Overexpression of Proteins Containing Tyrosine- or Leucine-based Sorting Signals Affects Transferrin Receptor Trafficking*

Targeting of many transmembrane proteins to post-Golgi compartments is dependent on cytoplasmically exposed sorting signals. The most widely used signals conform to the tyrosine- or the leucine-based motifs. Both types of signals have been implicated in protein localization to the same intracellular compartments, but previous results from both cell-free experiments and studies of transfected cell lines have indicated that the two types of signals interact with separate components of the sorting machinery. We have overexpressed several transmembrane proteins in stably transfected Madin-Darby canine kidney cells using an inducible promoter system. Overexpression of proteins containing tyrosine- or leucine-based sorting signals resulted in reduced internalization of the transferrin receptor, whereas recycling and polarized distribution was not influenced. Our results indicate that proteins with tyrosine- and leucine-based sorting signals can be transported along common saturable pathways.

The transferrin receptor (TfR) has been an outstanding tool in the elucidation of the early events of endocytosis because its ligand, holotransferrin (Tf), can easily be labeled and detected in experimental setups. After internalization, the TfR releases its ligand and is transported from sorting endosomes to a special recycling compartment from where it is sent back to the plasma membrane (for reviews see Ref. 6, page 768, and Ref. 26). Whereas efficient internalization of the TfR depended on the tyrosine-based signal in its cytoplasmic tail (7,(27)(28)(29), mutation of this motif did not have any effect on the recycling of the mutant TfR (28), which has been suggested to follow bulk membrane transport (30). In polarized MDCK cells the TfR is localized predominantly on the basolateral surface (31). The basolateral sorting information of the TfR is contained within its cytoplasmic tail (32,33) but is independent of the tyrosinebased internalization motif (34). In this report we used an inducible expression system to study the effect of overexpressing proteins containing tyrosine-or leucine-based sorting signals on TfR traffic in stably transfected MDCK cells.

EXPERIMENTAL PROCEDURES
Antibodies and Ligands-The mouse monoclonal IgG2a antibody 66Ig10 recognizes the extracellular domain of the TfR (35) and was obtained from Monosan (Holland). SC2 is a mouse monoclonal IgM antibody against the extracellular domain of CD8 and was a gift from G. Gaudernack (Norwegian Radium Hospital, Norway). D5 is a mouse monoclonal IgG2b antibody against the extracellular domain of CD1d (36) and was a kind gift from Dr. S. Balk (Harvard Medical School, Boston). The rabbit polyclonal HC20 antibody (37) and the two IgG1 and IgM mouse monoclonals BU45 (38) and BU43, respectively, recognize the luminal part of the invariant chain and were purchased from The Binding Site, Birmingham, UK. X22 is a mouse monoclonal IgG1 antibody against the clathrin heavy chain (39) and was obtained from Affinity BioReagents. A rat polyclonal antibody directed against mouse IgG2A was purchased from Zymed Laboratories Inc. Other secondary antibodies used in the study were Alexa488 goat anti-mouse IgG (Molecular Probes) and Texas Red (TR)-conjugated goat anti-mouse IgM and IgG2b antibodies and a fluorescein isothiocyanate-conjugated goat anti-rat antibody (Southern Biotechnology Associates, Inc.). TR-conjugated human Tf (TR-Tf) was obtained from Molecular Probes and BU45 was conjugated with Alexa594 from the same company. Ricin (40) was a kind gift from Dr. Kristian Prydz (University of Oslo, Norway).
Cell Culture-The Madin-Darby canine kidney (MDCK) strain II cell line stably transfected with the human TfR has been described previously (41) and was a kind gift from Dr. W. Hunziker, Switzerland. This cell line, here named TrC, was used for super-transfection experiments. The cells were grown in DMEM (BioWhittaker) supplemented with 9% fetal calf serum (Integro b.v, Holland), 150 g/ml hygromycin B (Saveen, Sweden), 2 mM glutamine (BioWhittaker), 25 units/ml penicillin, and 25 g/ml streptomycin in 6% CO 2 in a 37°C incubator. For polarized studies cells were plated at high density (1.5 ϫ 10 6 ) onto 24-mm diameter Transwell polycarbonate filter units with a pore size of 0.4 m (Costar). Cells were grown for 4 days to form tight monolayers prior to the experiments.
