|Home | About | Journals | Submit | Contact Us | Français|
Ubiquitin (Ub) attachment to membrane proteins can serve as a sorting signal for lysosomal delivery. Recognition of Ub as a sorting signal can occur at the trans-Golgi network and is mediated in part by the clathrin-associated Golgi-localizing, γ-adaptin ear domain homology, ARF-binding proteins (GGA). GGA proteins bind Ub via a three-helix bundle subdomain in their GAT (GGA and target of Myb1 protein) domain, which is also present in the Ub binding domain of target of Myb1 protein. Ubiquitin binding by yeast Ggas is required to direct sorting of ubiquitinated proteins such as general amino acid permease (Gap1) from the trans-Golgi network to endosomes. Using affinity chromatography and nuclear magnetic resonance spectroscopy, we have found that the human GGA3 GAT domain contains two Ub binding motifs that bind to the same surface of ubiquitin. These motifs are found within different helices within the three-helix GAT subdomain. When functionally analyzed in yeast, each motif was sufficient to mediate trans-Golgi network to endosomal sorting of Gap1, and mutation of both motifs resulted in defective Gap1 sorting without defects in other GGA-dependent processes.
Ubiquitin (Ub)1 can serve as a specific lysosomal sorting motif when covalently attached to membrane proteins (1). Ubiquitinated membrane proteins that have passed quality control at the endoplasmic reticulum are delivered to internal membranes of multivesiculated bodies where they can undergo complete degradation by lysosomal proteases. Ubiquitination occurs at many different compartments within the secretory/endocytic pathway, including the cell surface and trans-Golgi Network (TGN) and is likely to occur at endosomes as well (1). Ubiquitin may also act as a specific sorting signal at various intracellular locales, each of which contributes to the final delivery of cargo to the lysosome. Ubiquitin can serve as an internalization signal for proteins such as Ste2 and FcγRII receptor (2, 3). Recognition of Ub as an internalization signal is likely mediated in part by epidermal growth factor receptor pathway substrate 15 (Eps15) and Ent1 and Ent2. In yeast, mutation of the Ub binding domains of these proteins slows the rate of Ste2 internalization (4, 5). Ubiquitin also serves as a specific signal at the endosome for incorporating proteins into the lumenal membranes of multivesicular bodies. Ubiquitin is recognized at the endosome by the Vps27-Hse1 (Hrs-STAM in mammalian cells) complex, the ESCRT-I complex, and possibly the ESCRT II complex (7-9). Ubiquitin can also facilitate the sorting of proteins from the Golgi directly to endosomes, thus forcing ubiquitinated proteins to bypass the cell surface (1). The general amino acid permease, Gap1, and the uracil permease, Fur4, are well studied examples of proteins that undergo this type of sorting (10, 11). When their substrate is limited, these transporters are delivered to the cell surface where they are relatively stable. However, when substrate is available, these cell surface transporters are down-regulated and delivered to the lysosome/vacuole, and newly synthesized transporters are routed directly to endosomes where they are then sorted into the degradative multivesicular body pathway.
Recently, we have shown that the Gga proteins (Golgi-localizing, γ-adaptin ear domain homology, ARF-binding protein) contribute to the sorting of ubiquitinated proteins from the TGN to the endosome (12). Gga proteins are modular clathrin-associated proteins containing an N-terminal VHS (Vps27, Hrs, STAM) domain, an Arf-binding GAT (GGA and Tom1) domain, a clathrin binding hinge region followed by a domain resembling the C-terminal “ear” of γ-adaptin (13). The human genome encodes three GGA genes, GGA1, GGA2, and GGA3, with the latter found in both a short and long form because of alternative splicing (13). In mammalian cells, GGA1 and GGA2 proteins directly bind to dileucine motifs within the cytosolic tails of proteins such as the cation-independent mannose-6-phosphate receptor and facilitate their incorporation into AP-1-coated vesicles targeted for transport to endosomes. The short splice variant of GGA3, which is expressed at high levels (14), and the yeast Gga proteins lack a portion of the VHS domain required for interaction with dileucine motif-containing receptor tails (12), implying that these proteins may serve other functions. Saccharomyces cerevisiae has two GGA genes, GGA1 and GGA2, which are expressed at slightly different levels but are otherwise functionally interchangeable. Deletion of both GGA genes delays endocytic delivery of Gap1 to the vacuole and causes a Vacuolar Protein Sorting (Vps) phenotype characterized by secretion of carboxypeptidase Y (12, 15).
