The epidermal growth factor receptor juxtamembrane domain has multiple basolateral plasma membrane localization determinants, including a dominant signal with a polyproline core.

The epidermal growth factor (EGF) receptor is located predominantly in the basolateral membrane of polarized epithelia, where it plays a pivotal role during organogenesis and tissue homeostasis. We have shown previously that a 22-amino acid sequence in the EGF receptor juxtamembrane domain contains autonomous sorting information necessary for basolateral localization using the Madin-Darby canine kidney epithelial cell model. The goal of this study was to determine the molecular basis of EGF receptor basolateral membrane expression using site-directed mutagenesis to modify specific residues in this region. We now show that this sequence has two different, functionally redundant basolateral sorting signals with distinct amino acid requirements: one dependent on residues (658)LL(659) conforming to well-characterized leucine-based sorting signals, and a second containing a polyproline core comprising residues Pro(667) and Pro(670) ((667)PXXP(670)). Our data also suggest that Arg(662) contributes to the function of the proline-based signal. (667)PXXP(670) was the dominant signal when both motifs were present and was more effective than (658)LL(659) at overriding strong apical sorting signals located in the same molecule. Site-directed mutations at Arg(662), Pro(667), and Pro(670) were also associated with increased apical expression of full-length EGF receptors, demonstrating for the first time that the juxtamembrane region is necessary for accurate polarized expression of the native molecule.

The ability to establish and maintain plasma membrane asymmetry is a fundamental property of polarized epithelial cells that form physical barriers between various body compartments (reviewed in Refs. 1 and 2). Epithelial cell plasma membranes are organized into distinct apical and basolateral domains, which face the lumen or underlying cells and connective tissue, respectively. Each domain's unique molecular com-position facilitates a wide variety of organ-specific functions, including vectorial transport and compartmentalized cell signaling (1)(2)(3). The dynamic nature of plasma membrane asymmetry allows for epithelial cell plasticity and the ability to respond to a variety of physiological cues. Alterations in membrane polarity are often associated with epithelial cell dysfunction and pathophysiology, stressing the central role of spatial organization in normal cell function (reviewed in Ref. 4).
Epithelial membrane asymmetry is due in part to signalmediated domain-selective membrane protein sorting from intracellular compartments (reviewed in Refs. [5][6][7]. A number of relatively diverse apical sorting signals have now been identified, including glycophosphatidylinositol (GPI) 1 membrane anchors (8,9), ectodomain glycans (10,11), and transmembrane or ectodomain amino acid sequences (12). Basolateral sorting signals are generally composed of relatively short cytoplasmic amino acid sequences (6,7), and many are related to tyrosine (13,14)-or leucine (15,16)-based signals originally identified as clathrin-coated membrane localization signals (reviewed in Refs. 17 and 18). Tyrosine-and leucine-based signals are linked to clathrin via interactions with specific clathrin adaptor protein (AP) subunits (reviewed in Refs. 19 and 20). The three major classes of mammalian APs are AP-1 found predominantly at the trans-Golgi network (TGN), AP-2 at the plasma membrane, and AP-3 at the TGN and endosomes (19,21,22). Although a role for AP-facilitated transport in clathrin-dependent pathways is well-established (19), it is only recently that these molecules have been implicated in basolateral transport at the TGN (23,24). Other basolateral sorting signals have been identified, however, that bear no relation to known clathrin-coated membrane localization motifs. Other than clusters of essential charged amino acids, most of the signals in this category are devoid of recognizable motifs that might offer clues regarding physiologically relevant protein-protein interactions (13,25,26).
The EGF receptor is located in the basolateral membrane in many different epithelial cell types (reviewed in Refs. 1 and 27). Nowhere is the importance of restricted basolateral EGF receptor expression more apparent than in polycystic kidney disease (28 -30), where apically mislocalized receptors have a major role in disease progression (31,32). We have shown previously that newly synthesized EGF receptors are delivered directly to the basolateral surface in the MDCK cell model (33) and have identified sorting information located between juxtamembrane domain residues Lys 652 and Ala 674 necessary for domain-specific targeting of cytoplasmically truncated receptors (26) (see Fig. 1A). Importantly, this same sequence mediates basolateral targeting when transplanted to a heterologous reporter molecule, proof that is an autonomous, dominant signal (26). Residues Lys 652 through Ala 674 lack critical tyrosine residues (26) and do not overlap any of the EGF receptor sorting signals responsible for clathrin-mediated internalization located in the carboxyl terminus (34). In addition to MDCK cells, EGF receptors also exhibit a predominantly basolateral localization in LLC-PK 1 cells (35,36), which lack a novel epithelial cell-specific AP-1 subunit isoform necessary for correctly sorting basolateral membrane cargo with AP-dependent sorting signals (37). Hence, polarized EGF receptor sorting from internal compartments is likely mediated by novel proteinprotein interactions. Data presented in this study show that the EGF receptor juxtamembrane domain has a hierarchy of functionally redundant basolateral membrane localization signals with distinct amino acid requirements.

