Inactivation of Gα_z causes disassembly of the Golgi apparatus

Heterotrimeric G proteins are classically known to function as signal transducers on the plasma membrane (Kaziro et al., 1991; Neer,1995; Hamm, 1998). Each G protein consists of an α subunit that binds a guanine nucleotide,and a βγ complex. The interaction of a G protein with an activated receptor triggers guanine nucleotide exchange on the α subunit, leading to its dissociation from the βγ dimer. The activatedα-subunit and βγ dimer act on appropriate downstream effectors. Individual subunits of G proteins consist of multiple species with different tissue distributions. Gα family members are classified into four subfamilies, α_s, α_i,α _q and α_12/13, and each member is coupled with different receptors to mediate different signals(Simon et al., 1991).

Recently, the functions of heterotrimeric G proteins in various endomembrane systems have also been recognized(Lang, 1999). These include protein transport from the endoplasmic reticulum (ER) to the Golgi apparatus(Schwaninger et al., 1992),intra-Golgi transport (Helms et al.,1998), cargo sorting and vesicle budding from the trans-Golgi network (Leyte et al., 1992; Pimplikar and Simons, 1993), control of exocytosis(Ohara-Imaizumi et al., 1992;Aridor et al., 1993;Vitale et al., 1993;Ohnishi et al., 1997), and autophagic sequestration (Ogier-Denis et al., 1996).

In mammalian cells the Golgi apparatus is of fundamental importance for the secretion and posttranslational modification of secretory and membrane proteins. The Golgi apparatus consists of a set of highly dynamic membrane compartments that are maintained through the continuous anterograde and retrograde flow of proteins and lipids. The Golgi apparatus undergoes disassembly and reassembly during mitosis, and there is a cell cycle-dependent mechanism that controls the Golgi structure and thus ensures the fidelity and reliability on the partitioning of this organelle into daughter cells(Warren and Malhotra, 1998). Golgi disassembly at the onset of mitosis may be controlled by signaling proteins at the periphery of Golgi membranes(Nelson, 2000).

We and others have previously shown the involvement of heterotrimeric G proteins in regulation of the Golgi structure(Hidalgo et al., 1995;Jamora et al., 1997;Jamora et al., 1999;Yamaguchi et al., 1997;Yamaguchi et al., 2000). However, it is currently unclear as to which subunit of heterotrimeric G proteins is involved in this regulation, and little is known about the mechanism by which G proteins exert their actions. A study involving a semi-intact cell system demonstrated that Gβγ causes vesiculation of the Golgi apparatus through direct activation of protein kinase D(Jamora et al., 1999). On the other hand, we showed that the activation of G proteins by GTPγS or AlF₄^- protects the Golgi apparatus from disassembly caused by nordihydroguaiaretic acid (NDGA). Consistent with the involvement of Gα-mediated signaling in Golgi organization, the addition of Gβγ can reverse the effect of GTPγS on NDGA-induced Golgi disassembly through reformation of the heterotrimer(Yamaguchi et al., 1997). Furthermore, overexpression of Gα_i2 or Gα_z,both of which are α_i family members, results in attenuation of the NDGA effect (Yamaguchi et al.,2000). Although our data comprise circumstantial evidence that Gα_i2 and/or Gα₂ are involved in Golgi organization, it is not clear whether these G proteins actually regulate the Golgi structure in normal living cells.

Recently, a novel family of G protein regulators, i.e. regulators of G protein signaling (RGS) proteins, has emerged(Berman and Gilman, 1998;Hepler, 1999;De Vries et al., 2000). RGS proteins stimulate the intrinsic GTPase activity of heterotrimeric G proteins,thereby decreasing the concentration of active Gα in cells. In addition to this effect, RGS proteins prevent the binding of Gα to their effectors. Therefore, RGS proteins serve as negative regulators of G protein-mediated signaling pathways (Berman and Gilman, 1998; De Vries et al., 2000).

RGS proteins appear to be ideal tools for demonstrating the physiological relevance of our finding that active Gα subunits prevent NDGA-induced Golgi disassembly. In this study, we overexpressed various RGS proteins that preferentially inactivate a different set of G proteins, and examined the Golgi morphology. The results suggest that, indeed, active Gα_z, but not Gα_i2, is necessary for maintenance of the integrity of the Golgi structure.