Gga proteins also bind Ub via a subregion of their GAT domains, which is encompassed by a C-terminal three-helix bundle that is distinct from the N-terminal region required for Arf·GTP binding (12, 16). This three-helix bundle region, but not the Arf·GTP binding region, is also present in the Tom1 protein where it contributes to Ub binding (17, 18). Yeast bearing GGA2 carrying a deletion of this three-helix bundle as their sole copy of GGA are specifically defective in diverting Gap1 toward endosomes under nitrogen replete conditions, whereas other GGA-dependent functions remain unaltered (12). Gga proteins are also required to divert ubiquitinated mutant forms of the Pma1 ATPase from the TGN to endosomes (19). These data support a model whereby Gga proteins bind ubiquitinated cargo at the TGN and usher it into transport vesicles targeted to endosomes, thus subverting the delivery of ubiquitinated cargo to the plasma membrane. Of the human GGA proteins, hGGA3 binds most strongly to Ub (12, 20). Interestingly, a point mutation in the three-helix bundle of the hGGA3 GAT domain (L276A) that compromises Ub binding also perturbs the transport of ubiquitinated epidermal-derived growth factor receptor from the cell surface to late endosomes, possibly indicating another role for the recognition of Ub by GGAs (20).
Previous mutagenesis studies have mapped the putative site within the human GGA3 GAT domain that interacts with Ub. This region corresponds to a small area within the third helix centering around Leu-276 and Leu-280 in the long form of hGGA3 (17, 21). However, our previous studies with the yeast two-hybrid assay demonstrated that deletion of this region in yeast Gga1 did not completely block Ub binding (12). Furthermore, failure to find point mutations that abolished Ub binding by random mutagenesis coupled with a “reverse two-hybrid” screen indicated that other regions of the GAT domain might contribute to Ub binding. Using hGGA3 as a model, we have shown that GAT domains can have two distinct Ub binding motifs. In the context of hGGA3 GAT, each is sufficient for Ub binding and each binds to a similar surface on Ub. Using chimeras of yeast Gga2 containing wild-type and mutant human GGA3 GAT domains, we have shown that each motif can contribute to Ub-dependent Gap1 trafficking.
The YAB538 MAT α gga1Δ::TRP1 gga2Δ::HIS3 leu2 ura3 strain derived from BY4735 and BY4704 as previously described (22). Yeast with stably integrated GGA2-HA or gga2-ΔGAT-HA as their sole copy of GGA (PLY3084 and PLY3085) were made as previously described (12). The apl2Δ::Kanr gga1Δ::TRP1 gga2Δ::HIS3 leu2 ura3 strain was made by replacing the APL2 open reading frame with a KanMX cassette in YAB538 cells carrying the URA3-GGA2 plasmid pAB491. To make ggaΔ end3Δ cells, PLY3085 cells were selected for Ade+ Leu- to remove the gga2-ΔGAT-HA allele. These cells grew more slowly than their parental strain and were temperature-sensitive for growth at 37 °C.
C-terminal HA epitope-tagged yeast Gga2p were expressed under their endogenous promoter using the previously described plasmids (pAB491) or versions converted to a LEU2-based plasmid (22). The gga2-ΔArf allele containing mutations in the Arf·GTP binding region was previously described (12, 22). Mutations and fragments of human GGA3 were based on the long isoform (GenBank™ NP 619525). Plasmids expressing GAT domains were made by cloning PCR amplified regions corresponding to hGGA3 residues 166–318 (133–285 of the short hGGA3 isoform) into the pET151D-TOPO vector (Invitrogen) to produce pPL2277 as described (12). Mutant hGGA3 GAT domains were made by PCR recombination and also cloned into the pET151D-TOPO vector (Invitrogen). GST·ARFGTP and GST·ARFGDP plasmids were made by subcloning the open reading frame of human Arf containing a Gln-70 or Asn-31 mutation into pGEX-3X.
Plasmids encoding GGA2 chimeric proteins containing the GAT domains of hGGA3 were made using gap repair recombination of pPL2117 cut with StuI and PstI cotransformed with hGGA3 GAT domains PCR-amplified with the following oligos: GGTAAACCTGAAGATTTGAGGGAAGCTAACAAATTAATGAAAATCATGGTGAAGGAAGACGAGGC, AGCAGAAACATGACTTGGATGTATCTGCGAAGCAGCGTTGGAGTCCCCTTCAATAATTGTTTTGT. pPL2117 is similar to GGA-2-HA (pAB491) except that it contains the hGGA1 GAT domain. This replaces residues 226–321 in yeast Gga2 with residues 200–298 of hGGA3 corresponding to the three-helix bundle C-terminal region of the GAT domain. Plasmids were rescued from yeast and verified by sequencing.
