The Function of Yeast Epsin and Ede1 Ubiquitin-Binding Domains During Receptor Internalization

Ubiquitin binding by epsins and Ede1 is not required for receptor internalization

Previous studies from our laboratory suggested that Ede1 and the epsin UIMs share a common function in the process of receptor internalization. Specifically, we demonstrated that deletion of Ent1 UIMs caused a significant internalization defect in cells lacking wildtype epsins and Ede1, but not in cells lacking only epsins (21). To determine whether the functional redundancy between Ede1 and the UIMs of the yeast epsins is due to the Ede1 ubiquitin-binding domain, we analyzed the effect of mutating the Ede1 UBA domain on receptor internalization in ent1Δ ent2Δ ede1Δ (3Δ) cells expressing Ent1 lacking functional UIMs (Ent1^UIMΔ).

To measure receptor internalization in mutant yeast strains we employed an assay that measures the internalization of [³⁵S]-labeled α-factor pheromone by its G protein-coupled cell surface receptor, Ste2. When we performed these assays on 3Δ cells expressing Ent1^UIMΔ and an Ede1 clone harboring a deletion of the C-terminal UBA domain (ede1^UBAΔ), we observed no defect in Ste2 internalization as compared to cells expressing Ede1^WT (Figure 1A). To confirm that deletion of the UBA domain eliminated the ability of Ede1 to bind ubiquitin in this strain background, we tested whether Ede1^UBAΔ expressed in ent1^UIMΔ ent2Δ cells could bind ubiquitin immobilized on Sepharose beads. Although both wildtype Ede1 and Ede1^UBAΔ bound ubiquitin in the presence of a wildtype epsin, Ede1^UBAΔ did not bind ubiquitin in ent1^UIMΔ ent2Δ cells (Figure 1B). In the presence of wildtype Ent1, Ede1^UBAΔ presumably bound to ubiquitin indirectly through the formation of an Ent1-Ede1 complex. However in cells in which neither epsins nor Ede1 carry a UBD, binding of Ede1 to ubiquitin was abolished. These findings indicate that Ent1^UIMΔ and Ede1^UBAΔ are able to mediate the internalization of Ste2 without the ability to bind ubiquitin.

The results described above suggest that the UBA domain is not the region of Ede1 that functionally overlaps with epsin UIMs in receptor internalization. Further support for this idea came from the analysis of the Ede1^{EH 3W-A} protein that carries mutations inactivating each of the three EH domains (26, 27). Unlike Ede1^UBAΔ, Ede1^{EH 3W-A} was unable to rescue Ste2 internalization in 3Δ cells expressing Ent1^UIMΔ (Figure 1A). These surprising results suggest that Ede1 compensates for the function of epsin UIMs not by binding ubiquitin, but by forming protein-protein interactions mediated by EH domains.

The results described above indicate that epsin and Ede1 ubiquitin-binding is dispensable for the internalization of receptors that use ubiquitin internalization signals. To more thoroughly investigate the role of epsin UIMs in receptor internalization, we analyzed both ent1 and ent2 mutants carrying mutations that impair ubiquitin-binding by UIMs, ent1^{UIM S-D} and ent2^UIMΔ (18). As we observed with 3Δ+ent1^UIMΔ, 3Δ cells expressing Ent1^{UIM S-D} were defective for Ste2 internalization. However, when we performed these assays on 3Δ cells expressing Ent2^UIMΔ, we observed only a minor defect in internalization relative to cells expressing wildtype Ent2 (Figure 1C). This finding suggests that there is a functional difference between Ent1 and Ent2 and provides further evidence that efficient receptor internalization can occur in the absence of ubiquitin-binding by Ent1, Ent2 and Ede1.

Internalization of receptors carrying ubiquitin-independent signals requires epsin UIMs

In addition to ubiquitin, yeast plasma membrane proteins can be internalized by a linear peptide signal with the consensus sequence NPFX_(1,2)D, where X is any amino acid (17, 28). Wildtype Ste2 carries a weak version of this signal, GPFAD, that can be converted to a strong signal by mutation to NPFAD (17). We used an ent1Δ ent2Δ ede1Δ strain with a precise deletion of the STE2 open reading frame to test the endocytosis of variants of Ste2 internalized by different sorting signals. Ste2^NPFAD contains the NPFAD linear peptide sorting signal and mutations of all the C-terminal-tail lysine residues (K→R) that serve as sites of ubiquitination. In contrast, Ste2^{All Lys} contains an F394A point mutation that disrupts the linear peptide sorting signal, but leaves ubiquitination sites intact. In 3Δ cells expressing Ent1^WT, both Ste2 variants are internalized to the same extent as exogenously expressed wildtype receptor (Ste2^WT). In 3Δ+ent1^{UIM S-D} cells, internalization of Ste2^WT is defective as observed previously. Furthermore, the internalization of both Ste2^{All Lys} and Ste2^NPFAD is likewise defective in these cells (Figure 2). These results indicate that receptors carrying both ubiquitin-dependent and ubiquitin-independent internalization signals are equally reliant on the epsin UIMs. Thus mutation of the epsin UIMs affects a general step in receptor internalization required for both Ste2 sorting pathways.

