the same extent in control cells and in cells expressing wild-type TBC1D20 (Fig. 3E). Interestingly, under the same conditions, the pool of membrane-associated p115 was lost in TBC1D20-expressing cells (Fig. 3E). This effect is not simply due to blocking ER-Golgi membrane trafficking or the disruption of the Golgi complex, since BFA, which specifically inactivates COPI, and the GTP-restricted form of Sar1, which blocks the COPII pathway by preventing vesicle uncoating, did not cause loss of p115 from the membrane pellet or change the distribution of GM130 (Fig. 3E). Together, these results suggest that Golgi proteins are still associated with membrane structures after TBC1D20 expression, but their fates may be different. Similar to BFA treatment, Golgi enzymes are present in the ER, whereas Golgi matrix proteins such as GM130 are associated with a haze of small vesicular remnants. Significantly, the Rab1 effector p115 was lost from membranes in TBC1D20-expressing cells, indicating that there may be defects in the Rab1-dependent COPII vesicle-tethering pathway, and suggesting that TBC1D20 may be a regulator of Rab1.

To identify candidate target Rabs for TBC1D20, an unbiased two-hybrid screen was performed, based on the use of mutants locking the specific Rab-GAP interaction (Haas et al., 2005). This approach revealed that Rabs 1 and 2 are potential candidate targets of TBC1D20 (A.K.H. and F.A.B., unpublished). To directly test this possibility, biochemical GTP-hydrolysis assays were performed (Fig. 4). When tested against an extensive selection of Rabs, TBC1D20 strongly stimulated GTP hydrolysis by Rab1 and Rab2 (Fig. 4A). Although the majority of Rabs tested were not activated by TBC1D20, some increased activity of Rab8A, 13 18 and 35 was observed (Fig. 4A). These Rabs are closely related members of the Rab1-Sec4 subfamily, and this probably explains why a Rab1 GAP can activate them in vitro. TBC1D20 activity towards Rab1 was dependent on the TBC domain encoded by amino acids 1-317 but not other regions of the protein (Fig. 4B). This activity was abolished when either the catalytic arginine 105 in the TBC domain was mutated to alanine, or the conserved glutamine 67 in the GTP-binding site of Rab1 was mutated to leucine (Fig. 4B), in agreement with the catalytic mechanism proposed for Rab GAPs (Pan et al., 2006). Furthermore, TBC1D20 did not display GAP activity towards Sar1 and ARF1 (Fig. 4A), both non-Rab GTPases known to be required for protein trafficking and the maintenance of normal Golgi structure.

Rabs1 and Rab2 are important components of the ER-Golgi transport pathway in mammalian cells. Expression of a GAP for either Rab1 or Rab2 should inactivate the cognate Rab, and therefore block protein transport and disrupt Golgi structure. The screen described in Figs 1 and 2 revealed that only five TBC-domain proteins cause disruption of the Golgi complex in HeLa cells, and these were therefore tested against both Rab1 and Rab2 in biochemical GTP-hydrolysis assays (Fig. 4C). This revealed that only TBC1D20 promotes GTP hydrolysis by Rab1 or Rab2 to a significant extent. Consistent with these observations, target Golgi Rabs are already known for the TBC1D22 family, proposed to act on Rab33 (Pan et al., 2006), RN-tre, reported to act on Rab43 (Fuchs et al., 2007; Haas et al., 2005). TBC1D20 is therefore a specific GAP for Rab1 and Rab2 in vitro.

To further narrow down the target of TBC1D20, the effects of expressing dominant-negative Rabs were compared (Fig. 5A). Dominant-negative Rab43 T32N was taken as a positive control that caused Golgi fragmentation, but did not mimic the TBC1D20 phenotype (Fig. 5A,B). By contrast, dominant-negative Rab1 N121I, but not wild-type or activated Q67L mutant, caused a loss of Golgi phenotype similar to that caused by TBC1D20 (Fig. 5A,B). Rab2 and the Rabs that gave weak GTP-hydrolysis signals with TBC1D20 did not cause the loss of Golgi phenotype seen with dominant-negative Rab1 (Fig. 5A,B). In addition, whereas Rab2 depletion caused Golgi fragmentation, it did not cause the loss of Golgi phenotype seen with TBC1D20 or in cells expressing dominant-negative Rab1 (supplementary material Fig. S4A-C). Together, these findings support the idea that TBC1D20 acts as a Rab1 GAP in vivo.