Recombinant Ub was expressed and purified as detailed previously (23, 24). The 15N-labeling medium was from Spectra Stable Isotopes (Columbia, MD). 15N-Ub was mixed with either purified hGGA3 GAT domain or yeast GGA2 GAT domain in 40 mm NaPO4, pH 7.2, containing 10% D2O at 25 °C. Resonance assignments were used as described (25). GAT domains were produced from pET151D-TOPO-based vectors in BL21 bacterial cells upon induction with isopropyl 1-thio-β-d-galactopyranoside. Proteins were purified over nickel-agarose and dialyzed in 40 mm NaPO4, pH 7.2. The Ub surface was generated from Protein Data Bank accession number 1ogw. Peak assignments were made using the Sparky software package (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco). Chemical shift differences were calculated using the formula (0.2 N2 + H2)1/2. To represent the magnitude of chemical shift changes, the values derived from (0.2 N2 + H2)1/2 for each residue were normalized to 0.3439 ppm to represent 100%. This value represented the maximal chemical shift change in Ub bound to wild-type hGGA3 GAT at 64 μm. In subsequent figures, values greater than 75% of this value were plotted as maximal (red), values within 60–75% were plotted as orange, 50–60% yellow, 40–50% green, 30–40% dark blue, 20–30% light blue. A value of 0.068 or less was considered insignificant. Peaks that disappeared altogether were assigned red. As a rough estimate for the level of binding for each GGA-GAT mutant with Ub, we calculated a Total Shift Index (TSI). Although the change in chemical shift for each residue was proportional to the level of binding as determined by titration experiments with wild-type hGGA3 GAT, the results for the wild-type could not be used to estimate the binding affinity of Ub for mutant GAT domains because the relative chemical shift differences for individual residues in Ub varied among the different GAT mutants. In an effort to average out these variations, the TSI was calculated by summing the magnitude of chemical shift differences ((0.2 N2 + H2)1/2) induced in 15N-Ub upon GAT binding. Residues whose peaks disappeared were assigned a value of 200% of the maximal shift difference of wild-type GAT domain (0.3439 ppm). Summed shift changes were then expressed as a function of the number of residues analyzed. Using the peak positions of Glu-18, Ser-20, Thr-22, and Leu-56 in Ub, which do not undergo chemical shift changes when bound to GAT, we estimated a S.E. for measuring peak positions within all the data of 0.008 ppm or 0.75% of the normalized shift.
A GST fusion of Ub, GST·Arf, or GST alone was expressed in bacteria and purified as described (9, 12). GST affinity chromatography experiments were performed with bacterial lysates or purified proteins as described (9, 12).
For ADCB sensitivity assays, yeast were first grown in S.D. ammonia, serially (3-fold) diluted, and plated onto S.D. ammonia plates containing l-azetidine-2-carboxylic acid (ADCB; Sigma) as described previously (12). To control for cell density, cell dilutions were also plated onto S.D. ammonia plates in the absence of ADCB. Cells were photographed after 2 days of growth. For fluorescence assays, GGA2-HA end3Δ, gga2-ΔGAT end3Δ,or ggaΔ end3Δ cells carrying various GGA2-GAThGGA3 chimeric genes on low copy plasmids were transformed with a CUP1-GAP1-GFP plasmid or CUP1-FUR4-GFP plasmid made by homologous recombination from the CUP1-GAP1-GFP plasmid. Cells were grown overnight in S.D. media lacking leucine and uracil and containing 100 μm bathocupoinedisulfonic acid. Cells were pelleted and resuspended in 200 μm CuSO4 in either YPD (for Gap1·GFP analysis) or S.D. media containing 20 μg/ml uracil (for Fur4·GFP analysis). Cells were visualized after resuspension in 100 mm Tris, pH 8.0, 0.2% NaN3, and NaF.
A putative GGA3-GAT structure was made using Swiss-Model (www.expasy.org) using the crystal structure of human GGA1 GAT domain (26). Models were rendered with VMD software (www.ks.uiuc.edu/Research/vmd/). Manual interactive docking and energy minimization was performed with the SYBYL®/Base (Tripos Inc.) software to provide a model of the hGGA3 GAT domain in a complex with Ub.
