In S. cerevisiae synthesis of phosphatidylinositol (3,5)-bisphosphate [PtdIns(3,5)P2] by Fab1p is required for several cellular events, including an as yet undefined step in the ubiquitin-dependent trafficking of some integral membrane proteins from the trans-Golgi network to the vacuole lumen. AP-1 is a heterotetrameric clathrin adaptor protein complex that binds cargo proteins and clathrin coats, and regulates bi-directional protein trafficking between the trans-Golgi network and the endocytic/secretory pathway. Like fab1Δ cells, AP-1 complex component mutants have lost the ability to traffic ubiquitylated cargoes to the vacuole lumen – the first demonstration that AP-1 is required for this process. Deletion mutants of AP-1 complex components are compromised in their ability to synthesize PtdIns(3,5)P2, indicating that AP-1 is required for correct in vivo activation of Fab1p. Furthermore, wild-type protein sorting can be restored in AP-1 mutants by overexpression of Fab1p, implying that the protein-sorting defect in these cells is as a result of disruption of PtdIns(3,5)P2 synthesis. Finally, we show that Fab1p and Vac14p, an activator of Fab1p, are also required for another AP-1-dependent process: chitin-ring deposition in chs6Δ cells. Our data imply that AP-1 is required for some Fab1p and PtdIns(3,5)P2-dependent processes.
The type III phosphatidylinositol phosphate (PtdInsP) kinases are an evolutionarily conserved family of enzymes responsible for the synthesis of phosphatidylinositol (3,5)-bisphosphate [PtdIns(3,5)P2] from phosphatidylinositol 3-phosphate (PtdIns3P) (Cooke, 2002; Michell et al., 2006). In S. cerevisiae, Fab1p is the sole type III PtdInsP kinase, and production of PtdIns(3,5)P2 by Fab1p is essential for several processes, including retrograde transport from the vacuole (characterized by a single swollen vacuole) (Bryant et al., 1998; Dove et al., 2004; Yamamoto et al., 1995); vacuole acidification (Bonangelino et al., 1997; Yamamoto et al., 1995); and ubiquitin-dependent protein trafficking from the trans-Golgi network (TGN) to the vacuole lumen via a prevacuolar multivesicular body (MVB) (Dove et al., 2002; Odorizzi et al., 1998; Shaw et al., 2003). Complementation analysis in S. cerevisiae, and studies in other organisms would suggest that some, if not all of these functions are conserved throughout eukaryotes (Augsten et al., 2002; Cooke, 2002; Ikonomov et al., 2001; McEwen et al., 1999; Meijer et al., 1999; Morishita et al., 2002).
PtdIns(3,5)P2 functions via recruitment of PtdIns(3,5)P2-specific effector proteins (Cooke, 2002; Dove et al., 2004; Eugster et al., 2004; Friant et al., 2003). To date, several PtdIns(3,5)P2 effectors have been described that mediate some of the Fab1p-dependent functions in S. cerevisiae: Svp1p is required for retrograde transport from the vacuole; and Ent3p, Ent5p and Svp3p are required for protein trafficking to the vacuole lumen (Chidambaram et al., 2004; Dove et al., 2004; Eugster et al., 2004; Friant et al., 2003). Thus the functional outputs of Fab1p are both regulated independently and are mechanistically distinct. For example, svp1Δ cells are deficient in retrograde transport from their vacuoles; however, svp1Δ cells, unlike fab1Δ cells, traffic ubiquitylated cargoes correctly and have acidified vacuoles (Dove et al., 2004). The trafficking and vacuole acidification defects of fab1Δ cells are not an indirect consequence of loss of Fab1p-dependent retrograde transport from the vacuole, but the result of failures in other processes.
A corollary of the above observations is that the correct execution of Fab1p-dependent functions will require exquisite temporal and spatial regulation of PtdIns(3,5)P2 synthesis (Cooke, 2002; Dove et al., 2004; Michell et al., 2006). However, our understanding of the mechanisms leading to the activation of Fab1p is poor. Two S. cerevisiae proteins, Vac7p and Vac14p, are required for in vivo activation of Fab1p (Bonangelino et al., 2002; Dove et al., 2002; Gary et al., 2002). Vac7p seems to be unique to fungi; furthermore, its role in Fab1p activation in S. cerevisiae is equivocal, as the severity of the phenotypes displayed by vac7 mutants appears to be strain dependent (Dove et al., 2002; Michell et al., 2006). By contrast, Vac14p has homologues in other model organisms (Bonangelino et al., 2002; Dove et al., 2002; Sbrissa et al., 2004). Deletion of VAC14 causes an almost complete loss of in vivo PtdIns(3,5)P2 synthesis, and consequently vac14Δ cells are phenotypically indistinguishable from fab1Δ cells (Bonangelino et al., 2002; Dove et al., 2002). Furthermore, in mammalian cells a Vac14p homologue is required for PtdIns(3,5)P2 production (Sbrissa et al., 2004), and it is likely that the Vac14p homologues in other organisms will prove to be activators of type III PtdInsP kinases. In S. cerevisiae all PtdIns(3,5)P2-dependent processes appear to require Vac14p; indeed, no function for Vac14p has been found other than to activate Fab1p (Bonangelino et al., 2002; Dove et al., 2002).
