Importance of Cys, Gln, and Tyr from the Transmembrane Domain of Human α3/4 Fucosyltransferase III for Its Localization and Sorting in the Golgi of Baby Hamster Kidney Cells*

Human fucosyltransferase III (EC 2.4.1.65) (FT3wt) is localized in the Golgi of baby hamster kidney cells and synthesizes Lewis determinants associated with cell adhesion events. Replacement of the amino acid residues from the transmembrane domain (TM) Cys-16, Gln-23, Cys-29, and Tyr-33 by Leu (FT3np) caused a shift in enzyme localization to the plasma membrane. The mislocalization caused a dramatic decrease in the amount of biosynthetic products of FT3wt, the Lewis determinants. Determination of the expression levels on the surface with mutants of the enzyme, where one, two, or three of these residues were replaced by Leu, suggested that Cys from the TM was required for the localization of FT3 in the Golgi. Furthermore, Cys-23 and Cys-29 mediated the formation of disulfide-bonded dimers but not higher molecular weight oligomers.In vitro reconstitution of intra-Golgi transport showed that FT3wt was incorporated into coatomer protein (COP) I vesicles, contrary to FT3np. These data suggested that Cys, Gln, and Tyr residues are important for FT3wt sorting into the transport vesicles possibly due to interactions with other membrane proteins.

GTs are distributed along the Golgi from the cis to the trans regions as well as the trans-Golgi network (TGN). They are not committed to a single cisterna but their localization overlaps that of other GTs in adjacent cisternae. The mechanisms underlying the sequential distribution and localization of GTs along the exocytic pathway are not fully understood. The presence of polar residues inside the transmembrane domain (TM) is known to drive ␣-helix association (5) and to be important for the correct localization of protein M in the Golgi (6). The luminal domain of GTs has also been shown to be important for the formation of protein-protein complexes. It has been observed that cis/medial GTs form large insoluble oligomers, contrary to late Golgi GTs that usually exist as dimers (7)(8)(9). Furthermore, changing the targeting of cis/medial Golgi mannosidase II to the endoplasmic reticulum (ER) leads to accumulation of N-acetylglucosaminyltransferase I in the ER (7,10) in a kin-recognition mechanism. On the other hand, it has been observed that increasing the length of the TM of ␣2,6sialyltransferase caused a shift in localization from the Golgi to the plasma membrane (11,12). Recent evidence has also shown that the cytoplasmic tail was essential for the correct Golgi localization of fucosyltransferase I (13) and ␣3-galactosyltransferase (14). The cisternal maturation model, which explains several aspects of transport across the Golgi, proposes that anterograde cargo progresses through the Golgi within the cisternae, whereas Golgi resident proteins are concentrated into retrograde transport coatomer protein (COP) I-coated vesicles, which fuse with previous cisternae, thereby promoting Golgi maturation (15)(16)(17)(18)(19). Golgi resident GTs such as mannosidase II and N-acetylglucosaminyltransferase I (cis/medial Golgi), or ␤4-galactosyltransferase (trans-Golgi and TGN) have been shown to be concentrated into COP I vesicles in a process that is dependent on GTP hydrolysis by ADP-ribosylation factor (Arf)-1 and its effector Arf GTPase-activating protein (GAP) 1 (18,20).
FT3wt is localized in the trans-Golgi and TGN of Baby Hamster Kidney (BHK) cells (4). The molecular basis underlying its correct localization has not been elucidated until recently.
In the present work, we have observed that the replacement of Cys-16, Gln-23, Cys-29, and Tyr-33 from the TM of FT3wt by Leu residues (FT3np) caused a shift of the enzyme from the Golgi to the plasma membrane. This resulted in a decrease of the biosynthetic products of FT3wt, the Lewis determinants. Enzyme dimerization mediated by the TM Cys was observed and was abolished in the mutants where both Cys were mutated. Furthermore, TM Cys were required for the localization of FT3wt in the Golgi. Concomitantly, FT3wt was concentrated into COPI vesicles contrary to the mutant FT3np, indicating that the Cys, Gln, and Tyr residues are important for the sorting of FT3wt into the transport vesicles.

EXPERIMENTAL PROCEDURES
Materials-Dulbecco's modified Eagle's medium (DMEM) was obtained from Sigma. Fetal calf serum, penicillin/streptomycin, and geneticin were obtained from Invitrogen.