The mouse monoclonal anti-tetra-His antibody was from Qiagen. The mouse monoclonal anti-FLAG antibody (M2) and rabbit polyclonal anti-FLAG antibody were obtained from Sigma. The mouse monoclonal anti-mannosidase II (Man II)antibody (53FC3) was purchased from BabCo. The mouse monoclonal anti-GM130,anti-p115, and anti-membrin antibodies were purchased from Transduction Laboratories. The rabbit polyclonal anti-βCOP and anti-syntaxin 5 antibodies were raised in this laboratory. The rabbit polyclonal anti-transferrin antibody was from DAKO. The rabbit anti-NADH-cytochrome P-450 reductase antibody and mouse monoclonal anti-ERGIC-53 antibody were generous gifts from Dr Akitsugu Yamamoto (Kansai Medical University) and Dr Hans-Peter Hauri (University of Basel), respectively. The FITC-conjugated anti-mouse and anti-rabbit IgG antibodies were from Rockland and Chemicon International,respectively. The Texas red-conjugated anti-mouse and anti-rabbit IgG antibodies were from Kirkegaard & Perry Laboratories and Southern Biotechnology Associates, respectively.

The cDNAs for RGS4, GAIP, RGSZ1 and RGSZ2 were kindly donated by Drs. Elliott Ross and Tohru Kozasa (University of Texas Southwestern Medical Center), and subcloned into mammalian expression vectors pcDNA3 (Invitrogen)and pFLAG-CMV-6 (Sigma). The full-length cDNA encoding human RGS2 was amplified from a human leukocyte cDNA library (Clontech) by means of the polymerase chain reaction (PCR) with primers corresponding to the initiation and termination sites of RGS2. To obtain the full-length cDNA encoding RGS3,nucleotides corresponding to the N-terminal and C-terminal regions were independently amplified from the human cDNA library by PCR, and then cloned into pFLAG-CMV-6. The expression plasmid for the vesicular stomatitis virus ts045 G protein fused to the green fluorescent protein (VSVG-GFP) was a generous gift from Dr Jennifer Lippincott-Schwartz (National Institutes of Health, USA).

A rat cDNA containing the entire coding sequence of Gα_zwas cloned into a eukaryotic expression vector, pALTER-MAX (Promega). Gα_z(G204A/E246A/A327S), in which Gly-204, Glu246, and Ala-327 of Gα_z were replaced with Ala, Ala, and Ser, respectively,was generated by oligonucleotide-directed mutagenesis using an Altered Sites II Mammalian Mutagenesis System (Promega). A Gα binding deficient mutant of RGSZ1, RGSZ1(E116A/N117A), in which Glu-116 and Asn-117 were individually substituted by Ala, was prepared by PCR-mediated mutagenesis.

HeLa, BHK, and Clone9 cells were cultured at 37°C in α-minimum essential medium supplemented with 10% fetal bovine serum, 50 IU/ml penicillin, and 50 μg/ml streptomycin, under humidified air containing 5%CO₂. PC12 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 50 IU/ml penicillin, 50 μg/ml streptomycin, 10%fetal bovine serum, and 5% horse serum. Cells were transfected by lipofection using LipofectAMINE PLUS according to the manufacturer's instructions (Life Technologies).

Indirect immunofluorescence microscopy was performed essentially as described previously (Tagaya et al.,1996). Cells were grown on 15-mm diameter glass coverslips in 12-well tissue culture plates. At the indicated times after transfection,cells were subjected to fixation in 4% paraformaldehyde in phosphate-buffered saline (PBS) for 20 minutes at room temperature, followed by permeabilization with 0.2% Triton X-100 in PBS for 20 minutes. They were then incubated for 30 minutes in PBS containing 2% bovine serum albumin (BSA), and incubated for 1 hour at 37°C with appropriate primary antibodies in PBS-2% BSA. The cells were washed three times with PBS, and then stained with FITC- and/or Texas red-conjugated secondary antibodies. An Olympus BX50 fluorescence microscope was used for routine immunofluorescence analysis. Confocal images were obtained with an Olympus Fluoview 300 laser confocal microscope. For the transferrin internalization assay, transfected cells were incubated in the presence of 25 μg/ml of FITC-conjugated transferrin (Sigma) for 1 hour at 37°C. After extensive washing in PBS, they were fixed with 4%paraformaldehyde in PBS. Since the fluorescence intensity of FITC-conjugated transferrin was weak, the fixed cells were further stained with the anti-transferrin rabbit polyclonal antibody and the FITC-conjugated secondary antibody.

