Arl1p is involved in transport of the GPI-anchored protein Gas1p from the late Golgi to the plasma membrane

Arf-like protein 1 (Arl1p), a member of the ARF-like protein subfamily of small GTPases, is localized in the late Golgi compartment in both mammalian cells and yeast (Lu et al., 2001; Setty et al., 2003; Liu et al., 2005), and has been implicated in the regulation of Golgi structure and function (Lu et al., 2001). In yeast and mammals, Arl1p can recruit the GRIP-domain containing effectors, Imh1p or Golgin-245 and Golgin-97, to the Golgi membrane (Lu and Hong, 2003; Panic et al., 2003; Setty et al., 2003). Mammalian Arl1 and GRIP-domain proteins, Golgin-245 or Golgin-97, have recently been shown to regulate membrane trafficking between the TGN and endosomal system (Lu et al., 2004; Yoshino et al., 2005). Studies also indicated that ARL1 has genetic interaction with YPT6 (Bensen et al., 2001; Tong et al., 2004). Ypt6p, a small GTPase, is associated with the Golgi and required for the fusion of endosome-derived vesicles with the late Golgi (Lewis et al., 2000; Siniossoglou et al., 2000). Overexpression of Arl1p complements temperature sensitivity of a ypt6 mutation, and a double deletion of arl1 and ypt6 in yeast is lethal (Bensen et al., 2001; Tong et al., 2004). Therefore, Arl1p was hypothesized to regulate membrane trafficking between the TGN and endosomal system and may have redundant function with Ypt6p in bringing endosome-derived retrograde vesicles to the TGN (Graham, 2004). These data are consistent with the conclusion that Arl1p functions in transport pathways in the TGN; however, no endogenous cargo of Arl1p-derived vesicles has yet been identified in yeast.

Recently, Arl3p was found to act upstream of Arl1p and regulate membrane binding of Arl1p, which, in turn, regulates Golgi localization of Imh1p (Jackson, 2003; Panic et al., 2003; Setty et al., 2003). This regulation was hypothesized to result from the recruitment of Arl1p's guanine nucleotide exchange factor (GEF) by Arl3p (Jackson, 2003). In addition, the association of Arl3p with the Golgi membrane is dependent upon the acetylation of Arl3p at its N-terminus by an acetyltransferase and a Golgi-localized integral membrane protein, Sys1p (Behnia et al., 2004; Setty et al., 2004). Thus, a signaling cascade of Sys1p-Arl3p-Arl1p-Imh1p was proposed to regulate trafficking events in the TGN (Graham, 2004; Setty et al., 2004). Moreover, Gcs1p was identified as a GTPase-activating protein (GAP) for Arl1p that regulates Golgi association and activity of Arl1p (Liu et al., 2005). Hence, both Arl3p and Gcs1p regulate Arl1p activity.

Glycosylphosphatidylinositol (GPI)-anchored proteins are a diverse class of proteins that are anchored to the membrane by a post-translational lipid modification, the GPI-moiety. Previous studies of GPI-anchored protein trafficking have focused on transport between the endoplasmic reticulum (ER) and Golgi, and sorting and endocytosis at the plasma membrane (Muniz et al., 2000; Chatterjee and Mayor, 2001; Morsomme et al., 2003; Mayor and Riezman, 2004). By contrast, little is known about the molecular mechanisms by which GPI-anchored proteins move from the TGN to the plasma membrane. Gas1p, a GPI-anchored protein, represents the major cell-surface glycoprotein in yeast. In a Sar1p-dependent process involving Gas1p interacting with a cargo receptor, Emp24p complex, Gas1p is delivered from the ER through COPII vesicles to the Golgi complex (Barlowe et al., 1994; Doering and Schekman, 1996; Muniz et al., 2000). Furthermore, Gas1p is sorted and packaged separately from non-GPI-anchored proteins, such as Gap1p, in the ER membrane (Muniz et al., 2000; Muniz et al., 2001). Since GPI-anchored proteins are localized in the membrane microdomains, lipid rafts (Kubler et al., 1996; Bagnat et al., 2000; Bagnat and Simons, 2002) and membrane microdomains are required for GPI-anchored protein transport in the ER (Skrzypek et al., 1997; Sutterlin et al., 1997), we speculated that the transport of GPI-anchored proteins and non-GPI anchored proteins from the TGN to the plasma membrane is regulated by different sorting and packaging machinery (Simons and van Meer, 1988), which is similar to the separation from the ER. Since a Sar1p/SNARE-based vesicle sorting mechanism has been established for the exit of GPI-anchored proteins from the ER to the Golgi complex, it is possible that GPI-anchored proteins also use an ARF-family-member-based mechanism to traffic from the TGN. In this study, we show that Arl1p and its signaling cascade Sys1p-Arl3p-Arl1p-Imh1p are involved in the transport from the late Golgi to the plasma membrane of the GPI-anchored protein Gas1p, but not the cell-wall-localized, GPI-anchored proteins Crh1p, Crh2p and Cwp1p, or the non-GPI-anchored protein Gap1p.