To test if endogenous TBC1D20 is required for normal Rab1 regulation, HeLa cells were depleted of TBC1D20 using RNA interference (Fig. 5C,D). Western blot analysis showed that TBC1D20 could be efficiently silenced using specific siRNA duplexes (Fig. 5C). Endogenous TBC1D20 is a low abundance protein spread throughout the ER and hence difficult to detect (A.K.H. and F.A.B., unpublished). For this reason a GFP-tagged, catalytically inactive form of the protein was used to monitor the efficiency of RNA interference. Consistent with the idea that TBC1D20 is a negative regulator of Rab1 and hence its effectors, the intensity of p115 staining was increased in TBC1D20-depleted cells compared to the control (Fig. 5D). The typical extended ribbon-like staining pattern of GM130 became more compact, but did not increase in intensity (Fig. 5D). In addition, COPII-positive structures defined by Sec31 partially lost their perinuclear organization adjacent to the Golgi, and become scattered throughout the cytosol (Fig. 5D). Despite these changes, the transport of VSV G from the ER to the plasma membrane was not altered (A.K.H. and F.A.B., unpublished). TBC1D20 therefore contributes to the organization of COPII vesicles and the Golgi complex, but is not an essential component of the vesicle transport machinery.

If TBC1D20 is a GAP for Rab1, then it should be able to cause a block in ER to Golgi transport. To test this, VSV G transport assays were performed using cells expressing wild-type and catalytically inactive mutant forms of the TBC-domain proteins causing Golgi or ERGIC fragmentation (see Figs 1 and 2). Strikingly, only TBC1D20 prevented the appearance of the VSV G at the cell surface (Fig. 6A,B). This effect was dependent on the catalytic arginine, indicating that this is due to specific inactivation of Rab1. Furthermore, depletion of Rab2 did not block anterograde transport of VSV G from the ER to the cell surface (supplementary material Fig. S4D). Therefore, the effect of TBC1D20 expression is mainly due to the inactivation of Rab1.

Interestingly, in TBC1D20-expressing cells, VSV G was apparently unable to exit the ER after 60 minutes of transport, and did not accumulate in punctate structures characteristic of COPII vesicles or ERES (Fig. 6C). The effects of depleting components of ER-Golgi vesicle tethering complexes, Rab1, p115 and GM130, were then compared to those seen with TBC1D20 overexpression (Fig. 6D). Western blotting showed that these proteins were specifically depleted (see supplementary material Fig. S5A). Similar to TBC1D20, depletion of Rab1 or p115 prevented the appearance of VSV G at the cell surface (Fig. 6D). In contrast to p115 depletion, where VSV G left the ER and accumulated in a collapsed punctate structure reminiscent of GM130 staining in p115 depleted cells (supplementary material Fig. S5B), VSV G was unable to leave the ER in Rab1-depleted cells (Fig. 6D). Supporting this observation, similar results were obtained in cell expressing dominant-negative but not dominant active or wild-type forms of Rab1 (Fig. 6E). GM130 although required for normal Golgi-ribbon and COPII vesicle organization as expected (Puthenveedu et al., 2006), was not required for VSV G transport (see supplementary material Fig. S5B and Fig. 6D). The similarity of the TBC1D20 overexpression, Rab1 depletion, and Rab1 dominant-negative phenotypes indicates that TBC1D20 exerts its effects by inactivating Rab1. Although Rab1 binding proteins, such as p115 and GM130, may contribute to this phenotype, for example the loss of COPII vesicle organization, they cannot explain why secretory cargo proteins are blocked in the ER. However, this would suggest that TBC1D20 and Rab1 have some function at the level of the ER.

