Regulation of Golgi structure and function by ARF-like protein 1 (Arl1)

Vesicle-mediated transport plays a fundamental role in the secretory and endocytic pathways and is roughly divided into three steps: vesicle budding/formation mediated by the coat protein complex, which is intimately involved in cargo packaging; vesicle targeting/tethering mediated by specific tethering factors so that vesicles are delivered to the acceptor compartment; and vesicle docking and fusion mediated by the interaction of vesicle-SNARE (v-SNARE) and SNAREs on the target membrane (t-SNAREs) (Palade, 1975; Rothman and Wieland, 1996; Schekman and Orci, 1996; Hong, 1998; Mellman and Warren, 2000; Pelham and Rothman, 2000). GTPases of the heterotrimeric type, as well as members of the Ras-like small GTPase superfamily, participate in various steps of vesicular transport. Among the Ras-like small GTPase superfamily, the Rab/Ypt1/Sec4 family represents the largest family and its members are associated with distinct compartments of the secretory and endocytic pathways (Zerial and McBride, 2001; Novick and Zerial, 1997; Chavrier and Goud, 1999; Nuoffer and Balch, 1994; Schimmoller et al., 1998). The current view is that Rab proteins participate in targeting/tethering of vesicles to acceptor compartments through the action of their effectors, initiating the first layer of interaction between the vesicles and the target compartment and facilitating SNARE interactions (Zerial and McBride, 2001; Chavrier and Goud, 1999; Waters and Pfeffer, 1999).

Various GTPases are involved in vesicle formation. Sar1 regulates protein export from the ER mediated by the COPII coat protein complex (Springer et al., 1999; Aridor et al., 2001). The ADP ribosylation factor (ARF) family includes six highly homologous members (Chavrier and Goud, 1999; Boman and Kahn, 1995; Donaldson and Jackson, 2000; Moss and Vaughan, 1998). The key regulators such as guanine nucleotide exchange factors and GTPase-activating proteins for ARFs have been identified and studied extensively (Chavrier and Goud, 1999; Donaldson and Jackson, 2000; Moss and Vaughan, 1998; Jackson and Casanova, 2000). ARFs were originally identified as cofactors required for cholera-toxin-catalyzed ADP-ribosylation of the stimulatory component of adenylate cyclase Gs (Kahn and Gilman, 1984) and are important for membrane trafficking at several stages of the exocytotic/endocytotic pathway (Chavrier and Goud, 1999; Boman and Kahn, 1995; Donaldson and Jackson, 2000; Moss and Vaughan, 1998; Jackson and Casanova, 2000; Dascher and Balch, 1994; Zhang et al., 1994). ARF1 regulates COPI vesicle budding and is a component of the coat in the early secretory pathway (Serafini et al., 1991; Palmer et al., 1993). Specifically, ARF1 (in its active GTP-bound form) interacts directly with the β subunit of coatomer (Zhao et al., 1997) and also participates in the packaging of cargo proteins into budding vesicles (Malsam et al., 1999; Stephens and Pepperkok, 2001). ARF1 and COPI also function in the endocytic pathway (Daro et al., 1997; Gu and Gruenberg, 2000). In addition, ARF1 has also been shown to participate in vesicle formation mediated by AP-1 (Robinson and Kreis, 1992; Stamnes and Rothman, 1993) and AP-3 coat protein complexes (Ooi et al., 1998) in the trans-Golgi network (TGN) and/or the endosome. ARF6 is distributed within the plasma membrane and endosomal structures and has been implicated in traffic between the surface and the endosomes (D’Souza-Schorey et al., 1995; D’Souza-Schorey et al., 1998; Al-Awar et al., 2000). Phosphatidylinositol 4-phosphate 5-kinase α is a downstream effector of ARF6 (Honda et al., 1999).

In addition to COPI, AP-1 and AP-3, several other effectors have been identified for ARF1, including phospholipase D (PLD1) (Roth, 1999; Cockcroft et al., 1994; Brown et al., 1993), Arfaptin1, POR1/Arfaptin2 (Kanoh et al., 1997; Van Aelst et al., 1996), GGA1-3 (Boman et al., 2000; Dell’Angelica et al., 2000; Hirst et al., 2000) and MKLP1 (Boman et al., 1999). These effectors may interact with different residues of ARF1, and it is expected that more effectors for ARFs remain to be identified (Kuai et al., 2000).