The adaptins are a large, well-conserved family of proteins that form heterotetrameric adaptor protein (AP) complexes (Boehm and Bonifacino, 2001; Boehm and Bonifacino, 2002). The AP complexes link vesicle trafficking and protein sorting by binding to both coat proteins, typically clathrin, and cargo at the surface of sorting vesicles (Robinson, 2004). In S. cerevisiae there are three AP complexes: AP-1, AP-2 and AP-3 (Boehm and Bonifacino, 2001; Boehm and Bonifacino, 2002), of which only AP-1 has been shown to interact with clathrin: AP-1 complex gene disruptions exacerbate the defects seen in clathrin-heavy-chain mutants; and both Apl2p and Apl4p (the β- and γ-adaptin subunits of the AP-1 complex, respectively) have been shown to bind clathrin (Phan et al., 1994; Rad et al., 1995; Stepp et al., 1995; Yeung and Payne, 2001; Yeung et al., 1999), whereas S. cerevisiae AP-2 and AP-3 seem to function in a clathrin-independent manner (Cowles et al., 1997; Yeung et al., 1999). Most recent data would support a role for AP-1 in clathrin-coated vesicle (CCV)-dependent bi-directional transport between the TGN and the early endosome (EE) (Hinners and Tooze, 2003), and in S. cerevisiae this function of AP-1 ensures the proper localization of several proteins including Kex2p, Chs3p and Tgl1p (Ha et al., 2003; Valdivia et al., 2002; Yeung and Payne, 2001; Yeung et al., 1999). However, many of the molecular details of AP-1-dependent trafficking have yet to be determined.
The first indication that Fab1p might have a role in AP-1 function comes from yeast two-hybrid analyses showing that the Fab1p activator Vac14p binds to AP-1 (Dove et al., 2002). Here, we characterize further genetic and biochemical interactions between Fab1p and AP-1, and show that Fab1p and AP-1 are required for the transport of ubiquitylated cargoes to the yeast vacuole.
AP-1 complex deletion mutants fail to sort GFP-CPS to the vacuole lumen but are not deficient in other Fab1p-dependent functions
The S. cerevisiae AP-1 complex is a heterotetramer comprised of two large subunits, β-adaptin (Apl2p) and γ-adaptin (Apl4p); one of two medium subunits, μ-adaptin (Apm1p, or Apm2p); and one small subunit, σ-adaptin (Aps1p) (Boehm and Bonifacino, 2001). Since the Fab1p activator Vac14p has been shown to bind Apl4p, it is anticipated that AP-1 is required for one or more Fab1p-dependent functions (Dove et al., 2002), so we investigated whether the five AP-1 complex component deletion mutants are deficient in any of four processes associated with loss of Fab1p lipid kinase activity: ability to grow at 37°C – all AP-1 complex component deletion mutants were able to grow at 37°C, as shown previously (Cowles et al., 1997; Huang et al., 1999) (data not shown) – vacuole morphology, protein trafficking to the vacuole lumen and vacuole acidification (Odorizzi et al., 1998; Yamamoto et al., 1995).
The vacuolar hydrolase carboxypeptidase S (CPS) is an integral membrane protein that traffics from the TGN to the vacuole lumen via the MVB in a ubiquitin- and PtdIns(3,5)P2-dependent fashion – in fab1Δ cells GFP-CPS traffics incorrectly to the limiting membrane of the vacuole (Odorizzi et al., 1998). We investigated whether the AP-1 complex was required for correct sorting of GFP-CPS to the vacuole lumen (Fig. 1A). Similar to fab1Δ cells, GFP-CPS localizes to the limiting membrane of the vacuole in all the AP-1 component deletion mutants (Fig. 1A). All AP-1 component deletion mutants misdirect another cargo, GFP-Phm5p, which also requires PtdIns(3,5)P2 to traffic to the vacuole lumen (not shown) (Dove et al., 2002; Reggiori and Pelham, 2001).