Plasmid Constructs-A cDNA encoding the p33 form of human invariant chain (42) was subcloned into the heavy metal-inducible expression vector pMEP4 (Invitrogen). An Ii construct where the two leucines in the cytoplasmic tail of Ii was replaced by alanines (Ii 2LA ) has been described previously (43). Wild-type CD1d and a CD1d construct in which the tyrosine in the cytoplasmic tail was replaced by an alanine (CD1d YA ) in the pMEP4 vector has been described previously (36). The pMEP4 vector with chimeric DNA constructs, in which the cytoplasmic tail of CD8 was replaced by the counterpart of HLA-DM (HLA-DM/ CD8), or a mutated construct where the tyrosine in the cytoplasmic tail was replaced by an alanine (DM YA /CD8), 2 was a gift from Målfrid Røe (Oslo, Norway).
Transfection and Clone Selection-TrC cells were stably transfected by the DNA-calcium phosphate procedure as described elsewhere (44). Resistant clones were selected in the presence of hygromycin B (300 g/ml). Resistant clones were induced with 25 M CdCl 2 for about 16 h and assayed for expression by immunofluorescence microscopy.
Metabolic Labeling and Immunoprecipitation-Stably transfected MDCK cells grown in 35-mm wells were incubated to ϳ75% confluency in full medium supplemented with various concentrations of CdCl 2 and then starved for 45 min in DMEM lacking methionine and cysteine. Cells were then labeled at 37°C for 20 min in the same medium supplemented with 300 Ci of [ 35 S]methionine/cysteine (Amersham Pharmacia Biotech, Uppsala, Sweden) per milliliter, placed on ice, washed three times in ice-cold PBS, and lysed in ice-cold lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P-40) containing a mixture of protease inhibitors (4 g/ml phenylmethylsulfonyl fluoride, 2 g/ml antipain, 2 g/ml leupeptin, and 1 g/ml pepstatin A). The lysates were centrifuged for 10 min at 4°C at 10,000 ϫ g to remove cell nuclei and cellular debris and then incubated with BU45 for 1 h at 4°C. Dynabeads® rat anti-mouse IgG1 (M-450, Dynal, Norway) were then added (0.5 mg of Dynabeads® per lysate from 1 ϫ 10 6 cells) and incubated for an additional hour at 4°C. The precipitates were washed three times in 1.5 ml of ice-cold PBS, 1% Nonidet P-40 by use of a Dynal Magnetic Particle Concentrator (MPC®-M) to collect the beads after each step. The precipitates were resuspended in 25 l of sample buffer and boiled for 5 min. Protein released from the beads was analyzed by 12% SDS-polyacrylamide gel electrophoresis. The radioactive signal was amplified using Amplify TM solution (Amersham Pharmacia Biotech, Buckinghamshire, UK), and the bands were detected with the Molecular Imaging Screen CS (Bio-Rad) and analyzed with a Bio-Rad GS-250 PhosphorImager. The intensity of each band was quantified by the Molecular Analyst 2.0.1 software (Bio-Rad) and normalized to the total labeled protein in the corresponding lysate that was quantified with a 1500 Tri-Carb® Liquid Scintillation Analyzer (Packard Instrument Co.).
Radiolabeling-BU45 and D5 antibodies were iodinated with chloramine T according to manufacturer's recommendations. Briefly, antibodies (100 g) were incubated with Na 125 I (1 mCi) and chloramine T for 15 min on ice. Human Tf and ricin were iodinated using IODO-BEADS® (Pierce) as described by the manufacturer. Briefly, 100 g of protein was incubated with one bead for 10 min at room temperature. Iodinated proteins were separated from free Na 125 I and unincorporated 125 I 2 on Sephadex G-25M columns (Amersham Pharmacia Biotech, Uppsala, Sweden). The specific activity of the labeled proteins was determined by trichloroacetic acid precipitation. The amount of soluble radioactivity in the fractions used was generally less than 5% of total radioactivity.