In addition to the six ARFs, there exists a subfamily of small GTPases with homology to ARFs, referred to as ARF-like proteins (Arl). The first member (Arl1) was originally cloned from Drosophila (Tamkun et al., 1991) and is essential for normal development. The mammalian Arl1 was subsequently identified (Schurmann et al., 1994; Lowe et al., 1996) and shown to be associated with the Golgi apparatus (Lowe et al., 1996). In contrast to Drosophila Arl1, yeast Arl1 is not essential for cell growth (Lee et al., 1997). In addition, six other members (Arl2-7) of this subfamily have also been identified (Schurmann et al., 1994; Clark et al., 1993; Cavenagh et al., 1994; Jacobs et al., 1999). Arl2 appears to be cytosolic and has recently been shown to interact with a tubulin-specific chaperone known as cofactor D (Bhamidipati et al., 2000; Radcliffe et al., 2000) as well as Bart, a novel protein of unknown function (Sharer and Kahn, 1999). A recent study suggests that Arl4 is localized to nuclei and nucleoli (Lin et al., 2000). Interaction of Arl6 with Sec61β, a subunit of the core component of ER translocan, was recently reported (Ingley et al., 1999). The exact subcellular localization of Arl3 and Arl5-7 and the function of Arl1 and Arl3-7 have not been established. In this report, we provide morphological and biochemical evidence that Arl1 regulates the structure and function of the Golgi apparatus.

The coding region of rat Arl1 (Lowe et al., 1996) was PCR cloned into the tetracycline-inducible mammalian expression vector pSTAR (Zeng et al., 1998) between the EcoRI and BamHI sites. The Arl1(T31N) and Arl1(Q71L) mutants were generated by standard PCR mutagenesis and cloned into pSTAR. To create the N-terminal EGFP-tagged Arl1 (EGFP-Arl1), Arl1 was first PCR cloned into pEGFP-C1 (Clontech) between the XhoI and BamHI sites. The digestion of the resulting construct by NheI (end blunted) and BamHI gave rise to an insert containing EGFP and Arl1, which was then subcloned into EcoRI-(blunt ended) and BamHI-digested pSTAR. To generate the C-terminal EGFP-tagged Arl1 (Arl1-EGFP), a small linker with multiple cloning sites (MCS) made by annealing two synthetic oligonucleotides (5′-AAT TCG CTA GCG GAT CCG ATA TCG CGG CCG CA-3′ and 5′-GAT CTG CGG CCG CGA TAT CGG ATC CGC TAG CG-3′) was first ligated into EcoRI- and BamHI- digested pSTAR to construct a modified pSTAR vector that has more multiple cloning sites. The BamHI and NotI fragment containing EGFP from pEGFP-N1 vector (Clontech) was then cloned into the modified pSTAR. The coding sequences of wild-type Arl1, Arl1(T31N) and Arl1(Q71L) were also PCR cloned into this EGFP-containing pSTAR vector using EcoRI and BamHI sites to generate C-terminal EGFP-fused Arl1: Arl1-EGFP, Arl1(T31N)-EGFP, Arl1(Q71L)-EGFP and Arl1(G2A, Q71L)-EGFP. All constructs generated by PCR were verified by sequencing and found to be correct.

The Arl1 peptide-specific rabbit polyclonal antibody E6P1 has been previously described (Lowe et al., 1996) and specifically recognizes endogenous rat, hamster and human Arl1 but does not crossreact with ARF1, Arl2 or Arl3. The mouse monoclonal antibody (Mab) against a Golgi SNARE, GS28, has been described (Subramaniam et al., 1996). Mabs against β-COP (maD) and VSV-G (P5D4) were gifts from T. Kreis (Pepperkok et al., 1993). Rabbit polyclonal antibodies against human β-1,4-galactosyltransferase (GT) were described previously (Subramaniam et al., 1992). Mabs against γ-adaptin and α-adaptin were purchased from BD Transduction Laboratories. Sec31 mouse polyclonal antibodies have been described (Tang et al., 2000). The mouse monoclonal anti-ARFs antibody (clone 1D9) (Cavenagh et al., 1996), which recognizes all ARFs but does not react with Arls, was from Affinity Bioreagents, Inc.

A431 and CHO cells were grown in DME and RPMI media, respectively, supplemented with 10% fetal bovine serum at 37°C. The transfection was performed using either Lipofectamine^– (Gibco) for stable transfection of Arl1 and Arl1(Q71L) or Effectene^– Reagent (QIAGEN) for transient transfection of other constructs. The stable transfectants of CHO cells with pSTAR-Arl1 or pSTAR-Arl1(Q71L) were maintained in RPMI supplemented with 10% tetracycline-free fetal bovine serum (Clontech) and 1 mg/ml G418 (Gibco).

Transfected CHO cells were incubated in the presence of doxycycline (Clontech) at 8 μg/ml for 12 hours to induce the expression of transfected constructs. For Brefeldin A treatments, cells were incubated with 10 μg/ml Brefeldin A (Epicentre Technologies) at 37°C for different periods of time before fixation. Cells were washed with PBS containing 1 mM CaCl₂ and 1 mM MgCl₂ (PBSCM) and then fixed with methanol at –20°C for five minutes. Cells were then washed with PBSCM and incubated with antibodies in fluorescence dilution buffer (PBSCM with 5% fetal bovine serum and 2% BSA) for one hour at room temperature. After extensive washing with PBSCM, cells were incubated with the FITC-, Rhodamine- or Texas-red-conjugated secondary antibodies in fluorescence dilution buffer at room temperature for one hour. The cells were then mounted with Vectashield (Vector Laboratories) after washing. Confocal microscopy was performed with a Zeiss Axioplan II microscope equipped with a Bio-Rad MRC1024 confocal scanning laser (Bio-Rad). Typically, images shown are combinations of three to four optical sections 0.3 μm apart.

