A physically interacting network of evolutionarily conserved clathrin adaptors are involved in traffic between TGN and endosomes. Here, as an approach to define functional relationships between these adaptors in vivo, we analyzed genetic interactions between mutations affecting different adaptors and determined whether particular adaptor localization is dependent on other members of the network. Our results support a model in which Ent3p function is primarily dedicated to Gga-mediated traffic, whereas Ent5p acts with both AP-1 and Gga proteins (1). Localization mechanisms contribute to this functional distinction; Ent3p requires Gga proteins for localization, but Ent5p can localize to a significant extent without either Gga proteins or AP-1.
Our previous coimmunoprecipitation analysis of Ent3p and Ent5p interaction partners revealed Ent5p association with Gga proteins and AP-1, but Ent3p association only with Gga proteins (Duncan et al., 2003
). However, in vitro assays revealed similar levels of Ent3p and Ent5p binding by γ-ear domains of AP-1 and Gga proteins (Duncan et al., 2003
). The synthetic genetic interactions reported here are entirely consistent with the coimmunoprecipitations, providing evidence for the functional significance of the difference in binding partners. Genetic interaction analysis of ent3Δ
indicated primary function in Gga-mediated transport; ent3Δ
deletions did not exacerbate defects in α-factor maturation, but ent3Δ
combined together with apl2Δ
caused α-factor maturation defects not apparent in either single mutant. In contrast, ent5Δ
interactions revealed that Ent5p acted to different extents in both AP-1– and Gga-mediated traffic. Introduction of ent5Δ
into gga1Δ gga2Δ
cells dramatically decreased α-factor maturation, consistent with strong inhibition of AP-1–mediated transport. A subtle but reproducible effect on α-factor maturation was also observed in apl2Δ ent5Δ
cells, suggesting that Ent5p also provides some function in Gga-mediated transport. Overlapping function of Ent3p and Ent5p with Gga proteins could explain the mild synthetic phenotypes of individual ent
mutations combined with apl2Δ
. Supporting this possibility, the triple apl2Δ ent3Δ ent5Δ
mutant was severely compromised for growth. The suggestion that both Ent3p and Ent5p act in vivo with Gga2p is also consistent with physical association of Ent3p and Ent5p with Gga2p detected by coimmunoprecipitations (Duncan et al., 2003
It is somewhat less clear why introducing ent3Δ
into gga1Δ gga2Δ ent5Δ
further debilitated growth if Ent3p acts solely with Gga proteins. One possibility is that Ent3p does not normally function with AP-1 but can substitute to some extent when Ent5p is absent. Consistent with this notion, Ent3p has the ability to bind to the γ-ear domain of Apl4p in vitro (Duncan et al., 2003
). Alternatively, Ent3p could function with AP-1 in a way that might not be revealed by assaying α-factor maturation, for example, as a cargo-specific adaptor.
The synthetic phenotypes of ent3Δ
combined with ent5Δ
also fit predictions from a model in which Ent3p function is primarily dedicated to Gga-mediated transport, whereas Ent5p acts with both Gga proteins and AP-1 (Black and Pelham, 2000
; Costaguta et al., 2001
). Neither deletion alone significantly affects growth or α-factor maturation, presumably because impact of individual deletions on the Gga pathway is compensated by functional redundancy between Ent3p and Ent5p, and inhibition of AP-1–mediated transport is innocuous. However, the double deletion would impede both Gga- and AP-1–mediated transport, thereby accounting for the growth inhibition and maturation defects that were stronger than those caused by complete inactivation of Gga function in gga1Δ gga2Δ
cells. It is worth noting that the growth phenotype of ent3Δ ent5Δ
cells is not as severe as full inactivation of both Gga proteins and AP-1 in the apl2Δ gga1Δ gga2Δ
cells (Costaguta et al., 2001
; Duncan et al., 2003
), suggesting that some level of traffic can occur through Gga- and/or AP-1 in the absence of the Ent adaptors. This finding offers support for the view that AP-1 and Gga adaptors serve a more central role than Ent proteins in TGN/endosome traffic. The inability of overxexpressed Ent3p or Ent5p to suppress the growth phenotype of apl2Δ gga1Δ gga2-ts
cells at the fully restrictive temperature of 38°C is also consistent with the primacy of AP-1 and Gga adaptors.
