Competition between the golgin Imh1p and the GAP Gcs1p stabilizes activated Arl1p at the late-Golgi

The ADP-ribosylation factor (ARF) family, a subfamily of the RAS small GTPases superfamily, is composed of Arf and Arf-like (Arl) proteins and functions as a regulator of vesicular trafficking and cytoskeletal organization (D’Souza-Schorey and Chavrier, 2006; Gillingham and Munro, 2007b; Kahn et al., 2006). Similar to other small GTP-binding proteins, ARF cycles between the cytosolic GDP-bound and membrane-associated, GTP-bound forms. The nucleotide switching of ARF is facilitated by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs) that stimulate GTP exchange and hydrolysis, respectively (Donaldson and Jackson, 2011; Jackson and Casanova, 2000). Activated ARF recruits numerous proteins to the membrane or activates lipid-modifying enzymes to facilitate the vesicle maturation and budding from the donor membrane. These include coat complexes, lipid transporters and lipid-modifying enzymes, as well as long coiled-coil proteins that form the coated vesicles, generate membrane curvature, alter the lipid composition, and mediate the docking of vesicle to acceptor membrane (Gillingham and Munro, 2007a).

Arl1p, the best studied Arl protein, localizes to the trans-Golgi and regulates Golgi function in both mammalian cells and yeast (Behnia et al., 2004; Liu et al., 2005; Lowe et al., 1996; Lu and Hong, 2003; Lu et al., 2001; Setty et al., 2004; Shin et al., 2005; Zahn et al., 2008). In yeast, the GEF Syt1p activates a subset of Arl1p, and the ARFGAP Gcs1p regulates its GTP hydrolysis (Chen et al., 2010; Liu et al., 2005). In addition to the GAP and GEF, another Arl protein, Arl3p, also play an upstream role to regulate activation and Golgi localization of Arl1p. Although active Arf1p has been known to recruit or activate numerous effector proteins (Donaldson and Jackson, 2011), only one protein, the golgin Imh1p, is known to be the downstream effector of yeast Arl1p to date (Jackson, 2003; Setty et al., 2003).

Golgins are a family of coiled-coil proteins that are associated with the Golgi apparatus and are necessary for tethering events in membrane fusion and maintaining the architecture of the Golgi (Derby et al., 2007; Lieu et al., 2007; Lieu et al., 2008; Ramirez and Lowe, 2009; Reddy et al., 2006). GRIP-domain containing golgins are a group of golgins that are peripherally associated with the cytoplasmic face of the Golgi membrane by their GRIP domain. Other than the GRIP domain, the primary structural component of these proteins is a coiled-coil region, which is responsible for its membrane-tethering role. There are four GRIP-domain-containing golgins in mammalian cells, Golgin-97, Golgin-245/p230, GCC185, and GCC88, whereas Imh1p is the only GRIP-domain-containing golgin in yeast (Ramirez and Lowe, 2009). Golgin-97 and golgin-245 have been suggested to participate in anterograde transport from the TGN to the plasma membrane (Brémond et al., 2009; Kakinuma et al., 2004; Lieu et al., 2008; Lock et al., 2005). Although double deletion of IMH1 and YPT6 in yeast leads to a severe growth defect and vacuole fragmentation (Tsukada and Gallwitz, 1996; Tsukada et al., 1999), the exact physiological function of Imh1p remains unclear.

Golgin-97, Golgin-245/p230, and Imh1p are effectors of Arl1 and are targeted to the Golgi via interaction of their GRIP domain with GTP-bound form of Arl1 (Lu and Hong, 2003; Panic et al., 2003; Setty et al., 2003). Based on crystal structure analysis, it has been hypothesized that two Arl1-GTPs and two GRIP domains form a tetrameric complex on the TGN membrane (Wu et al., 2004). Because Golgi cisternae are not stable entities (Glick and Nakano, 2009), the surrounding golgin must be renewed constantly. It has been suggested that ArfGAP1, a membrane curvature sensitive GAP, controls the binding of golgins to active Arf1 or Arl1 through its GTP hydrolysis stimulating activity (Bigay et al., 2005; Drin et al., 2008). Similarly, yeast Gcs1p, an ortholog of ArfGAP1, also contains a membrane curvature sensitive ALPS motif (Bigay et al., 2005) and regulates the Imh1p Golgi targeting through the stimulation of Arl1p GTP hydrolysis (Chen et al., 2010; Liu et al., 2005). However, the mechanism by which Gcs1p accesses the GTP-bound form of Arl1p on the TGN and stimulates GTP hydrolysis has not been explored.

In this study, we report a novel bottom-up mechanism to modulate the inactivation of Arl1p through its effector Imh1p, which determines the timing of Arl1p GTP hydrolysis. Physiologically, deletion of IMH1 results in a decrease of GTP-bound Arl1p and, hence, less association with the Golgi. Biochemically, Imh1p competes with Gcs1p to interact with Arl1p; thus, Imh1p attenuates Gcs1p GAP activity toward Arl1p. We further show that the dimerization of Imh1p is required for its ability to regulate GTP hydrolysis of Arl1p. Overall, we present an unexpected interplay of Arl1p and its effector Imh1p upon their localization to the Golgi through the regulation of accessibility of Arl1GAP, Gcs1p.

