Multiple Signals Regulate Trafficking of the Mannose 6-Phosphate-uncovering Enzyme*

The “uncovering enzyme,” which catalyzes the second step in the formation of the mannose 6-phosphate rec-ognition marker on lysosomal enzyme oligosaccharides, resides primarily in the trans -Golgi network and cycles between this compartment and the plasma membrane. An analysis of green fluorescent protein-uncovering enzyme chimeras revealed that the transmembrane seg-ment and the first 11 residues of the 41-residue-cytoplas-mic tail are sufficient for retention in the trans -Golgi network. The next eight residues ( 486 YAYHPLQE 493 ) facilitate exit from this compartment. Kinetic studies demonstrated that the 488 YHPL 491 sequence also medi-ates rapid internalization at the plasma membrane. This motif binds adaptor protein-2 in glutathione S -transfer-ase-uncovering enzyme-cytoplasmic tail pull-down assays, indicating that the uncovering enzyme is endocytosed via clathrin-coated vesicles. Consistent with this finding, endogenous uncovering enzyme was detected in purified clathrin-coated vesicles. The enzyme with a Y486A mutation is internalized normally but accumulates on the cell surface because of increased recycling to the plasma membrane. This residue is required for efficient return of the enzyme from endosomes to the trans -Golgi network. These findings indicate that the YAYHPLQE motif is recognized at several sorting sites, including the trans -Golgi network, the plasma membrane,

N-Acetylglucosamine-1-phosphodiester-␣-N-acetylglucosaminidase, also known as "uncovering enzyme" or UCE, 1 catalyzes the second step in the formation of the Man-6-Ptargeting signal on lysosomal hydrolases. These residues mediate high affinity binding of the hydrolases to mannose 6-phosphate receptors in the trans-Golgi network (TGN), a necessary step in the transport of the hydrolases to lysosomes (1). We have recently shown that UCE itself resides in the TGN and cycles constitutively between this compartment and the plasma membrane (2). This previous study (2) also establishes that the 488 YHPL 491 and C-terminal 511 NPFKD 515 motifs present in the cytoplasmic domain of UCE were important in the trafficking of the enzyme. Mouse L-cells expressing human UCE with a Y488A mutation had 63% of total enzyme activity on the cell surface at steady state compared with 1% of wildtype enzyme activity. The His 510 stop mutant was intermediate with 7% of the enzyme activity at the cell surface. Because this analysis only involved steady-state measurements of enzyme activity, it was not possible to distinguish whether the mutations affected the rate of internalization, recycling of internalized UCE to the cell surface, enhanced escape of the enzyme from the TGN, or some combination of these processes. Furthermore, the molecular mechanism(s) by which the mutations alter trafficking were not explored.
In this study, we have examined the kinetics of internalization of these and additional mutant forms of UCE. We have also analyzed the ability of the cytoplasmic tail of UCE to interact with the plasma membrane adaptor complex AP-2 and with Eps15, an accessory protein that binds NPF sequences and the ␣ subunit of AP-2. Finally, we have examined the role of the cytoplasmic tail in the exit of UCE from the TGN. Our findings indicate that multiple signals regulate the trafficking of UCE at different compartments along the plasma membrane and endosomal and TGN trafficking pathway.

EXPERIMENTAL PROCEDURES
Distribution of UCE Activity-The cellular distribution of UCE activity using the standard assay with [ 3 H]GlcNAc-P-Man␣Me as substrate was performed on a variety of parental and human UCE-expressing mouse L-cells plated in 24-well tissue culture dishes as described previously (2). Assays were performed on cells in buffer that contained either 1% Triton X-100 (total UCE activity) or no detergent (surface UCE activity).