fab1Δ cells contain a single un-lobed, swollen vacuole (Yamamoto et al., 1995). We investigated the vacuole morphology of the AP-1 component deletion mutants (Fig. 1B). Although the vacuoles of both the apl2Δ and apl4Δ mutants are un-lobed, they are not swollen to the same extent as those of fab1Δ cells. The vacuoles of apm1Δ, amp2Δ and aps1Δ cells are wild-type in appearance. Finally, we investigated whether AP-1 is required for vacuole acidification (Fig. 1C). Unlike fab1Δ cells, all AP-1 complex component deletion mutants accumulate quinacrine in their vacuoles, implying no gross defect in vacuole acidification. This result, in combination with the vacuole morphology data, suggest that AP-1 is required for Fab1p functions at the vacuole, so we subsequently concentrated our efforts on investigation of AP-1 and/or Fab1p-dependent trafficking.
UbGFP-CPS and GFP-Sna3p traffic to the vacuole lumen in apl2Δ and apl4Δ cells
Ubiquitylation of CPS and Phm5p is required for their proper trafficking. Fusion of ubiquitin to the N-terminus of GFP-CPS (UbGFP-CPS) and GFP-Phm5p (UbGFP-Phm5p) bypasses the requirement for Fab1p and PtdIns(3,5)P2 for trafficking of these cargoes to the vacuole lumen (Katzmann et al., 2001; Katzmann et al., 2004; Reggiori and Pelham, 2001; Reggiori and Pelham, 2002). In addition, another cargo protein, GFP-Sna3p, sorts to the vacuole lumen via the MVB independently of Fab1p and PtdIns(3,5)P2 and with no requirement for ubiquitylation (Dove et al., 2002; Reggiori and Pelham, 2001). If AP-1 and Fab1p function on the same trafficking pathway, it would be anticipated that UbGFP-CPS, UbGFP-Phm5p and GFP-Sna3p traffic to the vacuole lumen in AP-1 complex mutants. Unlike GFP-CPS, UbGFP-CPS is indeed delivered to the vacuole lumen in fab1Δ, apl2Δ and apl4Δ cells (Fig. 2A). Similar results were also seen using UbGFP-Phm5p (not shown). GFP-Sna3p also traffics to the vacuole lumen in fab1Δ, apl2Δ and apl4Δ cells (Fig. 2B). These data demonstrate that the TGN-to-vacuole sorting machinery is intact in fab1Δ, apl2Δ and apl4Δ cells. Furthermore, the trafficking defects seen in fab1Δ, apl2Δ and apl4Δ cells are identical: these cells incorrectly direct the ubiquitylated cargoes to the vacuole membrane.
The protein trafficking defect of fab1Δ, apl2Δ and apl4Δ cells is different from that of class-E mutants
A crucial step in trafficking to the vacuole lumen is the sorting of cargo proteins into intra-lumenal vesicles (ILVs) at the MVB; mutants of components of the MVB sorting machinery predominantly fall into the class-E vacuole protein sorting (vps) category (Babst, 2005; Katzmann et al., 2003; Odorizzi et al., 1998). Class-E mutants are characterized by an enlarged MVB, referred to as the class-E compartment, in which cargoes such as GFP-CPS are trapped (Babst, 2005; Odorizzi et al., 1998; Reggiori and Pelham, 2001). We investigated the trafficking of GFP-CPS, UbGFP-CPS and GFP-Sna3p in two class-E mutants, vps4Δ and vps27Δ. As previously reported, the majority of GFP-CPS is retained in the class-E compartment in both these mutants, with a small fraction of cargo reaching the vacuole membrane (Fig. 3) (Odorizzi et al., 1998). Unlike AP-1 and fab1Δ mutants, UbCPS and GFP-Sna3p also fail to reach the vacuole lumen in vps4Δ and vps27Δ mutants (Fig. 3), as reported elsewhere (Reggiori and Pelham, 2001; Reggiori and Pelham, 2002; Yeo et al., 2003). It could be argued that the class-E phenotype is masked in fab1Δ cells, so we also looked at trafficking in fab1Δ/vps4Δ cells (Fig. 3). Again, GFP-CPS, UbGFP-CPS and GFP-Sna3p all fail to reach the vacuole lumen, stalling in the class-E compartment. These data demonstrate that the trafficking defects of fab1Δ and AP-1 mutants are distinct from those of class-E mutants, and would imply that neither Fab1p nor AP-1 are absolutely required for correct function of the MVB sorting machinery.