Cell Culture-BHK-21B cells were grown in DMEM supplemented with 10% fetal calf serum, 100 units/ml penicillin, 0.1 mg/ml streptomycin. Semi-confluent cells were transfected by the calcium phosphate precipitation method with 5 g of plasmid DNA and selected using several medium exchanges with DMEM containing 10% fetal calf serum and 1.5 mg/ml geneticin, for 2-3 weeks. Single clones were obtained after serial dilution of the selected cell pools and further propagated in the presence of 1.5 mg/ml geneticin. Cells were grown in a humidified incubator at 37°C, under a 5% CO 2 atmosphere.
Plasmid Construction-The mutants pCRFT3np1 to pCRFT3np14 ( Fig. 1) were generated by PCR-based site-directed mutagenesis of the previously described vector pCRFT3 (4) encoding the full-length wildtype enzyme. PCR was performed using the Expand High Fidelity DNA polymerase mixture (Roche Molecular Biochemicals) and the supplied buffer, according to the manufacturer's protocol, at standard concentrations of 0.2 mM of each deoxynucleotide and 0.3 M of each forward primer (TIB MOLBIO, Berlin, Germany), indicated in Table I, and of the reverse mutagenesis primer 5Ј-GGCCTCTCAGGTGAACCAAGCC-GCTATGCT for pCRFT3wt, FT3np, FT3np2, FT3np5, FT3np7, or 5Ј-T-CACTTGCCGCTGTTTGCGACGTAATTTTTGTCGAATCCAGCTCCG-GTGAACCAAGCCGC for the other mutants. PCR conditions were 30 cycles with 15 s of denaturation at 94°C, 20 s of annealing at 50°C, 2 min of elongation at 72°C, and a final elongation for 8 min, at 72°C. DNA fragments were cloned into the eukaryotic expression vector pCR3 using a TA cloning kit (Invitrogen). Positive clones were identified using standard techniques, and mutations were verified by automated DNA sequencing.
Cell Surface Biotinylation-Cell surface proteins were biotinylated essentially according to Tang et al. (21). Cells were grown to halfconfluency, washed three times with PBS, incubated with 0.5 mg/ml sulfo-NHS-SS-biotin (Pierce) in PBS containing 1 mM CaCl 2 and 1 mM MgCl 2 (PBSCM) for 2 ϫ 20 min. Cells were washed two times with PBSCM containing 50 mM NH 4 Cl followed by DMEM with 10% fetal calf serum, and extracted with PBS containing 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 5 mM EDTA, 0.5 g/ml leupeptin, 1 g/ml pepstatin, and 10 mM lysine, centrifuged at 10,000 ϫ g for 15 min at 4°C, and the supernatant was incubated with 50 l of streptavidinagarose suspension for 2 h at 4°C. The agarose beads were washed five times with 10 mM Tris-HCl buffer, pH 8.0, containing 0.5 M NaCl, 1 mM EDTA, 1% Nonidet P-40. The biotinylated proteins were eluted by boiling in SDS sample buffer and analyzed by SDS-PAGE followed by Western blot.
Immunofluorescence Microscopy-Half-confluent cells grown on glass coverslips were washed with PBS containing 0.5 mM MgCl 2 , fixed with 3% (w/v) paraformaldehyde in PBS for 30 min, and permeabilized with 0.1% (w/v) Triton X-100 for 15 min. Fixed cells were blocked with 1% bovine serum albumin in PBS for 1 h, incubated at room temperature for 2 and 1 h with primary and secondary antibodies, respectively. Antibodies were diluted in PBS containing 1% bovine serum albumin and washes were done with PBS. Coverslips were mounted in Airvol and examined on a Bio-Rad MRC1024 confocal microscope or a Leica DMRB microscope. Where indicated, cells were incubated with DMEM  5Ј-CC ATG GAT CCC CTG GGT GCA GCC AAG CCA CAA TGG CCA TGG CGC CGC CTG CTG GCC GCA CTG CTA  TTT CTG CTG CTG GTG GCT GTG CTG TTC TTC TCC CTC CTG CGT GTG TC-3Ј   FT3np1   5Ј-CC ATG GAT CCC CTG GGT GCA GCC AAG CCA CAA TGG CCA TGG CGC CGC TGT CTG GCC GCA CTG CTA  TTT CAG CTG CTG GTG GCT GTG TGT TTC TTC TCC CTC CTG CGT GTG TC-3Ј   FT3np2   5Ј-CC ATG GAT CCC CTG GGT GCA GCC AAG CCA CAA TGG CCA TGG CGC CGC TGT CTG GCC GCA CTG CTA  TTT CAG CTG CTG GTG GCT GTG CTG TTC TTC TCC TA- containing 750 g/ml cycloheximide for 4 h at 37°C prior to fixation. Subcellular Fractionation and Vesicle Preparation-For Golgi isolation, cells were grown to confluency in three 175 cm 2 flasks, inoculated on a 250-ml spinner and grown at 37°C for 2 days. Golgi isolation was done according to