While we were studying whether yeast golgin Imh1p, similarly to mammalian golgins, is involved in the structural organization of Golgi, we observed an unexpected result from two late-Golgi markers, Arl1-RFP and GFP-Sft2, in imh1Δ cells. The late-Golgi SNARE GFP-Sft2 showed similar punctate distributions in both IMH1 and imh1Δ cells, indicating that the late Golgi structure was intact even in the absence of Imh1p (Fig. 1A). To our surprise, however, Arl1-RFP was largely distributed throughout the cytoplasm in imh1Δ cells, similar to the effect of deletion of the ARL3 gene. The fluorescence intensity line-scans are shown in the right panels of Fig. 1A to emphasize the loss of enrichment of Arl1-GFP at the GFP-Sft2 foci. A detailed quantification of Arl1-RFP and GFP-Sft2 punctuate fluorescence signals showed that the amount of Arl1p localized to the Golgi dropped from approximately 35% to 11% in imh1Δ cells. Interestingly, the ratio of Golgi-associated Arl1p in imh1Δ is greater than in arl3Δ cells (Fig. 1B). A similar phenomenon was also observed with immunofluorescence staining of Arl1-HA expressed via the ARL1 promoter. The Arl1-HA signal in imh1Δ cells showed a distribution that was more cytosolic than that in IMH1 cells but not as diffuse as in arl3Δ cells (Fig. 1C).

To determine which domain of Imh1p is responsible for this effect, we introduced different fragments of N-terminally GFP-tagged Imh1p into the imh1Δ cells to test their ability to restore the Golgi association of Arl1-RFP (Fig. 1D). Both the full-length Imh1p and the C-terminal 177 amino acids of Imh1p (Imh1-C177), which contains the GRIP domain and localizes at the Golgi (Panic et al., 2003), can rescue the Golgi association of Arl1p in the absence of Imh1p. However, Imh1-C177 without the GRIP domain (Imh1-C177ΔGRIP) was unable to restore the Golgi localization of Arl1-RFP in imh1Δ cells (Fig. 1D). These results indicate that the GRIP domain, but not the major part of the N-terminal coiled-coil region, is required for the effect of Imh1p on Arl1p localization. In addition, we found that the disruption of IMH1 did not affect the Golgi localization of either Arl3p or Syt1p, the upstream regulators of Arl1p (supplementary material Fig. S1A,B). Together, these results suggest that Imh1p has a novel and direct effect on the Golgi localization of Arl1p.

The association of Arl1p with the TGN is dependent on its guanine-nucleotide-binding status. The GTP-bound active form Arl1p associates with Golgi membranes, whereas the GDP-bound inactive form is distributed in the cytosol. To determine whether the diffuse distribution of Arl1p in imh1Δ cells was a result of the reduction of GTP-bound Arl1p in the cell, we measured the amount of active Arl1p in wild-type and imh1Δ cells using an Imh1p pull-down assay (Chen et al., 2010). The C-terminal region of Imh1p (GST–Imh1-C177) binds specifically to the active form of Arl1p but not to the inactive form of Arl1p. Therefore, we used GST-tagged Imh1-C177 to pull down the active form of Arl1p from either wild-type or imh1Δ cells and measured the amount of Arl1p with western blotting. The results showed that there was less GTP-bound Arl1p in imh1Δ cells than in wild-type cells (supplementary material Fig. S2A). To examine whether an increase of the GTP-bound, active form of Arl1p could reverse the diffuse distribution of Arl1p in imh1Δ cells, we expressed the constitutively active mutant Arl1-Q72L, which is defective in GTP hydrolysis. As in wild-type cells, Arl1-Q72L was localized to the Golgi in imh1Δ cells (supplementary material Fig. S2B), indicating that the distribution of Arl1p in imh1Δ cells was due to the decrease in the GTP-bound form of the protein. This result suggests that the binding site on the Golgi for Arl1p localization is intact in imh1Δ cells, thus Imh1p may be involved in regulating the GDP/GTP cycle of Arl1p in vivo.

We next asked whether the decrease of GTP-bound Arl1p in imh1Δ cells was due to a decrease in GTP loading or an increase of GTP hydrolysis. A previous study indicated that Arl3p can activate multiple GEFs to promote GTP loading of Arl1p; hence, the disruption of the ARL3 gene resulted in the diffuse distribution of Arl1p (Chen et al., 2010; Panic et al., 2003). Given that the deletion of IMH1 did not alter the Golgi localization of Arl3p and Syt1p and that Arl1-Q72L remained localized to the Golgi in imh1Δ cells, we hypothesized that the upstream activation pathway of Arl1p remains intact. Gcs1p is a GAP protein for Arl1p, and overexpression of Gcs1p in yeast promotes Arl1p GTP hydrolysis and Golgi dissociation (Liu et al., 2005). Disruption of GCS1, in contrast, delays GTP hydrolysis of Arl1p and increases the fraction of Arl1p associated with the Golgi membrane (Liu et al., 2005). We, therefore, asked whether the less active form Arl1p in imh1Δ cells was due to increased GTP hydrolysis of Arl1p facilitated by Gcs1p.