AP-1 complex component deletion mutants have compromised PtdIns(3,5)P2 synthesis in vivo
AP-1 complex component deletion mutants show fab1-like trafficking defects, and AP-1 binds the Fab1p activator Vac14p, so AP-1 might potentially regulate PtdIns(3,5)P2 synthesis. We investigated whether the five AP-1 complex component deletion mutants were compromised in their ability to synthesize PtdIns(3,5)P2. Results from a typical experiment are shown (Table 1), and the combined data for PtdIns(3,5)P2 levels from all our experiments are in Table 2. All AP-1 complex component deletion mutants have a reduced PtdIns(3,5)P2 complement ranging from 70% (apl4Δ) to 81% (aps1Δ) of wild-type levels.
In yeast cells, PtdIns(3,5)P2 synthesis is provoked by both hyperosmotic and alkaline stress (Dove et al., 1997; Mollapour et al., 2006). In response to challenging with 1.1 M sorbitol all AP-1 complex component deletion mutants are compromised in their ability to synthesize PtdIns(3,5)P2, with stress-induced PtdIns(3,5)P2 levels being approximately 50% of that seen in wild-type cells (Table 3). Although the exact role of Fab1p in response to osmotic stress is uncertain (Michell et al., 2006), these data further confirm that the presence of AP-1 is required for full activation of Fab1p lipid kinase activity in vivo.
The reductions of PtdIns(3,5)P2 in AP-1 complex component deletion mutants appear to be relatively modest; however, as already discussed, the various PtdIns(3,5)P2-dependent functions described to date are clearly regulated independently (Cooke, 2002; Gary et al., 1998; Michell et al., 2006). Thus, the modest global loss of PtdIns(3,5)P2 seen in AP-1 complex component deletion mutants could be the result of the complete loss of PtdIns(3,5)P2 at one of its sites of action. If this is the case, it would be anticipated that AP-1 component deletion mutants are deficient in one or more, but not all, PtdIns(3,5)P2-dependent functions (if AP-1 component deletion mutants had lost their entire PtdIns(3,5)P2 complement, it would be expected they would display all the phenotypes of fab1Δ cells) – i.e. exactly the result we find from our phenotypic analyses. To further establish a link between AP-1 function and PtdIns(3,5)P2 we investigated the effects of overexpression of Fab1p in apl2 and apl4 mutants.
Overexpression of Fab1p restores wild-type protein sorting in apl2Δ and apl4Δ cells
Vac14p was first postulated to be an activator of Fab1p lipid kinase activity from the observations that not only do vac14 mutants present with an identical phenotype to fab1Δ cells, but that overexpression of Fab1p overcomes the vacuole morphology and acidification defects of vac14 mutants (Bonangelino et al., 1997). Subsequent work has shown that vac14 mutants have very low levels of PtdIns(3,5)P2 and, because the only known function of Vac14p is to activate Fab1p, it is assumed that overexpression of Fab1p reverts vac14 phenotypes through increased PtdIns(3,5)P2 synthesis (Bonangelino et al., 2002; Dove et al., 2002). As expected, overexpression of Fab1p in vac14Δ cells restores trafficking of GFP-CPS to the vacuole lumen (Fig. 4A). In apl2Δ and apl4Δ cells overexpression of Fab1p also restores trafficking of GFP-CPS to the vacuole lumen (Fig. 4B). Thus, protein trafficking defects of apl2Δ and apl4Δ cells appear to be a result of failure in PtdIns(3,5)P2 synthesis, and would indicate that AP-1 acts upstream of Fab1p.
Clathrin binding to AP-1 is required for trafficking of GFP-CPS to the vacuole lumen
At least some functions of AP-1 require binding to clathrin coats (Yeung and Payne, 2001; Yeung et al., 1999). To investigate whether clathrin binding to AP-1 is required for trafficking of GFP-CPS to the vacuole lumen, we mutated the clathrin-binding sites in APL2 and the putative clathrin-binding sites of APL4 to produce the mutants apl2-1 and apl4-1 (Fig. 5A). The non-clathrin-binding apl2-1 mutant has been characterized previously (Yeung and Payne, 2001). As clathrin-binding motifs are highly conserved, we used two lines of evidence to identify the clathrin-binding sites in Apl4p. First, it has been shown that two di-leucine (DLL) motifs in the hinge region of human and mouse AP-1 γ-adaptin are crucial for clathrin binding (Doray and Kornfeld, 2001). Second, the clathrin-binding sites in Apl4p have been mapped to the hinge region between residues 609 and 678 (Yeung and Payne, 2001), within which there are two DLL motifs at residues 657-59 and 661-63 that are excellent candidates for clathrin-binding sites. We have substituted alanine residues at both DLL motifs to generate the apl4-1 mutant (Fig. 5A). Both apl2-1 and apl4-1 mutants expressed as full-length proteins when assayed by western blotting (not shown). As expected, wild-type genes are able to restore GFP-CPS trafficking to their respective deletion mutants; however, neither apl2-1 nor apl4-1 is able to do so (Fig. 5B). The fact that both apl2-1 and apl4-1 mutants fail to traffic GFP-CPS correctly is interesting, because previous studies would suggest that disruption of clathrin binding to both Apl2p and Apl4p is needed for complete loss of AP-1 function (Yeung and Payne, 2001). In anticipation of this, we constructed an apl2-1/apl4-1 mutant that has the same GFP-CPS trafficking defects as the single apl2-1 and apl4-1 mutants (data not shown). Perhaps the GFP-CPS trafficking assay is more sensitive to loss of AP-1 function than other assays for AP-1 function. Nevertheless, it appears that correct trafficking of GFP-CPS requires AP-1 and clathrin, suggesting that it is a CCV-mediated process.
