Rabs and Arfs/Arls are Ras-related small GTPases of particular relevance to membrane trafficking. It is thought that these proteins regulate specific pathways through interactions with coat, motor, tether and SNARE proteins. We screened a comprehensive list of Arf/Arl/Rab proteins, previously identified on purified Golgi membranes by a proteomics approach (37 in total), for Golgi or intra-Golgi localization, dominant-negative and overexpression phenotypes. Further analysis of two of these proteins, Rab18 and Rab43, strongly indicated roles in ER-Golgi trafficking. Rab43-T32N redistributed Golgi elements to ER exit sites without blocking trafficking of the secretory marker VSVG-GFP from ER to cell surface. Wild-type Rab43 redistributes the p150Glued subunit of dynactin, consistent with a specific role in regulating association of pre-Golgi intermediates with microtubules. Overexpression of wild-type GFP-Rab18 or incubation with any of three siRNAs directed against Rab18 severely disrupts the Golgi complex and reduces secretion of VSVG. Rab18 mutants specifically enhance retrograde Golgi-ER transport of the COPI-independent cargo β-1,4-galactosyltransferase (Galtase)-YFP but not the COPI-dependent cargo p58-YFP from the Golgi to ER in a photobleach assay. Rab18-S22N also potentiated brefeldin-A-induced ER-Golgi fusion. This study is the first comprehensive application of large-scale proteomics to the cell biology of small GTPases of the secretory pathway.
The primary organelles of the secretory pathway are the endoplasmic reticulum (ER) (Porter et al., 1945) and the Golgi complex (Golgi, 1898). The trafficking of membranes between major elements of the secretory pathway is largely mediated by discontinuous carriers (Palade, 1975). The specificity of trafficking pathways depends on the selective incorporation of cargo into newly forming transport intermediates, movement of these transport intermediates to the target membrane and subsequent recognition and fusion steps (Palade, 1975).
Virtually all of these steps are regulated by small Ras-family proteins in the Arf/Arl (Donaldson et al., 2005; Kahn et al., 2006) and Rab (Stenmark and Olkkonen, 2001; Zerial and McBride, 2001) subfamilies. Many Rabs are known to regulate distinct aspects of trafficking pathways, often by regulating tethers, SNARE (soluble NSF attachment receptor) proteins and motors (Zerial and McBride, 2001). For example, Rab5 binds to early endosome antigen 1 (EEA1) and forms a complex with SNARE proteins to mediate homotypic fusion of early endosomes (Christoforidis et al., 1999; McBride et al., 1999). Rab1 interacts with other tethers including p115 required for the fusion of pre-Golgi intermediates with Golgi membranes (Sztul and Lupashin, 2006). Rab6 regulates a microtubule/kinesin-dependent pathway originating at the trans-Golgi network (TGN) (Echard et al., 1998; Girod et al., 1999; White et al., 1999). Thus, many Rab proteins possess a characteristic organellar distribution and regulate a specific transport step (Chavrier et al., 1990; Zerial and McBride, 2001).
There are in excess of 60 Rab-family (Pereira-Leal and Seabra, 2001) and 29 Arf/Arl (Kahn et al., 2006) proteins in humans compared with 11 Rab/Ypt proteins (Lazar et al., 1997) and five Arf/Arls in yeast (Behnia et al., 2004; Huang et al., 1999). Studies in mammalian cells have provided evidence for organization of specific Rabs in `Rab domains' (Sonnichsen et al., 2000), which possess distinct lipid compositions but may be defined in part also by Rab effector proteins capable of interacting simultaneously with different Rabs (De Renzis et al., 2002). Studies in yeast also find recruitment of Rabs to lipid domains and it has been proposed based on these data that lipids define in part the localization of Rab proteins (Sciorra et al., 2005). Arfs are found in pits coated with COPI (coat protein complex) or clathrin which may indicate the existence of other spatially delimited domains.
Because of their central roles, Arf/Arl and Rab proteins provide appealing initial starting points for molecular characterization of these trafficking pathways. The increased number of these proteins in mammals is believed primarily to derive from the larger number of trafficking pathways, particularly between the Golgi complex, cell surface and endosomes, in addition to the variety of cargo requiring transport. As a consequence, in more complex organisms, many Rab/Arf/Arl proteins would be expected to associate with the different transport intermediates involved in post-Golgi trafficking. However, a smaller number of Arf/Arl/Rab proteins would be required to regulate trafficking between the ER and Golgi where many distinct protein-protein interactions are required (e.g. tethers, SNARE proteins and motors) on a small number of distinct transport intermediates.
To further investigate the molecular mechanisms of small GTPases on the Golgi complex, we identified and GFP-tagged the entire subset of Golgi-localized Ras-family GTPases (primarily Arf, Arl and Rab proteins). Of the poorly characterized proteins identified, four proteins (Arl6, Rab18, Rab30 and Rab43) showed clear localization to the Golgi and ER. Of these, all but Arl6 predominately localized to pre-TGN locations within the early secretory pathway. Further detailed characterization of two of these proteins (Rab18 and Rab43) yielded data consistent with roles in retrograde and anterograde Golgi-ER pathways, respectively.