The block in cargo exit from the ER on TBC1D20 overexpression, Rab1 depletion and Rab1 dominant-negative N121I expression suggested a defect in the COPII vesicle transport pathway. Rab1 and its effector proteins are required for COPII vesicle tethering, and this is an essential step in ER to Golgi transport (Allan et al., 2000). This is most easily visualized by analysis of COPII clustering adjacent to the Golgi complex in the perinuclear region. Low-level expression of TBC1D20 (Fig. 7A) or expression of dominant-negative Rab1 (Fig. 5B) resulted in a loss of perinuclear COPII staining, leaving only a scattered peripheral pool of COPII. This effect was not seen when the catalytically inactive R105A mutant of TBC1D20 (Fig. 7A) or dominant active Rab1 were used, indicating that this effect is likely to be due to inactivation of endogenous Rab1. TBC1D20, therefore, appears to interfere with tethering of COPII vesicles by the Rab1-p115 pathway. However, since depletion of p115 does not mimic the effects of Rab1 inactivation, defective tethering cannot explain the observed requirement for active Rab1 in cargo exit from the ER (Fig. 6D).

One possibility is that Rab1 inactivation is interfering with the assembly of the COPII vesicle coat, and thus trapping cargo in the ER. However, this is unlikely to be the case for two reasons. First, TBC1D20 shows no activity towards Sar1 in vitro. Second, fluorescence recovery after photobleaching (FRAP) studies, using previously published methods (Watson et al., 2006), indicate that COPII coat turnover is the same in both control and TBC1D20-expressing cells (Fig. 7B).

The other possibility is that cargo selection is defective in TBC1D20-expressing cells. To further investigate this, the behaviour of ERGIC53 and p24, two proteins that cycle between the ER and the ERGIC, was investigated. In cells expressing wild-type TBC1D20 both endogenous ERGIC53 and GFP-tagged p24 were present in large punctate structures (Fig. 8A). Significantly, p24 was absent from the punctate scattered COPII-positive structures seen in wild-type TBC1D20-expressing cells (Fig. 8B, enlargement). This is in contrast to the distribution of p24 and ERGIC53 in cells expressing catalytically inactive TBC1D20, or in untransfected cells, where they are present in the perinuclear ERGIC and small punctate COPII-positives structures in the cell periphery (Fig. 8B). Rab1 inactivation by TBC1D20 therefore causes a defect in COPII vesicle tethering, and a transport defect where cargo molecules are no longer recruited into COPII positive structures, despite normal COPII turnover.

The results described so far indicate that TBC1D20 may control the activity of Rab1, and that interfering with this regulation prevents the exit of secretory cargo molecules from the ER. This raises the question of where TBC1D20 acts, and this was therefore investigated further (Fig. 9). First the topology of TBC1D20 was examined using a combination of carbonate extraction and proteinase K digestion experiments (Fig. 9A). These showed that TBC1D20 behaves as an integral membrane protein in carbonate extraction, and is accessible to proteinase K. This was similar to the type II transmembrane golgin, golgin84. This suggests that TBC1D20 is a type II membrane protein with the TBC domain oriented towards the cytosol. As expected, mapping experiments showed that the first 317 amino acids of TBC1D20 containing the TBC domain localize to the cytosol and cause disruption of the Golgi complex, and that deletions into this region result in a loss of this activity (Fig. 9B). These effects on the Golgi make it difficult to assess where TBC1D20 is localized, and the catalytically inactive form was therefore used for targeting experiments (Fig. 9C). This approach showed that full-length TBC1D20 was found to target to the ER and nuclear envelope, and that this targeting information was contained in amino acids 336-403 (Fig. 9C). Further deletions of the C terminus, amino acids 364-403, resulted in a slightly altered distribution to the nuclear envelope and Golgi complex, and a loss of the reticular ER staining pattern (Fig. 9C). Inspection of the TBC1D20 sequence revealed that it has a hydrophobic stretch from 364 to 387 contained within the putative ER-targeting region that could form a membrane-anchoring sequence or transmembrane domain (Fig. 9D). These observations fit with older findings that Ypt1 (yeast Rab1) GAP activity is found in the particulate membrane fraction (Jena et al., 1992).