Internalization of Radiolabeled Compounds-Stably transfected MDCK cells grown in 35-mm wells were incubated to ϳ75% confluency in full medium supplemented with various concentrations of CdCl 2 and then for 1 h with DMEM supplemented with 0.1% BSA. The cells were cooled on ice and incubated with iodinated BU45, Tf, or ricin for 1 h and then washed extensively with PBS/BSA (0.1%), supplemented with 0.1 M lactose in ricin experiments. The cells were subsequently transferred to a 37°C water bath and incubated for various time intervals (0, 1, 5, 15, or 30 min in duplicates) in full medium, supplemented with 100 g/ml cold ligand in Tf experiments. The cells were next cooled on ice and incubated twice with 0.5 M acetic acid in 0.15 M NaCl, pH 2.5, for 7 min to remove the surface-bound labeling, and lysed with 1 M NaOH. The amount of radioactivity in the samples was quantified with a Cobra® Auto-Gamma counter (Packard Instrument Co.). Internalized radiolabeled protein was calculated as the ratio of radioactivity present in the lysate to the total activity of the lysate and acid washes. Nonspecific binding of 125 I-BU45, 125 I-D5, and 125 I-Tf was determined in non-transfected cells and was typically less than 1% of the specific activity bound to transfected cells. Cell-surface binding at increasing concentrations of iodinated antibodies was analyzed by the method of Scatchard. In order to quantify the number of surface molecules per cell, the values were corrected for both the immunoreactive fraction as described previously (45,46) and the fraction of double positive cells, determined by fluorescence microscopy. The given values represent the mean Ϯ 1 S.E., obtained from at least three independent experiments, each performed in duplicate.
Uptake of Horseradish Peroxidase (HRP)-Stably transfected MDCK cells grown in 35-mm wells were incubated to ϳ75% confluency in full medium supplemented with various concentrations of CdCl 2 and then for 1 h with DMEM/BSA (0.1%). HRP (5 mg/ml) was added to the medium and allowed to be taken up by the cells for various amounts of time (0, 1, 5, 15, or 30 min, performed in duplicate). The cells were cooled on ice, washed extensively with PBS/BSA (0.1%), and lysed with 1 M NaOH. HRP uptake was quantified with the Immunopure® TMB Substrate Kit (Pierce) as described by the manufacturer, and the absorbance at 590 nm was detected by a Titertek Multiscan® Plus (Flow Laboratories, Switzerland). The given values represent the mean Ϯ 1 S.E., obtained from at least three independent experiments, each performed in duplicate.
Immunofluorescence Confocal Microscopy-Stably transfected MDCK cells grown on filter supports were induced to express the proteins of interest and processed for immunofluorescence microscopy essentially as described by Berod et al. (47) and permeabilized by 0.05% Triton X-100. The total cellular protein distribution was visualized by labeling fixed and permeabilized cell monolayers with the appropriate antibodies for 1 h at 37°C in a humidified chamber. In Ii-expressing cells Ii was detected with HC20/goat anti-rabbit Alexa594 and TfR with 66Ig10/goat anti-mouse Alexa488, whereas in CD1d-expressing cells CD1d was detected with D5/goat anti-mouse IgG2b TR, and TfR was detected with 66Ig10/rat anti-mouse IgG2a/goat anti-rat fluorescein isothiocyanate. Images were acquired with a Leica TCS-NT digital scanning confocal microscope equipped with a 60/1.2 water immersion objective (Leica, Germany). The pinhole diameter was kept below 1 m. The images were exported to Adobe Photoshop (Adobe Systems Inc.), processed for presentation, and printed on a Tektronix Phaser 450 dye sublimation printer (Tektronix Inc.).
Quantitative Immunofluorescence Microscopy-Stably transfected MDCK cells were grown on coverslips, induced with 25 M CdCl 2 overnight, and incubated with DMEM, 0.1 mg/ml BSA for 1 h. Cells were incubated with 20 g/ml TR-Tf or BU45-Alexa594 for 3 min and then fixed with 3% paraformaldehyde for 10 min on ice, washed, and incubated for 30 min with the anti-clathrin antibody X22 diluted in PBS, 0.1% saponin, washed again, and finally incubated with Alexa488 goat anti-mouse diluted in PBS, 0.1% saponin. Images were acquired using a Leica TCS-NT digital scanning confocal microscope equipped with a 100/1.4 oil immersion objective (Leica). The motifs of plasma membrane domains were selected by a random procedure. The images were magnified 4 times using the zoom function, adjusted to saturation by the glow-over function, and averaged 8 times to reduce the background noise. The pinhole diameter was 0.5 m. Images were exported as 1024 ϫ 1024 pixels, 8-bit RGB TIFF files, giving a dynamic range of 256 intensities per channel. Co-localization analysis was performed on a Macintosh computer (Apple Computer) using the Density Slicing and Automation modules of the Improvision OpenLab 2.0.2 core suite (Improvision, UK). First, binary image files were made displaying all pixels in the 150 -254 range of the green (clathrin) layer and of a superimposed image where both the red value (Tf) and the green value (clathrin) of the pixels had to be in the 150 -254 range in order to be selected. The binary files were used to calculate the total area of all regions of interest (ROI) larger than 4 pixels not touching the edge of the images. 25 different images were analyzed per experiment. Images were processed for presentation as described above.