Snc1p is a v-SNARE protein that recycles between the Golgi and plasma membrane, and its endosome-Golgi recycling requires the presence of Ypt6p (Lewis et al., 2000; Siniossoglou et al., 2000). To address whether Arl1p functions in a pathway parallel to the Ypt6p, we examined the recycling of Snc1p in arl1-null cells. In ypt6-deleted cells, most of the GFP-Snc1p was localized intracellularly, with loss of its plasma membrane and polarized distribution as seen in wild-type cells (Fig. 1A). In an arl1 mutant, however, unlike the findings in the ypt6 mutant, GFP-Snc1p was localized to the plasma membrane and polarized to the daughter cells, suggesting that Arl1p is not required for the recycling of GFP-Snc1p from the Golgi to the plasma membrane or the fusion of GFP-Snc1p vesicles from the endosomal system to the Golgi. These data indicate that Arl1p and Ypt6p function in different pathways in the TGN.

Transport of carboxypeptidase Y (CPY) and alkaline phosphatase (ALP) from the TGN to the vacuole occurs by two well-characterized pathways (Cowles et al., 1997). Arl1p and Arl3p were reported to be involved in the transport of CPY, but not ALP, from the Golgi compartment (Bonangelino et al., 2002). In our studies, there was no significant defect in an arl1 mutant in transport of CPY or ALP from the TGN to vacuole at 30°C and 37°C (our unpublished data) (Huang et al., 2002) (Fig. 1B,C), although, the transport of CPY and ALP in an arl1 mutant exhibited a moderate delay at 16°C. These data suggest that Arl1p is not directly involved in the transport of these vacuolar enzymes from the TGN to vacuoles.

We next examined the transport pathways towards the periplasmic region and plasma membrane. To observe the transport from the TGN to the periplasmic region, the secretion of biologically active mature α-factor was followed in a halo assay. No significant difference between wild-type and arl1-mutant cells was observed (data not shown), indicating that Arl1p is not required for α-factor secretion to the periplasmic region.

Plasma membrane proteins can be divided into lipid-raft associated or non-lipid-raft associated proteins; these two kinds of surface proteins might have distinct transport pathways from the TGN. To examine whether Arl1p is involved in a specific transport of plasma membrane proteins, we examined the distribution of Gas1p- and Gap1p-derived GFP-tagged proteins, GFP-GPI (Gas1p) and Gap1^K9K16-GFP, expressed in arl1-deleted yeast. GFP-GPI (Gas1p) is derived from Gas1p, a GPI-anchored protein and the major cell surface glycoprotein in yeast. Gap1p is a general amino-acid permease that does not possess a GPI anchor and does not localize to lipid rafts (Soetens et al., 2001; Bagnat and Simons, 2002). The Gap1^K9K16 mutation abolishes its ubiquitylation, thereby preventing its sorting to the vacuole; thus, the protein remains fully stable at the plasma membrane (Soetens et al., 2001). We found that GFP-GPI (Gas1p) accumulated in Golgi-like structures in arl1-null yeast, but no intracellular accumulation of Gap1^K9K16-GFP in the arl1-mutant cells was observed (Fig. 2A). We also examined the distribution of these plasma membrane proteins in arl3- and gcs1-mutant yeast. Similar to the arl1-mutant cells, in both the arl3- and gcs1-mutant yeast only the raft-associated GFP-GPI (Gas1p), but not Gap1^K9K16-GFP, showed intracellular accumulation (Fig. 2A). These data suggest that the Arl3p-Arl1p pathway plays a role in the sorting of GPI-anchored proteins to the plasma membrane.

