Distinct reading of different structural determinants modulates the dileucine-mediated transport steps of the lysosomal membrane protein LIMPII and the insulin-sensitive glucose transporter GLUT4.

Leucine-based motifs mediate the sorting of membrane proteins at such cellular sites as the trans-Golgi network, endosomes, and plasma membrane. A Leu paired with a second Leu, Ile, or Met, while itself lacking the ability to mediate transport, is the key structural feature in these motifs. Here we have studied the structural differences between the leucine-based motifs contained in the COOH tails of LIMPII and GLUT4, two membrane proteins that are transported through the secretory pathway and are targeted to lysosomes () and to a perinuclear compartment adjacent to the Golgi complex (), respectively. LIMPII and GLUT4 display negatively (Asp(470)/Glu(471)) and positively (Arg(484)/Arg(485)) charged residues, respectively, at positions -4 and -5 upstream from the critical Leu residue. The change in the charge sign of residues -4 and -5 results in missorting of LIMPII and GLUT4. We note that the acidic Glu residue at position -4 is critical for efficient intracellular sorting of LIMPII to lysosomes, but is dispensable for its surface internalization by endocytosis. Efficient intracellular sorting and endocytosis of GLUT4 require an Arg pair between positions -4 and -7. These results are consistent with the existence of distinct leucine-based motifs and provide evidence of their different readings at different cellular sites.

Their structural determinants are, however, poorly characterized: a Leu paired with a second Leu, Ile, or Met is the only constant structural feature in these motifs, but the pair by itself lacks the ability to mediate transport (2,8). Acidic residues located at positions Ϫ4 and Ϫ5 upstream from the Leu/ Leu(Ile) pair have been demonstrated to be required for efficient internalization of invariant chain (Ii) and CD4 chimeras in mammalian cells (see Table I) (10,11) and for transport of the t-SNARE Vamp3 to yeast vacuoles (12). In addition, phosphorylation of a Ser residue at position 4 or 5 upstream is critical for the surface internalization of some membrane proteins (13)(14)(15)(16)(17). On the other hand, some of the Leu-based motifs lack upstream acidic residues in intracellular protein sorting and endocytosis, and this also strongly suggests the existence of a family of leucine-based motifs with distinct structural determinants and activities. The possibility of distinct specificity determinants that modulate the recognition of leucinebased motifs is also evident when the ability of specific clathrin adaptors to discriminate between leucine motifs is studied (11, 18 -22).
Here we have studied the structural determinants involved in the Leu-mediated intracellular sorting and surface internalization of LIMPII and GLUT4, two membrane proteins with distinct cellular distributions. The results of our experiments indicate the existence of different Leu-based motifs endowed with distinct structural determinants. Furthermore, they show that Leu-based motifs use different structural determinants at different transport steps.

MATERIALS AND METHODS
Cell Culture-3T3-L1 fibroblasts were cultured for 48 h on plastic dishes or glass coverslips in normal medium (Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 4 mM glutamine, 50 mg/liter gentamycin, 100 mg/liter streptomycin, 100 IU/liter penicillin, and nonessential amino acids) in a humidified CO 2 incubator at 37°C. 3T3-L1 fibroblasts stably transfected with wild-type GLUT4 or GLUT4 mutants, when required, were converted into adipocyte-like (ADL) cells by incubation for 6 -7 days in IDBX medium (normal medium supplemented with 10 g/ml insulin, 10 g/ml biotin, 0.25 M dexamethasone, and 500 M 1-isobutyl-3-methylxanthine). Clonal ADL cells displaying numerous lipid droplets of medium and large size in their cytoplasm were used before the onset of endogenous GLUT4 expression.
DNA Constructs-Wild-type rat GLUT4 and LIMPII cDNAs were cloned into the M13mp19 vector. All mutants were made by site-directed mutagenesis as described (23). For development of clonal cell lines of 3T3-L1 fibroblasts, constructs and mutants were cloned into a modified pPUR vector (CLONTECH) carrying the spleen focus-forming virus long terminal repeat promoter in the ApaI/EcoRI sites.
Transfection of Mammalian Cell Lines-To monitor the effects of the mutations on the distribution of LIMPII and GLUT4, COS-7 cells grown to near confluency on glass coverslips for 48 h were transiently transfected using the DEAE-dextran procedure (24). Transfections were performed for 18 h only to prevent overexpression of the transfected proteins and therefore saturation of the sorting mechanisms. Furthermore, transfected cells were incubated for the last 3 h with 0.1 mM cycloheximide to deplete the secretory pathway of transfected proteins. Mutants displaying abnormal patterns of distribution were then studied in stably transfected 3T3-L1 fibroblasts. For stable protein expression, 3T3-L1 fibroblasts, grown to 70% confluency on 10-cm dishes, were transfected using the calcium phosphate precipitation method (25) and were selected with 5 g/ml puromycin 48 h post-transfection. Individual clones were isolated within 2 weeks of transfection with the help of Teflon rings and, following their expansion, were screened by immunofluorescence microscopy. Established clonal cell lines were homogeneous with regard to the expression and distribution of the transfected proteins. Clones were maintained thereafter with 5-7.5 g/ml puromycin. Antibodies-The rabbit polyclonal antibodies (pAb) OSCR6 (provided by Dr. A. Zorzano, Universitat Barcelona) (26) and 828 were raised against a peptide representing the entire cytoplasmic COOH tail of GLUT4. The development and characterization of mouse mAb 29G10, raised against the luminal domain of LIMPII, has been described (27). pAb OSCR6 was able to immunoprecipitate detergent-solubilized GLUT4 and reacted with the protein blotted onto nitrocellulose and therefore was used in immunoprecipitation studies and Western blot analysis. pAb 828 reacted better with methanol-fixed cells and was used to study the cellular distribution of GLUT4 by microscopy.