Recycling of Tf-Recycling of iodinated Tf was assayed as described by Jing et al. (27). Briefly, stably transfected MDCK cells grown in 35-mm wells were first incubated for 1 h with DMEM/BSA (0.1%) and then incubated for 30 min with 1 g/ml 125 I Tf in DMEM/BSA. The cells were then washed three times with PBS/BSA on ice and incubated for 15 min with Tf buffer 1 (150 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 20 mM NaAc, pH 5.0, 50 M deferoxamine mesylate). Next, cells were washed once with PBS/BSA, incubated for an additional 20 min with Tf buffer 2 (PBS/BSA, 50 M deferoxamine mesylate, 100 g/ml Tf), and washed three times with PBS/BSA followed by a chase period at 37°C. The medium was collected, the cells were lysed, and the amount of radioactivity in the samples was determined. The given values represent the mean Ϯ 1 S.E., obtained from at least three independent experiments, each performed in duplicate.
Determination of the Polarized Distribution-The polarized surface distribution of the transfected proteins was analyzed by binding of iodinated ligands or antibodies to cells grown on Transwell units. Cells were induced with 25 M CdCl 2 and cooled on ice and then 125 I-BU45, 125 I-D5, or 125 I-Tf was added to either the apical or the basolateral side of the monolayer for 1 h. Cell monolayer tightness was assessed by measuring the radioactivity in the medium of both chambers, and generally less than 0.1% of the added radioactivity had diffused through the monolayer. Unbound activity was removed by extensive washing (5ϫ) in PBS supplemented with 1 mM CaCl 2 and 1 mM MgCl 2 and 2% fetal calf serum. The filters were then excised, and bound radioactivity was measured in a ␥ counter. Nonspecific binding was determined by labeling of non-transfected cells, and the values, which generally represented less than 1% of the total binding, were subtracted from the total bound radioactivity to give the specific binding. The given values represent the mean Ϯ 1 S.E., obtained from at least three independent experiments, each performed in duplicate.
Materials-All materials, unless specified otherwise, were purchased from Sigma.

Less Efficient Ii Internalization upon Its Overexpression-To
study the role of the Ii expression level in its sorting, the p33 form of Ii was put under control of the metallothionein promoter PhMTII. Expression was induced with Cd 2ϩ ; cells were labeled with 35 S, and Ii was immunoprecipitated with BU45. The precipitated samples were analyzed by SDS-polyacrylamide gel electrophoresis (Fig. 1A), and the bands corresponding to Ii monomers were quantified by a PhosphorImager (Fig.  1B). No detectable amount of Ii could be immunoprecipitated from cells transfected with empty vector (data not shown). In cells incubated overnight with 0 -40 M CdCl 2 Ii expression increased proportionally to the Cd 2ϩ concentration in the medium. Cd 2ϩ did not influence the expression of endogenous LAMP-1 or transfected TfR (data not shown).
Previous studies have revealed that Ii molecules appearing at the cell surface are rapidly internalized (43,48,49). Cells incubated with various concentrations of Cd 2ϩ were labeled with 125 I-BU45 on ice. 125 I-BU45 accumulated on the plasma membrane upon increased expression level (Fig. 1B), and the rate of internalization was reduced ( Fig. 2). High levels of Ii expression has previously been associated with delayed transport through the endocytic pathway (50,51). Overexpression of a mutant Ii construct unable to induce such a delay exhibited the same reduced rate of internalization as wild-type Ii (data not shown). These results indicated that the internalization process itself, rather than post-internalization events, was affected by the overexpression. We therefore decided to investigate whether overexpression of one membrane protein containing an internalization signal would affect the internalization of other proteins containing similar or different signals for endocytosis.