We first examined the effect of GCS1 deletion in imh1Δ cells on the localization of Arl1-RFP. As shown in Fig. 2A, Arl1-RFP displayed diffuse distribution in imh1Δ cells, and deletion of GCS1 in imh1Δ cells restored the Golgi localization of Arl1-RFP. In contrast, deletion of GLO3, an ARF-GAP protein that cannot activate Arl1p GTP hydrolysis (Liu et al., 2005) in imh1Δ cells did not alter the diffuse distribution of Arl1p (Fig. 2A). Furthermore, deletion of GCS1 in arl3Δ cells had no effect on the diffuse distribution of Arl1p-RFP (Fig. 2A), indicating that deletion of GCS1 could not rescue the loss of active Arl1p resulting from lack of GTP loading and further suggesting that the GTP loading of Arl1p in imh1Δ cells was not affected. We next introduced into imh1Δgcs1Δ cells either Gcs1-GFP or Gcs1-zn-GFP, a mutant that has impaired GAP ability due to its inability to bind magnesium (Goldberg, 1999; Poon et al., 1996). We observed that overexpression of Gcs1-GFP, but not Gcs1-zn-GFP, in imh1Δgcs1Δ cells reproduced the dissociation of Arl1p from the Golgi, and we learned that this effect depends on Gcs1p GAP activity (Fig. 2B). We further examined the amount of active Arl1p in these cells. We overexpressed Arl1p under the control of the ADH promoter in different yeast strains and they all showed similar protein levels of Arl1p (supplementary material Fig. S3). Fig. 2C showed that Gcs1p is consistently required for the decrease in GTP-bound Arl1p in imh1Δ, but not in arl3Δ cells. Together, these results indicate that GTP loading of Arl1p in imh1Δ cells was not impaired but rather that the Gcs1p-activated GTP hydrolysis of Arl1p was enhanced.

How could loss of Imh1p affect Gcs1p GAP activity? Both Imh1p and Gcs1p directly interact with Arl1-GTP (Liu et al., 2005; Panic et al., 2003); therefore, we hypothesized that Imh1p physically blocks access of Gcs1p to Arl1p. We, therefore, examined the interaction efficiency of Gcs1p with Arl1p in the presence of Imh1p by GST pull-down assay. We prepared and purified recombinant GST-tagged Arl1p mutants that mimicked the GTP-bound form (GST-Arl1Q72Ld17N), or GDP-bound form (GST-Arl1T32Nd17N). These proteins were immobilized on glutathione-Sepharose beads and incubated with purified His–Imh1-C177, His–Imh1-C177ΔGRIP, or His-Gcs1, to determine their ability to interact with Arl1p. As shown in Fig. 3A and Fig. 3B, His–Imh1-C177 and His-Gcs1, but not His–Imh1-C177ΔGRIP, interacted with Arl1Q72Ld17N. Consistent with previous reports (Lu and Hong, 2003; Panic et al., 2003; Setty et al., 2003), the GRIP domain of Imh1p is required for its interaction with GTP-bound Arl1p.

To examine whether the binding of Gcs1p or Imh1p to Arl1p is mutually exclusive, we performed GST pull-down assays in which GST-Arl1QLd17N was incubated with a fixed amount of His-Gcs1 and increasing amounts of His–Imh1-C177 or His–Imh1-C177ΔGRIP. The results showed that the more His–Imh1-C177 that was incubated with GST-Arl1QLd17N, the less His-Gcs1 bound to Arl1p (Fig. 3C,D). In addition, His–Imh1-C177ΔGRIP, which lacks the ability to bind to Arl1p, did not alter Gcs1p binding to Arl1p. These results suggest that Imh1p and Gcs1p compete for binding to GTP-bound Arl1p.

To further verify whether the direct binding of Imh1p to Arl1p would interfere with Gcs1p GAP activity toward Arl1p, we performed in vitro Gcs1p GAP assays for Arl1p in the presence of Imh1p. Recombinant His-Arl1-d17N was purified and loaded with radiolabeled GTP prior to incubation with the same amount of His–Imh1-C177 or His–Imh1-C177ΔGRIP. His-Gcs1 was added to the mixture to initiate GTP hydrolysis, and GTP hydrolysis was observed at different time points post-His-Gcs1 addition (0-30 min). In the absence of Imh1p, the t_1/2 for Gcs1GAP activity was 1.2 min, whereas, in the presence of Imh1-C177, the t_1/2 showed a delay to 3.6 min. The pre-incubation of Arl1p with Imh1-C177ΔGRIP had no effect on Gcs1p-stimulated Arl1 GTP hydrolysis (Fig. 4A). We also measured the steady-state GAP activity of Gcs1p with various amounts of Imh1p. As shown in Fig. 5B, the reduction effect of Imh1-C177 on His-Gcs1 GAP activity toward His-d17N-Arl1 was dosage dependent (Fig. 4B). These results demonstrated that direct binding of Imh1p to Arl1p interferes with Gcs1p GAP activity in vitro. Therefore, Imh1p might shield GTP-bound Arl1p from hydrolysis in vivo by physically blocking access of Gcs1p to Arl1p.