Balch et al. (22) with some modifications. Prior to harvest, cells were incubated with 12 g/ml cycloheximide for 10 min at 37°C. Cells were centrifuged at 500 ϫ g for 5 min and washed twice with ice-cold PBS and twice with homogenization buffer (20 mM Hepes-KOH pH 7.0, 0.25 M sucrose). Pellet was resuspended in four volumes of homogenization buffer containing a protease inhibitor mixture (1 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, 1 g/ml antipain, 1 mM benzamidine/HCl, and 40 g/ml phenylmethylsulfonyl fluoride) and homogenized through a 16-m gap on a ball homogenizer. The sucrose concentration was adjusted to 37% (w/w) with a 62% (w/w) stock solution in 20 mM Hepes-KOH pH 7.0. The homogenate was transferred to a SW40 tube and overlaid with 4.5 ml of 35% (w/w) sucrose solution in 20 mM Hepes-KOH pH 7.0 and 2.5 ml of 29% (w/w) sucrose solution in 20 mM Hepes-KOH pH 7.0. After centrifugation at 26,000 rpm for 2.5 h, the Golgi-enriched fraction was collected at the 29/35% (w/w) sucrose interface. Contamination for endoplasmic reticulum was monitored by Western blot with anti-calnexin antibody. The fraction was aliquoted, snap frozen in liquid nitrogen, and stored at Ϫ80°C.
Vesicles were prepared as described before for rat liver Golgi-enriched fractions (20). The reaction mixtures consisted of 250 l containing 20 g of the Golgi-enriched fraction, 5 mg/ml rat liver cytosol, an ATP-regenerating system (20 mM ATP, 100 mM creatine phosphate, and 100 units/ml creatine kinase), 0.5 mM GTP, 1 mM dithiothreitol, and the following protease inhibitors: 0.5 mM benzamidine-HCl, 0.5 mg/ml leupeptin, and 0.2 mg/ml soybean trypsin inhibitor. The mixture was buffered with budding buffer (25 mM Hepes pH 7.0, 115 mM potassium acetate, 2.5 mM MgCl 2 , 1 mM dithiothreitol). After a 30-min incubation, the mixture was centrifuged for 15 min at 23,000 ϫ g, the supernatant was loaded on top of 5 l of 50% (w/w) and 50 l of 30% (w/w) sucrose and further centrifuged for 45 min at 120,000 ϫ g on a TLA55 rotor. The vesicle fraction was found in the 30/50% (w/w) sucrose interface and processed for Western blot analysis or electron microscopy. Immunodepletion of coatomer from cytosol was performed as described before by Lanoix et al. (20).
Electron Microscopy-Vesicle pellets obtained from the budding assay were resuspended in budding buffer and were incubated for 5 min on a formvar and carbon-coated copper grids under a humid environment at room temperature. Grids were washed with PBS, fixed with 4% paraformaldehyde in 0.2 M phosphate buffer, pH 7.4 for 10 min and further washed with PBS and water. Grids were incubated for 10 min on ice with 0.3% uranyl acetate in 1.8% methyl cellulose. Samples were observed on a BioTwin Philips electron microscope.
Protein Extraction and Western Blot Analysis-For dimer detection, BHK cells grown to confluency were resuspended in MNT buffer (25 mM MES, pH 6.5, 1% (w/v) Triton X-100) containing 150 mM NaCl in the presence of 30 mM N-ethylmaleimide (NEM). After incubation for 1 h on ice, samples were centrifuged at 15,000 ϫ g for 15 min at 4°C. The pellet was resuspended in the same volume of MNT, and protein from the pellet and supernatant was precipitated with ethanol and analyzed under non-reducing SDS-PAGE followed by Western blot. Detection was performed by the ECL Plus method (Amersham Biosciences). Quantitation studies of fucosyltransferase on blot were based on a calibration curve of peak area versus FT3 concentration between 1 and 14 ng of standard protein (2). The Scion software was used for peak area determination. After examining the TM of FT3wt we have identified amino acid residues more likely to participate in protein-protein interactions within the hydrophobic environment of the membrane. The amino acids Cys-16 and Cys-29, possibly involved in intramembrane disulfide bond formation, Gln-23, a polar amino acid rarely found in TM (5, 29), Tyr-33, possibly involved in hydrogen bonding, have been identified as the most likely candidates and were mutated to Leu residues, creating the mutant FT3np (Fig. 1).

