The neural cell adhesion molecule expresses a tyrosine-independent basolateral sorting signal.

Transmembrane isoforms of the neural cell adhesion molecule, N-CAM (N-CAM-140 and N-CAM-180), are vectorially targeted from the trans-Golgi network to the basolateral domain upon expression in transfected Madin-Darby canine kidney cells (Powell, S. K., Cunningham, B. A., Edelman, G. M., and Rodriguez-Boulan, E. (1991) Nature 353, 76-77). To localize basolateral targeting information, mutant forms of N-CAM-140 were constructed and their surface distribution analyzed in Madin-Darby canine kidney cells. N-CAM-140 deleted of its cytoplasmic domain shows a non-polar steady state distribution, resulting from delivery from the trans-Golgi network to both the apical and basolateral surfaces. This result suggests that entrance into the basolateral pathway may occur without cytoplasmic signals, implying that apical targeting from the trans-Golgi network is not a default mechanism but, rather, requires positive sorting information. Subsequent construction and analysis of a nested set of C-terminal deletion mutants identified a region of 40 amino acids (amino acids 749-788) lacking tyrosine residues required for basolateral targeting. Addition of these 40 amino acids is sufficient to restore basolateral targeting to both the non-polar cytoplasmic deletion mutant of N-CAM as well as to the apically expressed cytoplasmic deletion mutant of the p75 low affinity neurotrophin receptor (p75NTR), indicating that this tyrosine-free sequence is capable of functioning independently as a basolateral sorting signal. Deletion of both cytoplasmic and transmembrane domains resulted in apical secretion of N-CAM, demonstrating that the ectodomain of this molecule carries recessive apical sorting information.

The neural cell adhesion molecule, N-CAM, 1 a member of the Ig superfamily, was the first molecule shown to mediate cellcell adhesion (2). N-CAM has since been found to be developmentally expressed as a family of isoforms in a variety of both neural and non-neural tissues including skin (3), muscle (4), and liver (5). N-CAM mediates cellular adhesion through both homophilic (6) and heterophilic contacts (7,8) and is capable of activating intracellular secondary messenger systems (9,10). These interactions are fundamental to the establishment of form and pattern in diverse developmental pathways including that of the central nervous system, heart, limb, and kidney. Within the central nervous system, morphogenetic events involving N-CAM include neurite outgrowth (11), muscle innervation (12), and axonal regeneration (13). N-CAM is also predicted to play a role in synaptic plasticity by directing morphological changes in synaptic structures that are associated with learning and memory (14).
Although encoded by a single gene, alternative splicing generates a large number of N-CAM isoforms. The three predominant neural variants, N-CAM-180, N-CAM-140, and N-CAM-120 share a common ectodomain, but whereas N-CAM-140 and N-CAM-180 are transmembrane proteins, N-CAM-120 is anchored to the surface via a glycosylphosphatidylinositol anchor. The polarized surface distribution of N-CAM isoforms in different cell types suggests that spatial segregation within the plasma membrane may play a role in regulating adhesive interactions. In the retinal pigment epithelium, N-CAM-140 is segregated to the apical surface in vivo (15). N-CAM is also localized to the motor end plate of the myofibril of striated muscle (16). Within the nervous system, N-CAM-180 is found polarized to growth cones (17), to sites of cell-cell contact (18), and to postsynaptic membranes (19). Additionally, N-CAM-180 is polarized to the somatodendritic domain and excluded from axons of mouse hippocampal neurons (20). Despite these observations, the mechanisms by which the polarity of N-CAM is achieved are not well characterized.