It is generally thought that yeast contain two groups of GPI-anchored proteins: the first is comprised of plasma membrane-resident proteins such as Gas1p, whereas the second contains proteins such as Crh1p, Crh2p and Cwp1p, which are covalently linked to the cell wall and have a role in cell wall maintenance (De Sampaio et al., 1999; Hamada et al., 1998; Dielbandhoesing et al., 1998; Rodriguez-Pena et al., 2000; Rodriguez-Pena et al., 2002). To determine whether Arl1p regulates both groups of GPI-anchored proteins, we next examined the location of GFP-tagged Crh1p, Crh2p, Cwp1p and mRFP-tagged Gas1p in arl1-mutant and wild-type cells. All of these cell-wall-localized GPI-anchored proteins had proper location: cell-surface targeting and growing site polarization, in both arl1-mutant and wild-type cells (Fig. 2B). However, mRFP-Gas1p lost its peripheral distribution in the arl1 mutant. We further determined the subcellular localization of mRFP-Gas1p in the arl1 mutant. We found that the mRFP-Gas1p accumulated in the arl1 mutant and partially colocalized with the late Golgi marker GFP-Sft2p (Fig. 2C). These data indicate that Arl1p is specifically involved in plasma-membrane-resident GPI-anchored protein transport pathway towards the cell surface.

We further investigated whether endogenous Gas1p transport is indeed regulated by Arl1p. Gas1p is a β-1,3-glucanosyltransferase that is important in cell wall integrity (Kopecka and Gabriel, 1992; Mouyna et al., 2000). Disruption of GAS1 caused defects in morphogenesis and cell wall integrity, leading to hypersensitivity to Calcofluor White and Congo Red. Congo Red binds to β-1,3-glucan, a component of the yeast cell wall, and interferes with its assembly into a multi-stranded ribbon in the cell wall. Mutants with cell wall defects show hypersensitivity to Congo Red (Kopecka and Gabriel, 1992; Tomishige et al., 2003). Therefore, we first examined whether an arl1 mutant was hypersensitive to Congo Red. Fig. 3A shows that arf1, arl1, arl3, and gcs1 mutants were all hypersensitive to Congo Red. The hypersensitivity to Congo Red of these mutant yeasts could be rescued by an osmolarity adjustment (addition of 1.2 M sorbitol), indicating that the phenomenon of hypersensitivity to Congo Red in these mutants resulted from a defect in cell wall integrity (Fig. 3A). Arf3p, which is localized at the plasma membrane and is involved in polarity development (Huang et al., 2003), is not involved in hypersensitivity to Congo Red, indicating that not all deletions of Arf family members cause hypersensitivity against Congo Red.

Gas1p is glycosylated in the Golgi, resulting in a ∼125-kDa mature form from a ∼105-kDa ER form. The accumulation of mature Gas1p inside the cell would indicate a transport defect beyond the Golgi. Next, we used a plasma membrane proteinase shaving assay (Morsomme et al., 2003) to quantify the ratio of plasma-membrane-localized Gas1p to mature intracellular Gas1p. Yeast spheroplasts were prepared and treated with proteinase K to remove cell-surface Gas1p and residual Gas1p was detected by western blotting. As shown in Fig. 3B,D, the ratio of intracellular Gas1p to that on the surface in arl1-, gcs1- and arl3-mutant cells is higher than that in wild-type cells (42%, 52% and 45% compared with the 10-11% in wild-type cells), indicating that a higher percentage of intracellular Gas1p was observed in arl1-, gcs1- and arl3-mutant cells than in wild-type cells.

To further demonstrate the role of Arl1p in anterograde transport, we used pulse-chase experiments to track the transport of de-novo-synthesized Gas1p from the TGN to the plasma membrane (Morsomme et al., 2003). After a 10-minute pulse and 80-minute chase, approximately 90% of Gas1p in wild-type cells reached the plasma membrane, whereas in arl1-deleted cells ∼50% of Gas1p remained intracellularly (Fig. 3C,E). This intracellular accumulation was also observed in arf1-, arl3- and gcs1-null cells. These results demonstrate that the anterograde transport of de-novo-synthesized Gas1p requires the presence of Arl1p. Accumulation of glycosylated-Gas1p combined with accumulation of GFP-GPI (Gas1p) and mRFP-Gas1 in the late Golgi compartment (Fig. 2) suggests that Arl1p and its regulators (Arl3p and Gcs1p) are involved in Gas1p transport from the late-Golgi to the plasma membrane. Moreover, although arf1-mutant cells accumulate similar levels of Gas1p as arl1-mutant cells (Fig. 3B-E), they are more sensitive to Congo Red (Fig. 3A). Arf1p is the main Arf family member and is important for Golgi-to-ER and intra-Golgi transport. Because deletion of ARF1 leads to transport defects in many transport pathways in yeast (Gaynor et al., 1998; Yahara et al., 2001), we infer that for many cell-wall-integrity-related proteins trafficking is affected in the arf1 mutant.