EXPERIMENTAL PROCEDURES
Cell Culture-Madin-Darby canine kidney (MDCK) epithelial cells were maintained in minimal essential medium (MEM) supplemented with 10% fetal bovine serum (FBS) and 2 mM glutamine. MDCK cells were seeded on polycarbonate Transwell filter inserts (0.4-m pore size) (Costar Corp., Cambridge, MA) at a density of 5 ϫ 10 5 cells per 12-mm filter, or 5 ϫ 10 6 cells per 75-mm filter, to generate electrically resistant monolayers with well-developed tight junctions suitable for domainspecific assays 4 -6 days later (26,33). Chinese hamster ovary (CHO) cells were maintained in the alpha formulation of MEM supplemented with 10% FBS and 2 mM glutamine.
Point mutations were introduced into the full-length EGF receptor using a cDNA template cloned in pBK-CMV⌬lac⌬DraIII. pBK-CMV⌬lac⌬DraIII is a derivative of a pBK-CMV phagemid (Stratagene Cloning Systems, La Jolla, CA), created by deleting nucleotides 1098 -1299 in the inducible lac promoter (38) and by eliminating a unique DraIII site at nucleotide 240 in the f1 (Ϫ) origin of replication. Neither modification affected kanamycin-resistant growth in Escherichia coli. Sequences were amplified using the same forward primer described in the previous paragraph and one of the following reverse mutagenic primers designed to introduce a specific amino acid substitution (bold italics), a DraIII site (underlined) to facilitate subcloning, and a silent SacI site (double underlined) to facilitate recombinant screening without changing amino acid sequence: R662T, 5Ј-CTTCTCCACTGGGTG-TAAGTGGCTCCACGAGCTCAGTCTCCTGCAGC-3Ј; P667A, 5Ј-CTT-CTCCACTGGGTGTAAGTGCCTCCACGAGCTCCC-3Ј; and P670L, 5Ј-CTTCTCCACTGAGTGTAAGAGGCTCCACGAGCTCCC-3Ј. PCR products were digested at an EcoNI site located at EGFR nucleotide 1876 and the DraIII site introduced at the 3Ј-end of the PCR products, and ligated to full-length EGF receptor sequences cloned in pBK-CMV⌬lac⌬DraIII digested with the same enzymes. When sequenced, the recombinant products had an internal deletion, due to a second previously unrecognized EcoNI site located 5Ј to the nucleotide 1876 EcoNI site. To create full-length products, recombinant molecules were digested with BstXI and XbaI, liberating an 1836-nucleotide product encoding the carboxyl half of the full-length EGF receptor, including the mutation. These fragments were ligated to a 2100-nucleotide XhoI-BxtXI fragment encoding the amino half of the wild-type EGF receptor, including the sequence deleted in the initial recombinants, and pBK-CMV⌬lac⌬DraIII digested at XhoI and XbaI sites located at 5Ј and 3Ј sites, respectively, in the polylinker. An E673A substitution was made using a forward mutagenic primer 5Ј-CTCTTACACCCAGTGGAGCAG-CACCCAACCAAGC-3Ј, designed to anneal to a DraIII site (underlined) located at EGF receptor nucleotide 2284 and to introduce the amino acid substitution (bold italics). The reverse primer, 5Ј-CAAACGGT-CACCCCGTAGCTCCAGACGTCACTCTCTGGT-3Ј, was designed to anneal to a BstEII site (underlined) located at EGF receptor nucleotide 2889, and to introduce a silent AatII site (double underlined) for recombinant screening. PCR products were digested with DraIII and BstEII and ligated to full-length EGF receptor sequences cloned in pBK-CMV⌬lac⌬DraIII digested with the same enzymes. cDNAs encoding the extracellular domain of decay accelerating factor (DAFex), and DAFex linked to EGF receptor transmembrane and cytoplasmic sequences to residue 651 (DAF-651) and to residue 674 (DAF-674), cloned in pCB6ϩ have been described previously. DAF-663 was made using a forward primer 5Ј-GTATCTCGAGGGCTGTCCAAC-3Ј, designed to anneal at the 5Ј-end of sequences encoding the EGF receptor transmembrane domain, and to incorporate a silent XhoI site (underlined) without altering the amino acid coding sequence for EGF receptor residues Leu 609 and Glu 610 (bold italics); and a reverse mutagenic primer 5Ј-GTTCGTCTAGATCACTCCCTCTCCTG-3Ј, incorporating a stop code in place of the code for Glu-663 (bold) and an XbaI site (double underlined) compatible with the pCB6ϩ polylinker. DAF-674 with a P667A substitution was made using the same forward primer and a reverse mutagenic primer, 5Ј-GCGATATCTCAAGCTTCTC-CACTGGGTGTAAGTGCCTC-3Ј, which was designed to introduce a P667A substitution (bold italics), a Pro675STOP (bold), and an EcoRV site (underlined) compatible with the pCB6ϩ polylinker. PCR products were digested with XhoI and XbaI or EcoRV and ligated directly to pCB6ϩ/DAFex digested with the same restriction enzymes.
PCR primers were designed using the DNASTAR software package (DNASTAR, Inc., Madison, WI). PCR amplifications were carried out using a RoboCycler 40 temperature cycler (Stratagene Cloning Systems, La Jolla, CA). All PCR products and religated recombinant products were sequenced by automated DNA sequencing (Cleveland Genomics, Cleveland, OH). Sequences were verified by analysis of genomic DNA recovered from permanent clonal cell lines.