Intracellular sorting and vectorial delivery provide an important means by which plasma membrane asymetry is generated. Different cell types including kidney (21), retinal pigment epithelium (22), and intestinal (23) cell lines, as well as presumptive myocytes (24) and hippocampal neurons (25) are all able to sort proteins intracellularly as demonstrated by the polarized surface delivery of viral proteins. Much of what is known about the mechanisms of protein sorting and targeting has been obtained from in vitro studies of epithelial cells. For example, in MDCK cells, most membrane proteins follow a direct route from the trans-Golgi network (TGN) to their resident membranes (21). The TGN discriminates between apical and basolateral proteins and sorts them into specific vesicles which are then targeted and delivered to the appropriate surface (21). This intracellular sorting of membrane proteins in MDCK cells is in many cases signal-mediated, and dependent on both lipid and proteinaceous signals. Whereas glycosylphosphatidylinositol linkage serves to target proteins apically, basolateral localization is dependent on discrete cytoplasmic amino acid sequences (reviewed in Ref. 26). Cytoplasmic determinants capable of targeting proteins from the TGN to the basolateral surface often overlap with tyrosine-based determinants required for clathrin-mediated endocytosis in several basolateral proteins (as found in the low density lipoprotein receptor (27), lysosomal glycoprotein 120 (28), lysosomal acid phosphatase (29), and the Fc receptor (28)). Detailed point mutational analysis within these overlapping signals (29,30) has shown that these two sorting activities can be separated, suggesting that basolateral targeting utilizes secondary structural motifs related to but not identical to the tyrosine-based endocytic signal which has been characterized to form a type 1 ␤-turn (31). Basolateral signals such as in the transferrin receptor (32) and polymeric Ig receptor (33) are not associated with endocytic activity. However, structural analysis of the polymeric Ig receptor basolateral signal has shown that it consists of a ␤-turn followed by a nascent helix, which, when compromised by point mutations, leads to decreases in sorting fidelity (34) further supporting the involvement of ␤-turns in the basolateral sorting of membrane proteins.
Previous work from our laboratory has demonstrated that differential isoform expression can dictate the surface membrane distribution of N-CAM. When expressed in MDCK cells, N-CAM-120 is targeted to the apical surface, whereas the transmembrane isoforms (N-CAM-140 and N-CAM-180) are sorted basolaterally (1). In this work, we extend these findings and characterize a cytoplasmic peptide directing the intracellular sorting of transmembrane N-CAM. The surface distribution of mutant constructs of N-CAM-140 expressed in MDCK cells reveals 40 amino acids in the cytoplasmic domain required for basolateral targeting. These 40 amino acids function independently as a basolateral sorting signal as indicated by its ability to re-direct both a nonpolar mutant of N-CAM and a heterologous apical protein to the basolateral surface. The function of this tyrosine-independent signal requires regions predicted to form ␤-turn structures. However, point mutational analysis within these putative turn sequences did not identify specific residues critical for basolateral targeting. Finally, the apical secretion of soluble forms of N-CAM demonstrates the presence of apical sorting information in the N-CAM ectodomain.
Construction of N-CAM Mutants-Chicken N-CAM-180 cDNA (pEC1401) (36) was obtained from G. Edelman and B. Cunningham (Scripps Research Institute, La Jolla, CA). The cytoplasmic deletion mutant, 712t, was generated using an AatII site 2 amino acids from the transmembrane domain of N-CAM. After AatII digestion, the fragment encoding the N-CAM truncation was blunt ended and ligated to XbaI linkers containing a stop codon. An XbaI site 5Ј to the start codon allowed the subsequent subcloning of the truncation mutant into the pCMV5 eukaryotic expression vector.
The N-CAM truncation mutants, 788t, 770t, 749t, 725t, and 691t (identified by the amino acid site of truncation as numbered in Cunningham et al. (37)) were constructed using polymerase chain reaction amplification with pEC1401 as template. In all cases, the 5Ј polymerase chain reaction primer GGG GGT ACC AGC TTG GGC TGC AGG TCG ACT was used to introduce a KpnI site 5Ј to the start of the N-CAM coding sequence. Unique 3Ј primers inserted a stop codon (TAG) and ClaI restriction site at the sites of truncation. The sequence of the 3Ј primers is as follows: 788t, GGG ATC GAT CTA TTC TGT TAG TGG  GGT GGT CTC; 770t, GGG ATC GAT CTA GGG GGT CCG CTC CTC  TTC AGT; 749t, GGG ATC GAT CTA GGC AGC TTT GCC CTC CTC  CAT; 725t, GGG ATC GAT CTA GCA CAT GAG CAG GCC ACA TTT; 691t, GGG ATC GAT CTA CCC CAG GCC TGA CGT AGG GCT. The amplification products were gel purified, digested with KpnI/ClaI, and subcloned into pCMV5. The sequence of these and all subsequent mutant constructs was verified by dideoxynucleotide sequencing (38) using the Sequenase kit (U. S. Biochemical Corp.) or performed by Cornell University DNA Services (Ithaca, NY). Sequencing of one DNA clone of 691t revealed a point deletion which resulted in a nonsense mutation, truncating N-CAM after tyrosine 615. The expression of both soluble forms of N-CAM (691t and 615t) was subsequently analyzed.