Arl1-transfected CHO cells were grown on coverslips and induced with 8 μg/ml Doxycycline for about 12 hours. Coverslips were transferred to 3.5 cm diameter petri dishes and washed with serum-free RPMI. 400 μl RPMI (without serum) and an appropriate amount of VSVts045 virus were added to the dish. After incubation at 32°C, plus rocking, for one hour, 1.5 ml RPMI containing 10% fetal bovine serum was added and the cells were incubated for two hours at 32°C without rocking. Cycloheximide (Sigma) was added to a final concentration of 10 μg/ml, and the system was incubated for another hour before processing for indirect immunofluorecence microscopy.

Arl1(Q71L) and (T31N) pSTAR constructs were digested with EcoRI and BamHI and ligated into the EcoRI/BamHI sites of GAL4 DNA binding domain (GAL4-BD) vector pGBKT7 (Clontech). The resulting plasmids were used to transform yeast strain AH109 (MATa). For screening, the Arl1(Q71L) transformed AH109 yeast was mated with a pool of Y187 (MATα) (Clontech), which carried the GAL4 activation domain (GAL4-AD) fused to the human brain cDNA library (Clontech). The interaction-positive diploid yeast cells were selected on SD/-Trp/-Leu/-His/-Ade (QDO) plates. All the positive clones turned blue on X-α-gal (Clontech) QDO plate. The GAL4-AD-cDNA hybrid in pACT2 vector (Clontech) was recovered from the positive yeast cells and sequenced.

The coding sequence of mouse ARF1 (EST clone, GenBank Accession Number: AA266176) was cloned into the EcoRI/BamHI sites of pGBKT7. The corresponding mutants ARF1(T31N) and (Q71L) were generated in pBGKT7 by PCR mutagenesis and subsequently transformed into AH109 yeast cells. The coding region of human GGA1 was derived from EST clone (GenBank Accession Number: BF347585) and cloned into the GAL4-AD fusion vector pGADT7 (Clontech). The coding sequence of POR1 in pACT2 was recovered from one of the positive clones through screening. Mutant Arl1- and ARF1-containing AH109 cells were mated to POR1 and GGA1 transformed Y187 yeast cells, respectively. After selection on SD/-Trp/-Leu medium, the diploid yeast cells were assayed on QDO plates containing 20 μg/ml X-α-gal (Clontech). The positive interacting diploid yeast cells grew and turned blue in three to four days.

To facilitate cloning, the MCS of pGEX-KG (Amersham-Phamacia) was modified to place the EcoRI site in the same reading frame as in pGBKT7 vector. A small linker with EcoRI, NheI and BamHI sites made by the annealing of two oligonucleotides (5′-GATCAGAATTCGCTAGCGGATCCTTAAC-3′ and 5′-AATTGTTAAGGATCCGCTAGCGAATTCT-3′) was ligated with EcoRI- and BamHI-digested pGEX-KG to make a MCS modified vector - pGEB. Arl1(T31N), Arl1(Q71L) and wild-type ARF1 were cloned into pGEB. The recombinant GST fusion protein was produced in Escherichia coli DH5α and affinity purified by Glutathione Sepharose 4B (Amersham-Pharmacia) as previously described (Lowe at al., 1996). The GDP and GTPγS exchange reactions were performed as previously described (Christoforidis and Zerial, 2000). The POR1 pACT2 construct was digested by EcoRI/XhoI (blunt ended) and subsequently ligated into EcoRI/BamHI (blunt ended)-digested pBSK vector (Stratagene). S³⁵-methionine (Amersham-Phamacia)-labeled POR1 and GGA1 were obtained by in vitro translation of POR1/pBSK and GGA1/pGADT7 constructs using the TNT^– T7 Quick Coupled Transcription/Translation System (Promega) according to the manufacturer’s protocol. 60 μg guanine nucleotide-exchanged GST fusion proteins were incubated in Nucleotide Stabilizing Buffer (NS Buffer) containing 2 mg/ml BSA and 100 μM GDP or GTPγS at 4°C for one hour to block the beads before the addition of 10 μl in vitro translated POR1 or GGA1. After over night incubation at 4°C, the beads of each reaction were washed three times with NS buffer. Proteins bound to the beads were eluted by boiling in 2 × SDS sample buffer and resolved by 12% SDS-PAGE. Autoradiography was performed with the PhosphoImager (Molecular Dynamics).

This was performed as described previously (Griffiths, 1993).