Microscopy Studies-Lipid droplets were stained with Nile blue for 5 min on cells fixed with 3% paraformaldehyde, and the fluorescence was photographed. For protein localization studies, 3T3-L1 fibroblasts and ADL cells washed twice with PBS were fixed and permeabilized for 3 min with cold (Ϫ20°C) methanol, washed three times, and incubated at 37°C for 1 h with the corresponding specific antibody in PBS. After being washed for 15 min with PBS, the cells were stained at 37°C for 1 h with FITC-conjugated goat anti-mouse or goat anti-rabbit secondary antibodies (Cappel, Durham, NC) prepared to 10 mg/ml in PBS containing 0.25 mg/ml goat preimmune serum. Stained cells were washed once more for 15 min with PBS, mounted on glass slides using Gelvatol (Monsanto Co., St. Louis, MO), and studied under an Axiovert 135M inverted microscope (Zeiss) or a confocal Radiance 2000 microscope (Bio-Rad), and the fluorescence was photographed.
Quantitation of LIMPII and GLUT4 Cellular Levels-Stably transfected 3T3-L1 fibroblasts grown to near confluency were washed with PBS and incubated with 0.1 M Na 2 CO 3 (pH 11.3) at 4°C for 30 min. Cell extracts were centrifuged at 150,000 ϫ g for 30 min using a TL-100 ultracentrifuge (Beckman Instruments), and membrane pellets were extracted by incubation at 4°C for 1 h with 10 mM Tris (pH 7.4), 150 mM NaCl, 1% Triton X-100, and 2 mM EDTA (extraction buffer). After removal of insoluble material by centrifugation at 150,000 ϫ g for 30 min and adjustment of the protein in the supernatants to 2 mg/ml with 10 mM Tris (pH 7.4), 20 l was mixed with an equal volume of 2ϫ Laemmli buffer containing 5% SDS, heated at 37°C for 10 min, and resolved by 10% SDS-PAGE. Resolved proteins were blotted onto nitrocellulose, immunoprobed with either mAb 29G10 or pAb OSCR6, and studied using the ECL technique (Amersham Pharmacia Biotech, Aylesbury, United Kingdom).
Subcellular Fractionation Studies-Low-density microsome (LDM)and plasma membrane (PM)-enriched fractions were prepared from stably transfected 3T3-L1 fibroblasts by a modification of the procedure previously described (28). Briefly, the cells were washed with PBS, scraped with a rubber policeman in 3 ml of cold HES buffer (20 mM HEPES (pH 7.4), 1 mM EDTA, and 255 mM sucrose), and disrupted by N 2 cavitation (2400 kilopascals). All manipulations were performed at 4°C. Post-nuclear supernatants were fractionated by three consecutive centrifugations at 19,000 ϫ g for 20 min, then at 45,000 ϫ g for another 20 min, and finally at 180,000 ϫ g for 90 min to yield pellets enriched in LDMs. The membranes collected at 19,000 ϫ g were resuspended in HES buffer, layered onto a 35% sucrose cushion in 10 mM Tris (pH 7.4) and 1 mM EDTA, and recentrifuged at 108,000 ϫ g in an SW 40 rotor for 1 h. The white fluffy band recovered at the sucrose interface was then diluted 10-fold in 20 mM Tris and 2 mM EDTA, treated with 0.1 mM phenylmethylsulfonyl fluoride, and centrifuged at 150,000 ϫ g for 30 min to yield a pellet enriched in plasma membranes. LDM and PM pellets were mixed for 1 h at 4°C with 300 l of extraction buffer, and the insoluble material was removed by centrifugation at 150,000 ϫ g for 30 min. The resulting supernatants were scrutinized for GLUT4 by Western blot analysis as described above.
Glucose Uptake Studies-Untransfected and stably transfected 3T3-L1 fibroblasts cultured on 10-mm coverslips were quickly washed Clonal 3T3-L1 fibroblasts stably expressing the wild-type (wt-LIMPII) and mutant LIMPII proteins indicated were fixed/permeabilized with cold (Ϫ20°C) methanol and studied by confocal microscopy using mAb 29G10 and an FITC-conjugated goat anti-mouse second antibody (see "Materials and Methods"). Four serial cell sections are shown for each clone (A-D). Their maximal (max.) distance to the plane of cell attachment to the glass surface is given in m. One section per clone is shown at high magnification to facilitate the study of stained lysosomes and plasma membrane. Bars ϭ 20 m.
with Krebs-Ringer HEPES buffer (KRHB) and incubated at 37°C for 5 or 10 min with 15 ml of KRHB containing 0.2% bovine serum albumin and 1 mM 2-deoxy[ 3 H]glucose (specific activity ϭ 6.66 Ci/mol; PerkinElmer Life Sciences). Nonspecific 2-deoxyglucose uptake was measured in cells preincubated for 5 min with 15 l of KRHB containing 50 M cytochalasin B and 40 mM glucose, and the uptake assay was performed in the presence of both cytochalasin B and glucose. Glucose uptake was stopped by washing the cells quickly three times with 1 ml of ice-cold PBS containing 50 mM glucose. Cells were lysed with 200 l of 0.1 N NaOH to measure protein, and radioactivity was estimated by scintillation counting. 2-Deoxyglucose uptake was normalized to cpm/g of protein/min.