Sys1p is a Golgi-localized integral membrane protein that serves as a receptor for acetylated Arl3p (Behnia et al., 2004; Setty et al., 2004). Since Arl3p is required for Golgi localization of Arl1p, the deletion of SYS1 abolishes Golgi targeting and function of Arl1p (Behnia et al., 2004; Setty et al., 2004). Thus, we examined whether Sys1p is also required for the transport of Gas1p. Like the arl1 mutant, a sys1-null mutant produced Congo Red hypersensitivity to the same extent as an arl3 deletion (Fig. 4A). GFP-GPI (Gas1p), but not Gap1^K9K16-GFP, accumulated in Golgi-like punctate structures in the sys1 mutant (Fig. 4B). The arf3 mutant serves as a control for this experiment because GFP-GPI (Gas1p) can properly localize to the plasma membrane in both the arf3-mutant and wild-type cells. These results are consistent with the current model for a Sys1p-Arl3p-Arl1p cascade in TGN-vesicular transport. The cascade of Arl3p-Arl1p to regulate Gas1p transport was further confirmed with Congo Red hypersensitivity analysis (supplementary material Fig. S1). Although the expression of wild-type Arl1p only partially rescued the Congo Red hypersensitivity of the arl3 mutant, the constitutive active Arl1Q72L fully rescued this hypersensitivity of the arl3 mutant. This observation supports the notion that Arl1p is downstream of Arl3p and the Arl3p-Arl1p cascade functions in Gas1p transport.

To examine whether the subcellular distribution of endogenous Gas1p was affected in the arl1 mutant, we first fractionated spheroplast-homogenized lysates into two fractions, plasma-membrane-rich (P13) and microsome-Golgi-rich and soluble fraction (S13), by sedimentation centrifugation. Fig. 5A shows that the level of Gas1p in the microsome-Golgi-rich and soluble fraction (S13) in the arl1 mutant is considerably higher than in the wild-type cells. We further investigated this by sucrose-density-gradient centrifugation to separate and analyze the fraction S13. Gas1p, Drs2p (late Golgi marker), Emp47p (early Golgi marker), Pgk1p (cytosol marker) and Pep12p (endosome marker) in each fraction were identified by western blot analysis and quantified (Fig. 5B). Although in the arl1 mutant, some Gas1p distributed at lighter fractions (2-6 fractions), most of the Gas1p was at fractions 7-9 and distribution pattern of Gas1p was more similar to that of the late Golgi protein Drs2p (Fig. 5B lower panel). These data indicate that, in the arl1 mutant, endogenous Gas1p mainly accumulated at the Golgi compartment and most likely at the late Golgi.

We next examined whether an effector of Arl1p, Imh1p, is involved in Gas1p transport. In the Congo Red hypersensitivity assay, imh1-null cells did not show a growth defect compared with the growth of arl1 mutant yeast on YPD and Congo Red plates (Fig. 6A). However, it did cause a mild intracellular accumulation of Gas1p (Fig. 6B,C), suggesting that Imh1p is involved in Gas1p transport.

Although anterograde transport of Gas1p is affected in the arl1 mutant, it is possible that this is an indirect effect resulting from a defect in retrograde transport (Graham, 2004; Lu et al., 2004). Therefore, we examined whether ypt6-mutant cells exhibit a similar defect as seen with the arl1 mutant. The ypt6-null cells did not show hypersensitivity to Congo Red when the growth of ypt6-null cells on YPD and Congo Red plates are compared (Fig. 6A). Furthermore, ypt6-null yeast did not exhibit an obvious defect in Gas1p transport in the plasma membrane proteinase shaving assay (Fig. 6B,C), indicating that the deficiency of retrograde transport mediated by Ypt6p did not cause a Gas1p transport delay.