Total RNA was isolated from cells using isogen reagent (Nippon Gene) and then used as the template for RT-PCR analysis. RT-PCR reactions were performed with a BcaBEST RNA PCR kit (Takara). PCR was performed for 45 and 20 cycles for Gα_z and glyceraldehyde-3-phosphate dehydrogenase (GAPDH),respectively, using the following sense and antisense primers:Gα_z, 5′-CACCTGGAGGACAACGCCGCT-3′ and 5′-TTTCGGTTTAGGTCCTCGAACTGA-3′ (500-bp product); GAPDH,5′-CATGGAGAAGGCTGGGGCTC-3′ and 5′-CTCAGTGTAGCCCAGGATGC-3′ (523-bp product), respectively. The number of cycles was pre-determined to fall within the linear range of amplification of each PCR product. The PCR products were electrophoresed on 1.6% agarose gels, stained with ethidium bromide, and visualized by UV irradiation.

We previously demonstrated that overexpression of Gα_z or Gα_i2 prevents NDGA-induced Golgi disassembly(Yamaguchi et al., 2000). To address the physiological significance of this finding, we examined the effect of overexpression of RGS proteins on the structure of the Golgi apparatus.

RGS proteins can attenuate G-protein mediated signaling pathways by acting as GAPs for the Gα_i, Gα_q, and Gα_12/13 families, but not the Gα_s one. To date, more than 20 RGS proteins in mammalian tissues have been identified. Certain RGS proteins exhibit high selectivity for a single or a few Gαspecies (Berman and Gilman,1998; De Vries et al.,2000). RGSZ1 (Glick et al.,1998; Wang et al.,1998) and RGSZ2 show high selectivity for Gα_z.

By immunofluorescence microscopy, we first examined the localization of a Golgi marker β-COP in HeLa cells expressing N-terminal hexahistidine-tagged (6×His-) RGSZ1. Although nontransfected cells showed a typical ribbon-like Golgi structure, a dispersed or punctateβ-COP staining pattern was frequently observed in cells overexpressing 6×His-RGSZ1 (Fig. 1). The degree of β-COP dispersal was dependent on the level of RGSZ1 expression. When RGSZ1 was expressed at high levels, β-COP was almost completely dispersed (Fig. 1a). On the other hand, expression of RGSZ1 at moderate levels yielded a punctate or large dot-like β-COP staining pattern (Fig. 1c), which might represent fragmented Golgi. Approximately 50% of the cells expressing a detectable level of RGSZ1 showed a completely or partially disrupted β-COP staining pattern, whereas less than 7% of the nontransfected ones did. The latter value may represent the percentage of cells in the late G₂ and M phases. It is known that the Golgi ribbon-like structure is reorganized into a punctate structure and exhibits a more perinuclear localization at the G₂/M transition, and then is completely dispersed (Shima et al.,1998) or transported back to the ER(Zaal et al., 1999) during mitosis. The effect of the RGS proteins on the Golgi structure is not peculiar to HeLa cells, similar results being obtained for BHK and Clone9 cells (data not shown).