Transient Transfections and Permanent Cell Lines-CHO cells were seeded at a density of ϳ2 ϫ 10 6 cells/100-mm tissue culture dish and transfected using DNA mixed with FuGENE 6 transfection reagent (Roche Diagnostics, Indianapolis, IN) 5 h later. Filter-grown MDCK cells were transfected using DNA mixed with Lipofectin transfection reagent (Invitrogen, Rockville, MD) added to both sides of the filter. Both cell types were assayed for transient membrane domain-specific protein expression 40 -48 h post-transfection. Exogenously expressed human EGF receptors are readily detected in both cell types in CHO cells, because they lack endogenous receptors (39), and in MDCK cells, because exogenously expressed human receptors can be distinguished from endogenous canine receptors with species-specific antibodies (33).
To make permanent MDCK cell lines, cells seeded at a density of ϳ7 ϫ 10 5 cells/100-mm tissue culture dish were transfected with DNA-Lipofectin complexes as described previously (33). Cells that had been grown in selection medium containing G418 (0.8 mg/ml Geneticin, Invitrogen) for 10 -14 days were labeled with the monoclonal antibody EGF-R1, specific for an external human EGF receptor peptide epitope (40,41), followed by FITC-conjugated goat anti-mouse IgG (Jackson Immuno-Research Laboratories, Inc., West Grove, PA), for enrichment by sterile sorting on a flow cytometer (Cytofluorograph IIs, Ortho Instruments, Westwood, MA).
Metabolic Labeling and Immunoprecipitations-Cells were rinsed twice and then preincubated in methionine and cysteine-free medium for up to 1 h, before metabolic labeling with [ 35 S]Express Protein Labeling Mix (1175 Ci/mmol, PerkinElmer Life Sciences, Wilmington, DE) diluted in the amino acid-deficient medium supplemented with 10% dialyzed FBS and 0.2% BSA. Filter-grown MDCK cells were labeled from the basolateral surface. Labeling medium was replaced with complete MEM supplemented with a 10-fold excess of non-radioactive methionine and cysteine (chase medium), and cells were incubated for additional periods of time. Cells were lysed with 1% (w/v) Triton X-100 in 0.1 M Tris, pH 7.4, supplemented with 2 mM EDTA, 1 mM EGTA, 0.2 mM phenylmethylsulfonyl fluoride, and 1 M leupeptin, for immunoprecipitation with antibodies recognizing external epitopes in the human EGF receptor (EGF-R1) or DAF (gift of Ed Medof, Case Western Reserve University) absorbed to protein A-Sepharose CL-4B beads (Sigma Chemical Co., St. Louis, MO). After extensive washing, immunoprecipitates were solubilized with Laemmli buffer for SDS-PAGE (42), and fluorographic detection. Radioactive quantitation was carried out by Phosphorstorage autoradiography (Molecular Dynamics Inc., Sunnyvale, CA).
Membrane Domain-specific Surface Immunoprecipitation-Filtergrown MDCK cells that had been metabolically labeled were rinsed twice with ice-cold MEM supplemented with 25 mM HEPES, pH 7.4, and 1% BSA (M/H/B) and then incubated with EGF-R1 ascites (10 l/ml) added to one side of the filter for 1 h on ice. Cells were rinsed four times with PBS supplemented with 1% BSA and then lysed with 1% (w/v) Nonidet P-40 in 0.1 M Tris, pH 6.8, supplemented with 15% (w/v) glycerol, 2 mM EDTA, 1 mM EGTA, and protease inhibitors. Cell lysates were added directly to protein A-Sepharose beads to capture surfaceexposed EGF receptors, and immunoprecipitates were washed, solubilized, and resolved by SDS-PAGE exactly as described in the previous section.
Membrane Domain-specific 125 I-EGF Cross-linking-Filter-grown cells were rinsed three times with ice-cold MEM supplemented with 0.2% BSA and then incubated with ϳ10 nM 125 I-EGF for 2 h at 4°C. Receptor-grade mouse EGF (Toyobo Biochemicals, Osaka, Japan) was labeled with 125 I (carrier-free, Ͼ350 mCi/ml, PerkinElmer Life Sciences) using chloramine-T. Cells were rinsed three times with the MEM/BSA solution and then incubated with 2 mM disuccinimidyl suberate (Pierce Chemical Co., Rockford, IL) in a solution of 0.1 M HEPES, pH 7.4, supplemented with 120 mM NaCl, 50 mM KCl, 8 mM glucose, and 1.2 mM MgSO 4 for 15 min at room temperature. The reaction was quenched by a 5-min incubation with 0.05 M Tris, pH 7.4, at room temperature. Cells were lysed with 1% Nonidet P-40 exactly as described above, and equal aliquots of total cell protein were separated by SDS-PAGE.