GFP-tagging and Golgi localization of Arf/Rab family candidate Golgi proteins
To determine the set of Golgi-associated Rab and Arf/Arl proteins, we compiled a list of Ras-family proteins identified from Golgi or COPI vesicle preparations (Gilchrist et al., 2006) by at least one unique peptide (see supplementary material Table S2). We additionally included Arl6 because a known Arl6-interacting protein Arl6IP2 (Ingley et al., 1999) was also identified. We expressed GFP fusions of each of these proteins in Vero cells, and stained with antibody against the Golgi marker GM130 and the nuclear stain DAPI (Fig. 1; see supplementary material Table S3).
ARFRP1, Arl6, Rab30 and Rab43 all gave a similar pattern of Golgi and ER staining in Vero cells (Fig. 1) and other cell lines including HeLa, NRK and COS7 (data not shown). We did not detect Golgi localization for Arl6-GFP in MDCK cells despite it showing consistent Golgi localization in the other cell lines (three or more independent experiments in each cell line). Rab18 consistently localized to an extended reticular structure resembling the ER and nuclear envelope in all transfected cell types examined, including HeLa, NRK, COS7 and Vero cells, while showing variable degrees of Golgi localization in Vero, COS7 and NRK cells.
Rab4 and Rab5 showed clear GFP labeling on the Golgi as well as punctae in Vero cells (see supplementary material Fig. S1), although not in NRK cells (data not shown). Of the poorly characterized proteins, only Rab35 was localized to numerous intracellular punctate structures resembling endosomes (Kouranti et al., 2006). However, it also showed strong cell surface localization (Fig. 1) and an absence of clear Golgi staining.
Rab24, identified by proteomic analysis on the ER but not on the Golgi (Gilchrist et al., 2006), stained the ER including nuclear envelope (data not shown) and not the Golgi. Arl1 and Rab10 showed diffuse staining, whereas Rab15 localized entirely to nuclei, possibly indicating failure of the GFP constructs to target. RabL3 localized to the cytoplasm and stress fibers (Fig. 1; and see supplementary material Fig. S2) but did not colocalize with β-tubulin (data not shown). However, 22 out of 35 GFP constructs localized at least partially to the Golgi (supplementary material Table S3), which correlated well with the results of the proteomic analysis.
Effects of mutations analogous to Ras dominant negatives on Golgi morphology
We next used overexpression assays to gain insight into possible roles of the Arl/Rab proteins in the secretory pathway. GFP-tagged wild-type and dominant-negative versions of all proteins were transfected into Vero cells (see supplementary material Table S3; Fig. 1 and Fig. 2A). Cells were subsequently immunostained for GM130 and DAPI to test for significant disruption of the Golgi complex, which was quantified by a measure of the Golgi area as described in Materials and Methods. By this measure, the strong dispersal of the Golgi complex produced by overexpression of Rab18 was unique among the tested wild-type constructs (Fig. 2A,C). Overexpression of dominant-negative Rab43-T32N also produced a strong dispersal of the Golgi complex (Fig. 2D). However, Rab1b-S22N, a GDP-locked mutant of Rab1b, which has been previously shown to disrupt the secretory pathway when overexpressed (Plutner et al., 1991), produced only a twofold increase in dispersal of the Golgi complex (Fig. 2D). This is probably related to the lower expression levels in our Vero cells.
The quantitative measurement of Golgi dispersal was repeated in COS7 cells, because this cell line was able to support higher expression levels (Fig. 2E,F), resulting in a fivefold increase in Golgi dispersal for the positive control Rab1b-S22N (Fig. 2F). Under these conditions, overexpression of Rab43-T32N produced highly visible dispersal of Golgi staining with GM130-labeled punctae scattered throughout the cytoplasm (Fig. 2B) resulting in a tenfold increase in Golgi dispersal relative to untransfected cells (Fig. 2F). Dominant-negative ARFRP1-T32N also produced a large quantifiable effect in COS7 cells, but this appeared to be due to elongation of the Golgi complex in expressing cells, which in turn produced an exaggerated measure of Golgi scattering in our automated assay. Overexpression of wild-type Rab18 produced a milder visible phenotype in COS7 compared with Vero cells, but this equated to an almost fourfold increase in Golgi dispersal, which was greater than any other Rab protein tested including Rab6a and Rab33b, overexpression of which has been previously shown to disrupt Golgi morphology (Jiang and Storrie, 2005; Martinez et al., 1997; Valsdottir et al., 2001). Additional functional assays were also used to assess the importance of these Rabs. We measured the cell surface arrival of the temperature-sensitive vesicular stomatitis virus glycoprotein (ts-O45-G mutant) (VSVG) in cells overexpressing the wild-type Rab proteins, in an automated assay (Liebel et al., 2003) (supplementary material Fig. S4). The strongest retardation in secretion was observed in cells overexpressing Rab18, which is consistent with the effects seen on the Golgi complex upon overexpression of this protein.
In summary, overexpression of two proteins, Rab18 and Rab43-T32N, resulted in the dramatic redistribution of the Golgi protein GM130, with Rab18 also influencing the secretory capacity of cells, suggesting potential roles for these proteins in Golgi-ER trafficking.