To identify partners for TBC1D20 that could explain its localization to the ER, or possibly act as regulators of its activity, a two-hybrid screen was performed using full-length TBC1D20 as a bait. This screen identified the C-terminal region of reticulon 1 variant 2 (reticulon) as an interaction partner of TBC1D20 (Fig. 10A; A.K.H. and F.A.B., unpublished). Mapping experiments revealed that the minimal region of TBC1D20 able to interact with reticulon was the C-terminal fragment from amino acids 336-403 that shows the same ER targeting as full-length TBC1D20 (Fig. 10A). However, the shorter fragment from amino acids 364-403 showing nuclear envelope and Golgi targeting, could not interact with reticulon. These results could be confirmed at the protein level using immune precipitation, where full-length TBC1D20, irrespective of its catalytic activity, was found to interact with reticulon (Fig. 10A). The first 337 amino acids of TBC1D20, lacking the ER-targeting region did not show any significant binding to reticulon under the same conditions (Fig. 10A). Finally, reticulon, like other family members, was found to localize predominantly to the ER (Fig. 10B). TBC1D20 therefore contains a specific ER-targeting motif that interacts with a known ER protein. Reticulon is particularly interesting in this regard, since reticulons have previously been shown to be excluded from the nuclear envelope and to play a role in the organization of the tubuloreticular ER (Voeltz et al., 2006).

These results suggested that reticulon might be important for controlling TBC1D20 function. As shown previously, wild-type TBC1D20 causes the apparent loss of GM130 when expressed alone (Fig. 10C,D). This effect can be antagonized by the expression of reticulon, judged by the recovery of GM130 staining into a fragmented perinuclear ribbon structure in cells expressing both TBC1D20 and reticulon (Fig. 10C,D). When co-expressed with a catalytically inactive form of TBC1D20, reticulon and TBC1D20 were both found in the ER, and GM130 showed a normal Golgi staining (Fig. 10C). If this effect of reticulon is specific, it should require the ER-targeting properties of TBC1D20. As predicted, amino acids 1-337, a TBC-domain containing fragment of TBC1D20 capable of causing Golgi fragmentation but lacking the ER-targeting domain, was resistant to the effects of reticulon (Fig. 10C,D). Importantly, the effect of reticulon was not mediated through alterations in the expression levels of the various TBC1D20 constructs tested, since these were equal under all conditions (Fig. 10E). Together these finding support the idea that reticulons help to functionally compartmentalize the ER, in this case by controlling the activity of TBC1D20, the GAP for Rab1.

We have identified functions at the Golgi complex for a subset of the large family of TBC domain proteins encoded in the human genome. One of the most interesting of these is TBC1D20, which is unique amongst the TBC-domain proteins in that it possesses a predicted transmembrane anchor that targets it to the ER and when overexpressed causes a complete disruption of the Golgi complex. TBC1D20 shows specific GAP activity towards Rab1 and Rab2, unlike the other GAPs tested. This agrees with previous studies on budding yeast Gyp8, the homologue of human TBC1D20, which was previously reported to have Ypt1 GAP activity, and to suppress the cold-sensitive defect in strains expressing the Ypt1 Q67L activated mutant (De Antoni et al., 2002). Further investigation revealed that TBC1D20 blocks protein transport at the level of ER exit, by inactivating Rab1. As expected, this phenotype was also observed in cells depleted of Rab1. This is consistent with older observations that extraction of Rab1 using the Rab chaperone GDI prevents the ER exit of VSV G (Peter et al., 1994). The common factor in these situations is that the activated GTP form of Rab1 is absent. Depletion of the Rab1 effector p115 did not cause the same effect, and as expected secretory cargo could exit the ER, but accumulated in a punctate cluster adjacent to the nucleus and did not reach the plasma membrane. Importantly, p115 depletion is not expected to alter the amount of activated Rab1, suggesting this is the difference between TBC1D20 expression and Rab1 depletion. These findings suggested that TBC1D20 and Rab1 have p115-independent functions in controlling the exit of secretory cargo and Golgi enzymes from the ER. Interestingly, a similar role has already been proposed for Ypt1 in budding yeast, which has been shown to play a role in the sorting of some cargo molecules exiting the ER (Morsomme and Riezman, 2002). Since neither TBC1D20 in human cells nor Gyp8 in yeast is essential for secretion, regulation of GTP hydrolysis by GAPs appears not to be critical for Rab1 or Ypt1 function under laboratory conditions (De Antoni et al., 2002; Richardson et al., 1998). However, this may be a simplification of what is required to maintain secretory pathway function under more physiological conditions.