To examine the effect of Arl1p activity on Gas1p transport, we introduced wild-type, constitutively active (Arl1Q72L), and constitutively inactive (Arl1T32N) Arl1p into arl1-null or wild-type cells. Ectopically expressed Arl1p, but not Arl1Q72L or Arl1T32N, repressed the Congo Red hypersensitivity observed in arl1 cells (Fig. 7A). In wild-type cells, overexpression of Arl1T32N, but not Arl1Q72L, caused hypersensitivity to Congo Red (Fig. 7B). The inability of constitutively active Arl1Q72L to complement the loss of endogenous Arl1p was also observed in our previous report (Liu et al., 2005). This result is consistent with the observation that a gcs1 deletion also disturbs Gas1p transport (Fig. 2A and Fig. 3). We next examined the localization of GFP-GPI (Gas1p) in wild-type or arl1-null cells overexpressing different forms of Arl1p. Consistent with Fig. 7A, in the arl1 mutant, only wild-type Arl1p could rescue the plasma membrane targeting of GFP-GPI. The expression of Arl1Q72L even caused GFP-GPI (Gas1p) accumulation in enlarged Golgi structures (Fig. 7C, upper panel). In wild-type cells, expression of Arl1T32N caused GFP-GPI (Gas1p) to accumulate in punctate structures in the absence of plasma membrane signals (Fig. 7C lower panel). Similar to Arl1T32N, expression of Arl1Q72L also caused accumulation of some GFP-GPI (Gas1p) in Golgi-like and unknown structures. However, it did not appear to thoroughly block the transport of GFP-GPI (Gas1p) to the plasma membrane. This result suggests that although Arl1Q72L is unable to hydrolyze GTP for proper vesicle trafficking in wild-type yeast, endogenous Arl1p still sustains basal level transport of its cargo molecules. Together, our data indicate that the transport of Gas1p needs functional regulation of Arl1p involving GTP binding, Golgi targeting and GTP hydrolysis.

Although Arl1Q72L could not restore the Congo Red hypersensitive phenotype in an arl1 mutant (Fig. 7A), it interestingly could rescue hypersensitivity in an arl3 mutant (supplementary material Fig. S1). Therefore, we asked what is different regarding the effect of Arl1Q72L in arl3 and arl1 mutant. Consistent with previous findings (Panic et al., 2003), we observed that endogenous Arl1p or ectopically expressed Arl1p lost its Golgi localization in the arl3 mutant (Fig. 8A). Our previous report (Liu et al., 2005) and Fig. 8B (upper panel) showed that Arl1Q72L caused the enlargement of late Golgi when expressed in ARL3 wild-type cells. However, Arl1Q72L maintained the Golgi-like distribution in the arl3 mutant (Fig. 8A) and these Golgi-like structures appeared normal both in size and number. We next examined the late Golgi structure of Arl1Q72L expressed in an arl3 mutant by co-expressing GFP-Sft2p and Arl1Q72L in an arl3 mutant (Fig. 8B). The Arl1Q72L partially colocalized with the late Golgi marker Sft2p in an arl3 mutant and, unlike that in ARL3 cells, did not cause enlargement of the Golgi. These results suggest that the proper location of Arl1Q72L and its normal morphology at the late Golgi in an arl3 mutant is correlated to its function in suppressing the Congo Red hypersensitivity in the arl3 mutant.

Here, we report that Arl1p participates in a class of GPI-anchored protein trafficking from the late Golgi to the plasma membrane. This is the first identification of an endogenous cargo molecule of Arl1p in yeast. Our study also supports the notion that regulators of Arl1p, such as Sys1p, Arl3p and Gcs1p, and effector Imh1p, all participate in the transport of Gas1p. The transport of Gas1p requires functional regulation of Arl1p involving GTP binding, Golgi targeting and GTP hydrolysis. Based on our findings, we conclude that the signaling cascade Sys1p-Arl3p-Arl1p-Imh1p facilitates the transport of a subset of GPI-anchored proteins from the late Golgi to the plasma membrane.

Our results indicate that Arl1p plays a role in the anterograde transport of Gas1p from the late Golgi to the plasma membrane, and rule out a redundant function of Arl1p with Ypt6p on the retrograde transport. Arl1p, Gcs1p and Arl3p (our unpublished data) are not required for Snc1p recycling that is mediated by Ypt6p. Although we could not exclude the possibility that Arl1p is involved in retrograde transport from the endosomal system to the TGN for cargo proteins other than Snc1p, we demonstrated that Ypt6p and Arl1p do not mediate the same cargo protein trafficking. Therefore, the genetic interaction and synthetic lethality of ARL1 and YPT6 might not come from redundant functions in bringing endosome-derived retrograde vesicles to the TGN. However, the observation that deletion of YPT6 did not retard the transport of Gas1p or show Congo Red hypersensitivity further supports the hypothesis that Arl1p and Ypt6p have distinct functions.

Although Arl1p and Arl3p were reported to be involved in the transport of CPY from the Golgi compartment (Bonangelino et al., 2002), our observation and another report have indicated that CPY is processed with kinetics similar to wild-type cells, but with moderate delay or mis-sorting to the extracellular medium in an arl1 mutant (Lee et al., 1997; Jochum et al., 2002). The partial secretion of CPY (Jochum et al., 2002; Rosenwald et al., 2002) and mild delay in CPY and ALP processing (Fig. 1B,C) suggest that Arl1p is not directly involved in vacuolar protein transport. The discrepancy between different reports of CPY processing in arl1-mutant cells might result from differing extents of indirect effects produced from Golgi function deficiencies, although the use of yeast strains with different genetic background may also be responsible for these differences.