The different degrees of adaptor colocalization are concordant with the synthetic genetic interactions and the physical interactions. Ent3p preferentially colocalized with Gga proteins compared with AP-1. Furthermore, Ent3p was mislocalized in ggaΔ cells and dispersed rapidly after inactivation of a gga2-ts allele, indicating dependence on Gga proteins for localization. Consistent with these results, the Ent3p ENTH domain lacking the region with γ-ear–binding motifs was not localized. From these findings we propose that Ent3p binding to Gga γ-ear domains is necessary for proper recruitment of Ent3p to sites of CCV formation in vivo. Ent5p, on the other hand, colocalized equally with both adaptors and localized to a substantial extent in cells lacking both AP-1 and Gga proteins, or as a truncated protein lacking γ-ear– and clathrin-binding motifs. Thus, interactions mediated by the Ent5p ANTH domain appear particularly important for Ent5p localization.
We favor the idea that Gga-mediated localization of Ent3p involves recruitment of cytoplasmic Ent3p to membrane-associated Gga proteins. However, we have not reproducibly detected changes in membrane association of Ent3p in mutant cells by differential centrifugation, which we attribute to differences between conditions before and after cell lysis. Thus, it is possible that the changes in localization observed by microscopy reflect alterations in the distribution of Ent3p between different membrane structures rather than between membranes and the cytoplasm. Regardless, the data indicate that Gga proteins are key determinants in Ent3p but not Ent5p localization.
The similar overall architectures of Ent3p and Ent5p do not provide obvious explanations for the differences in localization mechanisms; both proteins contain N-terminal E/ANTH domains capable of lipid binding followed by a region with multiple γ-ear–binding sites (Duncan and Payne, 2003
). In fact, Ent3p and Ent5p bind comparably to Gga and AP-1 γ-ears in vitro, suggesting that preferential association of Ent3p with Gga proteins is influenced by contributions from other factors in vivo (Duncan and Payne, 2003
). One factor could be cargo binding. The SNARE protein Vti1p binds to the Ent3p ENTH domain but not to Ent5p (Chidambaram et al., 2004
). Perhaps cooperative binding to Vti1p or other cargo and Gga γ-ears contributes to selective localization of Ent3p to Gga-containing structures.
Another factor contributing to differences in Ent3p and Ent5p localization could be phosphoinositide binding. Ent3p and Ent5p have been shown to bind phosphoinositides with some preference for PtdInsP and PtdIns[3,5]P2
(Friant et al., 2003
; Chidambaram et al., 2004
; Eugster et al., 2004
). The inability of the Ent3p ENTH domain alone to localize indicates that phosphoinositide binding by this domain is not sufficient for localization. However, if a particular phosphoinositide with affinity for the Ent3p ENTH domain is present at higher concentrations at membrane sites with Gga proteins than sites with AP-1, then Ent3p association with Gga-containing structures could be favored. Unlike the Ent3p ENTH domain, the Ent5p ANTH domain alone retains at least partial localization activity. Given the different modes of phosphoinositide binding by ENTH and ANTH domains, the differences between Ent3p ENTH domain and Ent5p ANTH domain could be attributable to either differences in affinity for a particular phosphoinositide or for different phosphoinositides. Another possibility is that protein binding by the Ent5p ANTH domain contributes to localization. Further analysis will be required to distinguish between these possibilities.