Although our previous studies have established that Arl1 is associated with the Golgi apparatus (Lowe et al., 1996), the exact distribution of Arl1 in the Golgi has not been examined ultrastructurally. In our present study, we first established that Arl1 is enriched on the trans side of the Golgi apparatus. Cryosections derived from CHO cells were immunolabeled with Arl1 antibodies followed by protein-A-conjugated gold particles and examined by electron microscopy. As shown (Fig. 1, upper panel), the gold particles are enriched in the Golgi apparatus. Often, the gold particles are preferentially detected on vesicular-tubular profiles on one side of the stack, suggesting that Arl1 is enriched either on the cis or trans side of the Golgi stack. In order to establish this, cryosections were double labeled using antibodies against Arl1 and a monoclonal antibody against γ-adaptin (a component of AP-1 coat), an established marker for the trans side of the Golgi apparatus (Kirchhausen, 1999). Importantly, Arl1 labeling is enriched on the same side of the Golgi stack as is AP-1 (Fig. 1, lower panel). Furthermore, Arl1 and AP-1 seem to be enriched in the same vesicular-tubular structures on the trans side of the Golgi apparatus, the significance of which remains to be investigated.

Consistent with our previous studies using NRK cells (Lowe et al., 1996) and the above immunogold labeling results, perinuclear Golgi localization of Arl1 was detected (Fig. 2A) when CHO cells were examined by indirect immunofluorescence microscopy. Arl1 labeling overlapped well with that of GS28 (Fig. 2B-C), a Golgi SNARE (Subramaniam et al., 1996). When CHO cells were transfected with a construct expressing Arl1 and processed for indirect immunofluorescence microscopy using a limiting amount of Arl1 antibody, Arl1 labeling was detected only in transfected cells (Fig. 2D-F). Golgi labeling was clearly detected when Arl1 expression levels were moderate. When expression was high, the majority of Arl1 was distributed in the cytosol. Similarly, when CHO cells were transfected with a construct expressing Arl1 tagged at its C-terminus with EGFP (Arl1-EGFP), Golgi association was clearly detected in cells expressing moderate levels (Fig. 2G-J). Again, the majority of the expressed protein was distributed in the cytosol when Arl1-EGFP was expressed at high levels (Fig. 2J). These results suggest that exogenous Arl1 is targeted to the Golgi apparatus and that this Golgi association is mediated by a saturable mechanism because overexpressed protein failed to associate with the Golgi. The saturable mechanism could be explained in two possible ways. The first is that Golgi association of Arl1 depends on its activation by nucleotide exchange of GDP for GTP, catalyzed by guanine nucleotide exchange factors, and that this activation step is rate-limiting. Alternatively, the activation step is not rate-limiting but there exists only a limited amount of a membrane receptor for the activated form of Arl1.

Arl1 contains a consensus myristoylation motif (Lowe et al., 1996) and has been shown to be myristoylated (Lee et al., 1997). To examine whether its Golgi association depends on myristoylation, two mutants of Arl1 were created. Arl1(G2A)-EGFP was produced by substituting Gly at position 2 of Arl1, which is essential for all known myristoylation, with Ala in the context of Arl1-EGFP. EGFP-Arl1 was formed by a fusion of the C-terminus of EGFP to the N-terminus of Arl1 so that the myristoylation signal is blocked. As shown in Fig. 3, both Arl1(G2A)-EGFP (Fig. 3A) and EGFP-Arl1 (Fig. 3D) were found in the cytosol and enriched in the nuclei with no detectable labeling of the Golgi marked by GS28, regardless of the expression levels. These results clearly established that N-terminal myristoylation of Arl1 is essential for Golgi association.

To understand the potential involvement of Arl1 in regulating Golgi structure and function, we have created mutant forms of Arl1 that exist either preferentially in the GDP- or GTP-bound forms. The mutant forms were created according to other reported mutants of ras-like small GTPases such as Rabs and ARFs (Chavrier and Goud, 1999; Dascher and Balch, 1994; Zhang et al., 1994). Arl1(T31N) was created by replacing Thr residue at position 31 with Asn and is expected to be restricted to the GDP-bound form and to function as a dominant-negative mutant of Arl1. Arl1(Q71L) was created by replacing Gln at position 71 with Leu and is expected to be restricted mainly to the GTP-bound active form owing to impaired GTP hydrolysis activity. When expressed after transfection, Arl1(T31N) resulted in complete disassembly of the Golgi apparatus marked by GS28 (Fig. 4A-C) and galactosyltransferase (GT) (Fig. 4D-F). This effect is similar to those observed for cells treated with Brefeldin A (Klausner et al., 1992) or cells expressing dominant-negative mutants of ARF1 (Dascher and Balch, 1994) or Sar1 (Storrie et al., 1998). Furthermore, perinuclear Golgi localization of γ-adaptin was also disrupted in cells expressing Arl1(T31N) (Fig. 4G-I), suggesting that Arl1 is either directly or indirectly involved in Golgi recruitment of AP-1. Recent studies suggest that unlike Golgi GT and GS28, Golgi matrix proteins, such as GM130, giantin and p115, remain associated with Golgi-like structures when the Golgi is disassembled by Brefeldin A or mutant Sar1 (Pelletier et al., 2000; Seemann et al., 2000). Consistent with these recent observations, the perinuclear Golgi-like labeling of p115 (Fig. 4J-L) and GM130 (data not shown) was not affected by the overexpression of Arl1(T31N). These results suggest that the expression of Arl1(T31N) also disassembled Golgi such that the Golgi matrix remained intact. As GDP forms of small GTPases are generally regarded as dominant-negative forms that compete with wild-type proteins for guanine nucleotide exchange factors essential for the activation step, the observations can be explained by proposing that the activation as well as the membrane recruitment of endogenous Arl1 was perturbed by the overexpressed Arl1(T31N), thus suggesting a role for Arl1 in maintaining the normal structure of the Golgi apparatus.