It has been suggested that the Golgi localization of GRIP-golgin in mammalian cells is dependent on the homo-dimerization of the GRIP domain, the interaction of the GRIP domain with activated Arl1, and the interaction of the GRIP domain with membrane phospholipids (Lu et al., 2006; Panic et al., 2003; Setty et al., 2003; Wu et al., 2004). In the current study, we found that the C-terminal region of Imh1p (Imh1-C177) is sufficient to maintain Arl1p at the trans-Golgi (Fig. 1E). To investigate whether the dimerization of Imh1p is required for its Golgi targeting as well as stabilizing of Arl1p on the Golgi, we generated Imh1p mutants with increasing large truncations of the coiled-coil region. The GRIP domain comprises the C-terminal 62 residues of Imh1p (Kjer-Nielsen et al., 1999). We therefore constructed Imh1-GRIP, Imh1-C89, Imh1-C177, and Imh1-C213, each of which possess an intact GRIP domain without or with part of the coiled-coil region (Fig. 5A). First, we performed two different yeast two-hybrid assays to verify the self-interaction of these polypeptides. The filter-lifting assay (lower panel) is more sensitive than the plate assay (upper panel). At similar expression levels (supplementary material Fig. S4), the self-interactions of Imh1-C89 or Imh1-GRIP were not prominent, whereas that of Imh1-C213 was stronger than that of Imh1-C177 (Fig. 5A). This result shows that the coiled-coil region facilitates the self-interaction of the Imh1p C terminus. Next, we examined whether these truncated Imh1p constructs could restore the Golgi localization of Arl1p in imh1Δ cells. In contrast to Imh1-C177 or Imh1-C213, neither Imh1-C89 nor Imh1-GRIP could restore the Golgi localization of Arl1-RFP in imh1Δ cells (Fig. 5B). These results indicate that GRIP domain alone is not sufficient to maintain Arl1-RFP on the Golgi and that the dimerization of Imh1p is required for this effect.

To verify that the dimerization of Imh1p, rather than the sequence elements within the deleted region, is required for the Golgi association of Imh1p and Arl1p, we forced the truncated Imh1p to dimerize using an N-terminally tagged DsRed protein that forms dimer spontaneously (Baird et al., 2000; Sacchetti et al., 2002). The dimer formation of DsRed–Imh1-C89 was confirmed by yeast two-hybrid assays and untagged or mCherry-tagged Imh1-C89 did not show self-interaction (supplementary material Fig. S5). When introduced into arl1imh1Δ cells, DsRed–Imh1-C89, but not mCherry–Imh1-C89, was targeted to the Golgi and restored the Golgi association of Arl1-GFP (Fig. 5C). In addition, neither mCherry-tagged nor DsRed-tagged Imh1p localized to the Golgi in arl1Δ cells, indicating these tagged Imh1p still required Arl1p to recruit them to the Golgi (supplementary material Fig. S6). These results indicate that homodimerization of Imh1p is required for the stabilization of Arl1p associated with the Golgi and for its Golgi targeting.

To further confirm that dimerization of the coiled-coil region is required for stabilizing Arl1p on Golgi, we tried to disrupt the dimerization of Imh1-C177. Structure prediction of the Imh1p C-terminus using the MultiCoil algorithm (Wolf et al., 1997) revealed a short dimeric, coiled-coil region from residue 734 to 820 (Fig. 6A, shaded area), whereas the GRIP domain does not conform to a coiled-coil structure. A helical wheel projection of the predicted coiled-coil between residue 760 and 787 identified it as an amphipathic helix, with the residues at heptad positions ‘a’ and ‘d’ expected to reside at the interface of a dimeric coiled-coil (Harbury et al., 1993) (Fig. 6B). Thus, to disrupt the dimer interface and destabilize the dimerization of Imh1p, two residues (N779 and L782, shown in red in Fig. 6B) were replaced with serines (Imh1-C177^{N779S, L782S}). To verify whether the self-interaction ability is weakened in Imh1-C177^{N779S, L782S}, we performed in vitro pull-down assays. As shown in Fig. 6C, recombinant GST–Imh1-C177 can pull down His–Imh1-C177, whereas GST–Imh1-C177^{N779S, L782S} cannot pull down His-tagged Imh1-C177^{N779S, L782S}. Moreover, both Imh1-C177 and Imh1-C177^{N779S, L782S} can interact with His-Arl1, indicating that the self-association-defective mutant of Imh1p can still interact with Arl1p in vitro (Fig. 6C). To investigate whether Imh1-C177^{N779S, L782S} is able to restore Arl1p to the Golgi, mCherry-tagged Imh1-C177^{N779S, L782S} was introduced to imh1Δ cells. As shown in Fig. 6D, mCherry-tagged Imh1-C177^{N779S, L782S} failed to localize to the Golgi in imh1Δ cells and was unable to restore the Golgi association of Arl1-GFP. Together, these results demonstrate that the dimerization of the N-terminal coiled-coil region of Imh1p is required for the Golgi localization of Arl1p.