We next examined whether or not other RGS proteins exert similar effects on the Golgi apparatus. RGSZ2 is another Gα_z-specific RGS that exhibits 62.7% amino acid identity with RGSZ1. GAIP, which is most similar to RGSZ1 (62.6% amino acid identity) and RGSZ2 (57.6% amino acid identity),exhibits GAP activity toward all Gα_i family members(De Vries et al., 1995;Berman et al., 1996). RGS2 exhibits high selectivity for Gα_q(Heximer et al., 1997). RGS3 is highly selective for the Gα_i family except for Gα_z (Scheschonka et al.,2000). RGS4 stimulates the GTPase activity of both Gα_i and Gα_q(Berman et al., 1996;Watson et al., 1996). These RGS proteins were expressed as FLAG-tagged proteins in HeLa cells, and then the Golgi morphology was examined. As shown in(Fig. 2A, expression of RGSZ2 as well as that of RGSZ1 induced the dispersal of β-COP, whereas expression of GAIP, RGS2, RGS3, or RGS4 had essentially no effects onβ-COP staining. The number of cells expressing a detectable level of RGS proteins was determined, and the percentage of cells with a disruptedβ-COP staining pattern was determined(Fig. 2B). More than 30% of the FLAG-RGSZ1- or FLAG-RGSZ2- expressing cells showed a dispersed β-COP staining pattern, whereas no significant dispersal of β-COP was observed for other RGS-expressing cells compared to control cells. These results suggest that the dispersal of β-COP by RGSZ1 and RGSZ2 is specific.

Since β-COP is a peripheral membrane protein that undergoes association and disassociation with the Golgi apparatus in an ARF1-dependent manner (Donaldson et al., 1992;Palmer et al., 1993), the results shown in Figs 1 and2 may imply that the expression of RGSZ1 merely induces the release of β-COP from Golgi membranes. To test this possibility, we examined the localization of several Golgi-associated peripheral and integral membrane proteins in RGSZ1-expressing cells. As shown in Fig. 3a-d,peripheral Golgi proteins GM130 and p115 showed scattered staining patterns and were distributed as distinct dots around the nucleus in RGSZ1-expressing cells. As shown in Fig. 3e-h,cis-Golgi integral membrane proteins membrin and syntaxin 5 exhibited similar fragmented patterns in RGSZ1-expressing cells. Man II, a Golgi-resident integral membrane protein, also showed a scattered distribution(Fig. 3i-l). These results demonstrate that RGSZ1 does not merely induce the release of peripheral proteins from the Golgi membrane, but also causes the fragmentation of entire Golgi stacks.

We wondered whether or not the fragmentation of the Golgi apparatus is a consequence of the disassembly of microtubules, which are known to play an important role in maintaining the Golgi structure(Lippincott-Schwartz, 1998;Thyberg and Moskalewski,1999). As shown in Fig. 4a,b, expression of RGSZ1 had no effect on the structure of microtubules. In addition, expression of RGSZ1 did not conspicuously affect the ER structure in most cells (Fig. 4c,d), although the formation of punctate aggregations was observed in some cells (data not shown). RGSZ1 expression also did not affect the uptake of FITC-transferrin (Fig. 4e,f). These results demonstrate that overexpression of RGSZ1 affects the Golgi structure without marked effects on other cellular structures or functions.

We next examined whether or not RGSZ1 expression perturbs anterograde vesicular transport. For this purpose, a temperature-sensitive VSVG mutant protein fused to GFP (VSVG-GFP) (Presley et al., 1997) was used. HeLa cells were co-transfected with plasmids for VSVG-GFP and FLAG-RGSZ1 and incubated at 40°C for 20 hours,and then the temperature was changed to 32°C to initiate transport from the ER. At 1 hour after the temperature shift, VSVG-GFP had moved from the ER to the plasma membrane through the Golgi in 80% of the control cells(Fig. 5). In RGSZ1-expressing cells, the transport of VSVG-GFP was significantly delayed. At 1 hour after the temperature shift, VSVG-GFP had reached the plasma membrane in 40% of the RGSZ1-expressing cells. In the other RGSZ1-expressing cells, VSVG-GFP was detected in dot-like structures that might represent fragmented Golgi membranes. Essentially the same results were obtained for BHK and Vero cells.

The inhibitory effect of RGSZ1 on the transport of VSVG-GFP might be underestimated. We noticed that the extent of Golgi dispersion was lower in cells expressing both RGSZ1 and VSVG-GFP than in those expressing RGSZ1 alone. In many cells, fragmented, large dot-like Golgi structures rather than completely dispersed ones were observed. It seemed likely that VSVG-GFP was transported to the plasma membrane through the perinuclear dot-like structures marked by Man II (Fig. 6a-c),β-COP (Fig. 6d-f), and ERGIC-53 (Fig. 6g-i). We speculate that the transport of VSVG-GFP would be severely inhibited if the Golgi apparatus were completely dispersed by high expression of RGSZ1.