As the Golgi apparatus disassembled when activation of endogenous Arl1 was (potentially) affected by exogenous Arl1(T31N), expression of the constitutively active species of Arl1, by restricting it to the GTP-bound, should have the opposite effect. This was indeed the case. Overexpression of Arl1(Q71L) resulted in an expanded perinuclear Golgi-like structure labeled by Arl1(Q71L) (Fig. 5A), and this expanded structure was positive for GS28 (Fig. 5B). Although less potent, expression of a fusion of C-terminus of Arl1(Q71L) with EGFP [Arl1(Q71L)-EGFP] similarly caused a visible expansion of the Golgi-like structure marked by the expressed protein (Fig. 5D) and GS28 (Fig. 5E). Unlike wild-type Arl1 or Arl1-EGFP, Golgi association of Arl1(Q71L) or Arl1(Q71L)-EGFP is not saturable because both were associated with the Golgi regardless of the expression levels. This result indicates that the observed saturable Golgi association of Arl1 is likely to result from a rate-limiting step in its activation. Once activated via guanine nucleotide exchange or point mutation (in the case of Arl1[Q71L] and Arl1[Q71L]-EGFP), Golgi association is no longer rate-limiting or saturable. Consistent with the fact that myristoylation is necessary for Golgi association of wild-type Arl1, the nonsaturable Golgi association of activated form of Arl1 was also dependent on myristoylation, because when the essential Gly at position 2 was replaced by Ala the resulting mutant protein Arl1(G2A, Q71L)-EGFP was not detected in the Golgi (Fig. 5G). Instead, the mutant was distributed in the cytosol and enriched in the nucleus. These results suggest that the Golgi apparatus is under tight regulation by Arl1 activities and that the role of Arl1 in regulating Golgi structure depends on its N-terminal myristoylation.

COPI not only participates in intra-Golgi movement of cargo proteins but also is responsible for controlling the export of traffic destined for retrograde transport back to the ER (Pelham and Rothman, 2000). Retrograde export occurs predominantly on the cis face of the Golgi. On the trans face of the Golgi, AP-1 participates in exporting cargo proteins to post-Golgi structures such as the lysosome via the endosome (Kirchhausen, 1999). The functions of COPI and AP-1 are both under the tight regulation of ARF1, whose activation depends on Brefeldin-A-sensitive guanine nucleotide exchange factors in many cell types (Helms and Rothman, 1992; Donaldson et al., 1992). In this regard, Golgi association of COPI and AP-1 and the formation of their respective vesicles are quickly inhibited (within five minutes) when cells are treated with Brefeldin A (Klausner et al., 1992). Previous studies have shown that expression of the activated form of ARF1 could confer Golgi association of COPI and AP-1 in a manner that is no longer sensitive to Brefeldin A (Dascher and Balch, 1994; Zhang et al., 1994). Since expression of the activated form of Arl1 resulted in an expanded Golgi-like structure, we have examined the effect of Arl1(Q71L) on the distribution of COPI and AP-1. As shown, the expression of Arl1(Q71L) caused a massive recruitment of both COPI (β-COP) (Fig. 6) and AP-1 (γ-adaptin) (Fig. 7) onto the expanded Golgi. Furthermore, the recruited COPI and AP-1 could not be dissociated by Brefeldin A treatment, even during prolonged treatments of up to 30 minutes (Fig. 6, Fig. 7). This observation is interesting because ARF1 has been established to be the key direct regulator for both COPI- and AP-1-mediated vesicle formation. Therefore, it is surprising to note that the activated form of Arl1 could similarly result in a massive recruitment of both COPI and AP-1 coat proteins in a Brefeldin-A-resistant manner. Several possibilities can explain these observations. The first is that the activated form of Arl1 independently drives the recruitment of COPI and AP-1. Alternatively, Arl1 acts together with or through ARF1 in mediating the recruitment of COPI and AP-1. In order to examine these possibilities, we checked the effect of overexpressed Arl1(Q71L) on the distribution of Golgi-localized ARFs recognized by monoclonal antibody 1D9 (Cavenagh et al., 1996). As shown in Fig. 8, 1D9 labels the Golgi region of CHO cells (Fig. 8B). In cells overexpressing Arl1(Q71L), extensive association of ARFs with the expanded Golgi-like structure was observed (Fig. 8A-C), suggesting that Arl1(Q71L) could stimulate Golgi recruitment of ARFs. ARFs redistribute from the Golgi to the cytosol within two minutes of Brefeldin A treatment (Klausner et al., 1992). When cells were treated with Brefeldin A for five minutes, ARFs were dissociated from the Golgi and distributed in the cytosol in cells not expressing Arl1(Q71L) (Fig. 8E). However, ARFs remained Golgi associated in Brefeldin-A-treated cells overexpressing Arl1(Q71L) (Fig. 8D-F). Association of ARFs with the expanded Golgi could still be observed after 30 minutes of treatment with Brefeldin A (Fig. 8G-I). These observations suggest that activated Arl1 [Arl1(Q71L)] can stimulate activation and Golgi recruitment of ARFs in a Brefeldin-A-resistant manner and this could be the molecular basis for the observed massive recruitment of COPI and AP-1 in cells overexpressing Arl1(Q71L).