Surface Biotinylation of UCE and Internalization Assay-The surface biotinylation of UCE was carried out by the method described previously (3) on 12-well plates of confluent mouse L-cells transfected with human UCE. The plates were chilled on ice, and the cells were washed with 1 ml of PBS at 4°C and then twice with 1 ml of PBSϩϩ (PBS supplemented with 0.7 mM CaCl 2 and 0.25 mM MgSO 4 ). 500 l of PBSϩϩ containing 1 mg/ml of Sulfo-NHS-SS-biotin (Pierce) were then added for 30 min on ice. After the biotinylation, the reagent was removed, and the cells were quickly washed twice with ice-cold Trisbuffered saline containing 50 mM glycine, pH 7.0, and then twice with Tris-buffered saline alone, pH 7.0. Control plates were kept on ice to allow the measurement of total cell surface-biotinylated UCE. The other plates were quickly brought to 37°C with prewarmed ␣-minimum Eagle's medium containing 10% fetal calf serum and incubated at 37°C for various times to allow uptake of biotinylated UCE followed by rapid cooling on ice to stop uptake and washing twice with ice-cold Trisbuffered saline. The remaining cell surface-biotinylated UCE was stripped of biotin with 2-mercaptoethane sulfonic acid (Sigma) as described by Moll et al. (4). The cells were treated with 500 l of 50 mM 2-mercaptoethane sulfonic acid in 50 mM Tris-HCl, pH 8.7, 100 mM NaCl, and 2.5 mM CaCl 2 at 4°C two times for 20 min before washing twice with ice-cold Tris-buffered saline, pH 7.0. One well was stripped at 0 time (never exposed to 37°C) to correct for the small amount of biotinylated UCE (usually Ͻ10%) that cannot be stripped. The control wells were never stripped. The cells in all wells were lysed in 250 l of lysis buffer (50 mM Tris, pH 7.0, 1% Triton X-100) containing protease inhibitor mixture tablets (Roche Molecular Biochemicals). The lysed cells were incubated 30 min at 4°C before sedimenting at 260,000 ϫ g for 20 min. An aliquot of the supernatant lysate was assayed for total UCE activity, and 200 l were added to a Microfuge tube containing 25 l of packed streptavidin-agarose beads (Pierce) preequilibrated with lysis buffer. The biotinylated UCE uptake was allowed to adsorb to the beads for 2 .5 h with turning at 4°C. The beads were removed by centrifugation at 835 ϫ g for 1 min and washed twice with lysis buffer before being assayed directly for UCE activity. The control beads represent the UCE biotinylated at the cell surface initially. The 0 time represents the UCE activity that could not be stripped and was used to correct the uptake at 37°C for various times. The uptake is expressed as the percentage of the total biotinylated UCE. Values for t1 ⁄2 of endocytosis were derived graphically from semilog plots of % biotinylated UCE remaining at the cell surface versus time using the early time points.
Exocytosis Assay-The exocytosis of internalized UCE was measured for the Y486A UCE mutant by allowing biotinylated UCE uptake as described above for 5 min followed by chilling to 4°C and stripping of the cell surface biotin with 2-mercaptoethane sulfonic acid. The cells were rewarmed to 37°C for various times to allow exocytosis of internalized biotinylated UCE, chilled at 4°C and stripped, and the biotinylated UCE within the cells was measured after cell lysis as described above.
Preparation of GST Fusion Proteins-Full-length cDNAs encoding wild-type, mutant Y488A, and H510Stop (the His 510 stop construct (H510Stop)) UCEs prepared as described previously (2) were used as templates in PCR reactions. The PCR reactions also contained a sense strand primer including an in-frame BamHI restriction site at the 5Ј end (CCTGTCCTGGATCCTGTCCAGAGC) and an antisense strand primer including an XhoI site at the 3Ј end (5Ј-CAAGCTTTCTCGAG-GTGCCACCCC-3Ј). The PCR fragments encoding Leu 473 through the stop codon at 510 or 516 were gel-purified, digested with BamHI and XhoI, and inserted into a similarly digested pGEX-5X-3 plasmid downstream of GST. Other mutations of the cytoplasmic tail in the GST fusion protein constructs were derived from the GST-UCE fusion constructs by QuikChange mutagenesis using the kit from Stratagene and the appropriate mutant primers. All coding sequences created by PCR were verified by sequencing. The fusion proteins were expressed in Escherichia coli BL21 cells and purified as described by Traub et al. (5).