Active Arl1p recruits cytosolic golgins to form a tetramer complex at the trans-Golgi (Lu and Hong, 2003; Panic et al., 2003; Wu et al., 2004). After the hydrolysis of Arl1-GTP, stimulated by its GAP, golgin and Arl1 recycle to the cytosol for another run of activation. It has been suggested that the dimerization of the coiled-coil region in golgins may be broken during this cycle (Ramirez and Lowe, 2009). However, whether the dissociation of golgin protein homodimers is required for their function is not known. We hypothesized that dynamic dimerization and recycling of Imh1p are physiologically regulated and are important for its ability to stabilize of GTP-bound Arl1p on the Golgi. Imh1p that is locked in dimer form may not be able to recycle and thus prevent Arl1p GTP hydrolysis by Gcs1p. DsRed forms high-affinity dimers that should not be separated to monomers under normal physiological conditions (Sacchetti et al., 2002). Therefore, we used DsRed-tagged Imh1p to test whether locked dimeric Imh1p could inhibit the recycling of Imh1p to the cytosol and prevent Arl1p GTP hydrolysis by Gcs1p. Hence, we overexpressed Gcs1p in imh1Δ cells that contained Arl1p-GFP and either mCherry–Imh1-C177 or DsRed–Imh1-C177. Our previous study has shown that the overexpression of Gcs1p results in Arl1p and Imh1p dissociation from the Golgi (Chen et al., 2010; Liu et al., 2005). Consistent with this finding, we observed the overexpression of Gcs1p promotes Arl1-GFP dissociation from the Golgi in imh1Δ cells expressing mCherry–Imh1-C177 (Fig. 7A); on the contrary, Gcs1p failed to promote the dissociation of Arl1-GFP from the Golgi in imh1Δ cells that expressed DsRed–Imh1-C177. Furthermore, full-length copies of Imh1p tagged with mCherry or DsRed were also examined in combination with the overexpression of Gcs1p in imh1Δ cells and showed similar results (supplementary material Fig. S7).

To further confirm that the dimerization of Imh1p can attenuate Arl1p-GTP hydrolysis by Gcs1p, we inserted an inducible dimerization DmrB domain into mCherry–Imh1-C177 to create mCherry–DmrB-Imh1-C177 and tested its effect on the Gcs1p-promoted hydrolysis of Arl1p-GTP. As shown in Fig. 7B, mCherry–DmrB-Imh1-C177, similar to mCherry–Imh1-C177, was co-localized with Arl1p in imh1Δ cells and dissociated to the cytosol in Gcs1p-overexpressing cells. In imh1Δ cells that overexpressed Gcs1p, we found that Arl1p and Imh1p located to the Golgi only after inducing the dimerization of mCherry–DmrB-Imh1-C177 (Fig. 7C). These results demonstrate that dimerization of Imh1p is critical for the attenuation of Arl1p-GTP hydrolysis and suggest that weakening the dimeric interaction of Imh1p might be an essential step to promote the hydrolysis of Arl1p-GTP by Gcs1p.

Many golgins are peripherally associated on the Golgi, potentially allowing their recycling from one part of the Golgi membrane to another via the cytosol, and thereby facilitating the remodeling of the Golgi apparatus as it fulfills its function. The yeast golgin Imh1p has been shown to be recruited to the TGN by GTP-bound Arl1p and act as an effector of Arl1p to mediate endosome-Golgi transport. Subsequently Gcs1p-stimulated GTP hydrolysis would release Arl1p and Imh1p into the cytosol for the next round of activation. However, the mechanism and timing for Arl1p GTP hydrolysis remains largely unknown. In this study, we report that Imh1p plays a novel role in modulating the level of GTP-bound Arl1p on the TGN. Imh1p competes with Gcs1p to interact with Arl1p in vitro, thus blocking Gcs1p access to Arl1-GTP. Mechanistically, Imh1p requires both its interaction with Arl1p and its self-interaction to maintain Arl1p at the TGN. Our findings reveal a critical role for Imh1p in modulating the spatiotemporal inactivation of Arl1p on the Golgi.

Our previous study demonstrated that Syt1p promotes the activation of Arl1p and the recruitment of Imh1p to the Golgi (Chen et al., 2010). Unlike the arl1Δ mutant, imh1Δ and syt1Δ did not show defects in Gas1p transport, cell wall integrity, and vacuolar biogenesis (Chen et al., 2010). Therefore, we learned that the activation of Arl1p is regulated in part by Syt1p and that active Arl1p, working through multiple GEFs, exerts distinct biological activities at the Golgi compartment. Consistent with this notion, our data show that deletion of IMH1, similar to the deletion of SYT1, results in a decrease in GTP-bound Arl1p and hence leads to the majority, but not all, of Arl1p distributing to the cytosol. A recent study showed that the deletion of ARL1 caused mislocalization of Gga2p (Singer-Kruger et al., 2008), and our preliminary data showed that, in contrast to the arl1Δ mutant, imh1Δ and syt1Δ did not show mislocalization of Gga2p. The multi-role and multi-regulator model of Arl1p is further strengthened by the distribution of Arl1p and Imh1p because the Imh1p puncta always contain Arl1p but not vice versa (Fig. 1D). Based on all of these observations, we hypothesize that Arl1p plays multiple roles at the Golgi network and that Imh1p modulates a subset of the GTP-bound Arl1p that is activated by Syt1p on the TGN.