Studies of LIMPII and GLUT4 Intracellular Sorting-Clonal 3T3-L1 fibroblasts cultured for 48 h to 80% confluency on 10-cm dishes were washed twice with methionine/cysteine-free Dulbecco's modified Eagle's medium and metabolically labeled for 30 min with 0.25 mCi/ml [ 35 S]methionine/cysteine (specific activity Ͼ 1000 Ci/ml; PRO-MIX, Amersham Pharmacia Biotech) prepared in Dulbecco's modified Eagle's medium and 2% dialyzed fetal calf serum. The 30-min pulse was adequate to metabolically label LIMPII and GLUT4 and also to begin to monitor their appearance at the plasma membrane since they required 45 and 20 min, respectively, to traverse the Golgi (27,29). Cells were chilled on ice at different times after labeling, and surface-exposed LIMPII and GLUT4 were biotinylated using sulfosuccinimidyl 2-(biotinamido)ethyl-1,3Ј-dithiopropionate (Pierce) and Bio-LC-ATB-BMPA, respectively, as described (30,31). Afterwards, the cells were extracted for 30 min with 1 ml of cold 0.1 M Na 2 CO 3 (pH 11.3), and membrane spins prepared by centrifugation at 150,000 ϫ g for 40 min were solubilized by incubation at 4°C for 30 min with 1 ml of extraction buffer. GLUT4 and LIMPII proteins were immunoprecipitated with mAb 29G10 and pAb OSCR6 bound to 10 ml of protein G-Sepharose (Amersham Pharmacia Biotech), and the total labeled proteins were measured in one-eighth of the immunospins, whereas the rest were boiled for 5 min in 0.2 ml of 0.5% SDS and precipitated again with 10 ml of streptavidin-Sepharose to measure the labeled protein deflected to the plasma membrane. Protein quantitation was performed by scanning the autoradiograms produced by SDS-PAGE.
Studies of Surface Internalization of LIMPII and GLUT4 -To study the rate of LIMPII internalization, intact clonal 3T3-L1 fibroblasts grown to 80% confluency were incubated at 4°C for 1 h with mAb 29G10 diluted 1:500 in PBS; washed four times with cold buffer; and after incubation for different time periods at 37°C, fixed with 2% paraformaldehyde and stained with an FITC-conjugated goat antimouse secondary pAb as described above. Cell fluorescence was studied with a Coulter cytometer and analyzed using the Epics Flow XL program. Dead cells were excluded from analysis by forward and side scatter measurements.
To study the rate of GLUT4 internalization, this protein in clonal fibroblasts was metabolically labeled as described above and then labeled for 1 min at 4°C with 1 mM Bio-LC-ATB-BMPA prepared in 250 ml of KRHB (31). Biotinylated samples were then washed and chased for different time periods at 37°C in KRHB. Quantitation of biotin-35 Slabeled GLUT4 retained on the cell surface was carried out using purified plasma membrane fractions (see "Subcellular fractionation Studies").
Studies of Transport of Newly Synthesized LIMPII Molecules to Lysosomes-Cells metabolically labeled for 15 min as described above were disrupted by N 2 cavitation, and the post-nuclear supernatant was mixed as described to give a final concentration of 20% Percoll (22). Fractions of 3.3 ml made to 1% Triton X-100 were incubated at 4°C for 30 min, and insoluble material and Percoll were removed by centrifugation. LIMPII was immunoprecipitated by incubation at 4°C for 2 h with 10 l of mAb 29G10/protein G-Sepharose and quantitated by scanning of the autoradiograms produced by SDS-PAGE.

RESULTS
The only constant determinant in the leucine-based motifs is a Leu paired with a COOH residue with a long aliphatic chain (Ile, Met, and Val). In addition, negatively charged residues at position 4 upstream have been implicated as being involved in membrane protein endocytosis in mammalian cells as well as in the targeting of Vamp3 to yeast vacuoles (10 -12).
Sorting of LIMPII and GLUT4, two membrane proteins with distinct cellular distributions, is mediated by leucine-based motifs, as shown by the accumulation of mutants with ablated leucine motifs in the plasma membrane (1, 2, 4). Interestingly, whereas in LIMPII, the residues at positions 4 and 5 upstream are negatively charged (Asp 470 /Glu 471 ), in GLUT4, they are FIG. 4. Cellular distribution of GLUT4 in stably transfected clonal 3T3-L1 fibroblasts. Clonal 3T3-L1 fibroblasts stably expressing the wild-type (wt-GLUT4) and mutant GLUT4 proteins indicated were fixed/permeabilized with cold (Ϫ20°C) methanol, immunostained with pAb 828 and an FITC-conjugated goat anti-rabbit second antibody (see "Materials and Methods"), and studied by confocal microscopy. Four serial cell sections are shown per clone (A-D). Their maximal (max.) distance to the plane of cell attachment to the glass surface is given in m. One section per clone is shown at high magnification to facilitate the study of the stained perinuclear GLUT4 storage compartment and the plasma membrane. Bars ϭ 20 m.
positively charged (Arg 484 /Arg 485 ). To further characterize the structural determinants contained in the leucine-based motifs, the Asp 470 /Glu 471 and Arg 484 /Arg 485 pairs were (a) substituted by pairs of opposite charge, (b) replaced by two uncharged residues, (c) split, and (d) separated from the downstream critical Leu residue (Fig. 1). The effect of these modifications on the cellular distribution, intracellular sorting, and surface internalization of LIMPII and GLUT4 was studied by microscopy and biochemical means.
To exclude any side effects of overexpression on the distribution of the transfected proteins, the clones selected by microscopy were studied for their levels in the transfected proteins by Western blot analysis. The selected clones expressed comparable or slightly lower levels of the LIMPII and GLUT4 mutants as compared with the wild-type proteins (Fig. 2, A and  B). Furthermore, the cellular levels of the GLUT4 mutants were comparable to those of endogenous GLUT4 in 3T3-L1 adipocytes and rat adipocytes (Fig. 2B).
Leu-based Signals in LIMPII and GLUT4 Comprise Different Upstream Structural Motifs-To study whether the Asp 470 / Glu 471 and Arg 484 /Arg 485 pairs could modify function indistinctly in the context of LIMPII and GLUT4, these residues were replaced by pairs of opposite charge (two Arg and two Glu residues, respectively), and the cellular distributions of the resulting mutants and the native proteins were compared by microscopy.