To construct 719t, pCMV5 containing 725t was digested with KpnI and AatII to release a 2.3-kilobase pair fragment encoding N-CAM amino acids 1-712. Two synthetic complementary oligonucleotides encoding amino acids 713-719 flanked with a 5Ј AatII site and a 3Ј stop (TAG) codon and ClaI site were synthesized (Department of Biochemistry, NYU School of Medicine, New York; sequence available upon request) annealed, and after digestion with AatII and ClaI, fused to the 2.3-kilobase pair fragment and cloned into pCMV5.
To construct the N-CAM/N-CAM fusion constructs, 712t/1-40 and 712t/1-29, nucleotides encoding cytoplasmic amino acids 749 -788 or amino acids 749 -777 were amplified and fused to the cytoplasmic deletion mutant, 712t. Both fragments were amplified using the 5Ј primer GCG GAC GTC GCC TTC TCG AAA GAT GAG TCC AAG introducing a 5Ј AatII site, and 3Ј primers GGG ATC GAT CTA TTC TGT TAG TGG GGT GGT CTC or CGC ATC GAT CTA GTG TTT TCC CCC ATC ATG GTT G, respectively, introducing a 3Ј stop codon and ClaI site. The purified amplification products were ligated in-frame to the 2.3-kilobase KpnI-AatII fragment described above, and cloned into pCMV5. 712t/1-21 was similarly constructed, using a synthetic complementary pair of oligonucleotides corresponding to amino acids 749 -769 (sequence available upon request).
Generation of the p75 NTR/ N-CAM Fusion Construct, p75t/1-40 -To create the fusion construct, p75t/1-40, a 950-base pair KpnI-PvuII fragment encoding the ectodomain, transmembrane domain, and first 5 cytoplasmic amino acids of p75 NTR was isolated from a full-length human p75 NTR cDNA clone in pBluescript (kindly provided by Barbara Hempstead, Department of Medicine, Cornell University Medical College). The basolateral N-CAM signal was amplified using the 5Ј primer GGT GAC GCG TTC TCG AAA GAT GAG TCC which introduced a silent point mutation creating a unique 5Ј MluI site and the 3Ј primer GG AGG TCT AGA CTA TTC TGT TAG TGG GGT GGT CTC introducing a 3Ј stop codon and XbaI site. The amplified fragment was digested with MluI/XbaI, gel isolated, and the p75 NTR and N-CAM fragments blunt end ligated and cloned into pCMV5 in a tripartite ligation reaction. Reading frame of the chimera was restored upon ligation. The pXba1 (39) mutant of human p75 NTR which we refer to as p75t, was obtained from B. Hempstead in the expression vector pMV7 and subcloned as an EcoRI-XbaI fragment into pCMV5.
Site-directed Mutagenesis of the N-CAM Basolateral Signal-A unique SplI site was added immediately 5Ј to the sequence encoding the putative ␤-turn regions in the construct p75t/1-40, by mutating the native sequence CGG ACC to CGT ACG using polymerase chain reaction (p75t/1-40/Spl1). Alanine point mutants were created using sets of synthetic oligonucleotides encoding native or mutated sequence which were annealed and ligated as above, and cloned into the SplI-XbaI restriction sites of p75t/1-40Spl1. The sequence of all oligonucleotides is available upon request.
Cell Culture and Expression of Mutant Constructs-MDCK cells strain II were grown and maintained in Dulbecco's modified essential media supplemented with 5% fetal calf serum (Hyclone) and antibiotics.
MDCK cells were transfected using the DNA-calcium phosphate procedure (40). Stably expressing clones were selected in 500 g/ml G418 or 200 g/ml hygromycin, isolated with cloning rings, and screened for expression by immunofluorescence. Briefly, cells grown on coverslips were fixed in 2% paraformaldehyde, permeabilized with 0.075% saponin, and incubated with the rabbit anti-chicken N-CAM polyclonal antibody (R527), mouse anti-chicken N-CAM monoclonal 5e (Developmental Studies Hybridoma Bank), or an anti-p75 NTR monoclonal antibody (41). Expression was visualized after incubation with the appropriate fluorescently labeled secondary antibody.
Immunofluorescence and Confocal Microscopy-5-day-old filter grown monolayers of transfected MDCK cells were fixed in 2% paraformaldehyde in phosphate-buffered saline containing Ca 2ϩ and Mg 2ϩ (PBS-CM) and processed for indirect immunofluorescence. Nonpermeabilized monolayers were incubated with the anti-N-CAM monoclonal antibody, mAb 5e (1:500), or anti-human p75 NTR monoclonal antibody (1:100). After permeabilization with 0.075% saponin in PBS-CM, monolayers were incubated with fluorescein-conjugated antimouse IgG and RNase, and nuclei subsequently labeled with propidium iodide. Immunofluorescent samples were examined using a Molecular Dynamics confocal microscope (Sunnyvale, CA).