In this study, we show that Gcs1p is responsible for the decrease of GTP-bound Arl1p in imh1Δ cells but not in arl3Δ cells (Fig. 2C). Our results prove that neither the GTP loading nor the Golgi targeting of Arl1p in imh1Δ cells was impaired but the GTP hydrolysis of Arl1p was enhanced. We further demonstrate that Gcs1p and Imh1p bind to Arl1p in a competitive manner; thus, the absence of Imh1p increases the accessibility of Arl1p for Gcs1p to stimulate GTP hydrolysis. The crystal structure of the Arl1-GTP/golgin-245 GRIP complex has been solved and shown that GRIP domain forms a homodimer in which each subunit interacts with one molecule of Arl1-GTP and becomes a heterotetramer (Wu et al., 2004). According to the structure, Arl1-GTP interacts with the GRIP domain through its switch II region together with additional residues from the switch I and interswitch regions (Lu et al., 2006; Wu et al., 2004). Although there is no related structure available for Arl1-GTP-Gcs1 complex, the co-crystal structure of Arf1-GDP and ArfGAP1 (Goldberg, 1999) indicated that ArfGAP1 binds to switch II and helix α3 of Arf1 to orient Arf1 residues for catalysis. These data support the idea that Gcs1p and Imh1p bind to the same region of Arl1p.

Competitive interactions between effectors and GAPs are in fact a characteristic of other members of the Ras superfamily of GTP binding proteins. For example, Rabphilin 3A (a Rab3A effector) and Rab GAP compete for binding to Rab3A (Clabecq et al., 2000). Likewise, the sites for binding of Raf1 kinase (an effector of Ras) and Ras GAP to Ras are overlapping (Scheffzek et al., 1997). However, whether these interactions affect the GTP hydrolysis of Rab or Ras are not clear. Interestingly, Bonifacino and colleagues (Puertollano et al., 2001) reported that Arf effectors GGAs and GAPs compete for interaction with Arf1-GTP and suggested a model for ARF1 regulation in which effectors delay dissociation of ARF1 from Golgi membranes. Consistent with this notion, Imh1p might shield GTP-bound Arl1p from hydrolysis in vivo by physically blocking the access of Gcs1p to Arl1p. We further examined how the Imh1p regulates the hydrolysis of Arl1p and found that dimerization of the Imh1p coiled-coil region inhibits GAP-induced GTPase activity of Arl1p but is not required for the interaction with Arl1p (Fig. 6Cl; Fig. 7). This mechanism differs from that the GGAs interact with ARF1 that overexpression of the GAT domain of GGAs is sufficient to stabilize the association of ARF1-GTP with TGN membranes in vivo, suggesting that they represent distinct modes of ARF/ARL regulation by Coat proteins and golgins.

Golgins are unique in their structure among various peripheral Golgi proteins in that they contain large regions of coiled-coil domain (Barr, 1999; Gillingham and Munro, 2003) and are proposed to be long rod-like molecules (Lu et al., 2006). A hierarchically organized three-tier interaction was suggested to govern the Golgi-targeting of Arl1-interacting GRIP golgins; first, self-interaction of the GRIP domain; second, interaction between the dimerized GRIP domain with the GTP-bound Arl1p; and third, the binding of the GRIP domain to membrane lipids through non-specific hydrophobic and electrostatic forces (Lu et al., 2006). However, based on the alignment of different GRIP-containing golgins (Lu et al., 2006), Imh1p does not possess the conserved C-terminal region in the GRIP domain, which interacts with phospholipids. This finding suggests that Golgi-targeting of yeast and mammalian Arl1-interacting GRIP golgins might not be identical. Consistent with this view, we found that the GRIP domain of Imh1p fails to localize at the TGN and reasoned that this failure may be due to the weakness or complete lack of direct interaction between the GRIP domain and phospholipids. Although the GRIP domain of Imh1p alone was able to interact with Arl1p (our unpublished data), our data showed that the self-interaction of the GRIP domain is not prominent and that the coiled-coil region is required for stronger self-interaction. Although a previous report showed that the dimerization of the GRIP domain is sufficient for the Golgi localization of mammalian golgins (Lu et al., 2006), it should be noted that the ‘GRIP’ domain used in their study, C-terminal 177 residues, contains part of the coiled-coil region in addition to the C-terminal 62-residue GRIP domain. Furthermore, the GRIP domain of Imh1p with part of the coiled-coil region, but not the GRIP domain alone, is sufficient to restore Arl1-RFP to the Golgi, indicating that the Golgi localization of Arl1p requires the self-interaction of Imh1p.

We also showed that dimerization of the Imh1p GRIP domain, but not the presence of sequence elements within the deleted coiled-coil region, is indispensable for the Golgi association of Imh1p and Arl1p. There is a requirement for displacing or reorganizing golgin to facilitate the dynamics of the Golgi architecture, and golgin complexes may be dissociated during the conversion from tethering to SNARE pairing (Ramirez and Lowe, 2009). Although golgins are extensively coiled-coil in nature, the predicted coiled regions may unfold under physiological conditions through the alterations of post-translational modifications, putative regulatory factors, or membrane environment. Together, we hypothesize that the regulation of the dimerization of Imh1p is critical for modulating the level of GTP-bound Arl1p on the Golgi.