Initially, we showed that Arl1p is required for GFP-tagged, GPI-anchored protein transport, but not non-GPI-anchored protein transport, from the Golgi to the plasma membrane. To analyze endogenous Gas1p distribution, we used immunofluorescent staining experiments to try to prove that endogenous Gas1p accumulates in the Golgi of an arl1 mutant. However, no apparent signal was observed for Gas1p using an anti-Gas1p antibody. Therefore, we used two biochemical methods, the plasma membrane proteinase shaving assay and subcellular fractionation, to monitor endogenous Gas1p distribution.

Although it has been reported that GPI-anchored proteins in mammalian cells are delivered to the plasma membrane by a trans-cytotic pathway and are not segregated with non-GPI-anchored proteins when they exit the TGN (Polishchuk et al., 2004), the transport of Gas1p regulated by Arl1p in yeast reveals a segregated pathway for GPI-anchored proteins from non-GPI-anchored proteins in the Golgi apparatus. In addition, our results are consistent with the recent report (Paladino et al., 2006) that sorting of GPI-anchored proteins occurs intracellularly before arrival at the plasma membrane. Our study further demonstrates that only the plasma-membrane-resident GPI-anchored protein Gas1p, but not cell-wall-localized GPI-anchored proteins, require Arl1p to facilitate transport from the Golgi to the cell surface. Consistent with the concept of cargo sorting in transport pathways (Muniz et al., 2000; Muniz et al., 2001), we conclude that different classes of GPI and non-GPI-anchored proteins would be sorted into different vesicles from the Golgi to the plasma membrane.

Using Gas1p distribution as a cargo marker, we further demonstrate that the Sys1p-Arl3p-Arl1p-Imh1p cascade is involved in a potential trafficking pathway for Gas1p from the late Golgi to the plasma membrane. Although the deletion of IMH1 could partially affect the transport of Gas1p (Fig. 6B), an imh1 mutant apparently had no hypersensitivity to Congo Red (Fig. 6A). This discrepancy between Arl1p and Imh1p might come from the fact that Imh1p is only one of the effectors of Arl1p. Thus, its absence may only partially delay the Gas1p-trafficking process. Interestingly, Arl1Q72L can suppress the Congo Red hypersensitivity in an arl3 mutant, which also supports the cascade of Arl3-Arl1 functions in Gas1p transport. Although our data support the model of a Sys1p-Arl3p-Arl1p-Imh1p cascade facilitating Gas1p transport, the detailed molecular mechanism of how Arl3p regulates Golgi targeting of Arl1p is still unknown. Overexpression of Arl1Q72L, but not Arl1p, could remain the normal late Golgi location in an arl3 mutant, indicating that Arl1Q72L could skip the requirement of Arl3p to target to Golgi membrane. This result is consistent with the notion that Arl3p might recruit Arl1p directly through interaction with its GEF (Jackson, 2003).

With the expression of ectopic mutant Arl1p in wild-type and arl1-null cells, the importance of GTP hydrolysis by Arl1p was observed (Fig. 7). The constitutively active Arl1p mutant Arl1Q72L could not rescue the Congo Red hypersensitivity in the absence of endogenous Arl1p. In addition, Arl1Q72L even caused substantial intracellular accumulation of Gas1p in wild-type and arl1-mutant cells, indicating that GTP hydrolysis is crucial for the function of Arl1p. Because GTP hydrolysis by Arl1p is decreased in the gcs1 mutant (Liu et al., 2005), we also observed a Gas1p transport defect in a gcs1 mutant (Fig. 3). It has been reported that GTP hydrolysis in the early stage of vesicle formation is important (Presley et al., 2002). Therefore, it is reasonable that the GTP-hydrolysis-deficient mutant Arl1Q72L cannot restore the transport defect of Gas1p in an arl1 mutant.

However, Arl1Q72L seems have normal function in the arl3 mutant and lose its effect on late Golgi morphology. We propose that although Arl1Q72L can target to late Golgi in the arl3 mutant, it may dissociate from membranes more easily than in ARL3-positive cells. Thus, Arl1Q72L maintains the transport function on the late Golgi in the arl3 mutant and will not cause the enlargement of the late Golgi.