Preparation of Additional Mutants of Full-length UCE and Their Transfection into Mouse L-cells-New mutants in the cytosolic tail of human full-length UCE cDNA were prepared as described previously (2) using the new Stratagene QuikChange XL mutagenesis kit and the appropriate nucleotide primers encoding the mutant sequence according to the instructions by the manufacturer. The mutant cDNAs were transfected into mouse L-cells and stably transfected cells isolated as described previously (2).
Preparation of Cytosol-Cytosol from bovine adrenal glands was prepared as described previously (6). The cytosol was depleted of AP-2 by adsorption with an immobilized anti-AP-2 antibody as described previously (7). Eps15 was depleted from the cytosol by adsorption with immobilized GST-␣ C (␣ C is the C-terminal domain of the ␣ subunit of AP-2) as described previously (5). E. coli BL21 cells expressing GST-␣ C were kindly provided by Dr. Linton Traub, and the fusion protein was purified as above. GST-␣ C can effectively remove Eps15, Epsin, amphiphysin, AP-180, and dynamin from cytosol (5).
GST Fusion Protein Binding Assays-The association of cytosolic proteins with the GST-UCE-cytoplasmic tail fusion proteins was carried out essentially as described previously (5) using bovine adrenal cytosol (5 mg/ml), and 125 g of GST fusion proteins adsorbed to 25 l of packed glutathione-Sepharose in a reaction volume of 300 l. After incubation at 4°C for 1 h, the glutathione-Sepharose beads were pelleted and washed twice with assay buffer (25 mM Hepes-OH, pH 7.2, 125 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM dithiothreitol) containing 0.1% Triton X-100 and twice with PBS before boil-ing in reducing SDS-PAGE sample buffer. An aliquot of the supernatant was also boiled in SDS-PAGE sample buffer, and 1-4% of the total supernatant and 10 -40% of the total pellet for each sample were run on an 8% reducing SDS-PAGE gel. The gel was blotted to nitrocellulose and probed with antibodies to Eps15 and AP-2␣ followed by exposure to appropriate horseradish peroxidase secondary antibody and detection with ECL using X-Omat film (Eastman Kodak Co.). For quantitation of Eps15 and AP-2 binding, the film was scanned using Adobe Photoshop 6.0 and analyzed using NIH Image 1.61 and SigmaPlot 1997 software.
Antibodies-Rabbit antiserum raised to a synthetic peptide of Eps15 was kindly provided by Dr. Ernst Ungewickell (Hannover Medical School, Hannover, Germany). Mouse monoclonal antibody 100/2 against the ␣-adaptin subunit of AP-2 and mouse monoclonal antibody 100/1 against AP-1/2␤ subunits were purchased from Sigma. Mouse monoclonal antibody TD.1 against clathrin heavy chain was kindly provided by Frances Brodsky (UCSF, San Francisco, CA), and rabbit antibody to ␣-mannosidase II was kindly provided by Kelley Moreman (University of Georgia, Athens, GA). The affinity-purified rabbit antibody to cation-independent mannose 6-phosphate receptor (CI-MPR) was a gift from Walter Gregory of this laboratory, and the affinitypurified rabbit anti-human UCE was prepared as described previously (2).
Isolation of Clathrin-coated Vesicles-Clathrin-coated vesicles were isolated by the method of Woodman and Warren (8) from mouse L-cells that were CI-MPR negative and had been stably transfected with bovine CI-MPR (Cc2 cells (9)). 2-ml fractions were collected from the second density gradient centrifugation (Ficoll-D 2 O) from the bottom to the top. Aliquots of each fraction were submitted to SDS-PAGE on an 8% gel, blotted to nitrocellulose, and probed with a mixture of antibodies to CI-MPR, clathrin heavy chain, ␣-Mannosidase II, and both AP-2 ␣ and AP-1/2 ␤. The blot was developed with a mixture of horseradish peroxidase anti-rabbit and horseradish peroxidase anti-mouse secondary antibodies and ECL detection. Another aliquot of 0.5 ml from fractions 3-11 was diluted in MES buffer and pelleted at 260,000 ϫ g for 20 min in a Beckman Optima TL ultracentrifuge. The pellet from each fraction was resuspended in 40 l of 20 mM Tris-HCl, pH 7.0, and 20 l were used for another SDS gel, and 20 l were assayed for UCE activity.