IMH1 is also called SYS3, which was identified from multicopy suppressors that complement the temperature-sensitive growth phenotype of the ypt6Δ-null mutant (Tsukada and Gallwitz, 1996). Deletion of YPT6 in yeast results in temperature-sensitive growth and vacuolar fragmentation. It had been shown that deletion of both IMH1 and YPT6 leads to a severely fragmented vacuole (Tsukada et al., 1999). These results indicated that Imh1p and Ypt6p might function in a similar pathway. However, physical interaction between Imh1p and Ypt6p has not been observed to date. Although recent studies showed that Arl1p and Imh1p were no longer localized to the Golgi in ypt6Δ cells (Benjamin et al., 2011), whether Ypt6p is involved in the Arl1p-Imh1p pathway through a transit interaction with Imh1p or indirect effect remains to be determined.

We have tried to overexpress different forms of Imh1p in ypt6Δ cells to see whether these protein variants can suppress temperature sensitivity. Our preliminary results indicated that neither Imh1-C177 nor GFP-Imh1 could suppress the temperature sensitivity of ypt6Δ cells, although both Imh1-C177 and GFP-Imh1 could stabilize Arl1p on the Golgi. We conclude that the C-terminal region of Imh1p is required for the stabilization of Arl1-GTP, but the N-terminus of Imh1p is required for the suppression of temperature sensitivity in ypt6Δ cells.

Previous studies indicated that Arl1p plays multiple roles at the Golgi network (Chen et al., 2010; Singer-Krüger et al., 2008) and the GTP exchange and hydrolysis of Arl1p is under precise temporal and spatial control. The central dogma of ARF-mediated vesicular formation is a unidirectional cascade in which GEF activates ARF, which in turn recruits its effector molecules and is deactivated by its GAP. Consistent with the dogma, yeast golgin Imh1p, the best-known effector protein of Arl1p, has been clearly proven to be recruited to the Golgi through its interaction with GTP-bound Arl1p. In this study, we found, for the first time, that the effector of ARF can act as a feedback regulator to spatiotemporally modulate GTP hydrolysis of ARF.

Previous studies have shown that dimers of mammalian TGN golgins and oligomers of yeast Imh1p are very stable and may not be dissociated into monomers under physiological conditions (Luke et al., 2005; Tsukada et al., 1999). Our preliminary data also show that Imh1p forms stable dimers/oligomers in cytosolic lysates obtained from wild-type, arl1Δ, or ypt6Δ null yeast (our unpublished data). However, we cannot exclude the possibility that these Imh1p dimers/oligomers may undergo further oligomerization or conformational change in response to Ypt6p and Arl1p binding in vivo. Therefore, we hypothesize that the conformation of the coiled-coil region of Imh1p may be regulated to allow Gcs1p access to GTP-bound Arl1p during vesicle formation at the TGN. Many membrane-associated proteins, including Gcs1p, are sensitive to membrane curvature (Bigay et al., 2005), thus the increase in membrane curvature may alter the interaction between the GRIP domain and the Golgi membrane, thereby not only weakening the interaction between Imh1p and GTP-bound Arl1p but also altering the conformation of the coiled-coil region of Imh1p. We propose a model (Fig. 8) in which, as an early event, Syt1p-activated Arl1p-GTP binds to the less curved membrane and recruits Imh1p to form a tethering complex. Subsequently, while the vesicle is mature and the membrane is highly curved, the membrane curvature may affect the conformation of the coiled-coil region of Imh1p, allowing Gcs1p to gain access to Arl1p-GTP and activate GAP activity, which, in turn, will stimulate GTP hydrolysis and disassemble the Arl1p-Imh1p complex. In the absence of Imh1p, Gcs1p would activate GTP hydrolysis of Arl1p early because of the lack of temporal control from Imh1p steric protection (Fig. 8, lower panel).

In conclusion, our findings identify an unexpected function of Imh1p in fine-tuning the levels of GTP-bound Arl1p on the Golgi. Our experiments also provide a link between the Syt1p-dependent activation and the Imh1p-Gcs1p-dependent inactivation of Arl1p on the Golgi. Defining the molecular mechanism of the spatial and temporal control of GTP hydrolysis for Arl1p will require considerable work. Further analysis of the regulation of Imh1p will improve our understanding of the mechanism of Arl1p activity and its functions at the Golgi.

The yeast strains used in this study are listed in supplementary material Table S1. The yeast culture media were prepared as described (Sherman et al., 1986). The YPD contained 1% Bacto yeast extract, 2% Bacto peptone and 2% glucose. The SD contained 0.17% Difco yeast nitrogen base (without amino acids), and 2% glucose. Nutrients essential for auxotrophic strains were supplied at specified concentrations (Sherman et al., 1986). The yeast strains were transformed using the lithium acetate method (Ito et al., 1983). The plasmids used in this study are listed in supplementary material Table S2. The GFP-Sft2p expression plasmid was a gift from Hugh R. B. Pelham (MRC-LMB, Cambridge, UK).

Yeast two-hybrid assays were performed using the ‘interaction-trap’ system (Golemis and Khazak, 1997). In this system, the bait was fused with the DNA-binding domain of LexA in pEG202 (bait plasmid). The pray was constructed using the vector pJG4-5 (pray plasmid), which uses the inducible yeast GAL1 promoter to express the protein fused to an acidic domain that functions as a transcriptional activation motif. The reporter yeast, YEM1α, expressing the interacting proteins, can transactivate two reporter genes, LacZ and LEU2, that allow for the expression of β-galactosidase and growth on minimal medium lacking leucine.