Predominant localization of Rab/Arl proteins within the Golgi complex
The proteins identified could potentially regulate transport pathways between the ER and early (cis) Golgi or between late (trans) Golgi and post-Golgi compartments. We have recently developed a quantitative light microscopic assay that permits us to localize the major pool of transfected proteins to early or late Golgi regions (Dejgaard et al., 2007) (Fig. 3A,B). In this assay, Arl6 gave a clear trans Golgi localization (Fig. 3C). Rab18 presented particular difficulties as its predominant localization in NRK cells was to the ER. Nevertheless, in those cells where some Golgi localization of Rab18 was observed, it was the most cis-localized of all the proteins tested (Fig. 3C). Rab43 also gave a cis localization by this assay (Fig. 3C). Similar results were obtained in a fully automated version of this assay (Dejgaard et al., 2007) which required nocodazole treatment of cells to simplify Golgi geometry (see supplementary material Fig. S2B). Consistent results were obtained in a second cell line (HeLa) stained with GM130 (cis) and TGN46 (trans) using both the manual and automated methods (see supplementary material Fig. S3A,B). Efforts were taken to image cells expressing relatively low levels of GFP-proteins. However similar results were obtained across a range of expression levels.
Although this method has limited resolution compared to electron microscopy (e.g. closely apposed ER or intermediate compartment might not be distinguished from cis-Golgi) and transfected rather than endogenous proteins were examined, results with the control proteins (cis for Rabs 1 and 2, trans for 6a and 6a' and intermediate for Rab33b) agreed closely with previous immuno-electron microscopic localizations. Rab18 and Rab43 clearly colocalized well with GM130, but only poorly with TGN38/46, consistent with the major pool of these proteins having an early Golgi localization. These proteins are thus excellent candidates for roles in ER/Golgi or intra-Golgi trafficking.
Arf/Rab proteins identified as Golgi associated are associated dynamically with the Golgi
Cytosolic Rabs are associated to GDP and require GTP exchange for recruitment to membranes, after which the Rab remains membrane-associated while GTP is bound. Rab proteins show very slow rates of spontaneous GTP hydrolysis, and must associate with GTPase activator proteins (GAPs) for efficient GTP hydrolysis. This is followed by rapid removal from membranes of Rab-GDP by Rab-GDI (Zerial and McBride, 2001).
A recent study identified a family of Rab-GAPs and suggested that only two Rab-GAPs (TBC1D20 and RN-tre), specific for Rabs 1, 2 and 43, are essential for Golgi function (Haas et al., 2007). To determine if other Rab/Arl proteins associated with Golgi membranes have a rapid association/dissociation cycle similar to Arf1 (Presley et al., 2002) and previously characterized Rab proteins (Vieira et al., 2003), suggesting interactions with Golgi GAPs and GEFs, we used a photobleaching approach (Presley et al., 2002). Analysis of recovery curves after photobleach of Golgi-associated proteins (Fig. 4) indicated all proteins tested, with the exception of ARFRP1, exchanged rapidly between the Golgi and cytosol with a t1/2 for recovery of Golgi-associated fluorescence to a plateau value in the range of 60-120 seconds (Fig. 4B,C). This indicates functional interactions of Rab GAPs with multiple Rabs on the Golgi complex.
Effects of Rab43-T32N on Golgi morphology and secretory function
To further explore the importance of Rab43 function in the secretory pathway, we transfected COS7 cells with GFP-Rab43-T32N. In contrast to cells transfected with wild-type GFP-Rab43 (Fig. 1), the GM130 staining pattern was highly punctate. These punctate structures were immediately adjacent to ER exit sites (visualized with anti-COPII (Sec23 subunit) antibody; Fig. 5A) and contained Golgi markers including p115, COPI, ManII and TGN46 (data not shown).
To determine if the punctate elements identified in cells expressing GFP-Rab43-T32N were functional Golgi mini-stacks, we again examined the trafficking of the VSVG cargo tagged with CFP/YFP (Presley et al., 1997). COS7 cells grown on coverslips were doubly transfected with GFP-Rab43-T32N and with VSVG-CFP or singly transfected with VSVG-YFP and incubated for 24 hours at 40°C (non-permissive temperature) to accumulate VSVG-CFP in the ER. Cells were subsequently shifted to 32°C (permissive temperature to exit ER) for varying times. We found that in cells expressing GFP-Rab43-T32N, VSVG-CFP was largely accumulated on the cell surface at 3 hours (Fig. 5B). This was apparent in >90% of cells examined and at both low and high levels of Rab43-T32N expression. Thus, the secretory pathway was functional in cells expressing Rab43-T32N despite the dramatic changes in appearance of the Golgi complex.
A similar phenotype (scattering of the Golgi complex to ER exit sites but with preservation of secretory trafficking) is found in cells in which microtubules are destroyed by nocodazole (Cole et al., 1996) or in which interaction of pre-Golgi intermediates with the microtubule motor dynein/dynactin is blocked (Burkhardt et al., 1997). Examination of Golgi fragmentation induced by Rab1b-S22N and Rab43-T32N indicated that Rab43-T32N was effective at considerably lower expression levels (Fig. 5C). Since motors such as the dynein/dynactin complex are low abundance molecules, this could indicate a specialized role for Rab43 in recruiting microtubule motors. When a GFP-tagged version of the p150Glued subunit of dynein/dynactin (Watson and Stephens, 2006) was co-expressed with CFP-Rab43, which did not redistribute the Golgi markers examined, p150Glued lost its microtubule localization and became largely cytosolic (Fig. 5D,E). These results taken together suggest a potential role for Rab43 in regulation of anterograde trafficking from ER to Golgi.