Epsin and Ede1 UBDs share an overlapping function with yeast AP180 adaptors

The previously described results suggest that epsin and Ede1 UBDs function to promote protein-protein interactions among components of the endocytic machinery, not primarily as receptors for ubiquitin internalization signals. We observed that the EH domains of Ede1, which bind to NPF motifs in several endocytic proteins (29-31), are important in the absence of epsin UIMs (Figure 1A). To identify proteins that might act with Ede1 EH domains in a protein-protein interaction that is functionally redundant with the epsin UIMs, we performed a high-throughput yeast 2-hybrid screen, using the EH domains (a.a. 1-376) of Ede1 as bait. We identified Ent1 and Ent2 in this screen, which were previously shown to bind EH domains (18). We also identified several additional yeast proteins implicated in endocytosis, including the other yeast monomeric adaptors, Yap1801 and Yap1802. As mentioned previously, the yeast AP180 adaptors function cooperatively with the epsins and harbor many of the epsin functional domains and motifs (14). A two-hybrid interaction between Ede1 EH domains and Yap1802 was previously reported (17).

To determine whether Yap1801 and Yap1802 are Ede1-interacting proteins that might substitute for Ent2 UIM function, we expressed Ent2^UIMΔ in an ent1Δ ent2Δ yap1801Δ yap1802Δ ede1Δ (5Δ) strain. We observed a ∼10 fold decrease in the rate of receptor internalization in 5Δ+ent2^UIMΔ cells as compared to 5Δ+ENT2 cells (Figure 3A). In addition, we tested 5Δ strains expressing wildtype epsins or Ent2^UIMΔ for growth at 30°C and 37°C on rich medium and found that the 5Δ+ent2^UIMΔ strain grew poorly at 30°C and not at all at 37°C (Figure 3B). The expression of wildtype Ede1 in the 5Δ+ent2^UIMΔ cells restored both receptor internalization and growth at 37°C (Figures 3A and 3B). This evidence indicates that Ede1 and the Ent2 UIMs share a common role that is crucial for receptor internalization. Furthermore, because receptor internalization mediated by Ent2^UIMΔ is significantly more rapid in 3Δ cells as compared to 5Δ cells, the yeast AP180 proteins and Ent2 UIMs are likely to perform overlapping functions that are necessary for receptor internalization and cell growth at high temperature.

The Ent2 UIMs and NPF motifs perform partially overlapping functions in endocytosis

Since the Ede1 EH domains and Ent1 UIMs are functionally redundant, we tested whether the Ede1 EH domains would also overlap in function with the Ent2 UIMs. We assayed Ste2 internalization in 5Δ+ent2^UIMΔ cells expressing Ede1^WT or Ede1^{EH 3W-A}. The 5Δ+ent2^UIMΔ ede1^{EH 3W-A} cells exhibited a clear defect in receptor internalization (Figure 4A), indicating that the Ent2 UIMs function redundantly with the Ede1 EH domains.

Ede1 EH domains bind to NPF sequences in Ent1 and Ent2 (18). Therefore, we predicted that Ent2 NPF motif mutations might cause a defect in receptor internalization in combination with Ent2 UIM mutations in the presence of wildtype Ede1. We generated a variant of Ent2 in which both NPF motifs were mutated to NAA (ent2^NPF-NAA), a mutation that eliminates binding of NPF motifs to EH domains (27). We expressed the Ent2^NPF-NAA mutant in ent1Δ ent2Δ yap1801Δ yap1802Δ (4Δ) cells and found that receptor internalization was slightly defective as compared to internalization by the same strain expressing wildtype Ent2. The internalization defect in 4Δ+ent2^NPF-NAA cells was the same as that in 4Δ+ent2^UIMΔ cells. Mutating both UIMs and NPF sequences together (ent2^{UIMΔ,NPF-NAA}) caused a further decrease in the rate of internalization (Figure 4B). These results suggest that epsin ubiquitin binding might serve a function similar to NPF-EH domain interactions in the organization of early endocytic machinery.