In marked contrast to COPI and AP-1, which mediate export from the Golgi, the distributions of COPII (Sec31), which mediates ER export, and AP-2 (α-adaptin), which mediates endocytosis from the plasma membrane, were not affected by the overexpression of Arl1(Q71L) (data not shown). These results suggest that Arl1 regulates membrane traffic selectively at the Golgi apparatus.

We next examined the effect of expression of Arl1(Q71L) on protein trafficking along the secretory pathway. Specifically, we analyzed the transport of the envelope protein (VSV-G) of VSVts045. Cells were infected with VSVts045 at 32°C for one hour and then incubated for another two hours to allow sufficient VSV-G protein to be produced, followed by an additional one hour incubation in the presence of cycloheximide to chase VSV-G along the pathway. As shown (Fig. 9B,E,H), VSV-G was almost completely transported to the cell surface in cells not expressing Arl1(Q71L). However, the transport of VSV-G to the surface was essentially abolished in cells expressing Arl1(Q71L) (Fig. 9A-C), and VSV-G now accumulated in the expanded Golgi-like structure marked by Arl1(Q71L). These results suggest that export of VSV-G from the ER mediated by COPII coat proteins occurred normally, although subsequent transport through the Golgi and/or from the Golgi to the surface was blocked. Similarly, VSV-G transport was arrested upon overexpression of Arl1(Q71L)-EGFP (Fig. 9D-F), although its effect was not as potent as that of Arl1(Q71L). The effect of the activated form of Arl1 on VSV-G transport was also dependent on myristoylation because VSV-G was transported normally to the cell surface in cells overexpressing Arl1(G2A, Q71L)-EGFP (Fig. 9G-I).

To gain a more detailed understanding of the effects of overexpressed Arl1(Q71L) on the organization of the Golgi, we examined the ultrastructure of the Golgi apparatus by immunogold labeling at the electron microscopy level. Cells were infected with VSVts045 at 32°C for one hour, followed by a two hour incubation in the absence of cycloheximide and a further one hour incubation in the presence of cycloheximide so that VSV-G is transported to the expanded Golgi-like structure in cells overexpressing Arl1(Q71L). Cryosections derived from these cells were then double labeled with antibodies against Arl1 and VSV-G to reveal the expanded structures (Fig. 10). In contrast to highly ordered stacks of Golgi cisternae observed in control CHO cells, the stacked cisternae of Arl1(Q71L) expressing cells, although still visible, were significantly disorganized. The cisternae became more dilated and expanded in cells expressing moderate levels of Arl1(Q71L) (Fig. 10, upper panel). Both Arl1(Q71L) and VSV-G (arrowheads) were detected in the dilated Golgi. Strikingly, the Golgi apparatus became transformed into an extensive vesicule-tubule network in cells expressing high levels of Arl1(Q71L) (Fig. 10, lower panel). Arl1(Q71L) decorated the entire vesicle-tubule network (smaller arrowheads) and VSV-G (larger arrowheads) was seen to be trapped in the network. Vacuole-like structures (V) were often seen, and Arl1(Q71L) was also detected on vacuolar membranes. This result indicates that the organization of the Golgi apparatus into an ordered stack of cisternae depends on the proper function of Arl1. As Arl1 is enriched in the vesicular-tubular profiles on the trans side of the Golgi, the observed Golgi transformation could be due to massive unproductive formation of vesicular-tubular profiles caused by activated Arl1(Q71L), leading to the formation of a tangled vesicle-tubule network.