The different genetic and physical interactions we have described for the two TGN/endosome Ent adaptors imply functional differences. One likely distinction is cargo binding. Ent3p can provide a cargo-selective role for Vti1p (Chidambaram et al., 2004
; Hirst et al., 2004
) and perhaps other cargo in Gga-mediated transport. Another difference may lie in the ability of the ENTH domains to deform lipid bilayers. The epsin1 ENTH domain has a capacity to promote membrane curvature that is not observed with the AP180 ANTH domain (Ford et al., 2002
; Stahelin et al., 2003
). If the ENTH and ANTH domains of Ent3p and Ent5p mirror this difference then Ent3p may provide membrane deformation activity that is particularly important in Gga-mediated transport. In this scenario, such activity would not be as critical for AP-1–mediated traffic. Considering this view, it is intriguing that combinations of AP-2 and AP180 lead to curvature of lipid monolayers characteristic of invaginated clathrin-coated pits (Ford et al., 2001
) that are not observed with either adaptor alone. Similarly, AP-1 and Ent5p may foster membrane curvature without substantial input from Ent3p. Ent3p and Ent5p also differ in clathrin-binding properties. Although Ent3p has been reported to bind clathrin (Friant et al., 2003
), direct comparison of clathrin interaction by coimmunoprecipitation of Ent3p and Ent5p revealed substantially more robust binding by Ent5p (Duncan et al., 2003
). We suggest that clathrin binding represents an important Ent5p activity contributed to both AP-1– and Gga-mediated transport processes that is not shared with Ent3p. The ability to bind clathrin and localize through the ANTH domain endows Ent5p with AP-1– and Gga-independent adaptor activity that is likely to account for the ability of overexpressed Ent5p to suppress the growth defect of apl2Δ gga1Δ gga2-47
Our characterization of gga2-ts
alleles revealed clusters of mutations in two regions within the VHS domain. The first, aa 73–102, corresponds to a region defined by nested N-terminal deletions as particularly important for Gga2p function in CPY transport (Mullins and Bonifacino, 2001
). Using SWISS-MODEL (Peitsch, 1996
; Guex and Peitsch, 1997
) and the crystal structure of the human GGA3 VHS domain (Misra et al., 2002
) to model Gga2p, aa 73–102 of Gga2p are predicted to span helices 4 and 5 in the right-handed 8-helix superhelix (unpublished results). Given this positioning in the center of the superhelix, mutations in this region could affect the overall fold of the VHS domain and have a general effect on the domain structure. The second region, aa 134–151, mostly encompasses helix 7, which lays between helices 6 and 8 that directly contact cargo in GGA3. Thus, mutations in this region have the potential to disrupt cargo binding by Gga2p, although the identity of such cargo remains to be established. Another possible consequence of mutations in this region is suggested by adjacent sequences from 152 to 165 predicted to lie in helix 8. This stretch of amino acids contains 5/7 appropriately spaced residues matching the ANTH domain phosphoinositide-binding consensus sequence described by Ford et al. (2002)
. Four of the five conserved residues, including the two lysines, are predicted to lie in a surface-exposed patch based on the GGA3 structure. The importance of the helix 8 sequences is emphasized by the strong conservation of Gga1p and Gga2p in this region (11/14 identical residues; B). Together these observations raise the possibility that Gga2p aa152–165 may interact with phosphoinositides and mutations in helix 8 or the adjacent helix 7 could disrupt this activity. Alternatively, as helix 8 in GGA3, the Gga2p helix 8 could be involved in cargo recognition. Future identification of Gga2p VHS-binding cargo and assays for lipid binding will address these possibilities and determine whether activities involving this region are shared with the less well-conserved sequences in human GGAs.
In summary, we have built a map of the physical and genetic interactions and characterized the localization relationships between members of the network of TGN/endosome clathrin adaptors. Our results support a model in which AP-1 and Gga proteins provide central adaptor function that is elaborated by Ent adaptors (1). Ent3p is recruited by, and interacts preferentially with, Gga proteins, whereas Ent5p functions with both AP-1 and Gga adaptors. This model provides a foundation to understand the contributions of specific adaptor functions to the process of CCV formation at the TGN and endosomes.