Fig. 1 shows a sequence alignment of the hGGA3 (human GGA3) GAT domain together with the corresponding region from yGga2 (yeast Gga2) and Tom1, both of which bind Ub. The aligned regions correspond to the C-terminal portion of the GAT domain that forms a three-helix bundle in the case of GGA1. Because hGGA3 is found in both a long and short splice variant, all numbering is referenced to the long form of GGA3 in contrast to our previous work where referencing was to the short form. Also shown is a predicted model of the three-dimensional structure of this region of hGGA3 GAT (Fig. 1A) generated using the known crystal structure of hGGA1, which also binds Ub. Previous studies identified a region within the C-terminal portion of the three-helix bundle subdomain of the hGGA3 GAT domain. Using the residue numbering corresponding to the long splice variant of hGGA3, mutation of Leu-276 to Ser or Ala, Leu-280 to Arg, or Asp-285 to Gly caused loss of Ub binding (16, 20). Similarly, mutation of the Tom1 GAT domain residues L285R and D289G, which correspond positionally to Leu-280 and Asp-285 in hGGA3, also resulted in a loss of Ub binding by Tom1 (17). These residues fall within the C-terminal α-helix of the GAT subdomain. Our previous yeast two-hybrid analysis showed that a VHS-GAT domain of yeast Gga1 lacking this region (residues 1–260) was still able to interact with Ub, albeit less strongly than a VHS-GAT fragment containing the entire GAT domain (12). This 1–260 Gga1 fragment would correspond to an hGGA3 fragment containing only residues 1–241 and thus lacking the C-terminal α-helix. Therefore, we reexamined the GAT domain to more precisely define the regions responsible for Ub binding. Using the alignment in Fig. 1B, we found two regions in separate helices that were relatively conserved among all the GAT domains. Throughout the course of this work we designated the N-terminal-most region Site1 and the other Site2. Site1 encompasses residues 218–232 and Site2 encompasses residues 272–286 of the long form of hGGA3. The modeled tertiary structure of the hGGA3 GAT domain helped predict which residues were likely to be exposed to the surface and thus potentially involved in Ub binding. Intriguingly, we found that both regions could be aligned such that exposed residues shared a high level of identity. Both helical regions showed an acidic residue (Glu) and a bulky hydrophobic residue (Leu) followed by another acidic residue (Glu or Asp) on the same side of the helix (Fig. 1C). The latter two residues in Site2 correspond to Leu-280 and Asp-284, which were found to be important for Ub binding in other studies. Also, the last four residues of both putative motifs end with Ser-Acidic-X-Leu. Because of the similarity of these regions and our previous two-hybrid results, we hypothesized that each formed a motif capable of binding Ub.
As a first test of this hypothesis we mutated conserved residues in Site1 or Site2 of the hGGA3 GAT domain. Site1 was altered with four alanine substitutions from 226RLLSE to 226AALAA and Site2 was altered with four alanine substitutions from 279ILQASD to 279AAQAAA. The mutant or wild-type hGGA3 GAT domains (residues 166–318, which include the N-terminal Arf binding region and the three-helix C-terminal region) were expressed in Escherichia coli behind a hexohistidine and V5 epitope tag and measured for binding to GST·Ub fusion and GST alone (Fig. 1D). As found previously, the wild-type hGGA3 GAT domain bound well to GST·Ub. The GAT domain with mutations in Site1 showed much less Ub binding (~10 times less than wild-type). Mutations in Site2 also diminished binding to Ub, although to a lesser extent than mutations in Site1. A mutant GAT domain carrying four alanine substitutions in both Site1 and Site2 showed no binding to Ub. These data showed that each Site was sufficient for Ub binding and that these regions together accounted for the Ub binding activity of the hGGA3 GAT domain. We next tested whether the residues predicted to be on the external surface of the helices were responsible for Ub binding. We mutated Leu-227 and Glu-230 in Site1 in the context of a mutant GAT containing the four alanine substitution mutations in Site2 that we believed completely inactivate Site2 binding. Fig. 1D shows that this mutant also showed no binding of GST·Ub, indicating the importance of these predicted surface residues in Site1.
From these results we then refined our mutational analysis to evaluate the contribution of the surface-exposed Leu-227 and Glu-230 of Site1 and the Leu-280 and Asp-284 of Site2. In GST·Ub binding experiments, loss of Ub binding was observed when Site1 or Site2 was mutated by alanine substitution of either Leu-227 or Leu-280 singly or in combination with the adjacent acidic residue (Glu-230 or Asp-284). In agreement with data obtained with the four alanine substitution mutants, mutations in Site1 caused a greater loss in Ub binding than those in Site2, although the effects were comparable overall. Combining mutations in Site1 and Site2 led to a complete loss of detectable Ub binding. No difference was observed between double Site1 and Site2 mutants containing alanine replacements of the leucine residues alone (L227A and L280A) or in combination with the acidic residues (E230A and D284A). Previous analysis has shown an L276A or L276S mutation in Site2 can also cause loss of Ub binding (16, 20). However, we did not test these corresponding residues because a hydrophobic residue is not present at this position in Site1 nor is this residue conserved between yGGA2 and Tom1 in Site2. Together, these data show that the surface-exposed residues of both Site1 and Site2 together are required for Ub binding.