To understand the molecular mechanism of Arl1 in regulating Golgi structure and function, we performed a yeast two-hybrid screen for proteins interacting with activated Arl1(Q71L). Initial characterizations (Fig. 11A) of 18 interacting clones revealed the presence of POR1/arfapatin-2, Golgin-97 and pericentrin as well as three unknown genes. POR1/arfapatin-2 has been shown to interact with the activated forms of Rac1, ARF1, ARF3 and ARF6 (Kanoh et al., 1997; Van Aelst et al., 1996; D’Souza-Schorey et al., 1997). We verified the interaction between Arl1 and POR1 (Fig. 11B). POR1 interacted only with the GTP-form of Arl1 and had no affinity for the GDP-form, a result similar to that observed for ARF1. Under identical conditions, the activated form of ARF1, but not Arl1, interacted with GGA-1, suggesting that GGA-1 is a specific effector for ARF1, whereas POR1 is an effector shared by several GTPases, including Arl1 and ARF1. Furthermore, in vitro translated POR1 was retained very efficiently by immobilized GST-Arl1 in its GTP-bound but not in its GDP-bound form (Fig. 11C). These results suggest that the GTP-bound form Arl1 could interact specifically with POR1. The specificity and significance of Arl1 interaction with other proteins are being investigated.

In our present study, we have investigated the subcellular localization of Arl1 using immunogold labeling of cryosections derived from CHO cells. We have shown that the majority of Arl1 is enriched in vesicular-tubular structures on one side of the Golgi apparatus. Double-labeling with γ-adaptin revealed that Arl1 is enriched on the side that is marked by γ-adaptin, suggesting that Arl1 is enriched on the trans side of the Golgi. As ARFs have been implicated in regulating traffic in the cis-Golgi, trans-Golgi and endosome (Serafini et al., 1991; Palmer et al., 1993; Zhao et al., 1997; Malsam et al., 1999; Stephens and Pepperkok, 2001; Daro et al., 1997; Gu and Gruenberg, 2000; Robinson and Kreis, 1992; Stamnes and Rothman, 1993), they may have a global role in regulating membrane traffic. On the basis of its known functions, ARF1 may not be restricted to particular subcompartments of the Golgi apparatus, although earlier localization studies suggested that ARF1 is enriched in the cis-Golgi (Stearns et al., 1990). The enrichment of Arl1 on the trans-side of the Golgi is of interest because it may suggest that one of its key functions may be to regulate traffic around the trans-side of the Golgi, although it remains possible that it may also regulate other traffic linked to the function of the Golgi apparatus. The molecular basis for Golgi association of Arl1 was thus examined, and we found that Golgi association of Arl1 is mediated by a saturable mechanism in intact cells. The basis for this saturable recruitment is likely to be a rate-limiting step in the activation (GDP-GTP exchange) of Arl1 catalyzed by guanine nucleotide exchange factors, because Golgi association of activated forms Arl1(Q71L) and Arl1(Q71L)-EGFP is not saturable. Furthermore, Golgi association of Arl1 depends on its N-terminal myristoylation and this is true not only for the saturable recruitment of wild-type Arl1 but also for non-saturable association of activated Arl1(Q71L). This notion is further supported by the fact that the inhibitory effect of Arl1(Q71L)-EGFP on VSV-G transport is abolished when the conserved Gly at position 2 responsible for myristoylation is replaced by Ala. We thus conclude that a concerted action of the N-terminal myristoylation and its activation via guanine nucleotide exchange is the underlying mechanism for Golgi association of Arl1 and that the activation represents the rate-limiting step for regulating its activity.

The functional importance of Arl1 was established by the expression of two different mutant versions of Arl1. The Golgi apparatus (marked by GT and GS28) disassembled when Arl1(T31N), which was expected to be restricted to the GDP-bound status, was expressed. Since the rate-limiting step in Arl1 Golgi recruitment is the activation step mediated by guanine nucleotide exchange factors, the likely explanation for the observed effect of Arl1(T31N) is that it competes with endogenous Arl1 for the limiting amount of guanine nucleotide exchange factors, resulting in an inhibition of activation and Golgi recruitment of endogenous Arl1. The failure to recruit Arl1 onto the Golgi may also lead to defective Golgi association of its putative effectors, which, together with Arl1, play a key role in maintaining the normal structure and function of the Golgi apparatus. On the other hand, expression of Arl1(Q71L), expected to be restricted to the GTP-bound activated form, led to an expansion of the Golgi apparatus. The ordered stacks of the Golgi were transformed into an extensive vesicular-tubular network upon expression at high levels. Although the molecular mechanism remains to be established, it is plausible that the uncontrolled Golgi association of Arl1(Q71L) leads to a massive recruitment of its effectors onto the Golgi and an alteration of the normal structure of the stacked cisternae of the Golgi apparatus. As the ordered stacked cisternae are believed to facilitate efficient traffic through this compartment and to govern effective sorting on both the cis as well on the trans side of the Golgi, the transformation of Golgi into an extensive vesicular-tubular network may interfere with the normal function and lead to an inhibition of protein transport through the compartment and/or to inhibition of export from the Golgi. In support of this interpretation, our experiments showed that transport of VSV-G from the ER to the cell surface was arrested at the level of the expanded structure upon overexpression of Arl1(Q71L). The accumulation of VSV-G protein in this vesicular-tubular network indeed indicates that transport through or export from the altered Golgi is blocked. On the basis of these results, we propose that one of the mechanisms for the cell to regulate the size and the stacked cisternae structure of the Golgi apparatus is by controlling the activation and Golgi recruitment of Arl1 and its effectors.