In clonal 3T3-L1 fibroblasts, the bulk of wild-type LIMPII was localized to lysosomes (Fig. 3), whereas GLUT4 was found in a reticular perinuclear compartment (GLUT4 storage compartment) and in vesicles scattered throughout the cytoplasm (Fig. 4), as shown by immunofluorescence microscopy. In 3T3-L1 fibroblasts as well as in other cell lines studied (COS, normal rat kidney, baby hamster kidney, and Chinese hamster ovary), anti-LIMPII antibodies did not stain the plasma membrane, even under conditions of protein overexpression, whereas anti-GLUT4 antibodies stained it weakly in a small number of cells.
Introduction of the Arg 470 /Arg 471 and Glu 484 /Glu 485 mutations in LIMPII and GLUT4, respectively, provoked a dramatic accumulation of both proteins at the plasma membrane (Figs. 3 and 4), indicating a strong inhibition of their normal sorting. Because phosphorylated residues upstream of leucine motifs have been shown to be critical for protein endocytosis (13)(14)(15)(16)(17) and Thr 468 was lost during the development of the LIMPI-I(Arg 470 /Arg 471 ) mutant, we checked whether the accumulation of LIMPII(Arg 470 /Arg 471 ) in the plasma membrane was pro- FIG. 5. Cellular distribution of GLUT4 in stably transfected clonal ADL cells. Clonal ADL cells stably expressing the GLUT4 proteins indicated were either fixed with 4% paraformaldehyde and stained with Nile blue to visualized cytoplasmic lipid droplets or fixed/permeabilized with cold (Ϫ20°C) methanol, immunostained using pAb 828 and an FITC-conjugated goat anti-rabbit secondary antibody, and studied by confocal microscopy (see "Materials and Methods"). Four serial cell sections are shown for each clone (A-D). Their maximal (max.) distance to the plane of cell attachment to the glass surface is given in m. One section per clone is shown at high magnification to facilitate the study of the stained perinuclear GLUT4 storage compartment and the plasma membrane. Use of pAb 828 in untransfected ADL cells produced no specific staining (not shown). Lipid droplets can be seen in the immunostained cells as large, round, negatively stained cytoplasmic structures. Bars ϭ 20 m. wt-GLUT4, wild-type GLUT4.
voked by the loss of Thr 468 . The answer was negative; a study of the distribution of the LIMPII(Ala 468 ) mutant showed that the replacement of Thr 468 by Ala did not provoke the accumulation of the protein in the plasma membrane (data not shown), therefore eliminating that possibility.
The effect of replacing the Asp 470 /Glu 471 pair by two uncharged residues on the distribution of LIMPII was also studied. The study, first performed in transiently transfected COS-7 cells, revealed a weak but significant staining of their plasma membrane (data not shown), suggesting a limited accumulation of the LIMPII(Ala 470 /Ala 471 ) mutant in that location. When the same experiment was repeated in stably transfected 3T3-L1 fibroblasts, the weak staining of the plasma membrane was again observed (Fig. 3; see also Fig. 6 below), thus confirming that the replacement of the Asp 470 /Glu 471 pair by two uncharged residues affected the trafficking and distribution of LIMPII. Altogether, these observations indicate the importance of the Ϫ4 and Ϫ5 charged residues in the Leumediated sorting of LIMPII and GLUT4 and show that the effect of charge sign on their distribution is conditioned by the protein context in which the Leu-based motif is expressed.
The Asp 470 /Glu 471 and Arg 484 /Arg 485 pairs were further manipulated by (a) inserting an Ala between the two paired residues and (b) by increasing their distance to the critical leucine. Splitting of the Asp 470 /Glu 471 pair in LIMPII, produced by replacing Ala 469 and Asp 470 by Asp and Ala, respectively ( Fig.  1), did not affect the distribution of the protein, which was confined within lysosomes in transiently transfected COS-7 cells immunostained with mAb 29G10 (data not shown). An acidic residue at position Ϫ5 was therefore not critical for LIMPII sorting. In sharp contrast, the splitting of the Arg 484 / Arg 485 pair in GLUT4, produced by replacing Phe 483 and Arg 484 by Arg and Ala, respectively (Fig. 1), provoked the deflection of GLUT4 to the plasma membrane both in COS-7 cells (data not shown) and in 3T3-L1 fibroblasts stably expressing GLUT4(Arg 483 /Arg 485 ) immunostained with pAb 828 (Fig. 4). We note that this effect was not produced by the replacement of Phe 483 by Arg since the distributions of the GLUT4(Ala 483 ) mutant and GLUT4 in transiently transfected COS-7 cells were identical (data not shown).
The differences between the Asp 470 /Glu 471 and Arg 484 /Arg 485 determinants were further exposed by studies of the distribution of LIMPII(Glu 471 /Leu 477 ) and GLUT4(Arg 485 /Leu 491 ), mutants in which the distance between Asp 470 /Glu 471 /Arg 484 / Arg 485 and the critical leucine was increased by the insertion of two Ala residues (Fig. 1). The results of these studies showed that moving the Asp 470 /Glu 471 pair away strongly affected the correct targeting of LIMPII to lysosomes (Fig. 3), as shown by its dramatic accumulation in the plasma membrane, whereas moving the Arg 484 /Arg 485 pair did not alter the distribution of GLUT4 (Fig. 4).