Effects of overexpression or inhibition of Rab18 on the Golgi-ER system
Previous studies have indicated that Rab18 can localize to lipid droplets under certain circumstances (e.g. Martin et al., 2005; Ozeki et al., 2005) and to endosomes (Lutcke et al., 1994). We were able to confirm that GFP-Rab18 partially localized to lipid droplets induced by oleic acid treatment in Vero cells (see supplementary material Fig. S5) and in HeLa cells (data not shown) although a major pool of GFP-Rab18 did not appear to be localized to lipid droplets (see supplementary material Fig. S5). Others (Martin et al., 2005) have also noted consistent localization of Rab18 to the ER, even after induction of lipid droplet formation.
To further explore the disruption of the Golgi complex caused by Rab18 overexpression, we examined the distribution of various other markers. We found that COPI, ManII and p115 colocalized with GM130 punctae, (data not shown). An activated mutant of Rab18 (Q67L) analogous to a GTPase-defective mutant of Ras (Feig and Cooper, 1988) gave a similar phenotype of scattering GM130 staining (data not shown). GM130 staining was often scattered to COPII positive punctae, suggesting relocalization of Golgi components to ER exit sites with high expression levels of Rab18 constructs (Fig. 6B) comparing to low expression levels (Fig. 6A) of this protein.
We had found that overexpression of GFP-Rab18 for 24 hours in a standardized screen using HeLa cells resulted in a reduction in VSVG trafficking to the cell surface (supplementary material Fig. S4), indicating a potential role for Rab18 in secretion. We conducted additional experiments, co-transfecting COS7 cells on coverslips with CFP-Rab18 and with VSVG-YFP. Coverslips were kept at 40° C for 48 hours after transfection to ensure high expression levels of Rab18 constructs followed by shifting to 32°C. In cells singly transfected with VSVG-YFP or in cells from double-transfected coverslips but lacking CFP-Rab18 fluorescence, staining with the VG antibody was readily visible on the surface of non-permeabilized cells (Fig. 6C). However, in non-permeabilized cells expressing CFP-Rab18 (Fig. 6D), or CFP-Rab18-Q67L (not shown) little or no VG staining was detected. This was confirmed by quantitative analysis (see Materials and Methods) of the fraction of total VG on the cell surface (Fig. 6F). Thus, CFP-Rab18 but not CFP-Rab18-S22N or a negative control (CFP-Rab5) blocks VSVG trafficking to the cell surface (Fig. 6F) The lesser inhibition by Rab18-S22N was surprising but as CFP/YFP Rab18-S22N shows some membrane association, one explanation is that it has residual GTP exchange activity and thus acts as a weakly active wild-type rather than a dominant negative. There was little effect of GFP-Rab18 expression (48 hours) on uptake of the endocytic markers transferrin or EGF (Fig. 6G), indicating that effects of Rab18 overexpression were specific to the secretory pathway. At shorter Rab18 expression times (24 hours) the block in VSVG trafficking was only partial, consistent with a dose-dependent overexpression effect (data not shown), similar to the results obtained in HeLa cells after 24-hour expression (supplementary material Fig. S4).
We next examined Golgi morphology and secretion of VSVG in cells depleted for Rab 18 with three independent siRNAs found to reduce the levels of the Rab18 mRNA by 60–94% (depending on the siRNA; Fig. 7A). We found that the siRNA giving the best knockdown (94%; siRab18-1) reduced VSVG secretion by 60% (Fig. 7B) and produced readily visible Golgi fragmentation with a six-fold increase in the total number of Golgi fragments detectable by light microscopy (Fig. 7C). This Golgi fragmentation was in excess of that produced by an siRNA directed against Rab1A (Fig. 7), which is well known to be required for secretory pathway function. Similar, albeit less pronounced, effects were produced by two independent siRNAs directed against Rab18, corresponding to the observation that these oligonucleotides showed a relatively lower reduction in the levels of Rab18 mRNA, as judged by RT-qPCR analysis (60-65% knockdown) (Fig. 7A). Golgi structure is normally maintained by interactions with the ER via trafficking pathways (Cole et al., 1996), thus disruption of the Golgi by either overexpression or down-regulation of Rab18 suggests a possible role for Rab18 in regulation of ER/Golgi trafficking.
Overexpression of Rab18 mutants enhances retrograde trafficking
To determine the effect of Rab18 and its mutants on cycling between the Golgi and ER (Zaal et al., 1999), we doubly transfected COS7 cells on glass-bottomed dishes with Galtase-YFP and with CFP-Rab18 or CFP-Rab18-S22N. The COPI-independent cargo Galtase-YFP has been localized to late Golgi compartments (Zaal et al., 1999) and its cycling between the Golgi and ER has been quantitated in HeLa cells using a photobleach assay (Zaal et al., 1999). COS7 cells singly transfected with Galtase-YFP were used as controls. Confocal image sequences were taken in which the Golgi was photobleached after acquisition of an initial prebleach image (Fig. 8A) as previously described (Presley, 2005).