Cys
Stable BHK cells expressing FT3wt or FT3np were constructed, and the mutant was localized by immunofluorescence microscopy. In non-permeabilized cells, FT3np was detected on the plasma membrane (Fig. 2D) contrary to FT3wt ( Fig. 2A). When FT3np cells were permeabilized with 0.1% Triton X-100, additional fluorescence with a perinuclear distribution characteristic of Golgi, which colocalized with the Golgi marker 58K was detected (Fig. 2E). As a positive control FT3wt showed colocalization with this marker (Fig. 2B). This labeling disappeared for FT3np (Fig. 2F) after incubation with 750 g/ml cycloheximide but not for FT3wt (Fig. 2C) indicating that it corresponded to a fraction of FT3np in transit to the PM. Thus, the mutant FT3np reached its final PM localization after transport through the GA. These results indicated that specific amino acid residues, Cys-16, Gln-23, Cys-29, and Tyr-33, from the TM of FT3wt played a critical role in the correct localization of the enzyme in the Golgi/TGN of BHK cells. Non-transfected BHK cells did not express fucosyltransferase III (data not shown) as previously reported (4).
In order to investigate the relative importance of each of the four mutated amino acids for Golgi localization of FT3wt, we have constructed and stably expressed in BHK cells the mutants FT3np1 to FT3np14 where one, two or three of these residues were replaced by Leu (Fig. 1). We selected cell clones that produced approximately similar amounts of total fucosyltransferase. We then determined their expression level on the cell surface by Western blot of the biotinylated PM proteins isolated with streptavidin-agarose (Fig. 3). FT3wt was detected at low levels on the PM, comparable to the background determined for the control calnexin, an intracellular membrane pro- tein resident in the ER. FT3np was highly detected on the PM thus corroborating the microscopy studies. From the single amino acid mutants FT3np1 to FT3np4, most striking was the presence of FT3np2 (C29L) at higher levels on the PM (88%), which indicated that Cys-29 played an important role for Golgi localization of the enzyme (Fig. 3C). Mutants where this residue was replaced by Leu (FT3np5, FT3np8, FT3np9, FT3np11, FT3np12, FT3np13) were all detected on the cell surface to a variable extent. FT3np6 (Q23L, Y33L) was detected on the PM at levels comparable to FT3wt, which indicated its localization in the Golgi. This suggested important roles for Cys-16 and Cys-29 for the retention process. In total cell extracts FT3np6 appeared as two bands, but on the PM only the heavier one was detected possibly because of more extensive post-translational modifications such as O-glycosylation. FT3np9 (C16L, C29L) appeared on the cell surface as three bands probably because of posttranslational modifications. The quantitation of the PM levels from all the mutants (Fig. 3C) suggested that the four amino acid residues were important for retention of FT3wt in the Golgi.
Plasma Membrane Targeting of FT3np Led to a Significant Decrease in the Biosynthetic Products of Fucosyltransferase III-BHK cells, when transfected with the wild-type form of fucosyltransferase III, acquired new glycosylation properties and started to express the Le a , sLe a , Le x , and sLe x determinants, as detected by immunofluorescence microscopy, using monoclonal antibodies anti-Le a PR5C5 (30), anti-sLe a CA19-9 (31), anti-Le x SH1 (32), and anti-sLe x FH6 (33) (Fig. 4). The Le a is the major de novo product observed since FT3wt has the predominant ␣4 fucosyltransferase activity (Fig. 4A). Though, in vitro, FT3wt preferentially fucosylates sialylated acceptors (2,4), in vivo, the amount of de novo sLe a formed was very low (Fig. 4B), probably because of competition between fucosyltransferase III and endogeneous ␣2,3-sialyltransferase. Probably the latter enzyme is not capable of sialylating fucosylated structures similar to its rat liver homologue (34). Therefore, if fucosyltransferase III acts on type I acceptors before ␣2,3sialyltransferase, Le a determinants are predominantly formed. The BHK-FT3wt cells also became capable of de novo synthesis of Le x and sLe x (Fig. 4, C and D) as previously observed from the detailed carbohydrate analysis of a reporter glycoprotein (35). Staining of Le a (Fig. 4A) where higher fluorescence intensities were observed, showed a vesicular-dispersed pattern through the cytoplasm. These vesicles were not identified; however, they did not consist of endosomes since they did not colocalize with the transferrin receptor (TfR) (data not shown). Both Le x and sLe x (Fig. 4, C and D) showed a perinuclear pattern characteristic of the Golgi and a few vesicles scattered through the cytoplasm, the staining levels being higher for the sialylated Le x motif.