Preparation, Transfection, and Microscopy of Green Fluorescent Protein (GFP)-UCE Transmembrane Domain-cytosolic Tail Fusion Proteins-Nucleotide sequences encoding the transmembrane domain and cytoplasmic tail of UCE were fused to the preprolactin signal sequence-GFP construct and transiently expressed in HeLa cells as described previously (2). Confocal fluorescence microscopy on GFP-UCE-transfected HeLa cells was carried out using stacks of images that were deconvoluted as described previously (2).

Effect of Cytoplasmic Tail
Mutations on UCE Distribution-As the initial approach to further define the determinants in the 41-residue-cytoplasmic tail of UCE ( Fig. 1) that modulate its trafficking, a series of additional mutations in this domain was generated, and their effect on cell surface expression of UCE activity was analyzed. The non-transfected parental mouse L-cells have 1.6% of their endogenous UCE on the cell surface (Table I). Cells transfected with wild-type human UCE have only a slightly greater percentage of activity on the cell surface (2.3-3.2%) despite expressing a 14 -21-fold more total activity. As reported previously (2), UCE with a Y488A mutation had 63% of total activity on the cell surface. The UCE-cytoplasmic tail contains a second tyrosine located two residues upstream from the 488 YHPL 491 sequence. Interestingly, the mutation of this tyrosine to alanine (the Y486A construct) resulted in 26 -36% of the enzyme activity being present on the cell surface in different clones. Mutation of both tyrosines to alanines (the Y486A/Y488A construct) had a small additional effect (68% of activity on the cell surface) over mutation of Tyr 488 alone. Deletion of the outer six residues of the tail (H510Stop) resulted in 7.8% UCE activity being at the plasma membrane, whereas the deletion of the outer 21 residues (the Met 494 stop construct (M494Stop)) did not increase cell surface expression (2.2% on cell surface). Furthermore, the mutation of Tyr 488 to alanine in the context of the M494Stop mutation resulted in 50% cell surface expression versus 63% with the Y488A mutation in the full-length tail. The mutation of Tyr 486 to alanine in the M494Stop background gave rise to only 5.9% of enzyme activity on the cell surface versus 26 -36% when this mutation is made in the full-length tail. When all of the cytoplasmic tail with the exception of 11 residues were deleted (the Tyr 486 stop construct (Y486Stop)), only 16% UCE activity was present on the cell surface, even though this mutant lacks both Tyr 486 and Tyr 488 .
Endocytosis of Biotinylated UCE from the Plasma Membrane-The steady-state level of UCE at the plasma membrane is determined by the rate of endocytosis of the enzyme, the fraction of internalized enzyme that recycles from endosomes back to the plasma membrane, and the rate at which the enzyme exits the TGN and is delivered to the plasma membrane. This delivery may be direct or occur following an initial transport to an endosomal compartment. Thus, mutations in the cytoplasmic tail that increase the steady-state level of UCE on the plasma membrane could be influencing one or more of these trafficking events. To begin to distinguish between these possibilities, we directly measured the endocytosis rate of wildtype UCE and a number of the mutants. The various cell lines were plated on a series of culture plates, and cell surface UCE was tagged with NHS-SS-biotin, a cleavable biotinylation reagent that does not permeate the membrane. Following biotinylation at 4°C, the total cell surface biotinylated UCE was measured by lysing the cells from one dish and capturing the biotinylated UCE on streptavidin-agarose beads. To measure the uptake of biotinylated UCE, the remaining culture plates were brought to 37°C for various times and then rapidly chilled, and residual cell surface-biotinylated UCE was stripped by cleavage of the S-S bond with the non-permeantreducing agent 2-mercaptoethane sulfonic acid. The internalized biotinylated UCE was captured from lysed cells on streptavidin-agarose beads, and UCE activity was assayed directly on the washed beads. Fig. 2 shows the results of the uptake experiments expressed as a percent of the total biotinylated UCE internalized, and Table I lists the calculated internalization rates.