To confirm these results we used NMR spectroscopy to monitor chemical shift differences in Ub upon binding to different mutant hGGA3 GAT domains. This was done not only to avoid any potential artifacts from the GST·Ub binding experiments but also to compare the binding surface of Ub used by Site1 and Site2. In principle, these experiments could distinguish whether Site1 and Site2 bound similarly to Ub, supporting the idea that they function as autonomous redundant sites, or whether each provides unique contacts with Ub to form a bipartite binding interface in the complete hGGA3 GAT domain. Purified wild-type and mutant hGGA3 GAT domains were mixed with 15N-labeled Ub (110 μm) and analyzed by 15N HSQC. Fig. 2A shows a titration experiment in which we followed the chemical shift differences Ub had upon binding increasing amounts of hGGA3 GAT domain. As expected, increasing amounts of hGGA3 increased the magnitude of chemical shift changes in Ub. Several residues showed a gradual increase in the titration scheme, and these changes were fitted to a curve from which we were able to estimate a Kd of 10–50 μm. As reported previously, the largest chemical shift changes were observed in three regions of Ub: residues 7–13, 45–48, and 68–72. These regions describe a surface of Ub that includes residues Leu-8, Thr-12, Lys-48, His-68, Val-70, and Arg-72. Several surface residues did not show large changes in chemical shift, including Ile-44 and Arg-42, which are involved in binding UIM domains, and Gln-62 and Glu-64, which are involved in binding TSG101 and Vps23.
We next looked at the binding of various mutant GAT domains in which a single alanine was substituted for the surface-exposed leucine either separately (Leu-227 or Leu-280) or in combination with the adjacent surface acidic residue (Glu-230 or Asp-284) (Fig. 2). The mutant GAT domains were analyzed at a concentration of 64 μm, at which the wild-type GAT induces large chemical shift changes in Ub. We then colorimetrically plotted the magnitude of chemical shift differencesalong the length of Ub and represented those changes on the three-dimensional surface of Ub. Consistent with the GST·Ub binding experiments, we found that GAT domains containing either single or double alanine substitution mutations in either Site1 or Site2 alone were still able to bind Ub. The trend was that GAT domains with an intact Site1 bound to Ub better than GAT domains with only Site2 intact. This is most evident in comparing the Site1 L227A mutant to the Site2 L280A mutant (Fig. 2). Mutation of both leucine residues in Site1 and Site2 resulted in very low binding. Only minor chemical shift changes were observed with this mutant, consistent with the GST·Ub binding experiments in Fig. 1. When both leucines (Leu-227 and Leu-280) and the acidic residues (Glu-230 and Asp-284) were mutated in Site1 and Site2 together, no significant chemical shift differences were observed. These data indicate that the acidic residues contribute to Ub binding because their loss in the context of the L227A,L280A mutant resulted in complete loss of chemical shift differences in Ub. Given the level of sensitivity measured in titration of the wild-type GAT domain, the affinity of the L227A,E230A,L280A,D284A mutant is less than 5% of wild-type.
In general, each of the Ub binding sites on the GAT domain induced similar chemical shift changes in Ub, indicating that they largely interact with the same surface of Ub. There were some differences between the two Sites, particularly in regard to the ratio of chemical shift changes induced in one residue versus another. These differences most likely reflect subtle differences in the geometry of the binding interface rather than radically different binding configurations. Because of the differences between Site1 and Site2 binding, however, we could not accurately use these data to estimate Kd. This was because the HSQC spectral changes induced by GAT domains containing either Site1 or Site2 alone did not match the chemical shift spectra of wild-type GAT. The likely explanation is that the chemical shift changes observed with wild-type GAT domain binding represent a mixture of both Site1 and Site2 bound to Ub. Furthermore, there may be a third configuration in which two Ubs could be bound to a single GAT. Using the magnitude of chemical shift differences observed on the surface of Ub as an indicator, we estimate that mutation of Site1 caused a stronger defect in Ub binding than mutation of Site2 (Fig. 2B). Summing the chemical shift differences across all residues of Ub to estimate a TSI indicated that the ordering of strongest binding to weakest binding was as follows: WT (TSI = 0.172 ppm) >Site2 L280A (TSI = 0.113 ppm) >Site2 L280A,D284A (TSI = 0.076 ppm) >Site1 L227A,E230A (TSI = 0.066 ppm) >Site1 L227A (TSI = 0.060 ppm) >Site1 + 2 L280A,L227A (TSI = 0.027) >Site1 + 2 L280A,D284A,L227A,E230A (TSI = 0.007): all values + 0.008 ppm.