In control cells, fluorescence was distributed between the Golgi complex and a dispersed pool which has been shown to be the ER (Zaal et al., 1999) (Fig. 8A). After photobleaching, Golgi fluorescence recovered on a timescale of 20-40 minutes (Fig. 8A). Recovery curves fit well to an inverted exponential with little or no immobile fraction (Fig. 8C). Thus, trafficking of Galtase-YFP between the ER and Golgi was considered to follow first order kinetics for analysis purposes. An ER-to-Golgi rate constant (kin) and Golgi-to-ER rate constant (kout) were calculated (see Materials and Methods, Fig. 8E).
Similar analysis was performed in cells double-transfected with Galtase-YFP and CFP-Rab18 or CFP-Rab18-S22N. Prebleach (i.e. steady-state) ratio of Golgi to non-Golgi fluorescence, ranged from roughly 70% Golgi/30% non-Golgi for control cells to 30% Golgi70% non-Golgi for cells expressing CFP-Rab18-S22N. Recovery curves could be fitted well to inverted exponentials with little immobile fraction as in controls. Golgi residence time (1/kout) was 22.8 minutes for Galtase-YFP in singly transfected cells (Fig. 8F). This was faster than the Golgi residence time of 57.2 minutes measured previously in HeLa cells by similar methods (Zaal et al., 1999). An ER residence time (1/kin) of 11.6 minutes was calculated for control cells (Fig. 8G).
Coexpression of Galtase-YFP with CFP-Rab18 for 48 hours led to a sharp reduction in both Golgi and ER residence times (to roughly 2 minutes each; Fig. 8F,G). These rates are too fast to explain by typical membrane trafficking pathways and are most compatible with the structures being Golgi remnants partly continuous with the ER [similar to `BFA bodies' (Orci et al., 1993)] rather than true Golgi. Diffusion of Galtase-YFP between connected membranes could occur on the timescale of this experiment. Such severe disruption of the secretory pathway would be consistent with the fact that VSVG trafficking to the cell surface is efficiently blocked under the same conditions (Fig. 6C,D).
Overexpression of CFP-Rab18-S22N reduced but did not completely block trafficking of VSVG-YFP, indicating the presence of functional Golgi complex. We repeated the Galtase-YFP cycling assay in cells co-transfected with CFP-Rab18-S22N in an attempt to identify selective effects on trafficking pathways. In co-transfected cells, there was a modest reduction in Galtase-YFP ER residence time from 11.6 minutes to 8.4 minutes (Fig. 8G). In contrast, Golgi residence times were reduced seven-fold from 22.8 minutes to 3.3 minutes (Fig. 8F). These results are most consistent with a massive selective enhancement of a Golgi-to-ER retrograde trafficking pathway in cells expressing CFP-Rab18-S22N.
As an alternative measure of retrograde trafficking, we treated COS7 cells transfected with Galtase-YFP and CFP-Rab18-S22N or singly transfected with Galtase-YFP with 5 μg/ml brefeldin A. Brefeldin A induces fusion of Golgi with ER membranes and collapse of the Golgi into the ER in easily observed discrete events termed `blinkouts' (Sciaky et al., 1997) (Fig. 9A). CFP-Rab18-S22N accelerated the rate of Golgi blinkout relative to single-transfected cells (Fig. 9B,E), consistent with enhancement of fusion of ER with Golgi membranes. In HeLa cells, addition of a pSilencer neo plasmid expressing Rab18-specific siRNA (hsiRNA1) had the reverse effect, inhibiting Golgi-ER fusion in this assay (Fig. 9C,F). CFP-Rab18-S22N-induced enhancement of Golgi/ER fusion still occurred in COS7 cells treated with nocodazole (Fig. 9D), which is more consistent with an increase in fusigenicity (e.g. from enhanced activity of tethers or SNAREs) rather than an increase in motor protein activity (e.g. Rab6/kinesin interactions).
Effects of Rab18 mutants on COPI-dependent retrograde trafficking pathways
Recycling of Galtase-YFP from the Golgi to the ER is not blocked by microinjection of antibodies against COPI, and likely employs a COPI-independent pathway regulated in part by Rab6 (Girod et al., 1999). In contrast, ERGIC53/p58 possesses a dilysine motif which interacts strongly with COPI (Andersson et al., 1999), and requires COPI for retrograde trafficking (Girod et al., 1999).
To monitor effects of overexpression of Rab18 and its mutants on COPI-dependent Golgi-to-ER trafficking of cargo, we examined the COPI-dependent cargo p58-YFP (Presley et al., 2002), the rodent analogue of ERGIC53 (Lahtinen et al., 1992). Cells were co-transfected with CFP-Rab18 or CFP-Rab18-S22N or transfected with p58-YFP alone (Fig. 8B) as a control. Anterograde and retrograde cycling between the Golgi and ER were monitored by photobleaching. P58-YFP was found to cycle rapidly between the ER and Golgi consistent with previous findings (Ben-Tekaya et al., 2004; Ward et al., 2001) (Fig. 8D). However, neither anterograde nor retrograde trafficking of p58-YFP was significantly affected by overexpression of CFP-Rab18 or CFP-Rab18-S22N (Fig. 8D,F,G). These data are most compatible with Rab18 function in a COPI-independent pathway from the Golgi complex to the ER.