The RGS family is defined by the RGS-box that binds to Gα subunits and is responsible for the GAP function. A point mutation in the RGS-box of RGS4 abrogates its ability to bind to Gα and therefore is not able to inactivate Gα (Druey and Kehrl,1997; Srinivasa et al.,1998). To determine whether or not the GAP activity of RGSZ1 is required for Golgi dispersion, RGSZ1(E116A/N117A), in which Glu-116 and Asn-117 substituted individually by Ala, was constructed and assessed as to its Golgi disassembling activity. This mutation was designed in analogy to the RGS4 mutation (Srinivasa et al.,1998). A similar mutation has been successfully used for assessment of the function of RGS3(Scheschonka et al.,2000).

We first tested whether RGSZ1(E116A/N117A), as expected, cannot bind to Gα_z. The wild-type or mutated FLAG-RGSZ1 was co-expressed in cells with an active mutant of Gα_z, Gα_z(QL),in which Gln-205 was replaced with Leu(Fig. 7A). The wild-type RGSZ1,which mainly remains in the cytosol (Fig. 3), was efficiently recruited to the plasma and internal membranes, and co-localized with expressed Gα_z(QL),indicating the interaction of RGSZ1 with Gα_z in cells. In contrast, the mutated RGSZ1 remained in the cytosol and was not colocalized with expressed Gα_z(QL). This indicates that the mutated RGSZ1 we constructed has the expected functional property in cells.

As shown in Fig. 7B, cells expressing wild-type RGSZ1 showed a dispersed β-COP staining pattern(Fig. 7Ba,b). In contrast,RGSZ1(E116A/N117A) had no detectable effect on the distribution of β-COP(Fig. 7Bc,d). The level of expression of the RGSZ1 mutant was comparable to that of the wild-type one(Fig. 7C). Similar results were obtained when Clone9 cells were used (data not shown).

As an alternative approach for investigating the effect of attenuation of the Gα_z function, we employed a dominant-negative mutant strategy. Several mutants possessing mutations in the conserved nucleotide-binding region of GTPases including the α-subunits of heterotrimeric G proteins have been characterized. The mutant we constructed was a triple one (G204A/E246A/A327S), in which Gly-204, Glu-246, and Ala-327 were replaced with Ala, Ala, and Ser, respectively. This mutant is equivalent to a triple mutant of Gα_s, which has decreased guanine nucleotide-binding ability, and dominantly inhibits receptor-mediated hormonal activation of Gα_s by sequestering Gβγ and activated receptors (Iiri et al.,1999).

The triple Gα_z mutant was not co-localized with RGSZ1 when co-expressed, suggesting that the Gα_z mutant is not in a GTP-bound active state in cells (data not shown). The triple Gα_z mutant plasmid was transfected into HeLa cells, and after 20 hours the cells were processed for immunofluorescence(Fig. 8). The triple Gα_z mutant was mainly distributed in intracellular membranous structures including those in the perinuclear region. Golgi protein membrin was dispersed in 39% of the cells expressing the triple Gα_zmutant, whereas it was not dispersed in wild-type Gα_z-expressing cells. Taken together, the effects of overexpression of RGS proteins and the dominant-negative Gα_zmutant strongly suggest a critical role of active Gα_z in maintenance of the Golgi structure.