When the GLUT4 distribution studies were repeated in 3T3-L1 fibroblasts that were treated for 6 -8 days with IDBX medium (see "Materials and Methods") and that developed large lipid droplets in their cytoplasm, but did not express endogenous GLUT4, the results (Fig. 5) were comparable to those with undifferentiated fibroblasts. To quantitate the accumulation of the LIMPII mutants in the plasma membrane of 3T3-L1 fibroblasts, cells stably expressing wild-type LIMPII or its mutants were fixed at 4°C for 1 h with 2% paraformaldehyde, permeabilized or not with 1% Triton X-100, and incubated with mAb 29G10 for 1 h at 4°C, and then their fluorescence was quantitated by flow cytometry. The results of these studies showed that 1.5% of wild-type LIMPII, 3% of LIMPII (Ala 470 /Ala 471 ), and Ͼ45% of LIMPII(Arg 470 /Arg 471 ) and LIMPII(Glu 471 /Leu 477 ) were exposed at the plasma membrane ( Fig. 6).
Attempts to quantitate the surface levels of GLUT4 in stably transfected 3T3-L1 fibroblasts by flow cytometry failed due to the lack of reactivity of a pAb antibody raised against the large exofacial loop of GLUT4 unless this was interrupted by the introduction of the hemagglutinin tag. 2 The surface levels of GLUT4 were therefore estimated (a) by quantification of its levels in subcellular fractions enriched in PMs and LDMs and (b) by measuring the glucose uptake by the cells (Fig. 7).
The steady-state levels of wild-type GLUT4 and GLUT4 mutants in PMs and LDMs were measured by Western blot analysis using pAb OSCR6 and the ECL technique. The results showed that the amounts of wild-type GLUT4 and GLUT4(Arg 485 /Leu 491 ) recovered with the plasma membrane were only 5 and 6% of the amounts recovered with the LDM fraction, whereas 20% of GLUT4(Glu 484 /Glu 485 ) and 45% of GLUT4(Arg 483 /Arg 485 ) were recovered with the PM fraction (Fig. 7A).
These results were confirmed by glucose uptake studies performed in clonal 3T3-L1 fibroblasts and ADL cells. It is impor-FIG. 6. Surface levels of wild-type and mutant LIMPII proteins with manipulated Asp 470 and Glu 471 residues. Clonal 3T3-L1 fibroblasts stably expressing transfected wild-type LIMPII or mutants with manipulated Asp 470 and Glu 471 residues were incubated at 4°C for 1 h with mAb 29G10 and, after their staining with an FITC-conjugated goat anti-mouse pAb, processed and studied by flow cytometry (see "Materials and Methods"). Mock control cells were incubated only with an FITC-conjugated goat anti-mouse pAb (arrows). The log of the surface fluorescence intensity (abscissa) is plotted against the relative cell number (ordinate) and is expressed in arbitrary units. Numbers indicate the percentage of LIMPII remaining at the plasma membrane.
Altogether, these observations confirmed the results of the immunofluorescence microscopy studies. They showed that an acidic residue at position 4 upstream from Leu 475 was critical for correct distribution of LIMPII. By contrast, correct distribution of GLUT4 required a pair of upstream basic residues whose distance to Leu 489 was not so critical. These results indicated the existence of important structural differences between the leucine-based motifs in the LIMPII and GLUT4 molecules.

Mutations of Residues at Positions 4 and 5 Upstream from Leu 475 and Ile 476 Reveal Differences in the Structural Requirements for Efficient Intracellular Sorting and Surface
Internalization of LIMPII-To investigate the role of the Asp 470 /Glu 471 pair in the intracellular sorting of LIMPII, we studied the effects of mutations on the rate of appearance of newly synthesized molecules on the cell surface. To monitor this appearance, clonal 3T3-L1 fibroblasts stably expressing wild-type LIMPII and the LIMPII(Arg 470 /Arg 471 ), LIMPII(Ala 470 /Ala 471 ), and LIMPII(Glu 471 /Leu 477 ) mutants were metabolically labeled for 30 min with [ 35 S]methionine/cysteine, and the molecules remaining on the plasma membrane at various chase times (20 min, 40 min, and 3 h) were biotinylated with the membraneimpermeable reagent sulfosuccinimidyl 2-(biotinamido)ethyl-1,3Ј-dithiopropionate and precipitated first with streptavidin-Sepharose and then with mAb 29G10. The analysis of these precipitates by autoradiography after 10% SDS-PAGE showed that the amounts of LIMPII(Arg 470 /Arg 471 ), LIMPII(Ala 470 / Ala 471 ), and LIMPII(Glu 471 /Leu 477 ) recovered after a 20-min chase were 3-4-fold higher than the amount of wild-type LIM-PII recovered (Fig. 8). This difference strongly suggested that the intracellular sorting of these three mutants was inhibited. Interestingly, a time course study revealed that although the amount of newly synthesized LIMPII(Ala 470 /Ala 471 ) recovered from the cell surface decreased with longer chase periods, the amounts of LIMPII(Arg 470 /Arg 471 ) and LIMPII(Glu 471 /Leu 477 ) were either retained at the cell surface or return there after internalization (Fig. 8). These observations agreed with the results of immunofluorescence microscopy studies showing that the plasma membrane of fibroblasts expressing LIMPII (Arg 470 /Arg 471 ) and LIMPII(Glu 471 /Leu 477 ) was stained more strongly than that of fibroblasts expressing LIMPII(Ala 470 / Ala 471 ). Together with the confinement of wild-type LIMPII in lysosomes, the results indicated that the acidic residue at position 4 upstream from Leu 475 was critical for efficient intracellular sorting of LIMPII.