Reduced Internalization of the TfR upon Overexpression of Proteins Containing Tyrosine-or Leucine-based Sorting Signals-An MDCK cell line stably transfected with the human TfR was super-transfected with a series of membrane proteins containing tyrosine-or leucine-based internalization signals (Table I and references therein). The TfR binds Tf and recycles rapidly and constitutively between the plasma membrane and early endosomes. We used iodinated Tf as a ligand for MDCK cells transfected with the human TfR. Untransfected MDCK cells bound less than 5% of the 125 I-Tf compared with the TrC cell line (data not shown). Internalization of 125 I-Tf was measured in a series of double transfected cell lines. As shown in Fig. 3A, 125 I-Tf was internalized rapidly in control cells trans-fected with TfR alone; about 50% of the cell-bound Tf was endocytosed after 1 min. However, upon co-expression of proteins containing tyrosine-based (HLA-DM/CD8, CD1d) or leucine-based (Ii) sorting signals, the internalization of Tf was significantly reduced after short chase periods. Approximately 25% reduction in the Tf uptake was observed after 1 min in each case. In a set of control experiments, co-expression of mutant proteins in which the critical tyrosines (HLA-DM YA / CD8 and CD1d YA ) or leucines (Ii 2LA ) were replaced by alanine, internalization of Tf was not reduced (Fig. 3B). It must be noted that the surface expression levels of the mutant proteins were adjusted to at least match the expression of their wild-type counterparts.
These results indicated that the internalization machinery was saturated upon overexpression of proteins containing intact sorting signals. If components of the sorting machinery were reaching saturation, the inhibitory effect should increase with the expression level of the competing protein. To study this, cells transfected with TfR and inducible CD1d or Ii were incubated with increasing concentrations of Cd 2ϩ . Increased expression of both CD1d and Ii resulted in correspondingly higher surface expression of the TfR (Fig. 4A), whereas the rate of TfR internalization after 5 min was reduced (Fig. 4B). Expression of similar, or up to 50 times higher, numbers of the internalization deficient mutant counterparts on the cell surface did not significantly influence the uptake of Tf.
Unaffected Internalization of HRP and Ricin upon Overexpression-We have demonstrated that overexpression of membrane proteins containing signals for endocytosis can inhibit internalization of other proteins. Next, we wanted to see whether such overexpression also affected nonspecific endocytosis. Cells were induced to express proteins with intact or mutated sorting signals and incubated with 5 mg/ml HRP at 37°C. After various periods, cells were transferred to 4°C and washed before lysis and HRP quantification of the lysates. As shown in Fig. 5A, the rate of HRP uptake was not significantly influenced by the overexpression. Ricin is a lectin that attaches  to terminal galactose residues on both glycolipids and glycoproteins and therefore binds to a large number of molecules per cell (for a review, see Ref. 40). 125 I-Ricin was added to the cells at 4°C for 30 min. After extensive washes, the cells were transferred to 37°C, and uptake was measured after different times. The rate of ricin uptake was also not affected by overexpression of proteins with intact sorting signals (Fig. 5B). These data show that unspecific uptake and bulk membrane internalization was not influenced by the overexpression of proteins with endocytosis signals.

Decreased Co-localization between Tf and Clathrin upon Overexpression of Proteins with Intact Sorting Signals-On
what level in the internalization process did the inhibition of endocytosis take place? Both tyrosine-and leucine-based motifs are signals for localization to clathrin-coated pits, so we asked whether clustering of Tf into clathrin-coated pits could be inhibited by overexpression of proteins with such signals. To obtain statistically reliable results, we chose an immunofluorescence assay that allowed analysis of large plasma membrane areas of many cells. Fig. 6 shows micrographs of cells incubated with the anti-Ii antibody BU43 or Tf-TR (red channel) and the total distribution of clathrin, labeled with the X22 antibody (green channel). The micrographs are parts of images with mean pixel values close to the average of the respective sets of images analyzed. The binary files (black and white images) were obtained by "density slicing" the two-channel digital confocal images and display pixels with medium-to-high green/red values of the corresponding micrographs above.
Cells expressing Ii (G) or Ii 2LA (A) were incubated with BU43 for 3 min at 37°C to obtain sufficient labeling and then fixed, permeabilized, and labeled with X22. High magnification confocal microscopy of plasma membrane regions revealed punctate labeling of both constructs, but the Ii 2LA positive spots also included larger regions. Both Ii and Ii 2LA co-localized with clathrin (visualized as yellow). Next, cells expressing Ii (H), Ii 2LA (B), CD1d (I), or CD1d YA (C) were labeled with Tf-TR for 3 min at 37°C and with X22 as explained above. The images revealed a relative decrease in colocalization between Tf-TR and clathrin upon overexpression of proteins with intact sorting signals. To verify this observation, in each case quantitative analysis of 25 images from randomly selected regions was done. We first analyzed co-localization between the anti-clathrin antibody X22 and the anti-Ii antibody BU43. Co-localization between clathrin and Ii/Ii 2LA was quantified by comparing "density sliced" medium-to-high intensity pixels of the two channels as described under "Experimental Procedures." The results are displayed in Table II.