A recent study on the mechanism of α-synuclein toxicity shows that this creates a stress situation resulting in a block of ER to Golgi trafficking (Cooper et al., 2006). This genomic screen in yeast revealed Ypt1 but not other Rabs can suppress α-synuclein toxicity and recover ER to Golgi trafficking, and that the TBC1D20 homologue Gyp8 significantly enhanced α-synuclein toxicity (Cooper et al., 2006). Interestingly, COPII vesicle components and the Ypt1 effector Uso1 were not recovered, suggesting that general enhancement of COPII vesicle formation and tethering is not the mechanism by which Ypt1 is acting. The explanation that we favour is that the activity of Ypt1, and Rab1, becomes limiting for the exit of secretory proteins from the ER under stress conditions, and this may be mediated through the action of the Gyp8/TBC1D20 family of Rab GAPs.

Recent work has uncovered a role for the reticulon family of transmembrane proteins in the organization and shaping of the ER, and the regulation of ER to Golgi transport (De Craene et al., 2006; Voeltz et al., 2006; Wakana et al., 2005). It is therefore interesting that the ER targeting region of TBC1D20 interacts with at least one member of the reticulon family, and that this interaction exerts a negative effect on the ability of TBC1D20 to disrupt the Golgi complex. These findings raise the question of why is ER exit defective in cells where Rab1 has been inactivated? One possible reason is that it may be important to couple the sorting of cargo and Rab1 into COPII vesicles to ensure that they will recruit the correct tethering factors and therefore dock with the correct destination. How Rab1 could exert such effects is less clear, and will require further investigation of its interaction partners, for example the TRAPP Rab1 GEF complex (Barrowman et al., 2000; Jones et al., 2000; Wang et al., 2000), and in mammalian cells the MICAL family of proteins (Weide et al., 2003). Interestingly, TRAPP has recently been shown to bind the COPII coat and may therefore be recruited to ER exit sites as COPII vesicles form (Cai et al., 2007; Yu et al., 2006). Thus, TRAPP might provide a means to ensure loading of Rab1 into the vesicle as it forms. A second explanation may relate to the proposed function of reticulons in the compartmentalization of the ER (De Craene et al., 2006; Voeltz et al., 2006), and the fact that vesicle formation from the ER is a highly controlled process restricted to specific ER exit sites. It is therefore conceivable that reticulon together with TBC1D20 may be important for defining these sites, or perhaps restricting them to subdomains of the ER.

One striking aspect of TBC1D20 expression is that unlike any of the other 38 Rab GAPs tested, it causes the apparent loss of cis-, medial and trans-Golgi markers. This effect is not simply due to blocking vesicle trafficking, since treatments that do this, such as the GTP-restricted form of Sar1, BFA or depletion of the vesicle tether p115, block ER to Golgi traffic, yet only cause partial fragmentation of the Golgi complex. As discussed, this may be due to the fact that these treatments will not directly alter the levels of activated Rab1, whereas TBC1D20 expression will. Why then, should blocking Rab1 function at the ER cause such extreme changes in Golgi structure?