We next evaluated the function of these Ub binding motifs in vivo. Previous experiments examined the role of yeast Gga proteins in the Ub-dependent trafficking of Gap1. When nitrogen is limiting, Gap1 is transported to the cell surface where it can facilitate the transport of a variety of nitrogen sources into the cell. However, when nitrogen is readily available, newly synthesized Gap1 is transported from the Golgi to endosomes, thus bypassing the cell surface (11, 27, 28). This transport step requires ubiquitination of Gap1. Consequently, in relatively rich nitrogen conditions, wild-type cells are resistant to the toxic proline analog ADCB, a substrate of Gap1. In contrast, ggaΔ cells are highly sensitive to ADCB not only because Gap1 is transported to the cell surface but also because its rapid endocytosis and delivery to the vacuole is compromised (12). As a result, there are high levels of Gap1 at the surface in ggaΔ cells. In cells carrying a mutant Gga2 protein lacking the entire three-helix bundle portion of the GAT domain (gga2-ΔGAT), Gap1 is endocytosed to the vacuole normally but newly synthesized Gap1 is delivered to the cell surface, rendering the cells moderately sensitive to ADCB. These data indicated that the Ub binding activity of Gga was specifically responsible for sorting ubiquitinated Gap1 from the TGN to endosomes. The structural data we obtained in the present study allowed us to more precisely correlate loss of Ub binding with loss of Ub-dependent Gap1 sorting to endosomes. This would serve as a more rigorous evaluation of the Ub binding function of Gga proteins as opposed to other functions that could be ascribed to the conserved GAT domain. Therefore, we assessed the function of chimeric yGga2 proteins, which contained either wild-type or mutant hGGA3 GAT domains (Fig. 3A). Using growth on ADCB as an index of Gap1 sorting, we found that mutation of both Site1 and Site2 was required to confer sensitivity to ADCB to the level observed for the gga2-ΔGAT allele. A very slight growth defect was seen when Site1 was mutated, but no defect was observed when Site2 alone was altered. This trend was observed regardless of whether each Ub binding site was altered by alanine substitution of one, two, or four residues. To confirm the effect of these chimeric yGga2-hGGA3 proteins on Gap1 sorting, we assessed distribution of Gap1·GFP in end3Δ cells that expressed chimeric Gga2 proteins with either wild-type or mutant hGGA3 GAT domains as performed previously (12). Gap1·GFP was expressed under the copper inducible control of the CUP1 promoter in cells grown in rich nitrogen conditions where Gap1 is normally sorted directly from the Golgi to the vacuole, thus bypassing the cell surface. As shown in Fig. 3C, wild-type Gga2-GAThGGA3 supported this direct sorting step because newly synthesized Gap1·GFP did not accumulate at the plasma membrane but rather the vacuole. In contrast, Gap1·GFP was easily detected at the cell surface of end3Δ cells when the hGGA3 GAT domain of the Gga2-GAThGGA3 chimera carried the L227A,E230A,L280A,D284A mutations (L·A, E/D ·A) or the more extensive 226RLLSE to 226AALAA and 279ILQASD to 279AAQAAA mutations (4·A). Similar effects were found on the sorting of Fur4 uracil permease (Fig. 3C), which also undergoes direct Ub-dependent sorting from the Golgi to the vacuole when cells are grown in excess uracil (10). In end3Δ cells with wild-type GGA2 (data not shown) or GGA2-hGGA3GAT alleles grown in the presence of uracil, Fur4 was efficiently delivered to the vacuole. However, when the ubiquitin binding sites were altered in the Gga2-hGGA3GAT protein, Fur4·GFP was found at the cell surface. Similar results were found using end3Δ cells carrying the gga2-ΔGAT allele missing the entire three-helix bundle. In addition to cell surface Fur4·GFP, Fur4·GFP was also found in the vacuole, implying that there are other factors besides the Gga that may contribute to the sorting of ubiquitinated Fur4 similar to what we observed previously for Gap1 (12). These phenotypes were not because of altered levels of expression of the different chimeras as all were expressed at the same level by immunoblot analysis (data not shown). Furthermore, mutations in the Ub binding regions of the hGGA3 GAT domain did not affect its ability to bind Arf-GTP (Fig. 3D). We also found that the GGA2-hGGA3GAT allele as well as the mutant chimeras supported other GGA functions. Previously, we found that the gga2-ΔGAT allele as the sole GGA resulted in normal endocytosis and sorting of vacuolar proteases. Other experiments have shown that deletion of GGA1 and GGA2 from yeast is lethal when cells also lack the AP-1 heterotetrameric adaptor complex, encoded in part by APL2 (29). These data imply that under particular circumstances, some Gga functions could be fulfilled by AP-1. Therefore, we also examined the function of the GGA2-hGGA3GAT chimeras in the absence of AP-1 to block this compensatory pathway. Each of the GGA2-hGGA3GAT chimeras was introduced on a LEU2-based plasmid into gga1Δ,gga2Δ,apl2Δ triple mutant cells carrying a URA3-based plasmid containing GGA2. Transformants were grown for several generations and then plated onto 5′-fluoroorotic acid to determine whether loss of URA3-based GGA2 plasmid could be tolerated (Fig. 3B). As expected, no cells could be recovered that lacked GGA and APL2. However, apl2Δ cells containing GGA2-hGGA3GAT chimeras with or without a GAT domain capable of binding Ub were viable and showed no growth defect even at higher temperatures. Furthermore, no vacuolar protein sorting defect as measured by secretion of carboxypeptidase Y was observed in these strains (data not shown). Interestingly, apl2Δ cells carrying a gga2-ΔARF allele incapable of binding Arf·GTP were also viable and showed no carboxypeptidase Y sorting or growth defects (data not shown). These data indicate that mutations that inactivate the Ub binding regions of Gga are very specific for the trafficking of Gap1 but do not affect other GGA-dependent functions.