Recent reports have found that the Arl1p homologues in Trypanosoma bruce (TbARL1, and the GRIP-domain-containing proteins Golgin-245 and Golgi-97) are involved in the exocytosis of GPI-anchored variant surface glycoprotein (VSG) transport of GPI-anchored marker protein (YFPSP-GPI) from the TGN to the cell periphery and transport of E-cadherin out of the TGN (Kakinuma et al., 2004; Price et al., 2005; Lock et al., 2005). Intriguingly, E-cadherin, like GPI-anchored proteins, is also lipid-raft-associated (Seveau et al., 2004). All of these reports indicate a potential function of Arl1p in regulation of lipid-rafts-associated protein transport from the TGN. However, the localization of Pma1p, another lipid-raft-associated membrane protein, is not affected by ARL1 disruption (our unpublished data), suggesting that not all lipid raft-associated plasma membrane protein transport is mediated by Arl1p.

In this study, we found that the molecular machinery used to transport the GPI-anchored protein Gas1p is Arl1p dependent. We further demonstrated that the signaling cascade Sys1p-Arl3p-Arl1p-Imh1p is involved in the transport of Gas1p from the Golgi to the cell surface. Defining molecular mechanisms of protein sorting and transport in the late Golgi in which Arl1p participates still requires considerable work. Through further analysis of Arl1p and its regulators and effectors, we may be able to gain a better understanding of the trafficking mechanisms that are used for GPI-anchored proteins.

Table 1 lists the yeast strains used in this study. Yeast culture media were prepared as described by Sherman et al. (Sherman et al., 1986). YPD contained 1% Bacto-yeast extract, 2% Bacto-peptone and 2% glucose. SD contained 0.17% Difco yeast nitrogen base (without amino acids and ammonium sulfate), 0.5% ammonium sulfate and 2% glucose. Nutrients essential for auxotrophic strains were supplied at specified concentrations (Sherman et al., 1986). Yeast strains were transformed by the lithium acetate method (Ito et al., 1983). Gene disruption was carried out as described by Lee et al. (Lee et al., 1994).

Plasmids used in this study are listed in Table 2. Plasmids were constructed according to standard protocols (Sambrook et al., 1989). GFP-GPI (Gas1p) (MBQ31) and Gap1^K9K16-GFP (TPQ88) were kindly provided by K. Simons (MPI-CBG, Dresden, Germany). Crh1-GFP, Crh2-GFP and Cwp1-GFP were kindly provided by J. Arroyo (Universidad Complutense de Madrid, Spain). mRFP-Gas1 (pMF608) was a gift from Y. Jigami (AIST, Ibaraki, Japan). The GFP-Snc1p and GFP-Sft2p expression plasmids were gifts from Hugh R. B. Pelham (MRC-LMB, Cambridge, UK). The anti-Gas1p antibody was a gift from H. Riezman (University of Geneva, Switzerland) and anti-Emp47p was from S. Schroder-Kohne (BioMedTec Franken e.V., Würzburg, Germany). The anti-actin, anti-ALP, anti-Arl1p, anti-CPY, anti-Pma1p, anti-Kex2p and anti-Drs2p antibodies were generated in our laboratory. Monoclonal anti-GFP was from (Sigma), anti-Pgk1p and anti-Pep12p antibodies were from Molecular Probes.

Images of live cells containing GFP-tagged or mRFP-tagged proteins were obtained after growth in synthetic medium to mid-log phase. All fluorescence protein-tagged chimeras were cultured in selection medium with 2% glucose, except GFP-GPI (Gas1p) and Gap1^K9K16-GFP, which were induced with 2% galactose. After overnight culture or induction, mid-log-phase cells were examined and images were captured using a Zeiss Axioskop microscope equipped with a cool Snap fx camera, and then processed with Image-Pro Plus software.

Yeast cells, grown overnight at 30°C in minimal medium with 200 mM (NH₄)₂SO₄ to an A₆₀₀ of 0.5, were incubated for 30 minutes at 37°C or 16°C, transferred to sulfate-free, minimal medium (final A₆₀₀=5) and incubated for 15 minutes at 37°C or 30 minutes at 16°C before addition of 30 μCi of Pro-mix L-³⁵S label (mixture of [³⁵S]methionine and [³⁵S]cysteine, 14.3 mCi/ml) per A₆₀₀ unit. After 10 minutes (37°C) or 20 minutes (16°C), labeling was terminated by addition of 5% (v/v) of chase solution containing 0.3% cysteine (w/v), 0.4% of methionine (w/v) and 100 mM (NH₄)₂SO₄. Samples (1 ml) were then removed at the indicated times and added to 1 ml of ice-cold 20 mM NaN₃ in double-distilled water. Cells were washed once with 10 mM NaN₃, followed by addition of glass beads (200 μl) and 200 μl of buffer containing 10 mM Tris-Cl (pH 7.5), 1% SDS, 1 mM EDTA, and 1 mM PMSF followed by vigorous mixing (vortex mixer) for 90 seconds at room temperature before boiling (95°C) for 6 minutes. Immunoprecipitation, electrophoresis and autoradiography were performed essentially as described (Huang et al., 1999), using anti-CPY or anti-ALP antiserum.