The rate of uptake of wild-type UCE was very rapid with a t1 ⁄2 of 0.65 min. Y486A UCE was endocytosed at approximately the same rate as wild-type UCE (t1 ⁄2 ϭ 0.4 min). This finding indicates that the increased amount of this mutant UCE on the cell surface (36%) is not the consequence of impaired internalization. The double mutant Y486A/M494Stop, which has much less cell surface enzyme (5.9%), had an endocytosis rate that was modestly slower than that of wild-type UCE (t1 ⁄2 ϭ 1.5 min). The rate of internalization of H510Stop UCE was also rapid (t1 ⁄2 ϭ 0.55 min), indicating that the small increase in surface expression of this mutant is not secondary to slow internalization. In contrast, Y486Stop UCE and Y488A UCE were taken up quite slowly (t1 ⁄2 Ͼ30 min). The decreased rate of internalization of Y488A correlates well with its high level of cell surface expression (63%). However Y486Stop accumulates to a much lesser extent on the cell surface (16%), indicating that it is delivered to the plasma membrane at a slower rate than Y488A UCE. Internalized Y486A Rapidly Recycles to the Plasma Membrane-The rapid uptake of Y486A UCE in the presence of an elevated level of cell surface molecules suggested that the internalized enzyme may be recycling from an endosomal compartment back to the plasma membrane. To estimate the exocytosis rate of Y486A UCE from the recycling endosomal compartment, cells expressing this mutant enzyme were biotinylated at 4°C, warmed to 37°C for 5 min to allow endocytosis of the biotinylated UCE, and then chilled at 4°C with subsequent stripping of remaining biotin from cell surface UCE. The plates of these cells were then rewarmed to 37°C for 2, 5, or 10 min before chilling and stripping the biotin from externalized UCE. The internal biotinylated UCE was captured on streptavidin-Sepharose beads and measured. The result shown in Fig.  3 and expressed as the percent of internal biotinylated UCE that was exocytosed does indeed indicate a rapid recycling of UCE back to the cell surface. Taking into account the rapid rate of reendocytosis of externalized biotinylated UCE, which is occurring during the experiment, it can be estimated that the t1 ⁄2 of exocytosis of Y486A UCE is between 1 and 2 min or 2.5-5 times slower than the t1 ⁄2 of endocytosis. The curve is already approaching steady state by the 10-min time point. This recycling experiment could not be performed with the wild-type UCE, because the low level of surface expression prevented adequate labeling.
Binding of AP-2 and Eps15 to UCE-cytoplasmic Tail-Tyrosine 488, which is essential for rapid endocytosis, is part of the YHPL motif that fits the well known endocytosis motif of YXX (where is a bulky hydrophobic residue) (10). This motif has been shown to interact with the 2 subunit of the plasma membrane adaptor complex AP-2 (11). Furthermore, the NPF sequence near the C terminus of the cytoplasmic tail is a candidate to interact with the cytosolic protein Eps15, which binds NPF sequences by its EH (Eps homology) domain (12). To study these potential interactions directly, various GST-UCE tail fusions (Fig. 4) were constructed and expressed and then tested for AP-2 and Eps15 binding using bovine adrenal cytosol as the source of AP-2 and Eps15 in pull-down assays. The GST-wild-type UCE-cytoplasmic tail displayed significant binding of both AP-2 and Eps15, whereas GST alone failed to bind either of the proteins (Fig. 5). The mutation of Tyr 488 to alanine marked decreased AP-2 binding (10% WT) and also impaired Eps15 binding (44% WT). The Y486A mutation resulted in modest decreases in both AP-2 and Eps15 bindings (43 and 55% WT, respectively). The H510Stop fusion protein, which lacks the NPF motif, exhibited almost no binding of both Eps15 and AP-2 (Ͻ1% WT).