We here utilize new proteomics data obtained from Golgi membranes in order to more systematically study GTPase function with respect to membrane traffic in the early secretory pathway. Of the 32 Arf/Arl/Rab proteins examined in this study, we initially focused on six poorly characterized proteins (Arl6, Rab30, Rab35, Rab43, RabL3 and Rab18). Subsequent more detailed analysis of two proteins, Rab43 and Rab18, now provides strong evidence of involvement in Golgi/ER trafficking pathways.
Role of Rab43 in anterograde trafficking
Further characterization of Rab43 suggested a role in anterograde trafficking of cargo from the ER to the Golgi since a dominant-negative mutant (GFP-Rab43-T32N) produced a striking Golgi redistribution into scattered punctae colocalizing with ER exit sites. Since trafficking of cargo from the ER to the cell surface was not disrupted despite the dramatic Golgi fragmentation, a simple explanation consistent with these data is that Rab43 regulates dynein/dynactin (Burkhardt et al., 1997) or similar proteins required for interaction between pre-Golgi intermediates and microtubules. A similar phenotype is seen when microtubules are disrupted with nocodazole (Cole et al., 1996; Storrie et al., 1998) or when the dynactin complex is disrupted by overexpression of the p50/dynamitin subunit of this complex (Burkhardt et al., 1997). Under these conditions, pre-Golgi intermediates accumulate at ER exit sites where functional Golgi mini-stacks are reconstituted. Some of our experimental results are suggestive of a possible interaction between Rab43 and dynein/dynactin (e.g. Fig. 5D,E).
Others (Haas et al., 2007) have recently published similar experimental results, showing fragmentation of the Golgi without a block in secretion in Rab43-T32N-expressing or Rab43-depleted cells. They also found a block in the transport of Shiga toxin from the cell surface to the ER when RN-tre, a proposed GAP for Rab43, is overexpressed (Fuchs et al., 2007) and suggested that Rab43 regulates retrograde trafficking from endosomes to Golgi. However, disruption of an endosome/Golgi trafficking pathway would not be expected to dramatically redistribute Golgi elements to COPII-positive ER exit sites (see Fig. 5A). Furthermore, a Rab regulating such a pathway would be expected to show TGN localization, whereas GFP-Rab43 shows an early Golgi localization (see Fig. 2). It is possible that the reduction in Shiga toxin transport could be a secondary consequence of the reorganization of the Golgi complex. Resolving these questions will require further studies including the identification and characterization of Rab43 effectors.
Role of Rab18 in Golgi to ER transport
Rab18 is not found in yeast but has been identified as a Golgi protein in trypanosomes (Jeffries et al., 2002). Rab18 has been localized to endosomes of mammalian cells by antibody staining in previous studies (Lutcke et al., 1994; Yu et al., 1993) and more recently to lipid droplets by multiple techniques including mass spectrometric analysis of purified lipid droplets (Brasaemle et al., 2004; Liu et al., 2004; Umlauf et al., 2004), immunofluorescence (Martin et al., 2005; Ozeki et al., 2005) and immunoelectron microscopy (Martin et al., 2005; Ozeki et al., 2005). Previous to this study, no fundamental generalized role has been proposed for Rab18 in the secretory pathway, although evidence has been provided for a role as a negative regulator of secretion in neuroendocrine cells (Malagon et al., 2005; Vazquez-Martinez et al., 2007).
However, Rab18 was identified by multiple peptides in all rat liver Golgi preparations in two independent proteomic studies (Gilchrist et al., 2006; Wu et al., 2004), consistent with it being an abundant protein on membranes efficiently co-purifying with the Golgi. Liver is a specialized secretory tissue in which ER lipids are transferred to the Golgi at a high rate, necessitating rapid recycling. In cultured cells, we find the predominant localization of Rab18 to be to the ER, consistent with previous studies showing Rab18 to be associated with ER membranes (Gilchrist et al., 2006; Martin et al., 2005). We propose a role in the regulation of membrane recycling back to the ER. Models consistent with such a role can be easily constructed: e.g. if Rab18 recruits a complex (tethers/SNAREs) to ER membranes required for the targeting and fusion of retrograde carriers generated in the Rab6-regulated COPI-independent retrograde pathway.
In this study, overexpression of Rab18 or of Rab18-Q67L resulted in failure of the model cargo protein VSVG to transit the secretory pathway and in severe disruption of the Golgi complex. Similarly, downregulation experiments with a number of independent siRNAs targeting Rab18, also resulted in reduced VSVG transfer to the cell surface and Golgi disorganization. However, the overexpression of Rab18-S22N not only failed to block VSVG transit to the cell surface but rather selectively upregulated the rate of Golgi-to-ER trafficking of Galtase-YFP, which cycles between the Golgi and ER. The overexpression of wild type Rab18 or its GDP/GTP mutants had little effect on Golgi/ER cycling of p58-YFP, which is believed to use a COPI-dependent pathway. Together these results suggest that one function, at least, of Rab18 is in the regulation of COPI-independent recycling to the ER.
We hypothesize that Rab18 acts in the same COPI-independent pathway as Rab6 (Girod et al., 1999), but plays a complementary role, regulating tethering or fusion (most likely on ER membranes) rather than interaction with microtubule motors. Such a role is also consistent with the recent identification of Rab18 on lipid droplets. Indeed, one role proposed for Rab18, to facilitate the association of ER membranes with lipid droplets (Ozeki et al., 2005), could potentially share common mechanisms (e.g. recruitment of tethering proteins by Rab18) with the interaction of Golgi-derived transport intermediates with the ER.