On the other hand, for BHK-FT3np it was observed that the Lewis determinants were formed only to a very minor extent (Fig. 4, E-H), even if the mutant form was active. FT3np had a comparable enzyme activity, 8.62 Ϯ 0.66 U/10 6 cells, to FT3wt, 8.67 Ϯ 2.50 U/10 6 cells. Since FT3np was localized on the PM it did not meet the GDP-Fuc donor substrate, which was concentrated intraGolgi, and, therefore, the biosynthesis of the Lewis determinants was severely inhibited. The low amounts of Lewis determinants that were observed were probably due to the fraction of FT3np in transit through the Golgi. tosyltransferase and ␣1,2 fucosyltransferase (8). However, the protein dimerizes and exists as a mixture of monomers and disulfide-bonded dimers (4). In the present work, we have analyzed the formation of dimers in FT3wt mutants where one cysteine (FT3np2 and FT3np4), both cysteines (FT3np9) or both cysteines, the glutamine and the tyrosine (FT3np) were replaced by leucines. It was observed that when both cysteine residues were changed (FT3np9 and FT3np), the dimer formation was abolished (Fig. 5). In the single cysteine mutants FT3np2 and FT3np4, it was observed that the presence of a single cysteine residue in the TM was enough for the formation of disulfide-bonded dimers. No other higher molecular weight complexes mediated by disulfide bonds were observed, probably due to steric hindrance of the catalytic domain.
FT3wt Is Incorporated into Transport Vesicles in Contrast to FT3np-The cisternal maturation model is currently accepted to explain some properties of the transport through the Golgi (16,36). It has been shown that cargo progresses through the Golgi together with the cisternae (17) whereas Golgi resident proteins, such as GTs, are retrieved into retrograde transport vesicles (20). We have isolated a Golgi fraction from the BHK-FT3wt and BHK-FT3np cells and induced the formation of vesicles, which we have separated in a sucrose gradient and collected from the 30/50% interface according to the method previously described (20). The assay was performed at 4 and 37°C, and the vesicle population was stained with uranyl acetate and observed by negative staining electron microscopy. At 37°C (Fig. 6A), a considerable number of vesicles was detected contrary to that observed at 4°C (Fig. 6B). These vesicles appeared irregular in shape because they were produced in the presence of GTP, and a large proportion of coatomer had already been released. The vesicles were analyzed by Western blot for several protein markers (Fig. 6C). It was observed that ϳ7% of the starting amount of FT3wt used in the assay was incorporated into the vesicles at 37°C. The medial-Golgi soluble NEM-sensitive soluble factor attachment receptor (SNARE) GS28 (18) was also incorporated (ϳ12% of the starting amount) into the vesicles, as well as the p24␥3, p24␣2, p24␦1, and p24␤1 subunits of the p24 complex (10 -20% of the starting material), known to recycle between the ER-Golgi intermediate compartment (ERGIC) and early Golgi compartments (37). In contrast, FT3np did not show a difference in behavior between 4 and 37°C and was not incorporated into vesicles at 37°C, similar to that found for the TfR or the soluble anterograde cargo ApoE.
Preliminary results have indicated that the isolated transport vesicles are COP I vesicles. When the budding assay was done in the presence of coatomer-depleted cytosol, it was observed that FT3wt was not incorporated in the vesicles at 37°C. These results were in agreement with those previously described for other Golgi resident proteins: ␣-mannosidase, N-acetylglucosaminyltransferase I, and ␤4-galactosyltransferase (20). DISCUSSION In this article, we present evidence that four amino acid residues, Cys-16, Gln-23, Cys-29, and Tyr-33, from the TM of FT3wt are required for its localization in the Golgi. The mutation of these amino acid residues to Leu caused a shift in fucosyltransferase III localization from the trans-Golgi/TGN to the plasma membrane, as detected by immunofluorescence microscopy and cell surface biotinylation. In general, there is no clear evidence for the influence of each of these domains, separately, for their correct localization. Our finding that for FT3wt, four amino acid residues from the TM are sufficient for its localization in the Golgi, provides a tool to elucidate the molecular mechanism underlying its Golgi retention/retrieval.