Although the number of ROI recognized by BU43 did not deviate much in cells expressing Ii compared with Ii 2LA , the total area of the Ii 2LA ROI was about 4 times higher compared with the Ii ROI. When intensities of lower magnitude were included, the fraction of Ii on the surface compared with Ii 2LA was found to resemble the values obtained with iodinated BU45 (data not shown). Comparison of both the ROI and the area values of the clathrin labeling showed a two times increase in cells expressing Ii compared with Ii 2LA . This indicated that the leucine-based signal was involved in the clustering of clathrin into defined areas. Furthermore, the amount of colocalization between X22 and BU43 was found to be somewhat higher in cells expressing Ii compared with Ii 2LA . Accordingly, when we compared the fractions of X22 that co-localized with BU43, there were no differences observed between Ii and Ii 2LA . However, when we compared the fractions of BU43 that colocalized with X22, we found that ratio to be 15 times higher for cells expressing Ii compared with Ii 2LA . We next analyzed whether overexpression of Ii or CD1d, or their internalization deficient mutant versions, would influence Tf co-localization with clathrin. The results show that the fraction of Tf-TR that co-localized with clathrin was significantly reduced when wildtype Ii or CD1d was expressed compared with the mutant versions (Fig. 7).
No Effect on TfR Recycling upon Overexpression of Proteins with Competing Internalization Signals-After internalization of the TfR into early endosomes, the receptor is recycled back to the plasma membrane via a specialized recycling compartment. The reduced Tf uptake observed above could in principle be caused by premature hyper-recycling of the TfR upon overexpression of proteins with internalization signals. To investigate whether recycling of the TfR was affected by the overexpression of membrane proteins, we assayed the kinetics of the retrograde transport. The recycling pathway was first saturated with radiolabeled Tf for 30 min, and the surface-bound 125 I-Tf was then stripped off. The release of 125 I-Tf into the medium was measured at various time points up to 1 h and compared with the cell-bound activity in the lysates. As shown in Fig. 8, the pool of internalized Tf was rapidly released into the medium with a half-time of approximately 10 min. The exocytic transport was not influenced by overexpression of proteins containing signals for endocytosis. Overexpression of proteins with mutated internalization signals also did not affect the recycling of 125 I-Tf (data not shown). These results indicate that the reduced uptake of Tf observed above was not due to increased recycling. It is also apparent that the retrograde transport steps of the TfR could not be saturated by overexpression of proteins not destined for the recycling pathway.
Basolateral Distribution of the TfR upon Overexpression of Proteins with Competing Internalization Signals-For many proteins the sorting signals for basolateral transport overlap with those involved in endocytosis. We wanted to study the effect of the expression level on the basolateral distribution of the proteins above. The two leucine-based motifs in the cytoplasmic tail of Ii are basolateral sorting motifs, whereas the tyrosine-based motif in CD1d is not critical for its basolateral sorting. To determine the polarized distribution of Ii and CD1d at various expression levels, cells were cultured on permeable filter supports to form tight monolayers and incubated with various concentrations of CdCl 2 prior to the experiments. Integrity of the monolayer after induction with Cd 2ϩ was confirmed by diffusion analysis of radiolabeled proteins and by electron microscopy (data not showed). 125 I-BU45 or 125 I-D5 was added to the basolateral or the apical side of the cell monolayer at 4°C. After removal of unbound antibodies the filters were excised, and bound radioactivity was detected in a ␥ counter. The fraction of specific binding to the apical and

TABLE II Colocalization between clathrin and invariant chain constructs
TrC cells expressing Ii or the mutant Ii 2LA in which both internalization signals are inactivated were labeled for confocal fluorescence microscopy with the anti-Ii antibody BU43 and the anti-clathrin antibody X22 as described in the legend to Fig. 6. Pixels with medium-to-high intensities of red (BU43), green (X22), and both red and green (colocalized) were selected as described under "Experimental Procedures." The number of regions made up by the selected pixels, and their average area, were estimated by computer analysis. basolateral domain was calculated for the different expression levels (Fig. 9A). At low-to-moderate expression levels more than 90% of both Ii and CD1d was localized on the basolateral membrane domain. Increased expression levels resulted in significant apical mistargeting of both Ii and CD1d, demonstrating that the basolateral protein targeting was saturable. The biochemical data were corroborated by confocal microscopy of total labeled polarized cells (Fig. 10A). We next examined the polarized distribution of the TfR upon overexpression of Ii or CD1d. Polarized cells were induced with Cd 2ϩ to express Ii or CD1d and incubated with 125 I-Tf, added to the apical or basolateral medium, at 4°C. The binding assays showed that about 90% of the iodinated Tf bound to the basolateral surface in control cells (Fig. 9B). As with non-polarized cells, overexpression of both Ii and CD1d resulted in increased surface expression of the TfR. However, the polarized distribu-tion of TfR was not affected (Fig. 10B), indicating that the TfR can utilize other components of the basolateral sorting machinery.