AP-1-dependent chitin ring deposition requires FAB1 and VAC14
Having established that correct trafficking of CPS requires both PtdIns(3,5)P2 and AP-1, we wanted to investigate whether other AP-1-dependent functions also required PtdIns(3,5)P2. Although several assays for AP-1 function have been established (Phan et al., 1994; Rad et al., 1995; Stepp et al., 1995; Valdivia et al., 2002; Yeung et al., 1999), only one is testable in fab1 mutants – for example, we could not use any assays based on temperature-sensitive mutants (Yeung and Payne, 2001) as fab1Δ cells lyse at high temperature (Yamamoto et al., 1995) – and that is the AP-1-dependent chitin-ring deposition in a chs6Δ strain background (Valdivia et al., 2002). In wild-type cells, the cell wall component chitin is deposited in ring-like structures and can be visualized by staining with Calcofluor White (Fig. 6A), and in apl2Δ, apl4Δ, vac14Δ and fab1Δ cells chitin rings are also clearly visible; by contrast, in chs6Δ cells Calcofluor White staining is diffuse, and chitin rings are absent (Valdivia et al., 2002) (Fig. 6A). Chitin rings can be restored to chs6Δ cells by deletion of AP-1 complex component genes (Valdivia et al., 2002) (Fig. 6B). Likewise, deletion of either FAB1 or VAC14 restores chitin rings to chs6Δ cells, implying that Fab1p and Vac14p, and hence PtdIns(3,5)P2, are required for this AP-1-dependent function. As an additional control we re-introduced wild-type APL2, APL4, FAB1 and VAC14 genes into their respective deletion mutants, and in all cases the chs6Δ phenotype was restored (Fig. 6C).
The data presented here provide the first evidence that Fab1p and AP-1 are required for the same trafficking pathway: trafficking of ubiquitylated cargoes from the TGN to the vacuole lumen. These data are supported by our biochemical analyses: all AP-1 complex component gene deletions have compromised PtdIns(3,5)P2 synthesis, and the trafficking defects of apl2Δ and apl4Δ cells can be recovered by overexpression of Fab1p. Furthermore, we show that another AP-1-dependent process, chitin-ring deposition in chs6Δ cells, also requires Fab1p. It remains a possibility that all AP-1-dependent processes require Fab1p.
Although many studies show that Fab1p and PtdIns(3,5)P2 is required for the correct trafficking of ubiquitylated cargoes via the MVB (Dove et al., 2002; Odorizzi et al., 1998; Reggiori and Pelham, 2001; Shaw et al., 2003), the exact role of Fab1p and PtdIns(3,5)P2 in this process is equivocal. Most mutants that show MVB trafficking defects fall into the class-E vps category (Odorizzi et al., 1998). The majority of the class-E genes code for proteins that assemble into three large protein complexes termed ESCRT I-III (endosomal sorting complexes required for trafficking) that, along with other class-E genes, mediate the recognition and sorting of cargo into ILVs at the MVB (Babst, 2005). All class-E mutants present with similar phenotypes: they have an enlarged MVB (class-E compartment), and all cargoes that normally transit via the MVB become trapped in the class-E compartment (Babst, 2005; Odorizzi et al., 1998; Raymond et al., 1992). Data clearly show that AP-1 and fab1 mutants are not class-E mutants, suggesting that neither Fab1p nor AP-1 has a role in the MVB sorting machinery.