Steady State Localization and Membrane Targeting Assays--To determine steady state protein distribution, 5-6-day filter grown monolayers of stably transfected MDCK cells were biotinylated with sulfo-NHS-biotin or NHS-LC-biotin two times, either apically or basolaterally as described previously (41). After biotinylation, filters were excised and monolayers solubilized in lysis buffer (150 mM NaCl, 20 mM Tris, pH 8.0, 5 mM EDTA, 1% Triton X-100, 0.2% bovine serum albumin, and protease inhibitors). Extracts were preincubated with fixed Staphylococcus aureus cells, centrifuged, and cleared supernatants incubated with R527 anti-N-CAM antibody (1:300) overnight. After 2 h incubation with protein A-Sepharose (10 mg/ml) the immunoprecipitates were washed as described (41). For p75 NTR immunoprecipitations, the anti-p75 NTR monoclonal antibody was first coupled to rabbit anti-mouse IgG bound to protein A-Sepharose. Bound proteins were released from the beads, separated by SDS-PAGE, and transferred to Immobilon-P membranes (Millipore, Bedford, MA) for visualization with radioiodinated streptavidin.
To examine polarized delivery to the cell surface, we used a biotin targeting assay (41). 5-6-day filter grown monolayers were pulse labeled for 15 min. At various time points during chase, filters were chilled to 4°C, biotinylated from the apical or basolateral surface, and immunoprecipitated as indicated above. Samples were subsequently re-precipitated with streptavidin-agarose, and the biotin-labeled immunoprecipitates separated by SDS-PAGE and visualized by fluorography.
Endocytosis Assays-Cells grown on Transwell filters for 5-6 days were biotinylated through a linkage sensitive to reduction either apically or basolaterally with NHS-SS-biotin as described (41) and incubated at 37°C for various times before cooling to 4°C to allow for protein internalization. Non-internalized biotin label was subsequently removed by incubation with 50 mM glutathione in 90 mM NaCl, 1 mM MgCl 2 , 0.1 mM CaCl 2 , 60 mM NaOH, and 10% fetal bovine serum followed by a 15-min incubation with 0.5 mg/ml iodoacetamide in PBS-CM. Monolayers were subsequently washed, lysed, and immunoprecipi-tated. After electrophoresis and transfer to Immobilon P in nonreducing conditions, biotin-labeled proteins were detected by 125 I-streptavidin blotting.

The Cytoplasmic Tail of N-CAM Contains Basolateral Sort-
ing Information-Both the 140-and 180-kDa isoforms of N-CAM have been previously demonstrated to be targeted to the basolateral surface upon expression in MDCK cells (35). To examine whether the cytoplasmic tail of N-CAM contained basolateral sorting information, a mutant form of N-CAM deleted of its cytoplasmic tail (712t) was constructed and stably expressed in MDCK cells (Fig. 1A). Indirect immunofluorescence and laser scanning confocal microscopy indicated that deletion of the cytoplasmic domain of N-CAM resulted in a loss of basolateral polarity ( Fig. 2A), with the truncated protein expressed on both the apical and basolateral surfaces of the cells. These vertical sections clearly show that the basolateral pool of N-CAM-140 is predominantly segregated to the lateral edges of the cells and not found diffusely throughout the basolateral surface, possibly reflecting retention at the lateral surface due to homophilic interactions between neighboring cells expressing N-CAM.
Domain specific biotin labeling of surface proteins and subsequent immunoprecipitation (see Fig. 2B) demonstrated that at steady state, whereas greater than 95% of surface N-CAM-140 was basolateral, a representative clone of 712t showed a basolateral polarity of 57.4% (Ϯ3.0, n ϭ 3). To examine whether this change in polarity reflected a loss of intracellular sorting fidelity or the possible recruitment of 712t into an indirect pathway to the apical surface, we utilized a pulse-chase membrane targeting assay. In Fig. 3A, newly synthesized 712t reaches the apical and basolateral surfaces within 10 -20 min of chase. Delivery to each surface shows similar kinetics and remains approximately 60% polarized to the basolateral surface throughout the chase period indicating that the deletion construct was directly transported to each surface domain. Therefore, the nonpolar distribution of 712t reflects a loss of basolateral sorting fidelity at the TGN and indicates a dependence on cytoplasmic sequences for the correct sorting of N-CAM to the basolateral surface in MDCK cells. This is not due to a general sorting defect of MDCK cells stably expressing 712t, as these cells sort correctly the basolateral marker E-cadherin (not shown) nor is it the result of clonal variation as more than 5 independently isolated clones expressing a range of different protein levels displayed similar sorting behavior. We considered the possibility that the nonpolar distribution of 712t might reflect a weakened association of 712t with the membrane due to the lack of charged residues in the cytoplasmic tail of 712t. However, the mutant protein exhibited basic characteristics of an integral membrane proteins as it was not extracted in either high salt or alkaline conditions (data not shown). Furthermore, pulse-labeled 712t was not detected in the media over an 8-h chase and all detectable protein remained cell-associated (data not shown) indicating that 712t is not shed from the cell surface. In these studies, we did not obtain clones expressing 712t at levels comparable to those expressing wild type N-CAM, suggesting that the non-polar targeting of this mutant is not due to saturation of the basolateral sorting machinery.