DISCUSSION
By using an inducible promoter the Ii expression level was increased up to 25-fold above background. An effect of higher Ii synthesis was a decreased rate of Ii internalization and increased surface expression, suggesting that the internalization machinery was saturated. To examine the saturation effect further, we used the TfR, whose internalization depends on a typical tyrosine-based motif, YTRF (7,(27)(28)(29), as an assay molecule. Uptake and recycling of radiolabeled Tf in MDCK cells stably transfected with the human TfR was measured, and we obtained results resembling those of previous studies (52,53). By use of an inducible expression system, we coexpressed proteins containing various tyrosine-or leucinebased sorting signals, and the results showed that internalization of the TfR was decreased when other proteins containing either tyrosine-or leucine-based internalization signals were overexpressed. Expression of proteins with inactivated internalization signals did not affect the Tf uptake, indicating that the saturation effect depended on intact sorting signals.
Cells must be able to cope with stress situations where the expression of proteins such as surface receptors is increased. In that perspective, it may be surprising that saturation effects could be observed when the expression of proteins bearing internalization signals was raised 10 -15 times above back- ground level. However, the background level was significant in our experiments, about the same magnitude as endogenous membrane protein expression. In addition, as we used nonnative promoters, expression of components of the sorting machinery was not co-stimulated, contrary to what happens in in vivo situations (19). Others (20,(22)(23)(24) have also reported saturation effects at expression levels similar to what we used in our experiments. Overexpression did not influence the uptake of the fluid-phase marker HRP and the membrane marker ricin, indicating that signal-dependent steps were specifically inhibited.
Both tyrosine-and leucine-based signals can be recognized for clustering into clathrin-coated pits, and we chose to examine the effect of overexpression of other proteins on Tf-clathrin co-localization by digital confocal image analysis. We found that both overexpression of Ii and CD1d resulted in reduced co-localization of Tf and clathrin at the plasma membrane, suggesting that clustering into clathrin-coated pits was inhibited. Analysis of the confocal images also revealed that overexpression of Ii resulted in increased numbers of regions labeled for clathrin. Recent results from a different analysis of confocal images indicated that endocytic clathrin-coated pit formation was independent on receptor internalization signal levels (54). However, overexpression of proteins with tyrosine-or leucinebased signals have previously been shown to promote recruitment of AP-1 to perinuclear structures (55)(56)(57) and overexpression of proteins with tyrosine signals; the TfR (58, 59) and HIV-1 Nef (60), has been associated with increased numbers of clathrin-coated pits on the plasma membrane. Our results suggest that leucine-based sorting signals can also promote formation of clathrin-coated structures at the cell surface.
Interactions between recombinant sorting signals and subunits of the adaptor complexes appear to be relatively weak in cell-free systems, typically in the micromolar range (61)(62)(63)67). Bremnes et al. (76) showed in a cell-free system that the membrane-distal leucine-based signal of Ii competed with a tyrosine-based signal for binding to adaptor chains. However, experiments in both cell-free systems and cell lines have been somewhat contradictory, and the in vivo significance of such interactions is currently unclear (for a review, see Ref. 6, page 764). Our results seemingly contradict those of Marks et al. (22) who reported that some leucine-based signals did not compete with certain tyrosine-based signals. However, efficient sorting of membrane proteins in live cells depends on numerous factors other than the recognition motif (for a review, see Ref. 79). Moreover, a recent study suggests the existence of functionally and biochemically distinct subpopulations of clathrin-coated pits on the plasma membrane (80), and a dual coated pit pathway hypothesis has been put forward by Weigel and Oka (81). It has been shown that some proteins with tyrosine-based signals, able to interact with adaptor complexes in cell-free systems, did not compete for internalization (82,83). In addition, sorting motifs may be differentially interpreted in different cells (84). These findings indicate that internalization of membrane proteins does not solely rely on interactions with AP-2 but may be a result of multivalent low affinity interactions. Our results suggest that tyrosine-and leucine-based sorting signals may compete for clustering into clathrin-coated pits, but we cannot distinguish if this was due to competition for common binding sites or steric hindrance at different sites on the sorting component(s) or maybe just for space within the pits.