A plasma membrane proteinase shaving assay was performed as described previously (Morsomme et al., 2003). After a 10-minute incubation with 10 mM MESNA (Sigma), the cell wall was digested in a 20-minute treatment with lyticase (100U/ml) at room temperature and spheroplasts were incubated with 200 μg/ml proteinase K (Sigma) for 30 minutes. After stopping the digestion with PMSF (1 mM) at 0°C for 10 minutes, proteins were TCA-precipitated and analyzed by immunoblotting. The signals on X-ray film were quantified using the AlphaImager 1220 v5.5.

Experiments were performed as previously described with some modifications (Morsomme et al., 2003). Briefly, proteins were labeled with 30 μCi of Pro-mix L-³⁵S label per A₆₀₀ unit and chased with 1% chase solution (0.3% cysteine, 0.4% methionine, and 100 mM (NH₄)₂SO₄) for 80 minutes, followed by the plasma membrane proteinase shaving assay procedure. After digestion, cells were lysed and Gas1p and actin were immunoprecipitated using specific antibodies. Proteins were then detected by autoradiography after separation by SDS-PAGE.

Cells were harvested by centrifugation from cultures (50 ml) grown in YPD medium to mid-exponential phase (A₆₀₀=1). About 50 of A₆₀₀-unit cells were washed by repeated suspension in ice-cold 10 mM NaN₃ in double-distilled H₂O and centrifugation, incubated with lyticase to form spheroplasts, suspended in 1 ml of ice-cold lysis buffer (20 mM HEPES pH 6.8, 1 mM EDTA, 0.1 M sorbitol, 50 mM KOAc and 1 mM DTT) containing protease inhibitors aprotinin, leupeptin and pepstatin (at 1 mg/ml each), 1 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride, and disrupted by passing through a 26G needle. The cell lysate was centrifuged (700 g) for 7 minutes to remove debris and unbroken cells. For velocity centrifugation, the cleared lysate was subjected to centrifugation (13,000 g) for 10 minutes at 4°C to generate the pellet (P13) and the supernatant (S13) fractions. Equal proportions of each fraction were subjected to SDS-PAGE. For sucrose-density-gradient centrifugation, S13 (0.8 ml) from velocity centrifugation was loaded on top of a manually generated five-step sucrose gradient (0.8 ml each on 30%, 25%, 20%, 15% and 10% sucrose in lysis buffer), which was then subjected to centrifugation (150,000 g for 3 hours at 4°C in a Beckman SW55 rotor). Fractions were collected manually from the top. Samples were precipitated with 10% trichloroacetic acid, resuspended in SDS sample buffer, separated by SDS-PAGE and analyzed by western blot analysis.

Cells grown to a density of A₆₀₀=0.5-1 in 3 ml of minimal selective medium with 2% glucose or YPD medium were prepared for indirect immunofluorescence essentially as described (Lee et al., 1997), except for replacement of the methanol (6-minute incubation)-acetone (30-second incubation) step with 0.2% SDS solution before Arl1p staining (Hagan and Ayscough, 1999). Antibodies included affinity-purified anti-Arl1p and commercial monoclonal anti-GFP antibody (Sigma). Secondary antibodies Alexa Fluor-488-conjugated goat anti-rabbit IgG and Alexa Fluor-594-conjugated goat anti-mouse IgG (Molecular Probes Inc.) were used at dilutions of 1:1000 and 1:2000, respectively. Preparations were inspected with a Zeiss Axioskop microscope equipped with a cool Snap fx camera, and then processed with Image-Pro Plus software.

We thank Jose M. Rodriguez-Pena, Hugh R. B. Pelham, Kai Simons, Stephan Schroder-Kohne, Takehiko Yoko-o, Yoshifumi Jigami and Howard Riezman for providing the expression plasmids and antibodies. We thank Randy Haun and Joel Moss for critical reviewing of this manuscript. This work was supported by grants from the National Science Council, Taiwan, R.O.C. (NSC-94-2320-B-002-117), the Program for Promoting Academic Excellence of University (NSC94-2752-B-002-007-PAE), and the Yung-Shin Biomedical Research Funds (YSP-86-019) to F.-J.S.L.