A Nested Set of N-CAM C-terminal Truncations Define Regions within the Cytoplasmic Domain Required for Basolateral
Localization-To localize the putative basolateral signal, a series of C-terminal deletion mutants, truncated at amino acids 788, 770, 749, or 725 ( Fig. 1A) was generated by polymerase chain reaction and stably expressed in MDCK cells. As demonstrated in Fig. 4A by immunofluorescence, these mutants show different membrane distributions. N-CAM-140 deleted of 22 C-terminal amino acids (788t) retained basolateral localization. However, deletion of 40 amino acids in the mutant 770t led to the nonpolar distribution of the protein, with staining evident on both the apical and basolateral surfaces of the monolayer. 749t and 725t, deleted of 61 and 82 C-terminal amino acids, respectively, show predominantly apical surface staining.
Immunoprecipitation from cells surface labeled with biotin confirmed this result (Fig. 4B). While 788t labeled predominantly from the basolateral side, 770t could be labeled from both the apical and basolateral surfaces of the monolayer. 749t and 725t were predominantly labeled from the apical surface. Targeting assays as described above demonstrated that the steady state polarized localization of 788t, 770t, 749t, and 725t resulted from the direct transport of these N-CAM constructs to their resident membrane domains (data not shown). The differential localization of the truncation mutants thus identifies (as summarized in Fig. 1) amino acids 749 -788 as a putative basolateral signal.
Amino Acids 749 -788 from the Transmembrane Domain in N-CAM Function Independently as a Basolateral Sorting Signal in MDCK Cells-Changes in basolateral polarity of the N-CAM truncation mutants suggested that amino acids 749 -788 were required for the appropriate sorting of N-CAM. To address whether this putative signal could indeed function as an independent basolateral sorting signal, a construct was generated (schematically represented in Fig. 1A) in which amino acids 749 -788 were fused to the nonpolar cytoplasmic deletion mutant 712t in an attempt to redirect the N-CAM mutant to the basolateral surface. Upon expression in MDCK cells, 712t/1-40 is localized to the basolateral surface, as indicated by the lateral N-CAM staining seen in Fig. 5A, and confirmed biochemically by domain specific biotinylation and subsequent immunoprecipitation (Fig. 5B). To ensure that this steady state localization was due to the intracellular sorting of the fusion protein, we monitored the arrival of newly synthesized 712t/1-40 at both the apical and basolateral surfaces, with the biotin targeting assay described above (see Fig. 3B).
We found the fusion construct to be delivered directly to the basolateral surface, demonstrating the ability of this region to function as a basolateral sorting signal. In an attempt to define the minimal amino acid requirements for basolateral sorting, we deleted 16 or 22 N-terminal amino acids from the cytoplasmic signal of 712t/1-40 by fusing the N-CAM ectodomain to amino acids 770 -788 (712t/23-40) or 764 -788 (712t/16 -40). Neither one of these regions of the cytoplasmic domain effectively bestowed basolateral localization (data not shown).
To confirm that the N-CAM basolateral signal could function in a transportable manner independently of any sorting information carried in the ectodomain of N-CAM, we fused the 40-amino acid basolateral signal to the ectodomain and transmembrane domain of an apically expressed heterologous protein, the p75 neurotrophin receptor (p75 NTR ) (schematically represented in Fig. 1). It has been previously demonstrated that a mutant form of p75 NTR truncated after the fifth amino acid of the cytoplasmic tail (p75t) is expressed apically (Ͼ80%) upon transfection of its cDNA in MDCK cells (42). Upon fusion of the N-CAM basolateral signal, the resulting chimeric protein migrated at the expected molecular mass of 65 kDa and was redistributed to the basolateral surface as shown by both immunofluorescence (Fig. 6A) and cell-surface biotinylation (Fig.  6B). This confirms that cytoplasmic amino acids 749 -788 of N-CAM are recognized as an independent basolateral sorting

FIG. 5. The region identified by analysis of C-terminal deletion mutants of N-CAM-140 functions independently as a basolateral targeting signal.