The rates of internalization of wild-type LIMPII and the three mutants, produced by manipulation of the Asp 470 /Glu 471 pair, were compared using a flow cytometry technique. Stably transfected 3T3-L1 fibroblasts were incubated for 1 h at 4°C with mAb 29G10, and after washing and incubation at 37°C for different time periods, the protein remaining at the cell surface was stained with an FITC-conjugated goat anti-mouse secondary pAb. From measurements of the fluorescence remaining at the cell surface by flow cytometry, we observed that wild-type LIMPII and LIMPII(Ala 470 /Ala 471 ) were rapidly internalized, whereas LIMPII(Arg 470 /Arg 471 ) and LIMPII(Glu 471 / FIG. 7. Surface levels of wild-type GLUT4 and mutants with manipulated Arg 484 and Arg 485 residues. A, clonal 3T3-L1 fibroblasts stably expressing either wild-type GLUT4 (wt-GLUT4) or the indicated mutants were lysed and fractionated, and the transfected proteins recovered with the plasma membrane were quantitated by Western blot analysis using pAb OSCR6. The experiment shown is representative of four independent experiments. B, plasma membrane fractions purified from 3T3-L1 fibroblasts and the ADL cells stably transfected with the indicated GLUT4 proteins were scrutinized for their content of GLUT1 by Western blot analysis. C and D, glucose uptake in 3T3-L1 fibroblasts and ADL cells, respectively, stably expressing the indicated GLUT4 proteins was measured as described under "Materials and Methods." Results are expressed as a proportion of the PM levels or the glucose uptake by clonal cells stably expressing wild-type GLUT4. Plotted values are the mean of two independent experiments and duplicate samples.
Leu 477 ) were retained at the cell surface (Fig. 9). These results indicated that the introduction of Arg at position 471 and moving Glu 471 upstream inhibited LIMPII internalization. More important, the intracellular missorting and the normal rate of endocytosis of LIMPII(Ala 470 /Ala 471 ) indicated important differences in the role of Glu 471 in the intracellular sorting and surface internalization of the protein. These results led us to compare the rates of transport of wild-type LIMPII and the mutants to lysosomes.
Transport of newly synthesized LIMPII molecules to lysosomes was studied in clonal 3T3-L1 fibroblasts pulse-labeled for 15 min with [ 35 S]methionine/cysteine and then chased for 45 min or 2 h with normal medium. The Golgi and plasma membrane were separated from lysosomes by centrifugation on 20% Percoll gradients, and their content was quantitated by autoradiography after immunoprecipitation with mAb 29G10. We observed that after 45 min of chase, 14% of wild-type LIMPII, 20% of LIMPII(Ala 470 /Ala 471 ), 6% of LIMPII(Glu 471 / Leu 477 ), and none of LIMPII(Arg 470 /Arg 471 ) were recovered with lysosomes (Fig. 10). These differences were roughly maintained after 2 h of chase, when the percentages of wild-type LIMPII and LIMPII(Ala 470 /Ala 471 ) recovered with lysosomes were 30 and 24%, respectively, as compared with 5-10% of LIMPII(Arg 470 /Arg 471 ) and LIMPII(Glu 471 /Leu 477 ) (Fig. 10). These results showed that the LIMPII molecules deflected to the plasma membrane reached the lysosomes at rates reflecting their rates of internalization from the plasma membrane.
Pairing of Arg 484 and Arg 485 Is Critical for Efficient Intracellular Sorting of GLUT4 -Recent evidence suggests that the Leu 490 /Leu 491 -based motif is involved in the targeting of GLUT4 from the TGN to the perinuclear compartment, where it is stored (4,32). The observation that the plasma membrane of cells stably expressing GLUT4(Glu 484 /Glu 485 ) or GLUT4(Arg 483 /Arg 485 ) was strongly stained led us to investigate the intracellular sorting of GLUT4 in these cells. The protocol used for this purpose was similar to the one used in the LIMPII studies, but the surface biotinylation of GLUT4 was performed with the specific impermeable biotin reagent Bio-LC-ATB-BMPA (31). After a 30-min pulse with [ 35 S]methionine/cysteine and a 40-min chase period, we found that the fraction of newly synthesized wild-type GLUT4 detected at the cell surface was 1.7%, whereas the fractions of GLUT4(Glu 484 / Glu 485 ) and GLUT4(Arg 483 /Arg 485 ) were 3.5 and 5.1%, respectively (Fig. 11). The comparatively faster appearance of GLUT4(Glu 484 /Glu 485 ) and GLUT4(Arg 483 /Arg 485 ) on the plasma membrane strongly suggested that their intracellular sorting was inhibited.
To study the effects of the Arg 484 /Arg 485 manipulation on the surface internalization of GLUT4, 3T3-L1 fibroblasts stably transfected with wild-type GLUT4, GLUT4(Glu 484 /Glu 485 ), and GLUT4(Arg 483 /Arg 485 ) were surface-biotinylated at 4°C for 1 min with 1 mM Bio-LC-ATB-BMPA. Cells were then washed and incubated at 37°C for 10, 20, or 40 min in normal medium, and then the membranes were fractionated. The biotinylated proteins were extracted from the PM-enriched fraction and precipitated with streptavidin-Sepharose, and GLUT4 was quantitated by Western blot analysis using pAb OSCR6. We observed that the effects of the Arg 484 /Arg 485 manipulation on the internalization of GLUT4 were similar to those on its intracellular sorting: GLUT4 was internalized faster than GLUT4(Glu 484 /Glu 485 ), which was internalized faster than GLUT4(Arg 483 /Arg 485 ) (Fig. 12). DISCUSSION Leucine-based transport motifs displayed in the cytoplasmic tails of membrane proteins have been implicated in their sorting at the TGN and endosomes and surface internalization by endocytosis (for reviews, see Refs. 8, 33, and 34). The reading of leucine-based motifs at different cellular sites and the distinct cellular distribution of the clathrin adaptors involved in their recognition (for reviews, see Refs. 35-37) strongly suggest the existence of related leucine-based motifs with different transport specificities. This possibility is also suggested by the fact that pairing of a Leu residue to a second Leu, Ile, Met, or Val (8), while critical, is insufficient for transport. In addition, the upstream acidic residues described in some dileucine motifs (Table I) (10 -12) are absent in others.