Although debated for many years, most recent findings support the view that the Golgi complex is not formed of static cisternae but rather that it is made up of highly dynamic and continuously maturing cisternae that receive and exchange material through vesicle trafficking pathways (Bonfanti et al., 1998; Losev et al., 2006; Matsuura-Tokita et al., 2006; Mironov et al., 1997). Exactly how new Golgi cisternae form is equally hotly debated, and under some conditions Golgi may form by division of pre-existing structures (Pelletier et al., 2002; Shima et al., 1997). However, there is also good evidence that the Golgi can arise directly from the ER (Bevis et al., 2002; Glick, 2002; Mironov et al., 2003), and that availability of cargo such as Golgi enzymes is a key factor in determining the size of the resulting Golgi cisternae (Guo and Linstedt, 2006). Furthermore, the organization of ERES and the COPII vesicle trafficking pathway is of key importance for normal Golgi biogenesis and function (Connerly et al., 2005). Our findings that ER exit of secretory cargo, ERGIC markers such as p24 and Golgi enzymes is blocked in cells where Rab1 is inactivated may be significant in this context. Interestingly, these cargo molecules fail to accumulate at ERES even though COPII turnover, as measured by FRAP experiments appears normal. By interfering with the sorting of Golgi enzymes and directly preventing the formation of Golgi precursors, Rab1 inactivation may therefore cause a much more serious defect in Golgi biogenesis than depleting individual tethering factors or Golgi matrix proteins.

The results presented here suggest that despite its greater size and morphological complexity the mammalian Golgi complex is more similar to its budding yeast counterpart in terms of Rab function than previously thought. In budding yeast, although additional Rabs are needed for transport from the Golgi to the cell surface, Ypt1 appears to be the sole Rab required for ER to Golgi transport, and transport though the Golgi (Brennwald and Novick, 1993). Surprisingly, the results presented here reveal a similar picture in mammalian cells. Rab1 depletion, inactivation by expression of TBC1D20, or expression of dominant-negative Rab1 all cause a block in ER to Golgi protein transport, and the collapse of the Golgi. Inactivation of other Rabs by expression of other TBC-domain proteins had no effect on secretion, despite the fact that a number of these caused Golgi fragmentation. This might indicate that other Rabs functioning at the Golgi are redundant, or not required for general anterograde cargo transport. These ideas and others discussed here, provide many avenues for further investigation.

Antibodies reagents were as follows: mouse anti-GM130, Sec31 and EEA1 (Becton Dickinson, Heidelberg, Germany); sheep anti-TGN46 (Serotec, Düsseldorf, Germany); sheep anti-GM130, p115 and GRASP55 (Barr et al., 2001; Shorter et al., 1999). Donkey secondary antibodies conjugated to HRP, AMCA, CY2 and CY3 were obtained from Jackson ImmunoResearch Labs (West Grove, PA).

Human Rab GAPs were identified by searching the GenBank database using the TBC-domain signature motifs defined by Goody and co-workers (Rak et al., 2000). The 38 human Rab GAPs identified by this method and expression constructs have been described previously (Fuchs et al., 2007). Yeast two-hybrid assays, protein expression and purification were performed as described previously (Haas et al., 2005).

hTERT-RPE1 cells (Clontech Laboratories) were grown at 37°C and 5% CO₂ in a 1:1 mixture of DMEM and HAMS F12 containing 10% calf serum, 2.5 mM L-glutamine, and 1.2 g/l sodium bicarbonate. HeLa cells were cultured at 37°C and 5% CO₂ in DME containing 10% FCS. HeLa or hTERT-RPE1 cells plated on glass coverslips at a density of 50,000 cells/well of a 6-well plate were used for plasmid transfection and RNA interference. All siRNA duplexes were obtained from Dharmacon Inc, Lafayette, CO and are described in Table SI in supplementary material. Where multiple siRNA duplexes are indicated the same results were obtained with all. For western blotting, cells from three wells of the six-well plate were washed in 2 ml PBS, then lysed in 70-80 μl 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% (wt/vol) Triton X-100. For each lane of a minigel 10 μg of the protein lysate was loaded.