The trafficking defects of AP-1, fab1 and vac14 mutants are very specific: cargoes such as CPS and Phm5p traffic incorrectly to the vacuole membrane, whereas cargoes fused to ubiquitin and other cargoes such as GFP-Sna3p traffic correctly (Dove et al., 2002; Katzmann et al., 2004; Reggiori and Pelham, 2001). To date, this precise phenotype has only been described in a small number of mutants, all with related functions: mutants that fail to synthesize PtdIns(3,5)P2 (Dove et al., 2002; Odorizzi et al., 1998; Reggiori and Pelham, 2001; Shaw et al., 2003), mutants that have defects in cargo ubiquitylation (Katzmann et al., 2001; Katzmann et al., 2004; Morvan et al., 2004; Reggiori and Pelham, 2001), mutants with defective AP-1 function (this study) and mutants of potential PtdIns(3,5)P2 effectors (ent3, ent5 and svp3) (Dove et al., 2004; Eugster et al., 2004; Friant et al., 2003). Furthermore, not only do ent3/ent5 mutants have defective CPS trafficking (Duncan et al., 2003; Eugster et al., 2004; Friant et al., 2003), Ent3p and Ent5p have been linked to AP-1 function: both have been shown to bind the γ-subunit of AP-1, and colocalize with clathrin (Duncan et al., 2003); ent3Δ/ent5Δ cells have compromised alpha-factor maturation (Duncan et al., 2003), an AP-1-dependent process (Yeung et al., 1999); and ent3 and ent5 mutants interact genetically with AP-1 mutants (Costaguta et al., 2006). Although their authenticity as PtdIns(3,5)P2 effectors has been questioned (Michell et al., 2006), Ent3p and Ent5p provide a functional link between AP-1, Fab1p, PtdIns(3,5)P2 and trafficking of ubiquitylated cargoes to the vacuole lumen. Thus, all available data point in the same direction: that Fab1p and PtdIns(3,5)P2 and AP-1 are needed for a common step in the transit of ubiquitylated cargoes from the TGN to the vacuole. Since our data also show that trafficking of ubiquitylated cargoes to the vacuole lumen also requires binding of clathrin to AP-1, we feel that this step is likely to be CCV-mediated trafficking between the TGN and the EE.
In fab1Δ/vps4Δ cells, like in vps4Δ cells, all cargoes are trapped in the class-E compartment, which suggests that Fab1p acts downstream of Vps4p at the MVB. However, in light of other data, we feel that this is unlikely. First, all cargoes, even UbGFP-CPS and GFP-Sna3p that traffic in a Fab1p-independent fashion, stall at the MVB in fab1Δ/vps4Δ and vps4Δ cells, implying that the class-E phenotype is both dominant and independent of fab1 phenotypes. Second, the other proteins identified in the Fab1p-dependent trafficking pathway seem to function upstream of the MVB: AP-1, Ent3p and Ent5p are all associated with trafficking at the TGN and ubiquitylation of cargo proteins must occur before sorting into the MVB, supporting earlier speculation that Fab1p acts upstream of the MVB (Hicke, 2003).
With no obvious role for either Fab1p or AP-1 in protein sorting at the MVB, we have adopted the following working model (Fig. 7). Although the exact role of AP-1 in trafficking to or from the TGN has yet to be elucidated, one current model describes two parallel clathrin-dependent pathways: the Golgi-associated gamma-ear-containing ADP-ribosylation-factor-binding protein (GGA)-dependent pathway and the AP-1-dependent pathway (Black and Pelham, 2000; Ha et al., 2003; Hinners and Tooze, 2003), the latter of which we propose also requires Fab1p and PtdIns(3,5)P2. GGA proteins bind and sort ubiquitylated cargoes, so it is possible that any cargoes ubiquitylated at the TGN will transit on the GGA-dependent route to the MVB (Pelham, 2004; Scott et al., 2004). Un-ubiquitylated cargoes could thus traffic from the TGN in an AP-1–Fab1p- and clathrin-dependent fashion to the EE to be ubiquitylated, almost certainly by Rsp5p (Katzmann et al., 2004; Morvan et al., 2004). On arrival at the MVB ubiquitylated cargoes are recognized by ESCRT I-III, sorted into ILVs and delivered to the vacuole lumen. Disruption of the AP-1–Fab1p pathway could cause cargo to spill over into the GGA- and clathrin-dependent pathway. By bypassing the EE, cargo arrives at the MVB without being ubiquitylated, is not recognized by the MVB sorting machinery, fails to sort into ILVs and is delivered to the limiting membrane of the vacuole. Artificially ubiquitylated cargoes and cargoes that do not require ubiquitylation will of course sort normally in AP-1/fab1 mutants regardless of their trafficking route to the MVB. Furthermore, disruption of the MVB sorting machinery will cause all cargoes to accumulate in the MVB, as seen in the class-E mutants, either in the presence or absence of Fab1p activity. A key feature of our model is that the effects of Fab1p and AP-1 on MVB trafficking are indirect: loss of AP-1 function causes a diversion of cargo away from the EE where it would otherwise become ubiquitylated, consequently causing the trafficking defect because of cargo non-ubiquitylation on arrival at the MVB; there is no loss-of-function at the MVB. Of course, our model assumes that cargo are ubiquitylated en route to the MVB and that there is no ubiquitylation of cargo at the MVB. Furthermore, it is possible that the GGA- and AP-1-dependent routes from the TGN are not as clearly independent as we have presented them, because AP-1 and GGA mutants interact genetically (Costaguta et al., 2006; Costaguta et al., 2001; Duncan et al., 2003), so perhaps the roles of AP-1 and GGA at the TGN are more complex than presented here. Nevertheless, we feel our model provides a useful and testable starting point from which to try to resolve the molecular details of Fab1p- and AP-1-dependent trafficking.