A, MDCK cells were stably transfected with the fusion construct, 712t/1-40, and processed for immunofluorescence as described (see Fig. 2A). The top panel presents a horizontal crosssection (XY-scan) revealing lateral expression of 712t/1-40 which is confirmed by confocal generated vertical section (bottom panel). Bar, 5 m. B, steady state localization of N-CAM/N-CAM fusion constructs was analyzed by domain-specific biotinylation and subsequent immunoprecipitation as described in the legend to Fig. 2. A, apical; B, basolateral. Deletions of cytoplasmic amino acid sequences predicted to form ␤-turns in 712t/1-40 abrogate basolateral sorting. While deletion of the membrane distal 11 amino acids of 712t/1-40 leads to nonpolar expression (712t/1-29), a deletion encompassing both putative ␤-turns (712t/ 1-21) is expressed predominantly on the apical surface. In all cases, at least 2 other clones were analyzed, and in all cases showed similar sorting behavior. The amino acids forming the C-terminal tails of these fusion proteins are listed in Fig. 1C.

FIG. 6. The 40-amino acid N-CAM basolateral signal redirects an apically expressed cytoplasmic deletion mutant of p75 NTR to the basolateral surface.
A, nonpermeabilized MDCK cells stably expressing the p75 cytoplasmic deletion mutant (p75t) or the fusion construct p75t/1-40 were stained for immunofluoresence as described in the legend to Fig. 2A, using a human p75 NTR specific monoclonal antibody (42). p75t is predominantly localized on the apical surface (i) of MDCK cells, and the chimeric construct, p75t/1-40 (ii) is expressed basolaterally. The lower panels present vertical confocal sections taken through the en face views shown above. B, domain specific biotinylation (as described in the legend to Fig. 2B) confirms that p75t is polarized to the apical surface while p75t/1-40 is re-directed to the basolateral surface. A, apical; B, basolateral. 1, p75t; 2, pool of p75t/1-40 transfected MDCK cells; 3-5, representative clones expressing p75t/1-40. In all cases, the fusion construct is expressed with Ͼ90% basolateral polarity. Although these clones express different levels of protein we found no changes in surface polarity indicating that sorting within this range is not affected by the level of protein expression. determinant in MDCK cells.
The N-CAM Basolateral Signal Mediates Slow Internalization-In the presence of saponin, we found by indirect immunofluorescence that the fusion constructs 712t/1-40 and p75/ 1-40 localized to intracellular vesicles as well as to the basolateral membrane. In subsequent double immunolabeling experiments, we found 712t/1-40 partially colocalized with the lysosomal marker AC17 (43) (data not shown), indicating its presence in the endocytic/lysosomal pathway. The endocytic activity of the N-CAM basolateral signal was therefore analyzed biochemically using a biotin internalization assay. As endocytosis can occur at different rates from the apical and basolateral surfaces (42,44) we took advantage of the large basolateral pool of mistargeted p75t which afforded sufficient signal for direct comparison of internalization rates to the basolaterally expressed p75t/1-40 chimera. As shown in Fig. 7, fusion of the N-CAM basolateral signal to p75t confers a change in endocytic behavior. While internalized p75t is not detected before 40 min of internalization, p75/1-40 is more rapidly internalized. After internalization, a pool of p75t remains detectable throughout the chase period while the pool of intracellular p75t/1-40 declines rapidly. When earlier time points were examined (0 -30 min) an approximate 2-3-fold increase in the initial rate of internalization of p75/1-40 over that of p75t was evident (not shown). These data indicate that the basolateral targeting signal mediates a modest increase in internalization, but at a rate much slower (Ͻ1% per min) than classical tyrosine-signal based internalization (ϳ10% per min) (45). It is interesting that although p75t/1-40 is internalized 2 to 3 times more efficiently than p75t, this increase is insufficient to explain the drastic change in protein stability. Pulse-chase analysis indicate that the half-life of p75t/1-40 is less than 1 h whereas the half-life of p75t is greater than 6 h in MDCK cells (42). Indeed, the rapid loss of the glutathione-insensitive pool of p75t/1-40 (Fig. 7) suggests a net movement of internalized protein toward a degradative compartment. These changes in endocytic ability and subsequent intracellular trafficking are independent of ectodomain information, as 712t/1-40 which contains the N-CAM ectodomain, behaves in an identical manner (not shown).