Previous studies showed that the expression of Gα_z is limited primarily to platelets, neurons and chromaffin cells, suggesting specific roles in these tissues (Matsuoka et al., 1988; Casey et al.,1990; Hinton et al.,1990). If Gα_z is involved in the organization of the Golgi apparatus, it should be expressed ubiquitously. To demonstrate the expression of Gα_z and to determine its localization in non-neuronal cultured cells, we performed indirect immunofluorescence microscopic analysis. Immunoreactivity to the anti-Gα_zantibody was mainly observed in the perinuclear region, with some in the plasma membranes of BHK and Clone9 cells(Fig. 9a,c). The perinuclear structure positive for the anti-Gα_z antibody most likely corresponds to the Golgi apparatus, because it was also stained by medial Golgi marker Man II (Fig. 9b,d). The staining with the anti-Gα_z antibody is specific because it was totally abolished when the antibody was preincubated with the peptide used for immunization(Fig. 9e,g). Similar Golgi labeling patterns for Gα_z were detected for other cells including PC12, Chinese hamster ovary, and NRK cells (data not shown). To confirm that the perinuclear Gα_z staining reflects the Golgi structure, we treated cells with brefeldin A (BFA). BFA is known to cause the redistribution of Golgi-resident proteins to the ER(Klausner et al., 1992). The perinuclear Gα_z staining as well as the Golgi marker Man II staining was dispersed upon the treatment of cells with BFA(Fig. 9i-l), suggesting that endogenous Gα_z is located in the Golgi apparatus.

Expression of Gα_z was also assessed by semi-quantitative RT-PCR analysis (Fig. 10). mRNA for Gα_z was expressed in non-neuronal Clone9 and NRK cells, although the neuron-like PC12 cells showed much higher expression. The control GAPDH mRNA level was approximately the same in these samples. The ubiquitous expression of Gα_z is consistent with the recent observation that Gα_z is detectable in various tissues of mouse (Hendry et al.,2000).

We and others previously demonstrated that the Golgi apparatus is disassembled by NDGA (Yamaguchi et al.,1997; Fujiwara et al.,1998; Drecktrah et al.,1998), and that this process is blocked by the addition of various G protein activators (Yamaguchi et al.,1997). Furthermore, overexpression of Gα_i2 or Gα_z attenuates NDGA-induced Golgi dispersal(Yamaguchi et al., 2000). However, these results were obtained under artificial conditions in which the Golgi apparatus was perturbed by NDGA. Therefore, the physiological implication of these findings remains to be examined. Here, we addressed the possibility that Gα_z is involved in the organization of the Golgi apparatus by examining the effects of RGS proteins and a dominant-negative Gα_z mutant on the morphology of the Golgi apparatus.

When Gα_z-selective GAPs such as RGSZ1(Glick et al., 1998;Wang et al., 1998) and RGSZ2 were expressed, various Golgi-resident proteins including β-COP and Man II showed dispersed distribution patterns in several types of cells. In contrast, the structures of the ER and microtubules were not markedly affected by the expression of these proteins. No significant Golgi disassembly was observed when RGS proteins other than RGSZ1 and RGSZ2 were expressed. These results suggest that the inactivation of Gα_z but not other Gα proteins, such as Gα_i2, induces Golgi disassembly. Chatterjee and Fisher demonstrated that RGSZ1, when expressed in COS-7 cells,is localized in the Golgi apparatus(Chatterjee and Fisher, 2000). In our experiments, expressed RGSZ1 was mainly distributed throughout the cytoplasm and nucleus. This discrepancy can be partly explained by the use of different cell lines. We also observed Golgi-like perinuclear localization of exogenously expressed RGSZ1 in COS-7 cells. However, similar to in other cell lines, a dispersed pattern of a Golgi marker β-COP was observed in cells expressing RGSZ1 at a high level (data not shown).

The idea that inactivation of Gα_z induces disassembly of the Golgi apparatus was supported by another finding with the use of a dominant-negative mutant, Gα_z(G204A/E246A/A327S). This mutant was designed in analogy with a dominant-negative Gα_s mutant in which three conserved residues are simultaneously mutated(Iiri et al., 1999). Since this type of mutant remains in a guanine nucleotide-free form, it can inhibit Gα-mediated signaling, probably by occupying activated receptors. As expected, overexpression of the triple Gα_z mutant induced Golgi disassembly.