Since Rab18-S22N potentiates BFA-induced Golgi-ER fusion even in cells in which microtubules have been depolymerized, and an siRNA against Rab18 inhibits this fusion, we hypothesize that the function of Rab18 is in the regulation of tethers/SNAREs (analogous to the role of Rab5 in the endocytic pathway). The apparent positive effects of Rab18-S22N may be due to binding of GTP, although at a lower level than wild-type, leading to a less severe phenotype. However, CFP-Rab18-S22N shows membrane association (Fig. 6D) unlike many analogous Rab mutants, which are efficiently extracted from membranes by Rab-GDI while in the GDP form, and it is thus likely that it possesses residual activity. Furthermore, overexpression of a protein required for Golgi-to-ER trafficking could be expected to disrupt the Golgi complex as has been demonstrated for constitutively active Rab6 (Martinez et al., 1997).
Although many years of classical biochemical and genetic studies have provided an excellent understanding of many aspects of secretory pathway function in mammalian cells, this study highlights the importance of embracing new approaches to further our knowledge of the machinery involved. Our identification of roles for Rab18 and Rab43 in ER/Golgi trafficking has been derived from a combination of detailed proteomic analysis of cellular membranes with live-cell imaging and underlines that further work is still necessary to completely elucidate the secretory pathway machinery.
Materials and Methods
Chemicals and supplies
All chemicals were obtained from Sigma-Aldrich (St Louis, MO) unless otherwise stated. Mouse monoclonal anti-actin (1:1000) and mouse monoclonal anti-acetylated tubulin (1:1000) were from Sigma-Aldrich, rabbit polyclonal anti-beta tubulin (1:1000, Abcam, Cambridge, MA), sheep anti-human TGN46 (1:1000; Serotec, Oxford, UK), mouse anti-TGN38 (Affinity Bioreagent, Golden, CO), mouse monoclonal anti-GM130 (1:300; BD Transduction Laboratories, San Diego, CA), rabbit polyclonal anti-mannosidase II (1:300; Chemicon Temecula, CA). Rabbit anti-COPII (Sec23 subunit; 1:300) was a kind gift of Xiaohui Zha (Ridsdale et al., 2006). Rabbit anti-β-COP (1:200) was a kind gift from J. Lippincott-Schwartz (National Institutes of Health, Bethesda, MD). Rabbit anti-p115 (1:100) was a kind gift from E. Sztul (University of Birmingham, AL). Rabbit anti-calnexin (1:50) was a kind gift from Ali Fazel (Montreal, CA), The anti-VSVG antibody VG was obtained from hybridoma cells previously described (Liebel et al., 2003). Secondary antibodies included goat-anti-rabbit/mouse conjugated to Cy3/Cy5 (Chemicon, Temecula, CA) and donkey anti-sheep IgG antibodies (Alexa Fluor 568; Molecular Probes, Invitrogen, Burlington, ON). Secondary antibodies were used at 1:300 for immunofluorescence, unless specified. Cy3-conjugated chrompure human transferrin (25 μg/ml) was purchased from Jackson ImmunoResearch, Mississauga, ON. Texas Red EGF (100 ng/ml), DAPI and TO-PRO-3 iodide (1:1000) and phalloidin (1:10,000) were obtained from Molecular Probes.
Human cDNA clones containing the open reading frames of GTPases not already GFP tagged were obtained from the RZPD (Berlin, Germany) and GFP tagged as summarized in supplementary material Table S1.
Human specific hairpin siRNA template oligonucleotide, hsiRNA1 (AATCGTGAAGGTCGATAGAAAT) was expressed in the pSilencer Neo™ plasmid (Ambion, Austin, TX). siRNA oligonucleotides siRab18-1 and siRab18-2 were from Ambion (siRNA 120633; 120635). siRab18-3 was from Qiagen (siRNA si02662709). Oligonucleotides were transfected using Oligofectamine (Invitrogen) according to the manufacturer's instructions.
Cell culture and transfection
COS7, HeLa, Normal Rat Kidney (NRK) and Vero cells were grown in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal calf serum, 2 mM glutamine, 150 mg/ml penicillin and 100 U/ml streptomycin (Invitrogen, Burlington, ON). Cells were kept in an incubator at 37°C with 5% CO2.
Cells were grown to 40-80% confluence on glass coverslips in six-well plates and transfected using FuGENE 6 reagent (Roche Diagnostics, Indianapolis, IN) or Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. All plasmids were purified using Qiagen plasmid purification kits (Qiagen, Mississauga, ON).
Immunofluorescence, live-cell imaging and microscopy
Cells were incubated for 24 hours after transfection at 37°C to allow expression of the GFP construct, fixed in 4% paraformaldehyde in PBS (pH 7.2) (PFA/PBS) for 10 minutes, and washed with PBS containing 25 mM glycine (10 minutes), with two subsequent washes with PBS (10 minutes each). Cells were then permeabilized with 0.1% Triton X-100 for 5 minutes then washed three times in PBS and placed in medium containing 3% heat-inactivated normal goat serum (NGS; Invitrogen, Burlington, ON) PBS (NGS/PBS) for 60 minutes. Cells were incubated with primary antibody diluted in NGS/PBS for 1 hour followed by three washes in PBS and subsequent incubation with secondary antibody. Cells were washed three times with PBS, and finally mounted on slides with Geltol mounting medium (Immunon, Shandon, Pittsburgh, PA) for immunofluorescence.