These results indicated that binding of the UCE-cytoplasmic tail to the 2 subunit of AP-2 and Eps15 was synergistic in the pull-down assays. This might be expected, because Eps15 also binds to the appendage of the ␣ subunit of AP-2 (13). To further explore this possible synergy, two additional experiments were carried out. In the first experiment, Eps15 was depleted from the cytosol by adsorbing it on immobilized ␣ appendage also referred to as the ␣ C fragment (5). As shown in Fig. 6, this caused a marked decrease in the binding of AP-2 to GST-UCE (27% WT). In the second experiment, AP-2 was depleted from the cytosol by prior adsorption on immobilized anti-AP-2 antibody. Fig. 7 shows that the depletion of the AP-2 was complete, and that the Eps15 was partially depleted perhaps by 2-fold, because it binds to AP-2 (see the intensity of the Eps15 in the supernatants). However, the interaction of GST-UCE wild-type with Eps15 was reduced to 23% of the amount bound in normal cytosol. These depletion experiments show that the binding of AP-2 and Eps15 to the cytoplasmic tail of UCE is synergistic under the conditions of the pull-down assays. UCE Is Found in Clathrin-coated Vesicles-The GST fusion protein pull-down experiments indicate that UCE is internalized at the plasma membrane via AP-2 containing clathrincoated vesicles. To directly test this notion, clathrin-coated vesicles were isolated from mouse L-cells, which express endogenous mouse UCE. After the second density gradient step as detailed under "Experimental Procedures," aliquots of the fractions were subjected to SDS-PAGE, blotted to nitrocellulose, and probed with antibodies to the clathrin heavy chain,

FIG. 2. Rates of endocytosis of wildtype and mutant human UCE expressed in mouse L-cells.
The rate of internalization of surface-biotinylated UCE was measured as described under "Experimental Procedures" and is expressed as the percentage of total biotinylated UCE that is endocytosed. AP-2, AP-1, the CI-MPR, and Golgi ␣-mannosidase II as shown in Fig. 8. In addition, aliquots of the fractions were concentrated and assayed for UCE activity as well as subjected to SDS-PAGE, blotted, and probed with antibody to UCE. It can be seen in Fig. 8 that the clathrin-coated vesicles were mostly present in fractions 3-5, which stained for clathrin, AP-1, AP-2, and the CI-MPR, whereas the remnant of Golgi membranes not removed in the first gradient was concentrated in later fractions as shown by ␣-mannosidase II staining. Uncovering enzyme fractionated primarily with the clathrin-coated vesicles. The result confirms that endogenous UCE is packaged into clathrin-coated vesicles but cannot distinguish AP-2 clathrincoated vesicles from AP-1 clathrin-coated vesicles.
Green Fluorescent Protein Fusion Proteins of Mutant UCE-To explore the effects of the various cytoplasmic tail mutations on the intracellular localization of UCE, we turned to visualization of a variety of GFP-UCE-TM-cytoplasmic tail fusion proteins (Fig. 9A). These GFP fusion proteins were transiently expressed in HeLa cells and localized by confocal imag-ing as shown in Fig. 9B. As previously reported, wild-type GFP-UCE is localized primarily in the TGN like the endogenous UCE (2). GFP-UCE M494Stop was also concentrated in the TGN, whereas GFP-UCE Y488A/M494Stop showed heavy surface staining with residual labeling in the TGN. The GFP-UCE Y486A/M494Stop exhibited less surface staining but more than observed with the GFP-UCE wild-type protein. Furthermore, there was much more peripheral staining, suggestive of endosomal localization. A similar pattern was obtained with GFP-UCE Y486A. The GFP-UCE Y488A/Q492A/E493A/ M494Stop construct was more concentrated in the TGN region than GFP-UCE Y488A/M494Stop, suggesting that Gln 492 -Glu 493 may play a role in the exit of the fusion protein from the TGN.

DISCUSSION
The data presented in this study demonstrate that the trafficking of UCE between the TGN and the cell surface is regulated by multiple signals. These findings are best discussed in terms of our current understanding of trafficking between these compartments. Proteins that exit the TGN may be delivered directly to the plasma membrane by the constitutive pathway or be routed to an intermediate endosomal compartment and then progress to the cell surface (14). Once at the plasma membrane, recycling proteins are mostly internalized via AP-2-containing clathrin-coated vesicles and delivered to early and/or sorting endosomes. Some recycling proteins, such as the CI-MPR and furin, continue on to late endosomes and then return to the TGN (15)(16)(17)(18). Other proteins including TGN38 pass from early endosomes to recycling endosomes from which they can go to the TGN or return to the plasma membrane (18,19). The movement from early and/or sorting endosomes to recycling endosomes and back to the plasma membrane is believed to be a default pathway, whereas retention in the endosomal compartment and delivery to late endosomes and the TGN is thought to be signal-mediated (20).