The C-terminal region (177 amino acids) of yeast Imh1p was N-terminally tagged with a histidine tag using the vector pET32a. His-tagged proteins were then purified using the Ni⁺ NTA resin (Qiagen, Chatsworth, CA) as previously described (Huang et al., 1999). Denatured, purified recombinant Imh1-C177 was isolated from an SDS-PAGE gel and used as an antigen for polyclonal antibody production in rabbits.

Images of live cells containing GFP-tagged or mRFP-tagged proteins were obtained after growth in synthetic medium to the mid-log phase. All of the fluorescent protein-tagged chimeras were cultured in selection medium with 2% glucose, except for cells expressing GFP-Imh1, which were induced with 2% galactose. After overnight culture or induction, mid-log-phase cells were examined and images were taken using a Zeiss Axioskop microscope equipped with a cool SnapSNAP fx camera. For easier detection of Arl1p in vivo, we used arl1Δ-null yeast overexpressing C-terminally RFP-tagged Arl1p as a wild-type strain. The ratio of Golgi association (RGA) values for Arl1p and Sft2p were calculated by measuring the fluorescence intensity values (arbitrary units) using Axio Vision Rel. 4.2 software: (1) punctate structure (PS) signal greater than 0.2 µm diameter; and (2) fluorescence signal of the whole cell (C). The RGA was then calculated as a percentage of the total fluorescence of both areas: RGS = (PS/C)×100. The measurements were performed on images that were acquired with the microscope set to the same intensity (Matheson et al., 2007).

Cells were grown to a density of A600 of 0.5-1 in 3 ml of minimal selective medium with 2% glucose and were prepared for indirect immunofluorescence as described (Lee et al., 1997). The antibodies used included commercially available monoclonal anti-HA (Covance) and polyclonal anti-GFP (Invitrogen) antibody. The secondary antibodies included Alexa-Fluor-488-conjugated goat anti-mouse IgG and Alexa-Fluor-594-conjugated goat anti-rabbit IgG (Molecular Probes) and were used at a dilution of 1∶1000 and 1∶2000, respectively. The preparations were visualized with a Zeiss Axioskop microscope equipped with a CoolSnap fx camera, and the images were processed using Image-Pro Plus software.

The activated Arl1p pull-down assay was performed as described by Chen et al. (Chen et al., 2010). Arl1p was expressed under the control of the ADH promoter in different yeast strains. Yeasts were lysed with glass beads at 4°C in 0.65 ml of 50 mM Tris, pH 7.5, 100 mM NaCl, 5 mM MgCl₂, and protease inhibitor cocktail (1 µg/ml aprotinin, 1 µg/ml leupeptin, 1 µg/ml pepstatin, 1 µM benzamidine, and 1 mM phenylmethylsulfonyl fluoride). The lysates were clarified by centrifugation at 10,000 g for 5 min. The clarified lysate (0.5 ml) was incubated for 5 hr at 4°C with 20 µg of either GST or GST-Imh1 containing the C-terminal 177 amino acids of Imh1p, each bound to glutathione-Sepharose beads (Amersham Pharmacia Biotech). The beads were washed three times with lysis buffer, and the bound proteins were eluted with 40 µl SDS-PAGE sample buffer. This sample was assayed for the presence of Arl1p by western blotting with Arl1p-specific antibodies.

The GAP activity was assessed essentially as described by Huber et al. (Huber et al., 2001). GAP activity was assayed by a single-round hydrolysis of Arl1d17N-bound GTP by using [γ-³²P]GTP and measuring the Arl1d17N-[γ-³²P]GTP after the GAP was added. Recombinant His-Arl1d17N protein was incubated in exchange buffer [1 mM EDTA, 0.5 mM MgCl₂, 100 mM NaCl, 1 mM DTT, and 25 mM HEPES (pH 7.4)], with [γ-³²P]GTP for 15 min at 30°C, resulting in the association of ∼50% of the labeled nucleotide with Arl1p proteins. Arl1p was further incubated with either recombinant His–Imh1-C177 or His–Imh1-C177ΔGRIP for 30 min at 4°C in exchange buffer after adding 8 mM MgCl₂. The GAP activity was determined at 30°C in 50 µl assay systems at different time points (up to 30 min) with recombinant His-Gcs1. Samples were collected at 0 and 5 min for the typical experiment. The samples were removed at various time points, diluted with 2 ml ice-cold stop buffer (25 mM Tris-HCl, pH 7.5, 100 mM KCl, 10 mM MgCl₂, and 1 mM DTT) and filtered on MF-membrane nitrocellulose filters (Millipore). The filters were washed three times in the same buffer, dried, and counted in a liquid scintillation counter (Beckman, LS6000IC). The GAP assays also show hydrolysis of [γ-³²P]GTP in a time-dependent manner.

Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.107797/-/DC1

We thank Drs Hugh R. B. Pelham, Kai Simons and Roger Y. Tsien for providing the expression plasmids. We also thank Drs Joel Moss, Randy Haun and Chun-Fang Huang for critical review of this manuscript.