For fractionation experiments HeLa cells were grown on 15 cm dishes to 70% confluence. Cells were either transfected using 15 μg of each plasmid DNA and left for 24 hours to express the protein of interest, or incubated for 30 minutes in growth medium containing 5 μg/ml BFA. The cells were then harvested and washed twice in ice cold PBS. The cell pellet was resuspended in 1 ml of 25 mM Tris-HCl pH 7.4, 130 mM KCl, 5 mM MgCl₂ containing a protease inhibitor cocktail (Roche Diagnostics). Cells were then broken open by passing them 40 times through a 27G needle using a 1 ml syringe. Nuclei and cell debris were removed by centrifugation for 5 minutes at 1000 g and 4°C. This post-nuclear supernatant was split into two aliquots, one was kept on ice and represented to the total material. The other half was centrifuged for 30 minutes at 100,000 g and 4°C. The supernatant, which was the soluble cytosolic fraction, was transferred to a fresh tube. The pellet, which was the membrane fraction, was resuspended in 100 μl sample buffer. Aliquots of the supernatant and total of each sample were precipitated by adding 0.5 μl of 10% (wt/vol) sodium deoxycholate and 30 μl of 100% (wt/vol) TCA. After 30 minutes incubation on ice, precipitated protein was collected by centrifugation at 20,000 g and 4°C. Pellets were washed with ice cold acetone, and then resuspended in 100 μl of sample buffer.

For proteinase K digestion experiments, HeLa S3 cells from two 70% confluent 15 cm dishes and one 10 cm dish transfected with GFP-tagged TBC1D20 constructs for 24 hours were washed twice in cold PBS, then scraped off the dishes, and pooled. The cell pellets were resuspended using six passes through a 21G needle in 50 mM Hepes pH 7.5, 200 mM sucrose, 1 mM MgCl₂, 1 mM EDTA, to give a final volume of 1 ml. This cell suspension was passed though an EMBL cell cracker (European Molecular Biology Laboratories, Heidelberg, Germany) fitted with an 8.002 mm diameter ball. The broken cell suspension was centrifuged twice at 1000 g and 4°C for 5 minutes to remove cell debris and leave a post nuclear supernatant. Proteinase K (New England Biolabs) was incubated at 37°C for 30 minutes prior to use to inactivate any contaminating lipase activity. The post nuclear supernatant was then adjusted to 10 mM CaCl₂, and 100 μl aliquots treated with 2 μg proteinase K, 0.5% (vol/vol) Triton X-100, or both proteinase K and Triton X-100 for 30 minutes on ice. To stop the reaction, a half volume of 100% (wt/vol) TCA was added, samples were then vortexed and incubated on ice for 30 minutes. Proteins were recovered by centrifugation at 20,000 g for 15 minutes, the pellets washed in 1 ml of –20°C acetone, resuspended in 100 μl of sample buffer, and 20 μl analysed by western blotting.

For carbonate extraction experiments, 100 μl aliquots of the post nuclear supernatant was centrifuged at 100,000 g for 30 minutes at 4°C to generate a membrane pellet. The pellet was resuspended in 100 mM Na₂CO₃ and incubated on ice for 30 minutes. To recover the membrane the sample was centrifuged at 100,000 g for 30 minutes at 4°C. The supernatant was precipitated using 25% (wt/vol) TCA, and equal amounts of the total, carbonate-extracted supernatant and membrane pellet fractions analysed by western blotting.

VSV G ts045 protein transport assays were carried out using an adaptation of a published protocol. HeLa cells plated on glass coverslips were treated with specific RNA duplexes for 56 hours at 37°C, then transfected with a plasmid encoding green fluorescent protein (GFP)-tagged VSV G protein for 2 hours at 37°C, then for 12 hours at 39.5°C. The cells were then incubated at 4°C to promote VSV G protein folding, and afterwards the growth medium was replaced with pre-warmed medium at 31.5°C. After the required chase period cells were fixed with 3% (wt/vol) paraformaldehyde in PBS. Cell surface VSV G was detected with a monoclonal antibody to the VSV G lumenal domain and a donkey anti-mouse secondary coupled to CY3, and total VSV G by GFP fluorescence. The ratio of surface to total measured fluorescence was used to calculate the extent of VSV G transport. Shiga toxin transport assays were performed using a published method (Fuchs et al., 2007). Cells were processed for immunofluorescence and images collected as described previously (Fuchs et al., 2007).