At steady state, the bulk of UCE is present in the TGN with 2% being on the cell surface and a small amount enroute between these two destinations. Because UCE cycles between the TGN and cell surface, this distribution can only occur if the rate of exit of UCE from the TGN is slower than its rate of return. Our data are most compatible with these processes being controlled by four types of signals, a TGN retention signal, a TGN exit signal, a plasma membrane internalization signal, and a signal(s) for the return of internalized UCE to the TGN. As previously shown, the GFP-UCE Y486Stop construct expressed in HeLa cells is mostly localized to the TGN, indicating that the transmembrane domain, perhaps in conjunction with the neighboring amino acids, serves as a TGN retention signal (2). A similar role for the transmembrane domain of TGN38 in TGN localization has been reported previously (21). Because it is difficult to quantitate the amount of fluorescent protein in subcellular compartments, we expressed the Y486Stop UCE construct in mouse L-cells and analyzed its distribution based on enzyme activity. The majority of the mutant enzyme was in an intracellular compartment with only 16% of the molecules being on the cell surface. As this construct lacks an internalization signal, any molecules that traveled to the cell surface would be expected to be trapped there. This indicates that Y486Stop UCE leaves the TGN very slowly, most likely because it lacks a TGN exit signal. In contrast, the Y488A/M494Stop UCE, which is also very slowly internalized, was found to have 50% of its activity on the cell surface, only slightly less than the 63% surface accumulation of Y488A UCE with a full-length cytoplasmic tail. This finding indicates that Y488A/M494Stop UCE leaves the TGN faster than Y486Stop UCE and suggests that the sequence 486 YAYHPLQE 493 plays a role in the exit from the TGN as well as in endocytosis at the plasma membrane. One potential mechanism is that these residues interact with the Golgi adaptor complex AP-1 that is known to bind to tyrosine-based motifs (22). The AP-1-derived clathrin-coated vesicles are believed to transport their cargo to endosomal compartments (14). In this scenario, UCE would traffic from the TGN to an endosomal compartment and then to the plasma membrane. However, if UCE does exit the TGN via an AP-1 clathrin-coated vesicle, it is surprising that the Y488A mutation does not have a more marked effect on slowing the movement of UCE from the TGN. The mutation of the Gln 492 -Glu 493 residues to alanines appears to slow the exit of GFP-UCE from the TGN, indicating that these amino acids may be either part of the exit signal or have an indirect effect on the signal. It is also probable that residues distal to the 486 YAYH-PLQE 493 sequence influence the rate at which UCE leaves the TGN.
The rate of internalization of wild-type UCE at the plasma membrane is quite rapid (t1 ⁄2 ϭ Ͻ1 min) comparable to the t1 ⁄2 for internalization of dimeric CI-MPR (0.75 min) (23). In comparison, the t1 ⁄2 for internalization of TGN38 is 4.6 min (19), and it is 3 min for the transferrin receptor (24). The fact that UCE is a homotetramer may be significant in this regard, because the complex would have four internalization signals (25). The rate of internalization of the CI-MPR is enhanced 5-fold when it is dimerized by ligand binding (23). The major determinant for the rapid endocytosis of UCE is the 488 YHPL 491 sequence, a typical tyrosine-based internalization signal (10). The mutation of Tyr 488 to alanine dramatically slowed the uptake of UCE, accounting for the striking accumulation of UCE with this mutation at the cell surface. This mutation also markedly decreased AP-2 binding in GST-cytoplasmic tail pull-down experiments, consistent with this interaction being essential for rapid internalization. The role of the C-terminal NPFKD motif in UCE trafficking is not clear. The cytoplasmic tail pull-down experiments showed that the NPF motif binds Eps15, which in turn enhances AP-2 binding through its interaction with the ␣-subunit appendage. However, the H510Stop UCE that is missing the NPFKD sequence is rapidly endocytosed. Thus, even though the UCE-cytoplasmic tail domain in the form of a GST fusion protein reacts synergistically with AP-2 and Eps15 in in vitro pull-down experiments, this three-part complex does not appear to be required for rapid endocytosis. The explanation for this apparent discrepancy may be that AP-2 binding to the tyrosine-based motif in vivo is of much higher avidity than occurs in the in vitro binding assays. As a consequence, synergistic binding with Eps15 may not be required for efficient AP-2 binding inside the cell. In this regard, it has been reported that a mutant of vesicle-associated membrane protein-4 that completely fails to bind AP-1 in an in vitro pull-down assay exhibits only a partial phenotype in vivo (26). Whereas the NPFKD deletion did not impair the rate of internalization, it did result in an increase in the percent of total molecules on the cell surface (7.8 versus 2.3% for the wild-type control). This result raises the possibility that Eps15 binding may facilitate the return of UCE from an endosomal compartment to the TGN. In the absence of Eps15 binding, UCE may undergo an increase in the frequency of recycling to the plasma membrane. Eps15 has been localized to endosomes as well as the plasma membrane, so it could serve functions in addition to facilitating endocytosis (27).