Although Gα_z is formally a member of the G_ifamily, it possesses several unique biochemical properties distinct from those of other Gα_i members(Fields and Casey, 1997). One unique character of Gα_z is its insensitivity to pertussis toxin-mediated ADP-ribosylation (Casey et al., 1990), a modification that inactivates other members of the G_i family. In addition to general G-protein activators such as GTPγS and AlF₄-, mastparan, a peptide that selectively activates the G_i family, also blocked the NDGA-induced Golgi disassembly (data not shown). In fact, the overexpression of either Gα_z or Gα_i2, both of which belong to the G_i family, has a similar protective effect against NDGA-induced Golgi disassembly (Yamaguchi et al.,2000). However, treatment of cells with pertussis toxin did not affect the Golgi morphology (data not shown), suggesting the involvement of a pertussis toxin-insensitive Gα, i.e. Gα_z, in the maintenance of the Golgi structure. The fact that RGS3 and RGS4, both of which can attenuate Gα_i activity, lack Golgi disassembly activity is consistent with this idea. Although overexpression of both Gα_i2 and Gα_z can block NDGA-induced Golgi disassembly, Gα_z may be involved in the maintenance of Golgi structure under physiological conditions.

Based on the results reported here, we envisage a mechanism by which Gα_z controls the structure of the Golgi apparatus. When Gα_z binds GTP, it activates a signaling cascade that is required for maintenance of the Golgi structure. When Gα_z is inactivated upon the hydrolysis of bound GTP, the Golgi structure undergoes disassembly as a consequence of loss of the signaling. Thus,Gα_z functions as a molecular switch that organizes the Golgi structure. This hypothesis predicts that Gα_z preferentially binds GTP, thereby being constitutively active, in interphase cells. From this point of view, Gα_z seems to be a favorable G protein because it exhibits a very slow intrinsic rate of GTP hydrolysis. Its k cat value is 200-fold lower than those of Gα_s and Gα_i (Casey et al.,1990).

Investigation of the mechanism by which Gα_z maintains the Golgi structure is a future challenge. The Golgi disassembly induced by inactivation of Gα_z does not appear to involve fast release of coat proteins (data not shown), as seen in cells treated with BFA. In addition, expression of RGSZ1 does not affect microtubule organization. Gα_z may act on Golgi stacking proteins or factors that can link this organelle to the cytoskeletal elements. However, our results do not exclude the possibility that Gα_z may control the export of secretory and Golgi-resident proteins from the ER. The transport of VSVG-GFP was partly inhibited in RGSZ1-expressing cells. Although our antibody failed to detect it, a portion of Gα_z may exist in the ER.

Our present view that active Gα_z is required for maintenance of the Golgi apparatus does not necessarily contradict the finding by Malhotra and colleagues that the βγ dimer causes Golgi disassembly (Jamora et al.,1997; Jamora et al.,1999). A similar situation was observed for trafficking from the ER to the Golgi apparatus. The addition of βγ dimer, which leads to inactivation of active Gα by shifting the equilibrium toward the formation of the trimeric complex, inhibits the formation of vesicles from the ER in semi-intact cells. Mastparan, an activator for heterotrimeric G proteins, also blocks vesicle formation(Schwaninger et al., 1992). The presence of multispecies of G proteins in organelles may account for these intricate and apparently inconsistent observations. It should be noted that the expression of Gα_z in non-neuronal cells is not high. The amount of the βγ dimer released as a consequence of the activation of Gα_z must be very small, and therefore may not be enough to cause Golgi disassembly. On the other hand, activation of major G proteins may yield a large amount of the βγ dimer, which results in disassembly of the Golgi apparatus. G_i family proteins may be more widely involved in the regulation of organelle structures in early secretory pathways. Wang et al. reported that treatment of rat hepatocytes with pertussis toxin caused the redistribution and fragmentation of the ER(Wang et al., 2000). Thus,different G proteins may be involved in more intricate cellular functions than presently assumed.

This work was supported in part by Grants-in Aid (#10215205 and #11480183)from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and by the Uehara Memorial Foundation. We thank E. Ross and T. Kozasa for providing the cDNAs for RGSZ1, RGSZ2, RGS4, and GAIP, H. Itoh for the Gα_z cDNA, A. Yamamoto for the anti-NADH cytochrome P-450 reductase antibody, H.-P. Hauri for the anti-ERGIC-53 antibody, and J. Lippincott-Schwartz for the VSVG-GFP construct. We also thank M. Anbiru, K. Miki, K. Yoshida and K. Yamazoe for their excellent technical assistance.