The massive recruitment of COPI and AP-1 onto the expanded Golgi apparatus upon overexpression of Arl1(Q71L) is of interest because a similar effect was observed upon expression of ARF1(Q71L) (Dascher and Balch, 1994; Zhang et al., 1994). As COPI and AP-1 are redistributed from the Golgi to the cytosol within two minutes of Brefeldin A treatment, whereas Golgi association of Arl1 remains intact after five minutes of Brefeldin A treatment (Lowe et al., 1996), it is unlikely that Golgi recruitment of COPI and AP-1 are directly regulated by Arl1. How do we explain the remarkable similarity of the effects observed for Arl1(Q71L) as compared to those observed for ARF1(Q71L) with regards to Golgi recruitment of COPI and AP-1? One possibility is that Arl1 and ARF1 share some common effectors in addition to effectors that are unique to each of them. Upon expression of the activated form of Arl1 and its Golgi recruitment, its effectors are subsequently recruited onto the Golgi membrane. As it is known that effectors for ARF1 have a strong positive feedback on ARF1 GDP-GTP exchange and activation (Zhu et al., 2000), some of the effectors recruited by Arl1(Q71L) will effect positive regulation of ARF1, resulting in activation and Golgi recruitment of ARF1 followed by its unique set of effectors.

POR1/Arfaptin2 is an effector of ARF1/3 (Kanoh et al., 1997) as well as activated Rac1 (Van Aelst et al., 1996). Our yeast two-hybrid screening using Arl1(Q711) as the bait retrieved POR1/Arfaptin-2. Using the yeast two-hybrid assay, it was shown that POR1/Arfaptin-2 interacted with GTP- but not GDP-bound forms of both Arl1 and ARF1. In addition, POR1/Arfaptin-2 could be retained by an immobilized, activated form of Arl1 and ARF1 but not by their respective GDP-bound forms. However, the activated form of Arl1 did not interact with GGA-1, suggesting that GGA-1 is a specific effector for ARF1. These results suggest that POR1/Arfaptin-2 is a common effector shared by Arl1, ARF1, ARF3 and Rac1. POR1/arfaptin-2 potently increase the affinity of ARF1 and ARF3 for GTP in vitro. The stoichiometry of GTP binding to ARF can be increased from about 0.05 mol/mol to almost 0.5 mol/mol of recombinant ARF. The observed effects of Arl1(Q71L) on massive COPI and AP-1 Golgi association could be potentially mediated by POR1/arfaptin-2 or other effectors that lead to activation and Golgi recruitment of endogenous ARF1/3. In support of this hypothesis, expression of Arl1(Q71L) also resulted in massive recruitment of ARFs onto the Golgi in a Brefeldin-A-resistant manner. Although it might be due to an indirect effect of Golgi disassembly, the observed dissociation of AP-1 from the perinuclear Golgi upon expression of Arl1(T31N) could result from indirect inactivation of Golgi-localized ARFs. Our results thus highlight a potentially important mechanism for cross-talk among different small GTPases in regulating membrane traffic.

During the preparation of this manuscript, a report from Kahn’s laboratory (Van Valkenburgh et al., 2001) appeared. It was different from our results in that they show that the Golgi localization of HA-tagged Arl1 is quickly disrupted within three minutes of Brefeldin A treatment, and HA-tagged Arl1(Q71L) is only partially resistant to Brefeldin A treatment, such that it also dissociates from Golgi within five minutes. The discrepancy is possiblly due to the tag that we think may interfere with their Arl1 function, as we observed that the C-terminal fusion of EGFP to Arl1 mutants makes them less potent than non-tagged versions in affecting Golgi structure and function. In their study, Van Valkenburgh et al. also uncovered an interaction of Arl1 with POR1/arfaptin-2 as well as with other proteins such as Golgin-245 and SCOCO, highlighting the point that Arl1 and ARF1 may have some common effectors in addition to effectors that are unique to each of them. Our unique establishment of Arl1 enrichment in the vesicular-tubular structures in the trans-Golgi, the demonstration that Golgi recruitment of Arl1 depends on both activation by guanine nucleotide exchange as well as on N-terminal myristoylation, and our investigation of the effects of various mutant forms of Arl1 on subcellular distribution of COPI, AP-1, and ARFs, and on the Golgi structure and function underscore the importance of regulation of Golgi structure and function by Arl1.

We thank the late T. Kreis for antibodies against β-COP and VSV-G, members of Hong’s laboratory for critical reading of the manuscript, and Y. H. Tan for his support. This work was funded by the Institute of Molecular and Cell Biology (to W.H.).