The most unexpected results came from the studies of Y486A UCE. Whereas 26 -36% of these molecules were on the cell surface in different clones, the kinetic studies showed that the mutant enzyme was internalized at the same rate as wild-type enzyme and then rapidly recycled to the plasma membrane.
The intracellular GFP-Y486A chimeric protein was found to be localized in both the TGN and endosomal compartments. These findings suggest that internalized Y486A UCE is not properly sorted in the endosomal system. Rather than being efficiently returned to the TGN as occurs with the wild-type enzyme, a large portion of the mutant molecules appear to enter a recycling pathway that delivers them back to the plasma membrane. The implication is that Tyr 486 is part of a signal for returning UCE from endosomes to the TGN.
In this regard, it is interesting to note that TGN38, which like UCE cycles between the TGN and the plasma membrane, contains a SDYQRL sequence in its cytoplasmic tail that is required for internalization at the plasma membrane and efficient retrieval from the recycling endosome back to the TGN (28 -31). Mutation of the serine residue to alanine or aspartate in full-length TGN38 results in a missorting of endocytosed TGN38 to late endosomes and/or lysosomes and, to a lesser extent, back to the plasma membrane (31). Roquemore and Banting (31) have suggested that the hydroxyl group of the serine has a direct or indirect effect on the ability of the cytoplasmic tail of TGN38 to interact with trafficking and/or sorting machinery at the level of the early endosome. Perhaps Tyr 486 located two residues upstream of the YHPL sequence in the UCE-cytoplasmic tail plays a similar role in endosomal trafficking. However, we also have evidence that in GFP-UCEtransfected HeLa cells, the vesicles in the cell periphery that stained for endogenous TGN46 (the human form of rat TGN38) were not colocalized with GFP-UCE vesicles, although both markers were perfectly colocalized in the TGN (2). Thus, it appears that the route of return of TGN46 and UCE from the plasma membrane to the TGN is not the same. It has also been shown that the cytoplasmic tail of the CI-MPR contains a tyrosine two residues upstream of the YSKV endocytosis motif, which enhances the rate of internalization of that receptor (32). A possible role of the upstream tyrosine in endosomal sorting has not been investigated.
In addition to TGN38, tyrosine-based motifs in the cytoplasmic tails of lamp1 (3) and the polymeric immunoglobulin receptor (33) regulate the sorting of these proteins in endosomes as well as in the TGN and at the plasma membrane. In the cytoplasmic domain of the Gp180/carboxypeptidase D protein, which cycles between the TGN and plasma membrane, the FXXL sequence plays the role of YXXL used for sorting in these other proteins (34). The findings with the 486 YAYHPLQE 493 motif in the UCE-cytoplasmic tail represent another example of a sorting sequence that operates at multiple sites including the TGN, plasma membrane, and endosomes. It will be important to identify the components of the endosomal sorting machinery that interact with this sorting signal to understand how UCE is returned to the TGN.