To investigate the possible existence of different leucinebased motifs, we first compared the sequences adjacent to the dipeptide in LIMPII and GLUT4, two membrane proteins sorted by leucine-based motifs to lysosomes and to a tubulovesicular storage compartment adjacent to the Golgi complex, respectively (1-3, 38 -40). The comparison of their dileucine motifs reveals that the pair of acidic residues at positions Ϫ4 and Ϫ5 in LIMPII, and shared by other membrane proteins with leucine-based motifs (see Table I), is replaced in GLUT4 by a pair of Arg residues or by the pair His/Arg (41). It should be pointed that although the presence of upstream Arg residues is not uncommon in leucine-based motifs (see Table I), there have been no studies regarding their role in transport.
The accumulation of LIMPII and GLUT4 at the plasma membrane after the exchange of their pairs of acidic and basic residues shows that these structural determinants built in their leucine-based motifs are recognized as being different. The distinct effects of their manipulation indicates that the character and position of the Asp 470 /Glu 471 and Arg 484 /Arg 485 pairs have distinct influence on the distribution of the two proteins. For example, the accumulation of LIMPII at the plasma membrane upon the insertion of two alanines between the Asp 470 /Glu 471 pair and Leu 475 and the lack of effect on GLUT4 distribution when they were inserted between Arg 484 / Arg 485 and Leu 489 indicate that an acidic residue at position Ϫ4 is critical for efficient targeting of LIMPII to lysosomes, whereas the basic Arg residue can be moved away from this position in the GLUT4 molecule without affecting its targeting. Moreover, the pairing of the two Arg residues is required for efficient sorting of GLUT4 to the storage compartment adjacent to the Golgi complex, as shown by its accumulation at the FIG. 9. Surface internalization of wild-type LIMPII and mutants with manipulated Asp 470 and Glu 471 residues. Stably transfected clonal 3T3-L1 fibroblasts expressing wild-type LIMPII or the mutants indicated were incubated at 4°C for 1 h with mAb 29G10. After washing free of excess antibody, the cells were surface-labeled with an FITC-conjugated goat anti-mouse pAb and then incubated at 37°C for the indicated times. The fluorescence remaining at the cell surface was quantitated by flow cytometry, and the log of its intensity (abscissa), expressed in arbitrary units, is plotted against the relative cell number (ordinate). Numbers indicate the percentage of LIMPII remaining at the plasma membrane. Note the contrast between the rapid internalization of wild-type LIMPII and LIMPII(Ala 470 /Ala 471 ) as compared with the internalization of LIMPII(Arg 470 /Arg 471 ) and LIMPII(Glu 471 /Leu 477 ). The experiment shown is representative of two independent experiments.

FIG. 10. Transport to lysosomes of newly synthesized wild-type LIMPII and mutants with manipulated Asp 470 and Glu 471 residues.
Clonal 3T3-L1 fibroblasts stably expressing either wild-type LIMPII (wt-LIMPII) or the indicated mutants were pulse-labeled for 15 min with [ 35 S]methionine/cysteine and chased for 45 min or 2 h at 37°C with normal medium. At each chase time point, the cells were disrupted by N 2 cavitation, and the post-nuclear supernatants were fractionated by centrifugation on 20% Percoll gradients as described (28). The 10-ml gradients were divided into three fractions, and LIMPII was immunoprecipitated with mAb 29G10, resolved by SDS-PAGE, and studied by fluorography. plasma membrane following its splitting by the insertion of one Ala residue, whereas the pairing of the Asp 470 and Glu 471 residues in the LIMPII molecule appears to be less critical for the sorting of LIMPII to lysosomes, as shown by the efficient targeting of the LIMPII(Asp 469 /Glu 471 ) mutant to lysosomes. It is important to note that although acidic pairs are often found at positions Ϫ4 and Ϫ5 in proteins with dileucine motifs (Table  I), at least in LIMPII, the acidic residue at position Ϫ5 appears to be dispensable.
With regard to the transport steps catalyzed by the two leucine-based motifs studied here, we note that the leucine motif has been implicated in sorting of LIMPII at the TGN (2,42) and in the surface internalization of chimeras bearing the entire cytoplasmic COOH terminus of the lysosomal protein (10). The study of the manipulation of the leucine motif in GLUT4 has revealed its involvement in its targeting from the TGN to the GLUT4 storage compartment (4,43), in the access of GLUT4 to a slow recycling compartment (44), and in the surface internalization of GLUT1 and transferrin chimeras constructed with the COOH-terminal 30 amino acids of GLUT4 (45,46). Thus, sorting at the TGN and surface internalization appear to be the two transport steps mediated by the leucine motif in LIMPII and GLUT4.
To study the effect of manipulating the upstream determinants on the intracellular sorting of LIMPII and GLUT4, probably at or near the TGN, we have compared the rate of appearance at the cell surface of newly synthesized molecules of the wild-type and mutant proteins. The faster appearance of all the LIMPII and GLUT4 mutants that accumulate at the plasma membrane is a strong indication of their inefficient intracellular sorting. Furthermore, the deflection to the cell surface of mutants developed by substituting the pairs of acidic and basic residues by pairs of opposite charge indicates the different structure of the two leucine motifs studied here. This observation suggests that though the intracellular sorting of LIMPII and GLUT4 is catalyzed by leucine motifs, the nature of the transport steps and that of the machineries involved in their recognition are probably different. The recent observations that AP-3 binds to a synthetic peptide from the COOH terminus of LIMPII, but not to GLUT4 (22), and that GLUT4 and CD3␥ synthetic peptides bearing the leucine-based sorting motif compete for binding to AP-1 (19) suggest their recognition by different clathrin adaptors. However, very little is known about the organization of TGN subcompartments or the sorting machineries operating there. It seems unlikely that the sorting of the two proteins studied here occurs in the same TGN subcompartment. The localization of AP-1 and AP-3 to what appear to be distinct compartments within the TGN area (47,48) suggests that the Leu-based sorting of LIMPII and GLUT4 could occur in different subcompartments. Further analysis of TGN subcompartments may show where and how LIMPII and GLUT4 are sorted after traversing the Golgi cisternae. The distinct sorting of LIMPII and GLUT4 is less striking when compared with soluble and membrane lysosomal proteins, which are sorted by Leu-based mechanisms into different clathrin-coated vesicles and yet have the same subcellular destiny (49,50).