Confocal Laser Scanning Microscopy-Filter-grown cells were rinsed three times with PBS and fixed with 3% paraformaldehyde in PBS for 10 min at room temperature. Some cells were stained with antibodies to external epitopes added to the apical or basolateral surface of nonpermeabilized cells. Others were permeabilized with 0.1% Triton X-100 in PBS for 10 min at room temperature and then stained with the ZO-1-specific rat monoclonal R26.4C antibody (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), or a ␤-cateninspecific mouse monoclonal antibody (Transduction Laboratories, Lexington, KY). Secondary antibodies were fluorophore-conjugated, species-specific Fab fragments that had been solid-phase absorbed to prevent cross-reactivity with primary antibodies made in other species (Jackson ImmunoResearch Laboratories, Inc., Fort Washington, PA). Cells were stained with primary or secondary antibodies for 1 h each at 37°C. Antibodies were diluted in PBS supplemented with 3% radioimmunoassay-grade bovine serum albumin, and cells were blocked with a solution containing 5% normal serum from the host animal used to generate the secondary antibody between incubations with primary and secondary antibodies. Filters were excised from plastic inserts and placed cell-side up on a glass slide. Coverslips were mounted on the cells using SlowFade anti-fade solution from Molecular Probes (Eugene, OR). Cells were examined with a Zeiss LSM 410 scanning laser confocal microscope (Zeiss, Gottingen, Germany) using the 488/568-nm wavelength lines of an argon-krypton laser. Cells were optically sectioned every 0.5 m. Image resolution using a Zeiss 100ϫ Neofluor objective and Zeiss LSM software was 512 ϫ 512 pixels.

RESULTS
Residues Arg 662 , Pro 667 , and Pro 670 Are Required for Accurate Basolateral Localization of Cytoplasmically Truncated cЈ-674 Molecules-The amino acid sequence of the originally identified a 22-amino acid juxtamembrane domain necessary for basolateral localization of cytoplasmically truncated EGF receptors is shown in Fig. 1A (26,33). We sought to characterize the molecular basis of basolateral localization, by focusing on several consensus amino acid protein interaction motifs located in this region. Examination of the sequence revealed the existence of a two consensus leucine-based motifs, located at Leu 658 -Leu 659 and Leu 664 -Val 665 (Fig. 1A). Although predominantly characterized in endocytic pathways (reviewed in Ref. 18), leucine-based motifs are implicated in basolateral sorting of at least two other membrane proteins (15,16) (also see Table I). Because leucine-based signals are impaired by dialanine substitution (18), we changed each of the leucinebased signals to alanines, either individually (P675STOP-LL-2xA or P675STOP-LV-2xA) or in combination (P675STOP-LL,LV-4xA), in the molecular setting of a cЈ-674 receptor. In each instance, the cytoplasmically truncated mutant receptors were delivered to the basolateral surface with approximately the same efficiency as wild-type cЈ-674 receptors (Fig. 2). In addition, all of the mutant receptors exhibited normal basolateral localization based on CLSM analysis (see Fig. 3 for P675STOP-LV-2xA; others not shown).
This region also has a proline-rich segment, residues 667 PLTP 670 , whose sequence is characteristic of protein interaction motifs with a polyproline core PXXP, where X denotes any amino acid (reviewed in Ref. 43). Recognition by cognate protein interaction partners is often dictated by the location of a positively charged residue relative to the PXXP core. Thus, proline residues Pro 667 and Pro 670 and the positively charged residue Arg 662 , as well as two negatively charged glutamic residues Glu 663 and Glu 673 , were altered by mutagenesis in cytoplasmically truncated cЈ-674 receptors (Fig. 1B). All of the residues were changed to alanines except for Arg 662 , which was changed to a threonine to mimic the residue found in the corresponding region of ErbB2 (44). The percentage of receptor mutants with individual substitutions at Arg 662 , Pro 667 , or Pro 670 that were delivered to the apical surface ranged from 25 to 30%, which was significantly higher than the 10% apical delivery characteristic of wild-type cЈ-674 receptors (Fig. 2) or full-length molecules (26,33). The same three mutations were also associated with higher than expected steady-state apical expression following domain-specific human receptor staining and CLSM (Fig. 3). These results suggest that a motif containing a polyproline core ( 667 PXXP 670 ) and an amino-terminal positively charged residue (Arg 662 ) has an important role in basolateral localization of cytoplasmically truncated receptors.
The EGF Receptor Juxtamembrane Domain Has Multiple Basolateral Determinants-Although individual mutations at residues Arg 662 , Pro 667 , or Pro 670 each had a modest effect on the efficacy and fidelity of basolateral sorting, none of the mutations was sufficient to completely inactivate the putative signal. Although not all basolateral signals are inactivated by a single point mutation (57), one possible explanation for this result is that the juxtamembrane domain has additional basolateral sorting information that is normally silent relative to the putative proline-based signal. To test that hypothesis, the polyproline core was removed by changing residue Leu 664 to a premature stop codon by site-directed mutagenesis (Fig. 1B).