Using both GST·Ub binding experiments and NMR spectroscopy with recombinant proteins, we found that the three-helix bundle in the C-terminal portion of the hGGA3 GAT domain possesses two similar Ub binding motifs located in separate α-helices. These motifs bind a similar surface of Ub and can work independently to mediate Ub-dependent sorting of the Gap1 transporter. Sequence analysis suggests that these two motifs are also present in yeast Gga2 and are composed of a central surface-exposed hydrophobic residue flanked by acidic residues on the same face of the α-helix. We focused our efforts on testing the role of the leucine (Leu-227 or Leu-280) and the proximal acidic residue (Glu-230 and Asp-284). We did not test the more N-terminal acidic residues (Glu-219 or Asp-273), which are conserved in the two motifs in hGGA3 and Tom1 GAT domain. Thus, all the rules that define this Ub binding motif have yet to be established. Interestingly, the hGGA3 Ub binding motifs partly resemble UIM domains. Both are α-helical, and both contain an exposed hydrophobic region flanked at least on one side with acidic residues. Also, in general terms, the same surface of Ub appears to be engaged in the binding of UIM and GAT domains. However, there are likely to be important differences. For instance, Arg-42 and Ile-44, and Leu-8, respectively, interact with the acidic residues and the hydrophobic middle of the UIM domain. These residues are important for binding UIMs and undergo significant HSQC spectral changes when bound to UIMs (30-32). Ubiquitin mutated at Ile-44 has been shown to abrogate GAT binding. However, there are conflicting reports on the binding ability of Ub mutated at Leu-8 (16, 20). Furthermore, we found only small chemical shift changes for Ile-44 or Arg-42 in our NMR experiments. Clearly the best strategy for characterizing these interfaces is to solve either a crystal structure or solution structure for each Site independently as well as the wild-type GAT domain bound to Ub. NMR experiments focused on this problem are currently underway. Using the information at hand, however, we proposed a working model for how the GAT domain binds Ub (Fig. 4). This model was constructed based on the assumption that a central exposed leucine in both GAT Ub binding sites would fit within the hydrophobic surface of Ub provided by Leu-8, Ile-44, and Val-70, whereas the acidic residues in the GAT domain that flank the exposed leucine would bind to Lys-48 and Arg-72. In this model, the central leucine of the stronger binding Site1 is oriented more toward Leu-8 and Val-70, whereas that of Site2 is rotated more toward Ile-44. Interestingly, in further support of the model presented, significant chemical shift perturbations of Ile-44 in Ub were only observed with hGGA3 GAT mutants in which only Site 2 was intact. One possibility we considered is that the GAT domain might be configured to bind favorably to Ub chains linked via Lys-63 because there is evidence that Lys-63-linked polyubiquitination may be required to mediate direct Golgi-to-vacuole sorting. In our model, a single GAT domain can bind two Ub molecules. However, assuming that both Site1 and Site2 bind Ub in similar orientations, the orientation of two Ubs bound to a single GAT would preclude a Lys-63→Gly-76 linkage.