The Ede1 UBA domain has a function distinct from that of the Ent2 UIMs

The data presented in Figure 1A suggest that the Ede1 EH domains, rather than the Ede1 UBA domain, are the part of Ede1 that is functionally redundant with the Ent1 UIMs in 3Δ cells. To test what part of Ede1 is functionally redundant with Ent2 in 5Δ cells, we compared receptor internalization in 5Δ+ent2^UIMΔ cells expressing Ede1^UBAΔ, Ede1^{EH 3W-A} or Ede1^WT. The 5Δ cells co-expressing Ent2^UIMΔ and Ede1^UBAΔ were able to internalize α-factor at the same rate as cells co-expressing Ent2^UIMΔ and Ede1^WT (Figure 5A). Therefore, deletion of the Ede1 UBA domain does not affect the ability of Ede1 to compensate for loss of Ent2 UIM function in 5Δ cells, consistent with the observation that the Ede1 UBA domain is not functionally redundant with the Ent1 UIMs.

Also consistent with results indicating a redundant function for Ent1 UIMs and the Ede1 EH domains, the Ede1^{EH 3W-A} mutant only partially rescued the receptor internalization defect of 5Δ cells expressing Ent2^UIMΔ. We generated a mutant of Ede1 lacking functional EH domains and the UBA domain (ede1^{EH 3W-A,UBAΔ}) to test whether deleting the UBA domain would intensify the defect caused by mutating the Ede1 EH domains. Surprisingly, the Ede1^{EH 3W-A,UBAΔ} mutant restored receptor internalization in 5Δ+ent2^UIMΔ cells to the same extent as wildtype Ede1 (Figure 5B). The effects we observed were not due to changes in mutant protein expression or stability, because the steady-state protein levels of wildtype Ede1 and Ede1 mutants did not correlate with defects in receptor internalization (Figure 5C). These observations indicate that the UBA domain does not functionally overlap with the epsin UIMs. Instead, the UBA domain appears to negatively regulate Ede1.

Ede1 is ubiquitinated

Mammalian Eps15 carries C-terminal UIMs instead of a UBA domain and is ubiquitinated. Ubiquitination of Eps15 is believed to negatively regulate the internalization of activated EGF receptor by binding to and inactivating the UIMs in an intramolecular interaction (32). Like Eps15, yeast Ede1 binds to ubiquitin via a domain at its C-terminus that appears to negatively regulate the protein's activity in receptor internalization.

To investigate whether Ede1 might also be regulated by an intramolecular interaction between its UBA domain and an attached ubiquitin, we tested whether Ede1 is ubiquitinated in yeast cells. We immuneprecipitated Ede1 from wildtype and ede1Δ cells expressing an epitope-tagged version of ubiquitin (HA-Ub) from a plasmid. Immunoblotting with anti-HA antibodies detected a precipitated protein that was specific to cells expressing Ede1 and HA-Ub, indicating that Ede1 is a ubiquitinated protein (Figure 6A). This protein migrated as a discrete band and was similar in molecular mass to Ede1, suggesting that the modification was monoubiquitin or a short oligo-ubiquitin chain. Although treatment of mammalian cells with a receptor ligand enhances Eps15 ubiquitination (33), stimulation of yeast cells with α-factor had no effect on the ubiquitination of Ede1 (Figure 6A). Ede1 immuneprecipitated from cells carrying a temperature-sensitive mutation in the catalytic domain of the Rsp5 ubiquitin ligase was not detectably modified at the non-permissive temperature (Figure 6B), providing more evidence that Ede1 is ubiquitinated. Together, these data demonstrate that Ede1, like Eps15, is modified by ubiquitin, consistent with the idea that Ede1 is negatively regulated by an intramolecular ubiquitin-UBA domain interaction.

Media and reagents

Yeast strains were grown in rich medium (2% bactopeptone, 1% yeast extract, 2% glucose supplemented with 20 mg/l adenine, uracil and tryptophan) or synthetic minimal media (YNB; US Biological, Swampscott, MA).

Monoclonal antiserum was prepared against the hemagglutinin (HA) epitope from the 12CA5 hybridoma cell line. Anti-Ede1 polyclonal antiserum was raised in rabbits against a recombinant Ede1 fragment (amino acids 1-400) expressed and isolated from E. coli. This antiserum specifically recognized a ∼200 kDa protein present in lysates prepared from wildtype but not ede1Δ∷TRP1 cells (our unpublished data).