One further problem with our model is that we also need to explain why there is ubiquitylation of cargo proteins in fab1Δ cells (Katzmann et al., 2004; Reggiori and Pelham, 2001; Reggiori and Pelham, 2002). Since cargo proteins are deubiquitylated prior to closure of the ILVs at the MVB, even in wild-type cells only a small proportion of cargo is probably ubiquitylated at any one time (Babst, 2005). In fab1Δ cells some cargo proteins could also be routed to the plasma membrane – there is good evidence that if clathrin-mediated transport from the TGN is blocked, cargo proteins traffic to the plasma membrane (Deloche and Schekman, 2002). Once at the plasma membrane, cargo proteins can be ubiquitylated by Rsp5p (Dunn and Hicke, 2001). Whether this cargo stalls at the plasma membrane or enters the endocytic pathway is uncertain, however, because only a small proportion of cargo would be needed to traffic via the plasma membrane to give the appearance of `wild-type ubiquitylation', it is possible that it would not be detected by fluorescence microscopy. Determining the ubiquitylation status of cargoes and the routes they take to the MVB will help to resolve this complex issue.
In conclusion, we propose that AP-1 is required for Fap1p- and PtdIns(3,5)P2-dependent trafficking of ubiquitylated cargoes to the vacuole lumen. Furthermore, it is possible that other functions of AP-1 require Fab1p. The exact role of Fab1p in AP-1 function will only be determined by further investigation, however, as Apl4p has been shown to bind the Fab1p-activator Vac14p (Dove et al., 2002), and we reproducibly see Fab1p binding to Apl2p by yeast two-hybrid analysis (data not shown), AP-1 possibly activates PtdIns(3,5)P2 synthesis by direct interaction with Vac14p and Fab1p. This would lead to the generation of the pool of PtdIns(3,5)P2 responsible for at least some of the functions of AP-1, and the subsequent recruitment of one or more PtdIns(3,5)P2-effector proteins.
Materials and Methods
Yeast and bacterial media was from Gibco BRL. Complete supplementary dropout medium (CSM) was from Q-biogene or Clontech. Trizma base (Tris), HEPES, PIPES and glycine were from Sigma. Amino acids for yeast culture were from Fluka or Sigma. dNTPs for PCR were from Roche. All DNA-modifying enzymes were from New England Biolabs. Oligonucleotides were from Qiagen. Unless otherwise stated, all other chemicals were from BDH or VWR International.
Yeast strains and plasmids
All yeast strains used for biochemical and cell biological assays were purchased from EUROSCARF (Johann Wolfgang Goethe-University Frankfurt, Germany) and were derivatives of BY4741 (MATa; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0). All single deletions were verified by PCR amplification from genomic DNA and sequencing. pBridge was from Clontech. YCplac111, YEplac181, YCplac33, vps4Δ and fab1Δ/vps4Δ cells were gifts from Stephen Dove. pRS313 was from New England Biolabs. Plasmids expressing GFP-CPS and UbGFP-CPS were gifts from Scott Emr (Katzmann et al., 2004; Odorizzi et al., 1998). Plasmids for expressing GFP-Phm5p, GFP-Sna3p and UbGFP-Phm5p were gifts from Hugh Pelham (Reggiori and Pelham, 2001). Plasmids expressing myc-tagged yeast proteins were constructed as follows. The MET25 promoter and phosphoglycerate kinase (PGK) polyadenylation-termination sequence were amplified by PCR from pBridge (Clontech), and sub-cloned in pZero Blunt (Invitrogen). The fidelity of the PCR was confirmed by sequencing. These fragments were excised by restriction enzyme digestion and ligated into YCplac33 (URA3). This plasmid was linearized, and oligonucleotides coding for the myc-tag (9E10) epitope inserted to give a plasmid with an MET25-Myc-PGK expression cassette. This cassette was then excised and ligated into YCplac111 (LEU2), YEplac181 (LEU2), pRS313 (HIS3) to give three expression plasmids with different auxotrophic markers. Apl2p, Apl4p and Fab1p were expressed from these plasmids as indicated.