Permanent MDCK cell lines expressing human receptors truncated to Glu 663 (or cЈ-663 receptors made with the L664STOP construct) were then evaluated for domain-specific expression of the recombinant molecule. Its membrane domain expression was compared with that of cЈ-651 receptors (made with the K652STOP construct), which lack the juxtamembrane domain and sort selectively to the apical membrane, and cЈ-674 receptors (made with the P675STOP construct), which exhibit high efficiency localization to the basolateral membrane (26,33). Biosynthetic delivery was evaluated by domain-specific surface immunoprecipitation to capture newly made human receptors following a pulsed metabolic label, and plasma membrane steady-state expression by domain-specific human 125 I-EGF chemical cross-linking or human EGF receptor staining and CLSM. Similar to full-length wild-type EGF receptors, relatively few cЈ-674 receptors are delivered to the apical mem-brane, compared with cЈ-651 receptors lacking juxtamembrane residues, where ϳ60% of the newly synthesized molecules were delivered to the apical surface during a 3-h interval (Fig. 4A). By comparison, cЈ-663 receptors exhibited an intermediate phenotype in the delivery assay, with ϳ30% of newly synthesized molecules targeted to the apical surface (Fig. 4A). At steadystate, the difference between cЈ-651 and cЈ-674 receptors membrane domain distribution was even more pronounced, with the vast majority of cЈ-651 receptors located at the apical surface based on results obtained by 125 I-EGF cross-linking (Fig. 4B) or by CLSM (Fig. 4C). Although cЈ-663 receptors also have a predominantly apical localization, basolateral expression was still evident in both of the steady-state assays (Fig. 4, B and C). These data suggest that residues Lys 652 through Glu 663 likely contain additional sorting information, because cЈ-663 receptors retain a limited capacity for basolateral expression. Residues Leu 658 and Leu 659 Are Critical Determinants When the Carboxyl Terminal Polyproline Core Is Deleted-To identify additional sorting motifs in the region between residues 652 and 663, a L658A,L659A dialanine substitution was engineered into a mutant receptor with a L664STOP lacking the proline-based carboxyl determinant (L664STOP-LL-2xA in Fig. 1B). Using CLSM to examine filter-grown cells that had been transiently transfected and then stained with an EGF receptor antibody to an external human-specific epitope added to both sides of the cell simultaneously, we observed that this mutation did cause an increase in the apical expression of receptor proteins compared with wild-type cЈ-663 receptors (Fig. 5B). Because mutation of the leucine-based motifs had no effect on polarized expression of cЈ-674 receptor receptors with an intact 667 PXXP 670 core (Figs. 2 and 3), these data suggest that the proline-based determinant is dominant over one or both of the leucine-based motifs in molecules with a P675STOP.
To verify that this region has two basolateral determinants, we created an additional set of receptor mutants with compound mutations in each of the putative sorting determinants in the molecular setting of a cЈ-674 receptor (P675STOP-LL,LV,PxxP-6xA in Fig. 5C). A mutant receptor with both proline residues changed to alanines was also made and analyzed by CLSM (P675STOP-PxxP-2xA in Fig. 5C). A majority of truncated molecules was present on the apical membrane when either of the compound mutations was expressed in combination with a P675STOP. Apical expression of the P675STOP-LL,LV,PxxP-6xA receptor, however, appeared to be slightly greater than that of the P675STOP-PxxP-2xA. To obtain quantitative data, three of these constructs (LL,LV-4xA, PxxP-2xA, and LL,LV,PxxP-6xA) were used to make permanent cell lines with uniform mutant protein expression. As shown in Fig. 6A, all of the mutant receptors exhibited elevated apical expression compared with cytoplasmically truncated wild-type cЈ-674 receptors, where ϳ5% of the total receptors located at the plasma membrane are localized to the apical surface, as judged by 125 I-EGF cross-linking. Approximately 20% of the receptors with both leucine-based motifs changed to alanines, and 33% of

FIG. 2. Membrane domain-specific delivery of cytoplasmically truncated EGF receptors with amino acid substitutions.
Membrane domain-specific delivery was determined using surface immunoprecipitation to capture newly synthesized EGF receptors following a 30-min pulse label with 35 S-labeled amino acids and a 2.5-h chase. Radioactive immunoprecipitates were resolved by SDS-PAGE and quantified by phosphorstorage autoradiography. Data are presented as the mean percentage of total surface radiolabeled EGF receptor delivered to the apical cell surface Ϯ S.E., n ϭ 3. Mean values for apical delivery of receptors with amino acid substitutions were compared with the mean value for molecules with a P675STOP (i.e. cЈ-674 receptors) using a standard two-tailed Student's t test, and statistically significant differences (p Յ 0.05) are denoted with asterisks.

FIG. 3. Membrane domain distribution of cytoplasmically truncated EGF receptors with amino acid substitutions in MDCK cells.
Filter-grown cells expressing receptor mutants (all in a P675STOP background) listed in the figure were fixed and stained with a monoclonal antibody specific for an external human-specific epitope added to either the apical (Ap) or the basolateral (BL) surface, followed by a FITC-conjugated secondary antibody. Cells were optically sectioned every 0.5 m, and vertical (x-z) optical sections perpendicular to the plane of the apical membrane were digitally compiled. Horizontal (x-y) optical sections are shown for sections just above the level of the tight junction (Ap staining) or for sections taken from the middle of the cell (BL staining). the receptors with a PxxP-2x substitution, were present on the apical surface. Apical expression of receptors with all three signals converted to alanines was ϳ65%, consistent with the results obtained by transient transfection. Permanent cell lines expressing the mutant receptors were also analyzed by domain-specific human EGF receptor staining followed by CLSM (Fig. 6B), to show that the entire cell population had increased apical expression. These results support the hypothesis that the failure to completely reverse polar sorting of cЈ-674 receptors by introducing mutations in the polyproline-based signal is due to compensatory activity of a functionally redundant signal that is normally silent.