³⁵S-labeled α-factor was purified from yeast as described previously (45).

Strains and plasmids

Strain transformations, growth and genetic manipulations were performed using standard techniques. Strains used are listed in Table 1. The construction of the 3Δ strain used in this study was described previously (21). The 4Δ and 5Δ yeast strains used in this study and the method for exchanging ENT2 plasmids in these strains were described previously (14). Because of an essential function performed by the epsin ENTH domain, strains must carry a form of epsin with a functional ENTH domain (15). The ent1Δ ent2Δ ede1Δ ste2Δ strain was generated by the targeted deletion of STE2 in the 3Δ strain.

The plasmids used in this study are listed in Table 2. The plasmids encoding ENT2^WT (LHP823), ent1^UIMΔ (LHP1431), ENT1^WT (LHP2703) and ent1^{UIM S-D} (LHP2774) were described previously (14). Plasmids encoding the Ste2 variants, STE2^WT (LHP424), ste2^NPFAD (LHP426) and ste2^{All Lys} (LHP427) were also described previously (46). The ent2^UIMΔ (LHP1412) plasmid was generated by precise deletion of sequences encoding both UIMs (a.a. 175-225) in LHP823 using Quikchange™ mutagenesis (Stratagene, La Jolla, CA). The ent2^NPF-NAA (LHP1924) and ent2^{UIMΔ,NPF-NAA} (LHP2790) plasmids were generated by mutating Pro348, Phe349, Pro492, and Phe493 to alanine by Quikchange™ mutagenesis in LHP878 and LHP1412, respectively.

LHP1531 carries a genomic fragment containing EDE1, including 485 base pairs of the promoter and 80 base pairs of the 3′ region, cloned into pRS426 (47). Truncation mutants of EDE1 were generated using Quikchange™ mutagenesis to introduce tandem stop codons at the indicated amino acid. Similarly, ede1^UBAΔ (LHP1532) was generated by the addition of tandem stop codons at amino acid 1334 to precisely delete the C-terminal UBA domain. EH domain point mutants in ede1^{EH 3W-A} (LHP1770) and ede1^{EH 3W-A,UBAΔ} (LHP1992) were generated by mutating Trp56, Trp176, and Trp319 to alanine using Quikchange™ mutagenesis. All mutations were verified by automated DNA sequencing. The expression of mutant proteins was checked by immunoblot analysis (see below).

α-factor internalization assays

³⁵S-labeled α-factor internalization assays were performed at 30°C using the continuous presence method, as described previously (48). Cells were grown to mid-log phase, harvested and pre-incubated at 30°C for 15 min prior to addition of ³⁵S-α-factor. Aliquots of cells (1 × 10⁸) were taken at the indicated time and incubated in either pH6 buffer (40 mM KH₂PO₄, 6.0 mM K₂HPO₄) or pH1 buffer (50 mM Na₃C₆H₅O₇). Samples were collected on glass-fiber filters (VWR, Batavia, IL) and analyzed using a Beckman LS6500 scintillation counter. The percent internalized α-factor for each time point was calculated by dividing the pH1 sample value by the pH6 sample value. Each data point represents an average of three or more assays and error bars represent the standard deviation.

Ubiquitin-binding experiments

Binding of proteins from yeast lysates to ubiquitin-Sepharose was performed as described previously (21) with the following modification: total yeast lysate proteins (2.0 mg) were incubated with 80 μl 25% ubiquitin-Sepharose slurry (Boston Biochem, Cambridge, MA) or Sepharose CL-4B (Pharmacia, Peapack, NJ) in 1.5 ml MES buffer at 4°C for 4 h.

Denaturing immuneprecipitations

Lysates and immuneprecipitations were carried out as described previously with minor changes (45). All buffers were supplemented with 5 mM NEM and 1 mM 1,10-orthophenanthroline to inhibit deubiquitinating enzymes. Following lysis by mechanical agitation using glass beads, SDS was added to 1% final weight by volume and lysates were incubated at 37°C for 10 min. After clarification, lysates were diluted to 2 ml. Immuneprecipitations were carried out at 4°C for 1 h, followed by incubation with Protein A Sepharose CL-4B (Amersham Biosciences, Piscataway, NJ) at 4°C for 30-60 min. Immuneprecipitates were washed as described, eluted by heating to 80-90°C in urea sample buffer and resolved by 6% SDS-PAGE. Immunoblotting was performed as described previously using α-Ede1 and α-HA antisera (39).