␤-Turns Are Not a Crucial Aspect of the N-CAM Basolateral Sorting Signal-Although the N-CAM basolateral signal does not mediate rapid internalization, computer modeling using both the Chou-Forsman (46) and Garnier-Robson (47) algorithms indicated that the signal contained two regions predicted to form turn structures (see Fig. 1). To address the involvement of these regions in basolateral sorting, we deleted either both putative ␤-turns (712t/1-21) or the membranedistal turn alone (712t/1-29) from the N-CAM fusion protein 712t/1-40 as schematically represented in Fig. 1C. As indicated in Fig. 5B, steady state biotinylation indicates that deletion of the distal turn (712t/1-29) leads to a loss of steady state basolateral polarity. Deletion of both regions (712t/1-21), however, leads to apical localization, demonstrating a dependence on these regions for efficient basolateral sorting. To further identify the precise amino acids requirements of the signal, we generated a number of point mutations in the basolateral signal of the p75t/1-40 chimeric protein. Since the results with the truncated signal suggested a role for the regions postulated to form ␤-turns, we concentrated on mutating amino acids that were most likely to affect the conformation of these regions. As such, we focused primarily on the asparagine and proline residues present in the two putative turn domains and looked at the effects of these mutations individually or in tandem. Distribution of the mutant proteins was assessed by domain specific biotinylation in pools of stably transfected MDCK cells. As summarized in Table I, none of the amino acid changes adversely affected basolateral sorting, implying that the ␤-turn may not be a critical feature of this basolateral signal and suggesting that some other characteristic of the entire 40amino acid stretch is required for basolateral targeting.
N-CAM Contains Apical Sorting Information-As demonstrated above, truncation of N-CAM-140 at amino acids 725 and 749, sites upstream of the basolateral signal, resulted in the apical expression of these constructs (725t and 749t, respectively), suggesting that N-CAM contained cryptic apical FIG. 7. The N-CAM basolateral signal mediates slow internalization. MDCK cells expressing p75t or p75t/1-40 were labeled at the basolateral surface with NHS-S-S-biotin at 4°C and then shifted to 37°C for various times to allow internalization to occur. Cells were incubated with glutathione (glut) and subsequently lysed and immunoprecipitated. The internalized glutathione-insensitive pool of protein was visualized by autoradiography using 125 I-streptavidin after separation by SDS-PAGE and transfer to polyvinylidine difluoride in nonreducing conditions. At 0 min of chase, all of the labeled protein is present on the surface and sensitive to glutathione. 75t/1-40 is detected more rapidly than p75t, indicating a more rapid initial rate of internalization. Additionally, the internalized pool of p75t/1-40 seems more labile as it is rapidly degraded after internalization. The gel presented is representative of at least triplicate experiments.

TABLE I Point mutational analysis of the N-CAM basolateral sorting signal
A series of point mutants was generated in the context of the fusion protein p75t/1-40 (see "Materials and Methods") to investigate the importance of amino acid residues suspected to form a ␤-turn. Cytoplasmic amino acids are numbered as in Fig. 1. A dash (-) indicates amino acids conserved from the parental construct, and amino acid changes are as indicated (A, alanine; Q, glutamine; E, glutamate). Expression of the constructs was analyzed in pools of stably transfected MDCK cells. The percentage of total surface protein present on the basolateral surface at steady state is indicated (% BL) and reflects the average of at least duplicate samples. Polarity was confirmed by immunofluorescence on at least three independently isolated clones. Point mutational analysis of the N-CAM basolateral signal failed to identify amino acid residues critical for basolateral sorting.
sorting information functioning recessively to the dominant basolateral signal. To localize this apical sorting information, we analyzed the expression of two soluble forms of N-CAM truncated at amino acids 691 or 615 (691t and 615t, respectively, shown in Fig. 8). MDCK monolayers stably expressing 691t and 615t were metabolically labeled and release of soluble N-CAM into the apical and basolateral media monitored by immunoprecipitation. As shown in Fig. 8B, 691t and 615t are expressed at the expected molecular masses of 120 and 105 kDa and are released predominantly from the apical surface, demonstrating that the ectodomain of N-CAM contains apical sorting information. Analysis of a representative clone of each construct demonstrated apical polarities of 72 Ϯ 4.0% (n ϭ 3) and 97.9 Ϯ 1.7% (n ϭ 3), 691t and 615t, respectively. Ater 3 h of chase, a smaller molecular weight form of N-CAM is immunoprecipitated from cellular lysates indicating the presence of the non-glycosylated precursor trapped intracellularly. Similar sorting fidelities were obtained from transient transfection experiments (data not shown). This evidence for apical sorting information in the ectodomain of N-CAM suggested that apical targeting in the absence of a dominant basolateral signal is mediated by a common lumenal targeting determinant. However, 712t is sorted in a nonpolar manner (Fig. 3), with a preference toward the basolateral surface. This suggested that juxtamembrane cytoplasmic sequences present in 725t and 749t but deleted in 712t may contain apical sorting information. Alternatively, the ectodomain may contain apical sorting information as suggested by the apical sorting of soluble mutants of N-CAM, which in the context of mutant 712t may no longer effectively be recognized by the lumenal sorting machinery in the TGN.