The Same Determinants Identified in Truncated EGF Receptors Also Regulate Basolateral Localization of DAF-EGF Receptor Chimeras-
We have shown previously that residues EGF receptor residues Lys 652 to Ala 674 constitute a dominant autonomous basolateral localization determinant in reporter molecules containing the luminal domain from decay accelerating factor (DAF) attached to transmembrane and cytosolic sequences from the EGF receptor. To test whether any of the amino acid substitutions associated with increased apical expression of cytoplasmically truncated EGF receptors had a similar effect in another molecular setting, two of the mutations, L664STOP and P675STOP-P667A, were expressed in the  (n ϭ 3). C, filter-grown cells expressing cytoplasmically truncated human receptors listed in the figure were fixed and stained with a monoclonal antibody specific for an external human-specific epitope added to either the apical (Ap) or the basolateral (BL) surface, followed by a FITC-conjugated secondary antibody. Cells were optically sectioned every 0.5 m, and vertical (x-z) optical sections perpendicular to the plane of the apical membrane were digitally compiled.

FIG. 5. Membrane domain distribution of cytoplasmically truncated EGF receptors with compound amino acid substitutions.
Filter-grown cells expressing mutant receptors listed in the figure were fixed and stained with a monoclonal antibody specific for an external human-specific epitope added to both sides of the cell, followed by a FITC-conjugated secondary antibody. Cells were optically sectioned every 0.5 m, and vertical (x-z) optical sections perpendicular to the plane of the apical membrane were digitally compiled. Horizontal (x-y) optical sections are shown for sections just above the level of the tight junction (Ap staining) or for sections taken from the middle of the cell (BL staining). A, cells expressing wild-type (WT) sequences in combination with a K652STOP; B, cells expressing WT sequences in combination with a L664STOP, or receptors with a L658A-L659A substitution and a L664STOP; C, cells expressing receptors with compound mutations in combination with a P675STOP or WT receptor sequences and a P675STOP.
protein chimeras (DAF-663 and DAF-674A, respectively, in Fig. 7A). Like other GPI-anchored membrane proteins, native DAF partitions into Triton-insoluble membrane rafts during polarized sorting to the apical surface of polarized cells (45). DAF is partially soluble in Triton, if cell extracts are incubated at 37°C instead of 4°C (Fig. 7B, left and middle panels) (46). Replacement of the GPI anchor with sequences encoding the transmembrane domain and seven amino acids from the EGF receptor (DAF-651) is sufficient to render the protein chimera Triton-soluble at 4°C (Fig. 7B, left panel) and to cause nonpolar membrane delivery in MDCK cells (26). Protein chimeras containing EGF receptor cytosolic sequences, including EGF receptor basolateral sorting determinants (DAF-674) are also Triton-soluble (Fig. 7B, left panel) and undergo selective transport to the basolateral surface (26). Both of the recombinant plasmids DAF-663 and DAF-674A encoded a stable protein product in transiently transfected cells (see Fig. 7B, right panel, for DAF-674A; data for DAF-663 are not shown). MDCK cells that had been transfected on filters were examined for transient surface expression of recombinant molecules by do-main-specific staining of non-permeabilized cells using an antibody recognizing an external DAF epitope. As shown in Fig.  7C, transiently expressed wild-type DAF exhibited a predominantly apical localization, compared with wild-type DAF-674, which was detected predominantly on the basolateral surface. Protein chimeras with either a L664STOP (DAF-663) or a P675STOP with a P667A substitution (DAF-674A), however, were detected on both sides of the cell, similar to results obtained when the same mutations, are expressed in the molecular context of a cytoplasmically truncated EGF receptor.
Arg 662 , Pro 667 , and Pro 670 Are Each Necessary for Accurate Basolateral Localization of Intact EGF Receptors-To determine whether residues with a dominant role in basolateral localization of cytoplasmically truncated EGF receptors and in DAF-receptor chimeras are also important in the native molecule, Arg 662 , Pro 667 , and Pro 670 , as well as Glu 673 , were individually mutated in cDNAs encoding full-length receptors (see Fig. 1C). To design a PCR mutagenic primer with favorable secondary structure, it was necessary to convert Pro 670 to a leucine residue instead of an alanine. Pro 667 and Glu 673 were FIG. 6. Quantitative analysis of cytoplasmically truncated EGF receptors with compound amino acid substitutions. Permanent cell lines expressing mutant receptors listed in the figure were isolated for further analysis. A, Ap and BL steady-state expression was determined using domain-specific 125 I-EGF binding followed by chemical cross-linking. Radioactive EGF receptors resolved by SDS-PAGE were quantified by phosphorstorage autoradiography. The numbers beneath each lane indicate the percentage of total radioactivity present at the Ap or the BL plasma membrane domain. B, filter-grown cells were fixed and stained with a monoclonal antibody specific for an external human-specific epitope added to either the Ap or the BL side of the cell, followed by a FITC-conjugated secondary antibody. Cells were optically sectioned every 0.5 m, and vertical (x-z) optical sections perpendicular to the plane of the apical membrane were digitally compiled. Horizontal (x-y) optical sections are shown for sections just above the level of the tight junction (Ap staining) or for sections taken from the middle of the cell (BL staining).

FIG. 7. DAF and DAF-EGF receptor chimera expression. A,
diagram showing structures of the GPI-anchored DAF protein, and chimeras containing the DAF luminal domain (gray) attached to sequences derived from the EGF receptor. DAF-651, DAF-663, and DAF-674 contain EGF receptor cytosolic sequences truncated to Arg 651 , Glu 663 , or Ala 674 , respectively. DAF-674A has a P667A substitution in the EGF receptor cytosolic tail sequence. B, metabolically labeled cells that had been extracted with Triton X-100 at either 4°C or 37°C were immunoprecipitated with a DAF-specific antibody, and immunoprecipitates were resolved by SDS-PAGE. Lysates were prepared from permanent MDCK cell lines (left and middle panels) or transiently transfected CHO cells (right panel). Native DAF and DAF chimeras are usually seen as doublets, due to processing of N-linked glycans (45). Molecular weight standards: phosphorylase B, 97,400; bovine serum albumin, 66,200. C, filter-grown MDCK cells transfected with plasmid DNAs encoding DAF or DAF chimeras, were fixed and stained with DAF luminal domain antibodies added to both the apical (Ap) and basolateral (BL) sides of the cell. Cells were optically sectioned every 0.5 m, and vertical (x-z) optical sections perpendicular to the plane of the apical membrane were digitally compiled. Horizontal (x-y) optical sections are shown for sections just above the level of the tight junction (Ap staining) or for sections taken from the middle of the cell (BL staining). converted to alanines, and Arg 662 to a threonine to mimic the residue found in the corresponding region of ErbB2 (44). Each of the recombinant molecules was first expressed in CHO cells lacking endogenous EGF receptors to verify that transiently expressed cDNAs encoded stable protein products with appropriate molecular weights (Fig. 8A). Filter-grown MDCK cells were then transfected with each of these constructs to assess membrane domain-specific human EGF receptor expression at steady state by CLSM. As shown in Fig. 8B, mutant receptors with individual amino acid substitutions at Arg 662 , Pro 667 , or Pro 670 all had higher than expected apical expression, compared with cells expressing an equivalent level of wild-type molecules or mutant receptors with an E673A substitution. Increased apical expression was not due to aberrant cell polarity, because cells expressing the mutant receptors exhibited normal localization of ZO-1 at the tight junction, as well as E-cadherin-associated ␤-catenin along lateral membranes (Fig.   8C). To obtain quantitative data, we produced permanent MDCK cell lines expressing one of the amino acid substitutions showing increased apical expression (P667A), as well as one without an apparent effect on receptor localization (E673A), in the transient assay. Fluorescence-activated cell sorting was used to obtain cell populations with approximately equivalent levels of plasma membrane-exposed EGF receptor proteins, and EGF receptor expression was analyzed by domain-specific surface immunoprecipitation of newly synthesized molecules (Fig. 9). The data in Fig. 9 indicates that receptors with a P667A substitution exhibit substantially elevated delivery to the apical membrane, compared with wild-type receptors and mutant receptors with an E673A substitution, where the vast majority is delivered to basolaterally. DISCUSSION We have shown that two different types of signals located in the juxtamembrane region regulate basolateral EGF receptor expression in polarized epithelial cells. One class of signals conforms to leucine-based signals already characterized in other molecules, based on its dependence on residues 659 LL 660 . The second is dependent on two proline residues, Pro 667 and Pro 670 , with a possible contribution by Arg 662 . The presence of functionally redundant signals is the most likely explanation for the fact that basolateral sorting is not completely reversed unless critical residues are mutated in both of the signals. Despite superficial functional similarities, however, the two signals are fundamentally different in at least two respects. First, the proline-dependent 667 PXXP 670 signal is dominant over 659 LL 660 when both signals are present. Second, 667 PXXP 670 is more effective than 659 LL 660 in overriding apical sorting signals located in the same molecule. Hence, it is unlikely that these signals are used interchangeably, but instead may act to regulate EGF receptor transport by distinct mechanisms. For instance, alternative signals may regulate polarized sorting from different membrane compartments (e.g. TGN versus endosomes) (6) or in other cell types where asymmetrically distributed proteins are sorted differently than in MDCK cells (47). Given the complexity of polarized membrane protein sorting, it is also possible that distinct sorting motifs mediate different transport steps along a common secretory pathway. In that event, the dominant phenotype of receptors with a defective 667 PXXP 670 signal suggests that it acts downstream of 658 LL 659 .
Although our data suggest that residues comprising the proline-dependent motif are necessary for restricting native EGF receptor expression to the basolateral membrane in polarized cells, additional signals not yet identified may also contribute to the final membrane phenotype. The additional signals may direct independent sorting events in the secretory pathway or specify post-TGN events such as vesicle docking or membrane retention (48). In fact, essential basolateral localization signals FIG. 8. Transient expression of full-length EGF receptors containing single amino acid substitutions. A, cells that had been transfected with cDNAs encoding wild-type human EGF receptor (EGFRwt) or intact receptors containing the amino acid substitutions listed in the figure, were metabolically labeled 48 h later. Cell lysates were immunoprecipitated with a human EGF receptor-specific antibody, and immunoprecipitates were resolved by SDS-PAGE. Molecular weight standards: myosin, 200,000; ␤-galactosidase, 116,300; phosphorylase B, 97,400; bovine serum albumin, 66,200. B, filter-grown cells were fixed and stained with EGF-R1 added to either the apical or the basolateral surface, followed by a FITC-conjugated secondary antibody. C, filter-grown cells were fixed, permeabilized, and stained with specific antibodies to the tight junctional protein ZO-1 or to ␤-catenin (abbreviated ␤-cat) associated with the cytosolic tail of the epithelial cell homotypic adhesion molecule E-cadherin. B and C, cells were optically sectioned every 0.5 m, and vertical (x-z) optical sections perpendicular to the plane of the apical membrane were digitally compiled. Horizontal (x-y) optical sections are shown for sections just above the level of the tight junction (Ap EGF receptor staining and ZO-1) or for sections taken from the middle of the cell (BL EGF receptor staining and ␤-catenin).
FIG. 9. Membrane domain-specific delivery of full-length EGF receptors containing single amino acid substitutions. Permanent MDCK cell lines expressing either a wild-type (WT) human EGF receptor or mutant receptors with a P667A or an E673A substitution were pulse-labeled for 30-min with 35 S-labeled amino acids followed by a 2.5-h chase. Apical (Ap) and basolateral (BL) surface delivery was determined using domain-specific surface immunoprecipitation to capture newly synthesized EGF receptors. have been identified in the Caenorhabditis elegans EGF receptor homologue (49), as well as the mammalian ErbB2 receptor (50), that bind to proteins with PDZ domains whose expression is critical for accurate receptor localization in polarized cells. Although the human EGF receptor lacks a consensus PDZbinding motif, other intrinsic signals could fulfill a similar role at the basolateral membrane. For example, an actin-binding domain mapped to EGF receptor carboxyl-terminal residues 984 -996 several years ago (51) could contribute to basolateral membrane retention, by tethering receptors to the actin cytoskeleton. This is consistent with our previous studies showing that expression of cytoplasmically truncated receptors lacking an intact actin-binding motif is not restricted to the basolateral domain, despite their highly efficient basolateral delivery in the secretory pathway (26). However, because the EGF receptor actin-binding motif is located in a domain that also contains at least two known internalization signals (34), understanding the molecular basis of any signals in this region that might be involved in polarized expression is not so clearcut. Clarifying the role of multiple sorting signals all contributing to basolateral localization will help illuminate the complex network of protein-protein interactions that organize and coordinate receptor trafficking and signaling. It may also provide a rationale for understanding how EGF receptor sorting is disrupted in different forms of inherited polycystic kidney disease, if individual disease gene products target distinct proteinprotein interactions acting sequentially in a common sorting pathway.
Although many different types of protein-protein interactions are mediated by proline-rich motifs, this represents the first basolateral localization determinant identified with these molecular characteristics (Table I). Only one other signal, residues RRPGAPESKCSR in major histocompatibility complex II invariant chain (Table I), has a potentially similar proline-rich core (underlined) (16). The best-studied proline-based interactions are those involving protein partners with SH3 or WW domains identified in numerous signal transduction and cytoskeletal protein adaptors (reviewed in Ref. 43). In addition to critical polyproline core residues, these interactions are also dependent on other residues in the core itself as well as in the nearby vicinity. SH3 ligands, for example, can be classified as class I or class II depending on the location of a positively charged residue relative to the polyproline core. Although the EGF receptor basolateral sorting signal could be classified as a class I SH3 ligand, based on its dependence on a positively charged residue (Arg 662 ) on the amino-terminal side of a polyproline core ( 667 PXXP 670 ), the diversity of protein interactions mediated by proline-rich sequences necessitates further investigation. The polyproline cores of many SH3 ligands also have critical serine or threonine residues, whose phosphorylation regulates binding affinity (43). Interestingly, the polyproline core of the dominant EGF receptor basolateral sorting signal contains the major EGF receptor mitogen-activated protein kinase substrate Thr 669 (52,53), raising the possibility that sorting signal recognition is regulated by phosphorylation.
Several key questions remain. First, what is the nature of the protein interactions mediated by these basolateral localization signals? AP-1-regulated sequestration of membrane cargo with tyrosine-dependent sorting signals into clathrin-coated transport vesicles is thus far the only mechanism implicated in regulated sorting of basolateral proteins (23,24). However, the nature of the EGF receptor sorting signals, as well as the accuracy of EGF receptor sorting in LLCPK 1 cells lacking the AP-1 isoform necessary for polarized sorting (37), suggests that these motifs underlie a novel mechanism. Whether EGF recep-tors are recruited to clathrin-coated transport vesicles independent of AP-1, or undergo sorting mediated by an entirely different set of molecules, remains to be seen. Second, what is the nature of the local structure of EGF receptor basolateral sorting signals allowing interaction with the sorting machinery? Although structural propensities can be predicted computationally (26), these algorithms do not take into account important features such as its proximity to the cytoplasmic face of the membrane. This information is essential to understand how sorting signal availability is regulated in the cell. In addition, the 667 PXXP 670 signal is located adjacent to ligandregulated lysosomal sorting signal at Leu 679 -Leu 680 (38,54). Thus in addition to its role in basolateral sorting, 667 PXXP 670 may also contribute to the structure of a nearby signal required for efficient transport of receptors to lysosomes following ligand stimulation. These questions are currently under investigation.