Gene cloning and site-directed mutagenesis
APL2 and APL4 genes were amplified by PCR from genomic BY4741 DNA using high-fidelity Pfu polymerase (Stratagene), with primers that introduced 5′ BamHI and 3′ XhoI restriction sites. PCR products were sub-cloned into pZero Blunt (Invitrogen). The fidelity of the PCR amplification was confirmed by sequencing.
Site-directed mutagenesis was performed using the QuikChange® site-directed mutagenesis kit (Stratagene), as directed by the manufacturer's instructions. All mutations were verified by sequencing.
A Leica DM RXA2 microscope was used for all microscopy work. Images were acquired with an ORCA digital camera (Hamamatsu, Japan) and processed in Open Lab (Improvision) and Adobe Photoshop (Adobe).
Yeast cells were grown to 8×106 cells/ml, washed in PIPES-buffered YEPD, pH 6.8, and 80 μM FM4-64 (Molecular Probes) added in 1 ml YEPD. Cells were incubated for 1 hour at 24°C in the dark, after which they were washed twice, and chased for 2 hours in 1 ml YEPD in the dark. Cells were viewed using a Texas Red filter cube.
Yeast strains were transformed with plasmids encoding GFP-CPS, UbGFP-CPS, GFP-Sna3p, GFP-Phm5p or GFP-UbPhm5p as described previously (Reggiori and Pelham, 2001), plated onto SD-ura medium and grown at 24°C. Cells were then inoculated into SD-ura liquid culture, grown overnight at 24°C to a density of 4-6×106 cells/ml and viewed using a GFP filter cube.
Yeast cells were grown to 8×106 cells/ml, harvested by centrifugation and incubated for 5 minutes in the dark at 24°C in 100 mM HEPES/KOH (pH 7.5), 3% (w/v) glucose, 200 μM quinacrine (Sigma). Cells were then washed in HEPES-glucose without quinacrine, re-suspended in 50 μl and viewed using a GFP filter cube.
Calcofluor White staining
Yeast cells were grown to a density of 8×106 cells/ml in YPD liquid media at 30°C, and 200 μl cells were pelleted and fixed with 1 ml 2% formaldehyde for 30 minutes. After centrifugation to pellet the cells, 20 μl of Calcofluor White (1 mg/ml in water) was added and the cells incubated for a further 30 minutes in the dark. Cells were washed twice in water, mounted and viewed using an A4 narrowband DAPI blue filter (340-380 nm excitation wavelength).
In vivo polyphosphoinositide (polyPI) measurement
In vivo polyPI measurements were performed as described previously (Cooke et al., 1998; Dove et al., 1997; Stephens et al., 1991). Briefly, yeast cells were grown in synthetic complete supplementary medium (SD) without inositol (Q-biogene). An overnight culture was diluted to 5×104 cells/ml and 10-20 μCi/ml [3H]-inositol (Amersham) added. Cells were grown to a density of 2-4×106 cells/ml (typically 12-16 hours), after which the cells were manipulated as outlined in the relevant figure legends, and killed by the addition of an equal volume of MeOH. Yeast cells were harvested by centrifugation, disrupted by vortexing with approximately 0.4 g glass beads (425-600 μm, Sigma) in the presence of 200 μl MeOH for 5 minutes, and lipids extracted and deacylated. The lipid head groups were analysed by high-performance liquid chromatography, and peaks detected by liquid-scintillation counting. All data points are presented as total counts per minute (cpm).
We would like to thank Hugh Pelham (MRC LMB, Cambridge, UK) and Scott Emr (Howard Hughes Medical Institute, La Jolla, CA) for reagents; Stephen Dove (University of Birmingham, UK) for reagents and useful discussion; Randy Schekman (Dept Molecular and Cell Biology, University of California, Berkeley, CA) and members of his lab for reagents and technical advice; David Gems (Dept Biology, UCL, UK) for use of equipment; and Kate Bowers (Dept Biochemistry and Molecular Biology, UCL, UK) for critical reading. F.T.C. and J.P.P. were funded by the Wellcome Trust.
- Accepted July 25, 2006.
- © The Company of Biologists Limited 2006