The sorting of LIMPII(Ala 470 /Ala 471 ) is of great interest. Although the replacement of the acidic residues at positions Ϫ4 and Ϫ5 by two uncharged alanines strongly inhibited the intracellular sorting of LIMPII, it did not affect its internalization, and as a result, its level of accumulation at the cell surface was relatively small. These observations clearly show that the requirements for the Glu residue at position Ϫ4 in the intra- FIG. 11. Surface appearance of newly synthesized wild-type GLUT4 and mutants with manipulated Arg 483 and Arg 484 residues. Stably transfected clonal 3T3-L1 fibroblasts expressing wild-type GLUT4 (wt-GLUT4) and the indicated GLUT4 mutants were pulselabeled for 30 min with [ 35 S]methionine/cysteine and, after 40 min at 37°C, chilled on ice, and GLUT4 exposed on the cell surface was derivatized at 20°C with Bio-LC-ATB-BMPA. A, GLUT4, GLUT4(Glu 484 /Glu 485 ), and GLUT4(Arg 483 /Arg 485 ) were immunoprecipitated from detergent lysates using pAb OSCR6 (T), and the surfacebiotinylated fraction (PM) was extracted using streptavidin-Sepharose. B, the protein bands shown in A were quantitated by densitometry scanning, and the percentage of the newly synthesized proteins deflected to the plasma membrane was calculated.
FIG. 12. Surface internalization of wild-type GLUT4 and mutants with manipulated Arg 483 and Arg 484 residues. Stably transfected clonal 3T3-L1 fibroblasts expressing wild-type GLUT4, GLUT4(Glu 484 /Glu 485 ), or GLUT4(Arg 483 /Arg 485 ) were photoaffinity-labeled with Bio-LC-ATB-BMPA as described under "Materials and Methods" and then incubated at 37°C for the indicated time periods. The affinity-labeled molecules remaining at the cell surface were precipitated by incubating the solubilized subcellular fraction enriched in plasma membrane with streptavidin-Sepharose. Precipitated GLUT4 was then quantified by Western blot analysis using pAb OSCR6. The relative decreases were measured by scanning the Western blots and are indicated (Percentage). The experiment shown is representative of two separate experiments. cellular sorting and endocytosis of LIMPII are different and that these two transport steps have distinct structural requisites. A search in the literature shows a few examples of the suggested involvement of acidic residues at position Ϫ4 in the endocytosis of membrane proteins bearing dileucine motifs (10 -12).
Our results show, however, that the LIMPII(Ala 470 /Ala 471 ) mutant is efficiently endocytosed. This is, however, not the only example since an Iip31 chimera constructed with seven of the last eight residues of LIMPII (10) and a VMAT2 mutant carrying an Ala at position Ϫ4 (51) have been shown to be efficiently endocytosed. In addition, CD4-mediated Nef down-regulation proceeds efficiently in cells expressing a CD4 molecule carrying an Ala at position Ϫ4 (21). On the other hand, the substitution of acidic residues at positions Ϫ4 and Ϫ5 precluded the endocytosis of the major histocompatibility complex II invariant chain Ii and the Ii/CI-M6PR and CD4/CD3␥ constructs (10,11). Altogether, these results suggest that the context in which a leucine motif is expressed may condition the structural determinants required for its activity. It remains to be determined whether these structural differences result or not in sorting of the proteins by different machineries. The efficient binding of AP-3 to the synthetic peptide mimicking the COOH terminus of LIMPII, in a manner dependent on the Asp 470 /Glu 471 and Leu 474 /Ile 475 pairs, as well as the failure of AP-2 to bind to the same peptide (22) provide evidence for the distinct reactivity of leucine motifs with different clathrin adaptors and raise the question as to which adaptor is involved in LIMPII endocytosis.
The striking correlation between the efficiency with which the LIMPII mutants deflected to the plasma membrane are endocytosed and unloaded into the lysosomes indicates that LIMPII is efficiently transported from the plasma membrane to lysosomes. As evident from the behavior of the LIMPII mutants discussed above, the GLUT4(Glu 484 /Glu 485 ) and GLUT4 (Arg 483 /Arg 485 ) mutants deflected to the plasma membrane also on the cell surface. This is indicated by the strong surface staining of the fibroblasts and ADL cells stably expressing these mutants as well as by the enhanced capacity of these cells to take up glucose from the medium. Moreover, both mutants are endocytosed significantly more slowly than the wild-type protein. Thus, the pairing of two Arg residues (i.e. positively charged) upstream from Leu 488 appears to be equally critical for both the intracellular sorting and endocytosis of GLUT4. With regard to the latter, it is interesting that a leucine-based motif with an Arg at position Ϫ4 contained in the NH 2 terminus of the insulin-regulated aminopeptidase, a protein that shows an ample co-distribution with GLUT4 (52, 53) and whose trafficking is also regulated by insulin (54,55), appears not to mediate the surface internalization of a transferrin chimera, but slows its recycling back to the cell surface (56). On the other hand, substitution of Arg for the phosphoacceptor Ser has been described to slow down the internalization of a CD4/CD3␥ chimera (17).