Our discovery of two relatively equivalent Ub binding motifs within the GAT domain partly conflicts with previous studies that analyzed residues confined to Site2. Like us, Shiba et al. (16) showed that residues Leu-280 and Asp-284 of hGGA3 are important for Ub binding. Alteration of Leu-280 to Arg or Asp-284 to Gly blocked binding to Ub. However, even when both of these residues were changed to alanine, which in our analysis completely inactivates Site2 binding activity, we still clearly observed binding of the GAT domain via Site1. This apparent discrepancy is most likely because of a greater sensitivity for measuring binding in our studies. The L280R and D284G mutations identified by Shiba et al. were originally identified in a reverse yeast two-hybrid screen searching for point mutations that compromised Ub binding; the level of saturation in this screen was probably not sufficient to identify Site1. There are indications from previous studies that are more consistent with the two-Site model proposed here. For instance, one interesting observation for many proteins that contain Ub binding modules such as CUE or UIM is that they themselves can undergo ubiquitination (5, 33). The mechanism for this monoubiquitination is not known, but it may be because the ability to bind Ub allows these proteins to associate with proteins undergoing ubiquitination by E3 ligases. Thus, Ub-binding proteins become ubiquitinated themselves as bystanders owing to their avidity for Ub. Regardless of the exact mechanism, the correlation between Ub binding and ubiquitination is strong. It is therefore instructive that hGGA3 gets ubiquitinated when overexpressed in cells and it is still ubiquitinated when Leu-280 or Asp-284 in Site2 are mutated, albeit to a lesser degree as predicted (34). The observation that these mutant hGGA3 proteins are still monoubiquitinated indicates they still possess residual Ub binding activity.
A similar two-Site mechanism may operate for the Tom1 GAT domain in which the two motifs are conserved (Fig. 1). Mutation of Leu-285 to Arg or Asp-289 to Gly, which correspond to Leu-280 and Asp-284 of Site2 in hGGA3, causes loss of Ub binding by Tom1 (17). However, the inability to measure residual binding from Site1 could again be because it is below the threshold for detection in these studies. The Tom1-L1 protein also known as Srcasm has a GAT domain highly similar to that of Tom1. The putative position of Site2 fits well the consensus binding motif defined here and in other studies. However, it lacks the equivalent acidic residue of Glu-230 present in Site1 of the hGGA3 GAT domain and instead possesses Ala-221. Interestingly, the Tom1-L1 GAT domain still binds GST·Ub but binds Ub-agarose and ubiquitinated proteins from cell lysates at a much reduced level relative to Tom1, possibly because it lacks Site1 (17).
Based on our data, we can present a more refined model for how Ub binds the GGA GAT domains, which should facilitate better formulation of functional experiments in the future (Fig. 4). Precise definition of the Ub binding requirements for GAT domains is becoming increasingly important because the function of the GAT domain itself is more complex than previously appreciated. The GAT domain of Tom1 is not only sufficient for Ub binding but also mediates binding with Toll-interacting protein (Tollip), another endosomal protein that binds Ub via a CUE domain (17, 18). Different residues of the Tom1 GAT domain are used to bind these partners, but their binding is competitive, evoking a mechanism whereby ubiquitinated proteins may be passed sequentially from Tom1 to Tollip (17). The hGGA3 GAT domain binds to Ub and also interacts with TSG101 in a yeast two-hybrid assay (20). The hGGA1 GAT domain also can interact with Ub and TSG101 as well as Rabaptin-5, the latter of which can compete against Ub binding (21, 35). And although each of these GGA GAT interacting factors appears to require a subset of different GAT residues for interaction, mutations such as alanine substitution of Leu-277 in hGGA1 or Leu-276 of hGGA3 result in decreased or completely abrogated interaction of all three binding partners (21). This type of global effect on multiple partners complicates the interpretation of previous in vivo experiments where trafficking of ubiquitinated proteins is followed in cells bearing a hGGA3 L276A mutant or yeast bearing Gga2 lacking the three-helix bundle GAT domain (12, 20). The trafficking effects on Gap1 that we measured here by ADCB sensitivity did strictly correlate with Ub binding. No single point mutation caused complete loss of Ub binding, and mutation of only one Ub binding site resulted in only minor ADCB sensitivity. Only when Site1 and Site2 were mutated together did we see an increase of ADCB sensitivity comparable with deletion of the entire C-terminal portion of the GAT domain. Furthermore, this effect was observed regardless of how many alanine substitution mutations were used to inactivate both Sites. These results strongly indicate that the effects we observed on Gga-dependent Gap1 and Fur4 sorting are strictly because of loss of Ub binding and not loss of other protein-protein interactions. Our failure to detect any other defects with these mutants, even in the absence of the AP-1 adaptor complex that may partly compensate for GGA functions, strongly indicates that Ub binding of Ggas is a function strictly confined to sorting ubiquitinated cargo and is not required for other GGA-dependent processes.
We thank Pat Scott at the University of Minnesota Duluth for helpful comments and suggestions.
*This work was supported in part by National Institutes of Health Grants GM58202 (to R. C. P.) and GM46869 (to A. D. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1The abbreviations and trivial terms used are: