INTRODUCTION

Formation of epithelial cell polarity is a fundamental process during embryonic development and organogenesis (1). In the adult, polarized epithelia provide a barrier between internal compartments as well as to the external environment, allowing the organism to regulate ion and solute transport and maintain homeostasis (2, 3). Development of epithelial cell polarity requires a well orchestrated, domain-specific segregation of lipids, membrane proteins, and cytoskeleton (2-4) and is guided by two independent sets of physiological signals. The first set of signals is regulated by E-cadherin, a Ca²⁺-dependent homotypic adhesion molecule required for formation of structures comprising the lateral barrier between epithelial cells (5). E-cadherin also triggers assembly of submembranous cytoskeleton important for membrane domain localization of basolateral proteins that interact with cytoskeletal components (4). Orientation of the apicobasal axis is dependent on a second set of signals transduced by integrin-mediated cell-substrate interactions (6). Epithelial cells have also devised strategies for delivering resident membrane proteins from the trans-Golgi network (TGN)¹ to the correct plasma membrane location (7). The recent demonstration that fibroblasts utilize similar strategies (8) suggests that mesenchymal to epithelial cell conversion during development involves differential utilization of pre-existing sorting routes, rather than establishment of new ones.

The EGFR is localized at the basolateral membrane of epithelial cells both in vivo (2, 3) and in cell lines commonly used for the study of membrane protein sorting, including Madin-Darby canine kidney (MDCK) cells (9) and porcine kidney LLC-PK1 cells (10). Mice lacking EGFRs suffer from impaired epithelial development in several organs, including skin, lung, liver, kidney, and gastrointestinal tract (11-13). EGFR expression is therefore critical during epithelial organogenesis and may well play a role in polycystic kidney disease (PKD) where nonpolar EGFR distribution has been noted in epithelia lining cysts originating in certain tubule segments (14). PKD is an inherited set of diseases characterized by progressive development of large fluid-filled cysts. The ensuing interaction of apical EGFRs with EGF in cyst fluid is postulated to contribute to progressive cyst enlargement (14). According to the "maturation arrest" hypothesis, gene products important at some step during mesenchymal to epithelial cell conversion in kidney organogenesis are expressed abnormally in PKD, thereby locking renal cells in a permanent state of underdevelopment (15). Since EGFR is the only membrane protein consistently shown to be mislocalized in PKD (14), a putative gene defect(s) might specifically affect the ability of developing renal epithelial cells to establish polar sorting routes for EGFR. This hypothesis is strengthened by recent evidence showing that EGFRs exhibit apical as well as lateral localization during embryonic development of early proximal tubules and in collecting duct (16).

Intracellular transport to plasma membrane domains is regulated in part by intrinsic signals recognized at different steps in polar sorting pathways. A number of polar sorting determinants have now been identified in membrane proteins, and principles for both sequence and structural requirements are beginning to emerge. Although a general consensus sequence has not been identified, basolateral sorting determinants typically consist of discrete cytoplasmic amino acid sequences. Several basolateral signals require either a critical tyrosine residue or a di-leucine motif similar to those described in endocytic pathways (reviewed in Refs. 17-19). However, in molecules with overlapping basolateral and endocytic signals, each signal usually has distinct amino acid requirements suggesting recognition by different sorting receptors (20-22). The low density lipoprotein receptor has two basolateral sorting signals, one which overlaps its endocytic signal and a second which is independent of any known endocytic code (18). Basolateral sorting of the polymeric immunoglobulin A receptor is regulated independently of a known endocytic signal, instead depending on a core of charged residues near the cytoplasmic face of the membrane-spanning domain (23). Similar to some tyrosine-based motifs (17-19), the polymeric immunoglobulin receptor basolateral signal adopts a -turn conformation in solution (23), suggesting -turns may be a fundamental feature of basolateral sorting signals.

A number of studies suggest that glycosylphosphatidylinositol (GPI) anchors help regulate apical sorting (reviewed in Ref. 24). GPI-anchored proteins, as well influenza hemagglutinin protein which is apically targeted but lacks a GPI anchor, cluster in glycolipid-enriched, detergent-insoluble membrane domains in the TGN (25, 26). This suggests that apical sorting could be regulated by selective partitioning into specialized membrane microdomains. Another potential apical sorting determinant is N-linked glycosylation of residues in the extracellular domain of membrane and secretory proteins (reviewed in Ref. 27). Evidence in support of this mechanism is provided by experiments in which treatment with the N-glycosylation inhibitor tunicamycin or mutational removal of N-glycosylation sites results in randomized delivery of certain secretory molecules (27). It is assumed that some membrane glycoproteins utilize a similar mechanism, since apical membrane proteins lacking transmembrane and cytoplasmic sequences usually retain their capacity for apical transport. It is unclear, however, whether this modification confers specificity directly by mediating interaction with a specific lectin-like protein or if it allows folding necessary for correct protein-protein interactions during exit from the endoplasmic reticulum.

Because of the importance of EGFR expression during normal epithelial cell development and pathophysiology, we have undertaken this study to characterize determinants that regulate its polar membrane expression. The EGFR is a type I membrane glycoprotein with a 622-amino acid N-glycosylated extracellular ligand binding domain, a 23-amino acid transmembrane domain, and a 541-amino acid cytoplasmic domain (28). In addition to a kinase catalytic core and tyrosine autophosphorylation sites (28), the EGFR cytoplasmic domain has three known endocytic signals (29, 30) and two signals directing transport to lysosomes (31).² EGFR also has an actin binding domain located at residues 984-996 (33), which could anchor EGFR to the basolateral cytoskeleton and help maintain domain-specific localization. We have shown previously that EGFRs are transported directly from the TGN to the basolateral plasma membrane by a high capacity mechanism in MDCK cells (34). We have also shown that amino acid residues Lys-652 to Leu-723 in the cytoplasmic juxtamembrane region support basolateral EGFR steady-state localization (34). Truncation to residue Arg-651, which removes all but 7 amino acids from the cytoplasmic domain, results in predominant apical receptor steady-state localization (34). The purpose of this study was to define more precisely the EGFR basolateral sorting signal and investigate mechanisms by which hierarchical sorting signals determine EGFR membrane distribution. Our results indicate that residues Lys-652 to Ala-674 are necessary and sufficient for targeting EGFRs directly from the TGN to the basolateral plasma membrane and can also direct the extracellular domain of an apical membrane protein, decay accelerating factor (DAF), to the basolateral surface. The molecular requirements of the EGF basolateral sorting signal distinguish it from most basolateral sorting signals by three criteria. First, it does not have a critical tyrosine or di-leucine motif. Second, it is not colinear with any of the known EGFR endocytic signals. Third, computer-based modeling predicts that the juxtamembrane region between residues Lys-652 to Ala-674 has a propensity to form two short  amphipathic helices. Our data do suggest that although carboxyl-terminal sequences are not necessary for basolateral transport, they are important for maintaining polar membrane distribution of EGFR mutant proteins with intact tyrosine kinase catalytic domains. Our data also suggest that in the absence of a positive cytoplasmic basolateral sorting signal, information dependent on N-linked glycosylation of the extracellular ligand binding domain is sufficient to target EGFRs directly from the TGN to the apical plasma membrane.

EXPERIMENTAL PROCEDURES

EGFR cytoplasmic deletion and substitution mutants were made using PCR-based strategies to modify EGFR coding sequences cloned in the eukaryotic expression plasmid pCB6⁺. The pCB6⁺ plasmid has a polylinker region downstream of a cytomegalovirus promoter, transcription termination and polyadenylation signals from the human growth hormone gene, an SV40 origin of replication and early region promoter-enhancer, and ampicillin and neomycin resistance genes (35). With one exception (see next paragraph), forward primers were designed to anneal to sequences 5 to a unique restriction site in the EGFR cDNA, whereas reverse mutagenic primers contained codons for substituted amino acids and/or premature stops 400-500 nt downstream of the forward primer, as well as a restriction site compatible with the pCB6⁺ polylinker. Further details regarding construction of plasmids with stop codon substitutions at codons for Pro-675, Ala-698, Tyr-974, Leu-993, and Val-1023 will be published separately.² An E663STOP substitution was made using a forward primer 5-TGCGTCTCTTGCCGGAATGTCA-3, which annealed to sequences 5 to a BsmI site at nt 1792 in the EGFR coding region; and a reverse mutagenic primer 5-CTTGCCGATATCATCACTCCCTCTCCTG-3, which incorporates a stop codon in place of the codon for Glu-663 (underlined) and an EcoRV site (in bold) compatible with the pCB6⁺ polylinker. A L664A,V665A substitution in a P675STOP background was made using the same forward primer; and a reverse mutagenic primer 5-TCTCTGATATCATCAAGCTTCTCCGCTGGGTGTAAGTGGCTCCGCTGCCTCCCTCTCCTG-3, incorporating alanine (double underlined) and P675STOP (underlined) substitutions and an EcoRV site (bold). PCR products were gel-purified, digested at sites incorporated at the ends of PCR products, and ligated directly to pCB6⁺/EGFR digested with the same restriction enzymes. The cDNAs with cytoplasmic truncations were named based on the carboxyl-terminal amino acid residue in the EGFR coding region (i.e. c-674 has a P675STOP substitution).

A L658A,L659A substitution in a P675STOP background was made using the "megaprimer" PCR method (36). Briefly, an initial PCR reaction was performed using a forward mutagenic primer 5-GCACGCTGCGGAGGGCTGCACAGGAGAGGGAG-3 to convert codons for Leu-679 and Leu-680 to alanine residues (double underlined) and a reverse primer 5-TCAATCGATCTAAGCTTCTCCACTGGGTGTAA-3, which annealed to the juxtamembrane sequence ending with P675STOP (underlined) and a ClaI site (bold). A subsequent PCR reaction was carried out using the product of the first reaction and the forward primer which anneals to sequences 5 to a BsmI site at nt 1792 in the EGFR coding region described in the previous paragraph. The resulting PCR product was reamplified using forward and reverse primers from the second and first PCR reactions, respectively. This product was gel-purified, digested with BsmI and ClaI, and ligated directly to pCB6⁺/c-674 digested with the same restriction enzymes.

Chimeric molecules that fuse the extracellular domain of DAF to the transmembrane and cytoplasmic domains of EGFRs truncated at residues 651 and 674 were made using similar PCR-based strategies. The human DAF cDNA (37) was excised from pBluescript II KS and subcloned into pCB6⁺. Forward primers were designed to anneal to a unique restriction site in the DAF cDNA, whereas reverse mutagenic primers contained codons that anneal to sequences 5 to the GPI anchor signal, replacing it with an XhoI restriction site. The XhoI site was used to produce an in-frame fusion with the extracellular juxtamembrane domain of the EGFR at residue Leu-609. The forward primer for both EGFR segments contained an XhoI site 5 to the transmembrane domain, whereas the reverse primers contained premature stops 129 or 198 nt downstream of the forward primer, as well as a restriction site compatible with the pCB6⁺ polylinker. The DAF extracellular domain cDNA (DAFex) was made using a forward primer 5-GTAACGTATGCATGTAATAAAGGAT-3, which annealed to sequences surrounding an NsiI site at nt 756 in the DAF coding region, and a reverse mutagenic primer 5-GATGTCTCGAGGGTTGTCTCATGAAAATGCTTGG-3, which incorporates an XhoI site (bold) immediately after the codon for Thr-344 (underlined) compatible with the pCB6⁺ polylinker. The DAF PCR product was gel-purified, digested at sites incorporated at the ends of the PCR product, and ligated directly to pCB6⁺/DAF digested with the same restriction enzymes, producing pCB6⁺/DAFex. The EGFR transmembrane-cytoplasmic segment was made using a forward primer 5-GTATCTCGAGGGCTGTCCAAC-3 which annealed to sequences 5 to the transmembrane domain, incorporating an XhoI site (bold) without altering the amino acid sequence (Leu-609, Glu-610 underlined). The reverse primer for the EGFR segment of DAF-651, 5-CTCTCTAGACTGCAGCAGCCTCC-3, annealed to a sequence 3 to a premature stop codon for Arg-652 incorporating an XbaI site (bold) compatible with the pCB6⁺/DAFex polylinker. The reverse primer for the EGFR segment of DAF-674, 5-GTTCTCGATATCCTAAGCTTCTCCACTG-3, annealed to the sequence surrounding the codon for Pro-675, incorporating a stop codon at Pro-675 (underlined) and an EcoRV site (bold) also compatible with the pCB6⁺/DAFex polylinker. Both EGFR PCR products were gel-purified, digested at sites incorporated at their ends, and ligated directly to pCB6⁺/DAFex digested with the same restriction enzymes. The chimeric cDNAs were named based on both the DAF extracellular domain and the carboxyl-terminal residue in the EGFR coding region (i.e. DAF-651 and DAF-674).

PCR primers were designed using the DNASTAR software package (DNASTAR, Inc., Madison, WI). The same software package was used for protein structural analysis. PCR amplifications were carried out using a RoboCycler 40 Temperature Cycler (Stratagene Cloning Systems; La Jolla, CA). All plasmid constructs were checked for correct orientation by restriction digestion analysis, and all PCR-amplified DNA sequences were verified by dideoxy chain termination DNA sequencing using a Sequenase II kit from U. S. Biochemical Corp.

MDCK II cells are canine kidney epithelial cells that exhibit many physical and functional properties characteristic of transporting epithelia (38). MDCK II cells were transfected with plasmids containing human EGFR, DAF, or DAF/EGFR cDNAs using Lipofectin reagent (Life Technologies, Inc.) as described previously (34). Transfected cells were grown 10-14 days in media containing G418 (0.8 mg/ml Geneticin; Life Technologies, Inc.). Thereafter, cells were stained with the EGF-R1 or II H6 monoclonal antibodies (mAb) followed by FITC-conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) every 7-10 days to enrich for cells expressing human EGFRs by sterile sorting on a flow cytometer (Cytofluorograph IIs; Ortho Instruments, Westwood, MA). EGF-R1 recognizes a peptide core epitope in the extracellular domain of the human EGFR (39, 40) but does not cross-react with the canine receptor (34). II H6 recognizes an epitope in the extracellular domain of the human DAF molecule (41). In some cases, independent cell lines were propagated from single cells isolated from enriched cell populations by automated sterile sorting.

All MDCK cell lines were grown in minimal essential medium supplemented with 10% FBS, 0.1 mM nonessential amino acids (Life Technologies, Inc.), and 2 mM glutamine. Monolayers of polarized epithelial cells with well formed tight junctions were produced as described previously (34). Briefly, cells were seeded on Transwell polycarbonate filter inserts (0.4-µm pore size; Costar Corp., Cambridge, MA) at a density of 5 × 10⁵ cells per 12-mm filter or 5 × 10⁶ cells per 75-mm filter, refed 24 h later, and every 2 days thereafter. Experiments were conducted 6 days after seeding.

Receptor grade mouse EGF (Toyobo Biochemicals, Osaka, Japan) was labeled with ¹²⁵I (carrier-free, >350 mCi/ml; NEN Life Science Products) using the chloramine-T method (approximate specific activity = 2 × 10⁸ cpm/mg of protein). Filter-grown cells were incubated with 8 × 10⁶ cpm ¹²⁵I-EGF per ml of serum-free minimal essential medium supplemented with 0.1% (w/v) bovine serum albumin added to either the apical or basolateral surface for 2 h at 4 °C. Cells were rinsed twice with PBS and then incubated with 2 mM disuccinimidyl suberate (Pierce) in 0.1 M Hepes, pH 7.4, supplemented with 0.12 M NaCl, 0.05 M KCl, 8 mM glucose, and 1.2 mM MgSO₄, for 15 min at room temperature. DSS was quenched by a 5-min incubation with 0.05 M Tris, pH 7.4, at room temperature. Cell lysates were prepared using 1% (w/v) Nonidet P-40 in 0.1 mM Tris, pH 6.8, supplemented with 15% (w/v) glycerol, 2 mM EDTA, 1 mM EGTA, 0.2 mM phenylmethylsulfonyl fluoride, and 1 µM leupeptin. Equal aliquots were separated by SDS-PAGE (42). Radioactively labeled proteins were quantitated by phosphorstorage autoradiography (43). Digitized images were analyzed using the ImageQuant^TM software package (Molecular Dynamics, Sunnyvale, CA), which averages five measurements of light emission for each pixel location, to give a pixel value proportional to the amount of stored radiation.

Filter-grown cells were rinsed 3 times with PBS, fixed for 10 min in 3.7% paraformaldehyde, and then rinsed 3 times with PBS containing 10% FBS. Some cells were incubated with EGF-R1 (1:100) or II H6 (5 µg/ml) mAbs for 1 h at 37 °C without permeabilization. Other cells were permeabilized for 10 min with 0.2% Triton X-100, and rinsed 3 times with PBS containing 10% FBS. Permeabilized cells were then incubated with rat mAb R26.4C to the tight junction protein ZO-1 (culture supernatant, neat), or mouse mAb to -catenin (1:100) for 1 h at 37 °C. R26.4C (44, 45) was obtained from the Developmental Studies Hybridoma Bank maintained by the Department of Pharmacology and Molecular Sciences, The Johns Hopkins University School of Medicine, Baltimore, MD, and the Department of Biological Sciences, University of Iowa, Iowa City, IA, under contract NO1-HD-6-295. -Catenin mAb was obtained from Transduction Laboratories (Lexington, KY). After staining with primary antibodies, cells were rinsed 3 times with PBS containing 10% FBS, and nonspecific sites were blocked by incubation with 5% normal serum from the host animal used to raise the secondary antibody for 30 min at room temperature. Cells were incubated with appropriate FITC-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) diluted 1:100 in 5% normal serum for 30 min at 37 °C and then rinsed twice with PBS containing 10% FBS and once with PBS. Cells attached to membranes were excised from plastic inserts and mounted cell-side up on a glass slide. The membrane was covered with a coverslip and edges were sealed to prevent drying. Specimens were examined with a Zeiss LSM 410 scanning laser confocal microscope using the 488/568 nm wavelength lines of an argon-krypton laser. The cell monolayer was optically sectioned every 0.5 µm. Image resolution using a Zeiss 100x Neofluor objective and Zeiss LSM software was 512 × 512 pixels.

Filter-grown cells were rinsed twice and preincubated in cysteine-, methionine-free medium for 1 h, and then labeled with [³⁵S]Express Protein Labeling Mix (100 µCi/ml; >600 Ci/mmol; NEN Life Science Products) in cysteine-, methionine-free medium supplemented with 0.2% bovine serum albumin and 10% dialyzed FBS added to the basal chamber of the filter. Labeling medium was replaced with complete minimal essential medium supplemented with nonradioactive cysteine (1.2 µg/ml) and methionine (0.75 µg/ml), and cells were incubated for periods indicated in the figure legends. Cells to be analyzed by immunoprecipitation only were lysed with 1% (w/v) Nonidet P-40 exactly as described above. Cells to be further analyzed by domain-specific biotinylation were rinsed twice with PBS supplemented with 0.1 mM CaCl₂ and 1 mM MgCl₂ and then incubated with freshly prepared sulfosuccinimidobiotin (s-NHS-biotin; Pierce) diluted in the same solution (1 mg/ml) for 30 min on ice. The reaction was quenched with 50 mM NH₄Cl, and cells were lysed with a solution of 1% (w/v) Triton X-100 in 20 mM Tris, pH 8.0, 50 mM NaCl, 5 mM EDTA, and 0.2% (w/v) bovine serum albumin supplemented with protease inhibitors. EGFRs were isolated by immunoprecipitation with EGF-R1 mAb adsorbed to protein A-Sepharose CL-4B beads (Sigma). Immunoprecipitates were either solubilized immediately in Laemmli buffer or boiled in 100 µl of 10% SDS for isolation of biotinylated proteins. The SDS-protein solution was diluted with 1 ml of lysis buffer and then incubated with streptavidin-agarose (Pierce) for 16 h at 4 °C to bind biotinylated proteins. Proteins bound to the agarose slurry were solubilized with Laemmli buffer. Proteins were separated by SDS-PAGE, and gels were treated with En³Hance (NEN Life Science Products) for fluorography. Radioactivity was quantitated by phosphorstorage autoradiography. For experiments with tunicamycin (Boehringer Mannheim), cells were incubated in media containing the drug (5 µg/ml) for 1 h prior to metabolic labeling with [³⁵S]Express Protein Labeling Mix. Tunicamycin was also included at the same concentration during the labeling and chase periods, and vehicle (Me₂SO) was added to control cells.

RESULTS

We have shown previously that truncating the EGFR to residue Arg-651, which removes all but 7 amino acids from the cytoplasmic domain, results in predominant apical receptor steady-state localization using the MDCK epithelial cell tissue culture model (34) (Fig. 1A). Since receptors truncated to residue Arg-723 in the proximal region of the kinase catalytic core exhibited basolateral steady-state localization similar to wild-type EGFR (Fig. 1A), we hypothesized that the 72-amino acid sequence between residues Lys-652 and Arg-723 contained a basolateral sorting signal (34). To more precisely map the basolateral sorting signal, we first asked what effect progressive truncations within this 72-amino acid region had on EGFR membrane domain steady-state localization (Fig. 1B). Steady-state plasma membrane domain expression of human EGFRs bearing truncations in this region was determined by quantitation of radiolabeled proteins after domain-specific ¹²⁵I-EGF cross-linking and SDS-PAGE using phosphorstorage autoradiography. Two independently derived cell lines expressing each of the receptor mutants were examined. These analyses showed that truncations to residue Gly-697, Leu-683, or Ala-674 resulted in an apical:basolateral membrane steady-state distribution of approximately 1:9 (Fig. 1B), similar to what is observed for wild-type human EGFRs (i.e. c-1186) in transfected cell lines (Fig. 1A) and endogenous canine EGFRs in untransfected MDCK cells (34).

Steady-state plasma membrane domain expression of cytoplasmically truncated human EGFRs. A, Upper, diagram showing transmembrane (black), juxtamembrane (residues 651-688), kinase (residues 688-957), and CaIn (residues 973-1022) domains of the full-length human EGFR (c-1186) and cytoplasmically truncated receptors (c-723 and c-651) from Ref. 34. The extracellular ligand binding domain (not shown) is identical for all receptors. Lower, autoradiogram showing domain-specific ¹²⁵I-EGF cross-linking for two independently derived cell lines expressing each of the receptors in the upper panel. B, diagram of EGFRs with truncations in the juxtamembrane domain (upper), and autoradiogram showing domain-specific ¹²⁵I-EGF cross-linking for two independently derived cell lines expressing each of these receptors (lower). C, diagram of EGFRs with truncations in the carboxyl-terminal region (upper), and autoradiogram showing domain-specific ¹²⁵I-EGF cross-linking for two independently derived cell lines expressing each of these receptors (lower). Radioactive bands in all panels were quantitated by phosphorstorage autoradiography, and numbers below each lane indicate percentage of total radioactivity present at the apical (Ap) or basolateral (BL) plasma membrane. Molecular weight standards: myosin, 200,000; -galactosidase, 116,250; phosphorylase B, 97,400; bovine serum albumin, 66,200.

[View Larger Version of this Image (21K GIF file)]

Steady-state distributions of receptors with progressive truncations between residues Leu-723 and Arg-651 were also examined using CLSM. Filter-grown cells were labeled in a domain-specific manner with EGF-R1 mAb followed by an FITC-conjugated secondary reagent (see "Experimental Procedures"). EGF-R1 is specific for a peptide epitope in the extracellular domain of human EGFRs (39, 40) but does not cross-react with canine EGFRs (34). These analyses clearly showed that the c-651 receptor is located almost entirely on the apical surface, in contrast to receptors truncated to residues Gly-697, Leu-683, or Ala-674, which are present at the basolateral surface similar to wild-type c-1186 EGFRs (Fig. 2). Although there is some basal staining, these receptors appear to be mostly concentrated along the entire aspect of the lateral membrane. The tight junction protein ZO-1 as well as the E-cadherin-associated -catenin molecule are both correctly distributed in all of these cell lines (representative images shown in Fig. 3), indicating that overexpression of human receptors does not adversely affect the polar phenotype. Taken together with domain-specific ¹²⁵I-EGF cross-linking results, CSLM results indicate that residues Lys-652 to Ala-674 are necessary and sufficient for steady-state basolateral EGFR localization.

[View Larger Version of this Image (72K GIF file)]

CLSM analysis of ZO-1 and -catenin localization in cell lines expressing human EGFRs.

[View Larger Version of this Image (72K GIF file)]

To determine the surface delivery of receptors with progressive truncations between residues Arg-723 and Arg-651, filter-grown cells were pulse-labeled for 15 min, changed to chase medium containing nonradioactive amino acids, and then subjected to domain-specific biotinylation at 30-min intervals during the chase period. Newly synthesized receptors arriving at the cell surface could then be isolated by sequential EGFR immunoprecipitation followed by streptavidin affinity purification. Similar to wild-type c-1186 receptors, the bulk of EGFRs truncated to residue Gly-697 or Ala-674 was delivered directly to the basolateral surface within 1 to 1.5 h of synthesis (Fig. 4). This was in contrast to c-651 receptors, which were delivered directly to the apical plasma membrane (Fig. 4). The kinetics of apical delivery appeared to be slightly delayed compared with basolateral delivery, with c-651 receptors reaching the cell surface within 1.5 to 2 h of synthesis. These data indicate that residues Lys-652 to Ala-674 are necessary and sufficient for targeting EGFRs directly from the TGN to the basolateral plasma membrane. These data also suggest that in the absence of a cytoplasmic basolateral sorting signal, information in the luminal and/or transmembrane domain is sufficient to target EGFRs directly from the TGN to the apical plasma membrane.

[View Larger Version of this Image (36K GIF file)]

The results described above support our hypothesis that residues Lys-652 to Ala-674 contain a determinant necessary for basolateral delivery of the EGFR. One important test of a putative basolateral sorting determinant is demonstration that the signal is autonomous, by transferring it to a molecule that is normally found at the apical plasma membrane. To determine whether the EGFR basolateral sorting signal can function outside the context of the EGFR, chimeric molecules were constructed that fuse the extracellular domain of human decay accelerating factor (DAF) to the transmembrane and cytoplasmic domains of EGFR truncation mutants (Fig. 5A). The wild-type DAF molecule is a 70-kDa protein with a 28-amino acid cytoplasmic tail that is cleaved and replaced with a GPI anchor (Fig. 5A). MDCK cells were stably transfected with cDNAs coding for wild-type DAF and both DAF-EGFR chimeric constructs. We found that DAF-EGFR chimeric proteins but not wild-type DAF were extractable with cold Triton X-100 (not shown), in keeping with the Triton insolubility of proteins with GPI anchors (25). DAF membrane polarity was assessed by CSLM after staining filter-grown cells with a DAF-specific mAb added to the apical or basolateral side. As shown by other investigators (24), wild-type DAF was found predominantly at the apical surface (Fig. 5B). When the DAF extracellular domain was fused to the transmembrane and cytoplasmic domain of EGFRs truncated to residue Arg-651, the DAF-651 chimeric protein exhibited nonpolar membrane expression (Fig. 5B). In contrast, when DAF was fused to the transmembrane and cytoplasmic domain of EGFRs truncated to residue Ala-674, the DAF-674 chimeric protein was localized predominantly at the basolateral membrane (Fig. 5B). These results indicate that the sequence between Lys-652 and Ala-674 of the EGFR juxtamembrane domain behaves as an autonomous, dominant basolateral sorting signal.

[View Larger Version of this Image (78K GIF file)]

Examination of the 23 amino acid sequence between residues Lys-652 and Ala-674 revealed the presence of two di-leucine motifs, located at residues Leu-658, Leu-659, and Leu-665, Val-666 (Fig. 6A). Since di-leucine motifs confer specificity for basolateral sorting of at least one molecule, namely macrophage IgG Fc receptors expressed in MDCK cells (46), we hypothesized that Leu-658, Leu-659, or Leu-664, Val-665 might play a critical role in EGFR basolateral sorting. To test that hypothesis, each of the di-leucine motifs was changed to a di-alanine by site-directed mutagenesis. Both di-alanine substitutions were made within the context of a receptor truncated to residue Ala-674 (i.e. c-674). Cell populations enriched for c-674 receptors with a L658A,L659A substitution were judged to have normal basolateral localization, based on quantitation of domain-specific plasma membrane ¹²⁵I-EGF cross-linking, and CSLM analysis of filter-grown cells stained with EGFR-specific antibody added to the apical or basolateral surface (Fig. 6, B and C, left panels). Similar results were obtained with cell populations enriched for c-674 receptors with a L664A,V665A substitution (Fig. 6, B and C, middle panels).

[View Larger Version of this Image (43K GIF file)]

Since the basolateral signal mapped to this region also lacks a tyrosine residue, these data suggest that the EGFR juxtamembrane basolateral sorting signal does not conform to molecular requirements identified in most proteins thus far examined. To further characterize this region, we therefore determined its predicted secondary structure using computer-based modeling. Our analysis showed that this region has a propensity to form two short  amphipathic helices (see Fig. 10 under "Discussion"). We found that truncating the receptor at residue Glu-663, which removes the predicted amphipathic helix distal to the transmembrane domain including the Leu-664, Val-665 motif, resulted in predominant apical localization at steady state (Fig. 6, B and C, right panels). This result suggests two possibilities that are not mutually exclusive. Residues Leu-664 to Ala-674 are critical to proper function of the EGFR basolateral sorting signal. Alternatively, truncation to residue Glu-663 disrupts critical secondary structure and therefore weakens the EGFR basolateral sorting signal. These results also argue against Leu-658, Leu-659, and Leu-664, Val-665 representing redundant sorting signals, since expression of Leu-658, Leu-659 alone is not sufficient to compensate for absence of Leu-664, Val-665 in receptors truncated to residue Glu-663.

Three independent endocytic codes have been identified in the EGFR carboxyl terminus (47): 973-FYRAL and 996-QQGFF located in the CaIn domain, so-called because it mediates the EGF-induced calcium response and internalization (30); and residues Val-1023 to Ala-1186 at the carboxyl terminus which contain 3 consensus NPXY motifs. Since basolateral and endocytic signals are colinear and share a common critical tyrosine residue in a number of proteins, we therefore sought to determine whether the carboxyl terminus of the EGFR contained additional basolateral sorting information. Analysis of this region also permits evaluation of the potential role of the actin-binding domain located at residues Asp-984 to Gln-996 (33) in maintaining polar EGFR expression. Domain-specific ¹²⁵I-EGF cross-linking of cells expressing receptors truncated to residue Thr-1022 at the distal border of the CaIn domain resulted in normal basolateral localization (Fig. 1C). However, when the receptor was truncated to residue Tyr-992, which deletes the 996-QQGFF endocytic signal and carboxyl-terminal NPXY motifs, or Phe-973, which deletes all known endocytic codes, there was a significant increase in the percentage of receptors found at the apical surface. Steady-state apical localization of c-992 and c-973 receptors ranged from 20 to 37% of cell surface receptors (Fig. 1C). CLSM analysis using EGFR antibodies showed similar results for domain-specific localization of receptors truncated to residues Thr-1022, Tyr-992, and Phe-973 (Fig. 7). Cells expressing c-992 and c-973 receptors also exhibited unusual morphology, having lost the uniform cobblestone appearance typical of MDCK II cells (Fig. 7). Despite aberrant EGFR membrane localization and unusual morphology, however, cells with c-992 (Fig. 3) or c-973 (not shown) receptors had normal distributions of the tight junction protein ZO-1, as well as the E-cadherin-associated -catenin molecule, indicating that other proteins sort normally in these cells.

[View Larger Version of this Image (91K GIF file)]

Results showing unusually high apical localization of c-973 and c-992 receptors might indicate that residues Glu-724 to Ser-971 mask the juxtamembrane basolateral determinant and that additional sequences between residues Leu-993 and Thr-1022 are necessary for proper conformation and basolateral transport of the full-length receptor. Alternatively, c-973 and c-992 receptors might be targeted to basolateral membranes but fail to maintain a polar distribution. To distinguish these possibilities, we examined the plasma membrane delivery of c-973 and c-992 receptors. Similar to EGFRs truncated to residue Thr-1022, c-973 and c-992 receptors were both delivered directly from the TGN to the basolateral surface (Fig. 8). All three receptors with truncations in the carboxyl terminus were delivered with essentially the same kinetics as receptors with truncations in the juxtamembrane domain or the wild-type molecule (Fig. 4). These results indicate that receptors lacking CaIn domain sequences undergo correct targeting from the TGN to the basolateral surface but that a significant percentage of these receptors undergo transport to the apical surface via a transcytotic mechanism. This is consistent with the idea that the juxtamembrane sequence described in this study is the major determinant of basolateral sorting for the wild-type molecule.

[View Larger Version of this Image (28K GIF file)]

Since the extracellular domain of the EGFR contains 7-9 N-linked oligosaccharides (40, 48), we sought to determine whether this modification was necessary for apical localization of receptors lacking a cytoplasmic basolateral sorting signal (i.e. c-651 in Fig. 1A). The large number of N-glycosylation sites made mutational removal of all of the sites impractical. We therefore treated cells with tunicamycin to prevent N-linked glycosylation, to determine the role of this modification on apical sorting. However, since tunicamycin also prevents correct folding causing proteins to accumulate in the endoplasmic reticulum (49), care was taken to examine only those receptor molecules that reached the cell surface, as judged by their accessibility to membrane-impermeant biotin. Filter-grown cells were pretreated with tunicamycin for 1 h, metabolically labeled for 3 h, and changed to chase medium for 3 h, after which they were subjected to domain-specific biotinylation. Newly synthesized EGFRs delivered to plasma membrane domains were then detected by sequential EGFR immunoprecipitation and streptavidin affinity purification, followed by SDS-PAGE and fluorography. In all cases, the form of the receptor detected in cells receiving tunicamycin had a reduced molecular weight compared with receptors in cells receiving vehicle, consistent with lack of N-linked oligosaccharides (see arrowheads in Fig. 9). Nonglycosylated forms of c-1186 receptors and receptors truncated to residue Ala-674 exhibited normal transport to the basolateral plasma membrane. In contrast, the nonglycosylated form of the c-651 receptor was evenly distributed between the apical and basolateral surfaces (Fig. 9), indicating loss or weakening of signals necessary for apical targeting of its fully glycosylated counterpart. We conclude from these results that N-linked glycosylation of the extracellular domain plays a critical role in apical targeting of EGFRs lacking a positive cytoplasmic basolateral sorting determinant.

[View Larger Version of this Image (20K GIF file)]

DISCUSSION

The goal of this study was to determine the molecular requirements for EGFR membrane expression in polarized epithelial cells. This was achieved by examining the sorting behavior of a series of EGFR mutant proteins with progressive cytoplasmic truncations using the MDCK cell model. These analyses have revealed that EGFRs possess a hierarchy of sorting information, which is summarized in Fig. 10. Examination of the sorting behavior of receptors with progressive truncations in the juxtamembrane domain identified a positive basolateral sorting signal located 7-30 amino acids from transmembrane domain. When the transmembrane domain and basolateral sorting signal of EGFR were fused to the extracellular domain of the apical membrane protein DAF, the chimeric molecule was redirected to the basolateral surface. These results indicate that the sequence between Lys-652 and Ala-674 of the EGFR juxtamembrane domain behaves as an autonomous, dominant basolateral sorting signal.

[View Larger Version of this Image (27K GIF file)]

In contrast to many other well characterized basolateral sorting signals, the EGFR basolateral signal does not contain any tyrosine residues. In addition, specificity is not critically dependent on either of the di-leucine motifs located in this region. Since the EGFR basolateral signal lacks amino acid motifs common to most sorting signals, we turned to computer-assisted modeling to gain insight to other residues that might confer specificity. Modeling of this region according to the method of Eisenberg (50) predicts that it has a propensity to form two short  amphipathic helices (Fig. 10, B and C). The helix proximal to the transmembrane domain consists of hydrophobic residues Leu-655, Leu-658, and Leu-659 aligned on one side opposite to charged residues Arg-653, Arg-656, and Arg-657, and a second helix distal to the transmembrane domain has hydrophobic residues Leu-664, Val-665, and Leu-668 opposite to charged residue Glu-666 (Fig. 10C). Charged residues Glu-661, Arg-662, and Glu-664 could be aligned on the same side as the other charged residues, or alternatively could break the amphipathic nature of a more extended structure by looping out from the hydrophobic face. Deletion of the distal amphipathic helix caused truncated receptors to be delivered predominantly to the apical surface. The simplest interpretation of this result is that specificity of the EGFR basolateral activity is critically dependent on residues Leu-664 to Ala-674. However, given the predicted secondary structure of this region, truncating receptors to residue Glu-663 could also disrupt hydrophobic interactions with other proteins or nearby membrane phospholipids important for stabilizing this region. The finding that di-alanine substitutions for Leu-658, Leu-559 and Leu-664, Val-665 do not affect basolateral sorting is consistent with this idea. Our current model postulates that specificity of the EGFR basolateral signal is conferred by electrostatic interactions mediated by charged residues in this region whose conformation is critically dependent on secondary structure.

In addition to lacking common amino acid motifs, the EGFR basolateral sorting signal also differs from most basolateral signals because it does not overlap any of the three known EGFR endocytic signals located between residues Phe-973 and the carboxyl terminus (Fig. 10A). Although receptors truncated to residues Phe-973 or Tyr-992 are transported directly to the basolateral surface, our results show that polar distribution of these mutant proteins is not maintained. These truncations may expose a signal that affects polar sorting in endosomes, promoting transcytosis at the expense of recycling. We cannot exclude the possibility that EGFR endocytic signals have the capacity to direct basolateral transport, particularly if placed near the transmembrane domain or in chimeric molecules. However, our data argue that they do not behave as basolateral determinants at their normal distance from the plasma membrane.

EGFRs truncated to residue Arg-651 are transported directly to the apical plasma membrane. Inhibiting N-linked glycosylation with tunicamycin caused random apical and basolateral delivery of c-651 receptors, while having no effect on basolateral delivery of wild-type EGFRs or receptors containing a minimal cytoplasmic basolateral sorting signal (i.e. c-674). These findings suggest that in the absence of a positive cytoplasmic basolateral determinant, information dependent on N-linked glycosylation of the extracellular ligand binding domain is sufficient to target EGFRs directly from the TGN to the apical plasma membrane (Fig. 10A). Results from the DAF-EGFR chimeras also indicate that the apical sorting determinant of the EGFR is likely contained within the extracellular domain, since attaching the EGFR transmembrane domain to the extracellular domain of DAF did not provide positive apical sorting information. Removal of basolateral signals usually results in efficient apical targeting of mutant proteins (18), suggesting that most molecules have hierarchical basolateral and apical signals. The functional consequences of opposing sorting signals in the same molecule remain to be established. However, in the case of the EGFR, it may have important implications during development of certain renal tubule segments or in PKD, where the EGFR is known to have apical epithelial cell localization (14, 16). This is consistent with the idea that EGFR apical sorting determinants, while normally recessive, can be made physiologically active under some circumstances.

To summarize, we have identified a positive basolateral sorting signal in the EGFR juxtamembrane region that is necessary and sufficient to target this molecule directly from the TGN to the basolateral plasma membrane. This signal must be recognized by a sorting mechanism having high capacity, since substantially overexpressed human receptors are transported to the basolateral membrane with the same efficiency as endogenous canine receptors (34). In the absence of this cytoplasmic basolateral sorting signal, EGFRs gain entry to vesicles targeted for apical delivery by a mechanism dependent on N-glycosylation. In addition, the sequence between residues Lys-652 and Ala-674 acts as an autonomous basolateral signal, redirecting the extracellular domain of the apical membrane protein DAF to the basolateral membrane. The finding that the EGFR basolateral signal does not have the same molecular requirements as many other membrane proteins may help explain why the EGFR is the only membrane protein consistently reported to be mislocalized in PKD, if gene defects associated with PKD somehow arrest differentiation of an EGFR-specific basolateral sorting pathway. Mutations that weaken or inactivate the EGFR basolateral sorting signal may also play a role in oncogenic progression of epithelial tumors having partially disturbed membrane polarity (32), if mislocalized EGFRs have access to inappropriate substrates.