For live imaging, cells were plated onto coverslip bottom dishes (MatTek Inc, Ashland, MA) or LabTek chambers (Nalge Nunc, Rochester, NY). MatTek dishes were maintained during imaging at 37°C using a Zeiss CO2-controlled live-cell chamber. LabTek chambers were maintained at 37°C on the microscope stage using a Nevtek air stream incubator (Nevtek, Williamsville, VA).
Unless otherwise stated, images were taken using a Zeiss LSM510 confocal microscope using a NA 1.4 63× oil-immersion objective lens with the pinhole set at 0.7-1.0 Airy units and slice thickness matched in all channels. GFP and Alexa Fluor 488 were visualized using 488 nm excitation and a BP 505-530 nm emission filter. Cy3 was visualized using 543 nm excitation and a BP 560-615 nm emission filter. Cy5 was visualized using 633 nm excitation and a 650 nm long-pass for emission. CFP fluorescence was excited using 405 nm excitation and a 470-500 nm band-pass emission filter. A 25× Neofluar variable immersion lens (NA 0.8) with fully open pinhole was used for quantitative live cell imaging experiments where all fluorescence in the cell must be accounted for (i.e. all experiments involving photobleach).
Golgi fragmentation was quantified based on published methods (Liebel et al., 2003) and fragmentation normalized to nontransfected cells (1 by definition). To compute Golgi dispersal for a single cell, the average position of the GM130 fragments weighed by fragment intensity (`Golgi centroid') was computed. A radius (r) from the Golgi centroid encompassing sufficient GM130 fragments to account for 50% of the total Golgi intensity was calculated. The area was then calculated (πr2). This area was normalized to Golgi complex in nontransfected cells in the same experiment (1 by definition) to give a relative measure of Golgi dispersal. Absolute cellular fluorescence was calculated in equivalent molecules of fluorescein using published methods (Piston et al., 1999).
Quantification of VSVG trafficking to the cell surface
COS7 cells grown on coverslips were singly transfected with VSVG-YFP/CFP or doubly transfected with VSVG-YFP/CFP and a GFP/CFP-tagged Rab construct and incubated for 24 or 48 hours at 40°C to accumulate VSVG-YFP/CFP in the ER. Cells were then shifted to 32°C for varying times (0, 1 or 3 hours) allowing VSVG to traffic to the cell surface. Samples were subsequently fixed and processed for immunofluorescence as described above using a mouse monoclonal antibody directed against an epitope on the luminal portion of the VSVG-YFP/CFP molecule (VG antibody). Alternatively, cells were stained with the VG antibody (Liebel et al., 2003) as above but without permeabilization. Labeling was subsequently visualized in either case using a Cy3-labeled secondary antibody. Without permeabilization, VG fails to stain intracellular VSVG (Liebel et al., 2003). Total VSVG staining including the ER and Golgi pools can be visualized either by fluorescence of the YFP/CFP tag or by permeabilization prior to staining with VG. 12-bit confocal images were acquired from all samples within a single experiment using identical settings and a 25× NA 0.7 oil-immersion objective with fully open pinhole.
The boundaries of transfected cells were outlined using Metamorph, background corrected using a region of interest outside any cell, and the ratio of VG/YFP fluorescence calculated. This ratio was normalized to 1.0 for control cells transfected with VSVG-YFP alone.
Measurement of rates of trafficking of Galtase-YFP and p58-YFP between ER and Golgi
COS7 cells in MatTek chambers were transfected with Galtase-YFP or p58-YFP alone or in combination with a Rab18 CFP construct (CFP-Rab18 or dominant-negative CFP-Rab18-S22N), incubated for 48 hours and subsequently transferred to a live imaging chamber (Carl Zeiss GmbH, Jena Germany) with 5% CO2 and held at 37°C on the stage of a LSM Zeiss 510 confocal microscope. Photobleach sequences (45 minutes total length) were acquired and recovery curves obtained as described (Presley, 2005) using a 25× NA 0.7 objective with fully open pinhole. The methods described in detail (Presley, 2005) were used to obtain forward and reverse rate constants between the ER and Golgi.
We wish to thank David Stephens for plasmid expressing GFP-p150Glued, Elizabeth Sztul for antibodies directed against p115, Jennifer Lippincott-Schwartz for antibodies directed against COPI, Xiaohui Zha for antibodies directed against COPII, and Ali Fazel for antibodies directed against calnexin. We thank Marieve Picard for her help. We also wish to thank Richard Vo, Charlie Koop and Brigitte Joggerst for technical assistance. We wish to thank Eric Danek, Ryan Petrie, and Arturo Mancini for helpful discussion and critical reading of the manuscript and members of the Bergeron laboratory for pre-publication access to data from MASCOT analysis. Members of the Lamarche-Vane lab provided support and encouragement for which we are grateful. This research was supported by grants from the Canadian Institutes for Health Research (MOP-49590 and PRG-80153) and the National Institutes of Health (R21-GM070588).