Transfection of BHK cells with FT3wt, which is targeted to the trans-Golgi and the TGN (4), conferred the cells the new glycosylation capacity to perform the biosynthesis of the Lewis determinants, Le a , Le x , and sLe x . However, when the enzyme was targeted to the PM the biosynthesis of the Lewis determinants was dramatically reduced even if FT3wt and FT3np had similar activities and were expressed in similar amounts. Since FT3np was in transit through the Golgi, its time of residence in this compartment should be much lower than that of FT3, and therefore the efficiency of Lewis determinants biosynthesis observed was much lower for FT3np than for FT3wt. Similarly, previous results have shown that soluble secretory forms of glycosyltransferases with low residence times in the Golgi, such as fucosyltransferase VI (sFT6, Ref. 35) and soluble ␣3- galactosyltransferase (s␣3GT, Ref. 39) kept their glycosylation capacity in vivo but with a lower efficiency. A 20-fold overexpression of sFT6 only produced small amounts of fucosylated products (10% ␣3-monofucosylated and 7% ␣3-difucosylated) (35). Furthermore, a 2-fold overexpression of s␣3GT produced lower amounts of cell-associated ␣3-galactosylated glycoproteins than full-length ␣3-galactosyltransferase (39).
In this work, we have also showed that FT3wt dimerization occurred within the TM through disulfide bond formation involving Cys-16 and Cys-29. The dimerization did not inactivate the enzyme since FT3wt and FT3np had similar enzyme activities. The formation of one disulfide bond mediated by Cys-16 or Cys-29 prevented a second disulfide bond since higher molecular weight oligomers have not been observed by non-reducing SDS-PAGE; this probably is caused by steric hindrance. Intramembrane disulfide bond formation might consist of a regulated process. Other GTs from the trans- Golgi  ␤4-Galactosyltransferase has been shown to form homodimers mediated by Cys-29 and/or His-32 from the TM, and dimerization was associated with Golgi localization (28). On the other hand, ␣2,6-sialyltransferase has been shown to dimerize through Cys-24 from its TM, dimerization not being required for Golgi localization or enzyme activity, but possibly necessary to induce higher multiplicity oligomerization (9). For FT3, it was found that mutation of Cys-29 led to increased levels of expression on the PM (FT3np2). Furthermore, when Cys-16 and Cys-29 were not mutated (FT3np6), levels of expression on the cell surface comparable to those of FT3wt were detected. These results suggested that Cys from the TM might play a role in Golgi retention of FT3.
We cannot exclude that the introduction of Leu residues into positions 16, 23, 29, and 33 in the TM might have created a new targeting signal to the PM. However, this is not probable since substitution of the TM of ␣2,6-sialyltransferase by a peptide with the same number of Leu residues did not have an effect on Golgi localization of the enzyme (11). Furthermore, the substitution of a single residue to Leu (FT3np1, FT3np2, FT3np3, and FT3np4) was enough to induce a localization shift to the PM.
In vitro studies have indicated that the correct localization of Golgi resident proteins, namely GTs, is due to their concentration into retrograde COP I-coated vesicles in a process that is dependent on GTP hydrolysis (20). We have found that FT3wt, a late acting glycosyltransferase, was also incorporated into transport vesicles contrary to its related plasma membrane mutant FT3np. Such vesicles also contained the medial Golgi SNARE GS28 or the p24 family proteins (p24␥3, ␣2, ␦1, ␤1) but not anterograde cargo such as the plasma membrane TfR or soluble ApoE.
The results obtained strongly support that the composition of the Golgi is maintained constant because of the incorporation of resident proteins such as FT3wt into retrograde COP Icoated vesicles. FT3np, which was transiently detected in the Golgi, but whose final destination was the PM, was never detected in this transport vesicle population.
The molecular basis for FT3wt concentration in the retrograde transport vesicles involves Cys-16, Gln-23, Cys-29, and Tyr-33 from its transmembrane domain. These amino acid residues might mediate interactions with other proteins or specific lipids necessary for sorting into COP I vesicles. Crosslinking and immunoprecipitation experiments are presently in progress in order to detect and identify potential FT3wt-binding proteins important for the sorting process.