Several studies have shown that the precise requirements for a signal to work in endocytosis is not necessarily the same as for basolateral sorting (23,(85)(86)(87), direct sorting from the TGN to lysosomes (88), or post-endocytic lysosomal targeting and degradation (89,90). However, experiments with the vacuolar proton pump inhibitor bafilomycin A 1 indeed suggest a role for the internalization motif in post-endocytic sorting of the TfR (91,92). Furthermore, recycling of the low density lipoprotein receptor to the basolateral surface of polarized MDCK cells has been shown to depend on the same tyrosine-based sorting signals required for basolateral sorting in the TGN (93). Altogether, these data indicate that receptor recycling is not simply a bulk process.
After internalization, the TfR is sorted from sorting endosomes to a special recycling compartment from where it is transported back to the plasma membrane (for reviews, see Ref. 6, page 768, and Ref. 26). To investigate whether overexpression of other tyrosine-based sorting signals would affect TfR recycling, we assayed Tf secretion in cells expressing proteins with various intact or inactivated sorting signals. The results showed that TfR recycling was not affected by the overexpression of other proteins containing internalization signals. One explanation for this could be that the internalization signal of the TfR was not involved in its recycling (28,94) and that TfR recycling was dependent on some unknown signal or followed the bulk membrane flow (30). However, we cannot exclude that the YTRF signal of the TfR was somehow involved in recycling but that the interaction with the actual sorting components could not be saturated with the signals studied.
Both tyrosine-and leucine-based sorting signals can be recognized for basolateral sorting in polarized cells. Basolateral sorting signals are generally divided into two groups depending on whether they are co-linear with signals for clathrin-coated pit localization or not (for a review, see Ref. 95). It has previously been shown that basolateral distribution of the low density lipoprotein receptor, which had one basolateral sorting signal co-linear with the tyrosine-based internalization signal (23), and the cation-dependent mannose 6-phosphate receptor, whose basolateral sorting information was unrelated to the internalization motifs (24), were both sensitive to the expression level; upon overexpression the distribution was shifted toward the apical domain. Both Ii and CD1d have been shown to distribute mainly on the basolateral membrane domain of MDCK cells. Whereas both internalization motifs in the Ii cytoplasmic tail can function as basolateral sorting signals (96,97), the internalization motif is not required for basolateral distribution of CD1d (36). In this study we have demonstrated that the distribution of both proteins was shifted toward the apical domain upon overexpression. CD1d was not mislocated to same extent as Ii, possibly due to the abundant basolateral sorting information contained within the Ii tail; at least three separate signals are individually sufficient for efficient basolateral distribution (96).
The TfR is mainly distributed on the basolateral membrane domain of MDCK cells (31), and analysis of truncated TfR constructs have shown that the cytoplasmic tail was required for the basolateral distribution (32,33). In a recent report from Odorizzi and Trowbidge (34), it was shown that a region of the tail comprising the internalization signal was required and was sufficient for basolateral sorting. However, the internalization signal itself was not involved in basolateral transport of neither newly synthesized nor internalized TfR molecules. In this study we examined the polarized distribution of the TfR in cells overexpressing Ii or CD1d. More than 90% of the surface TfR was distributed on the basolateral membrane, in accordance with previous reports (41). Although the expression of TfR on the cell surface increased upon overexpression, indicating that internalization was also inhibited in polarized cells, the TfR was still efficiently distributed to the basolateral domain. Altogether, the results suggest that several saturable mechanisms are involved in polarized sorting in MDCK cells. It is evident that more detailed studies are required to elucidate the complex sorting of membrane proteins trafficking in the exocytic and endocytic pathways, both in terms of molecular interactions as well as characterization of sorting compartments and transport routes. However, with the increasing number of candidate molecules able to interact with the sorting signals, it might be possible in the near future to study directly the sorting events in live cells.