To begin to address these possibilities, we truncated N-CAM at amino acid 719. As indicated by immunofluorescence (not shown) and steady state biotinylation techniques (Fig. 8C) 719t is expressed predominantly on the apical surface. To ensure that the bulk of the newly synthesized protein reached the cell surface, confluent monolayers of transfected MDCK cells were exposed to trypsin (100 g/ml) added to either or both domains. This treatment resulted in cleavage of more than 90% of pulsechased wild type N-CAM and greater than 80% of mutant 719t, indicating that the bulk of these proteins reaches the cell surface.
Although not shown, we also expressed a mutant form of N-CAM in MDCK cells in which two lysine residues were fused to the cytoplasmic tail of 712t which lacks positively charged amino acid residues. However, we find that the addition of these two charged residues has no effect on the surface polarity of 712t, and that this mutant protein remains directly delivered to both the apical and the basolateral surfaces (data not shown).

DISCUSSION
In this study, we have identified a basolateral sorting signal in the cytoplasmic domain of N-CAM-140. Differences in the delivery and surface distribution of cytoplasmic truncation mutants demonstrated that the cytoplasmic amino acids 749 -788 contain basolateral targeting information. This signal is recognized intracellularly and can function independently as demonstrated by its ability to direct sorting to the basolateral surface when fused to both the nonpolar N-CAM cytoplasmic deletion mutant (712t) or to the normally apical cytoplasmic deletion mutant (p75t) of the p75 NTR . Additionally, this signal is tyrosine-independent and mediates a modest increase in the rate of internalization of chimeric proteins. Although basolateral sorting is impeded by deletions within regions of the basolateral signal that are predicted by computer analysis to form turn structures, site-directed mutagenesis within these regions did not identify amino acid residues critical for basolateral targeting.
Relationship to Other Characterized Basolateral Sorting Signals--The 40-amino acid basolateral sorting signal that we have identified in N-CAM contains neither the tyrosine based nor dileucine-based motifs characteristic of most basolateral signals. Basolateral signals overlapping tyrosine-based endocytic signals (low density lipoprotein receptor (48), lysosomal acid phosphatase (49)) as well as those unrelated to endocytic determinants (poly Ig receptor, (34)) have both been associated with type 1 ␤-turns, suggesting that this structure forms an important feature of tyrosine-based basolateral signals. Surprisingly, computer modeling predicted that two regions within the N-CAM signal are capable of forming turn structures. Deletion of these regions led to a loss of basolateral polarity suggesting that these structures were involved in the basolateral sorting of N-CAM. In light of these results we constructed a series of point mutations that we expected to disrupt the secondary structure of the basolateral signal. None of these point mutations were found to impede basolateral targeting, even though point mutational analysis has in most cases successfully identified critical features of sorting signals. These results suggest that the ␤-turn is not a crucial aspect of the N-CAM basolateral signal and that a more global characteristic of the signal is required for basolateral sorting. The ␤-turn is not an invariable trait of all cytoplasmic signals. For instance, TGN38, an integral membrane protein found to cycle between the plasma membrane and the TGN, contains a tyrosine-based internalization motif found to lie not within a ␤-turn but within a nascent helix (50). Furthermore, although evidence exists implicating the ␤-turn in basolateral sorting mediated by both endocytic and non-endocytic signals other signal motifs, such as the dileucine motif, may be recognized by a different mechanism at the TGN. The dileucine motif, first identified in CD3 (51) as mediating both internalization and lysosomal targeting, has since been implicated in the internalization and basolateral targeting of the low affinity IgG Fc receptor (52). Although initially believed to function in a manner similar to the tyrosine-based signals, it has been shown that the medium chains of the adaptor complexes AP-1 and AP-2 specifically interact with tyrosine-based signals of several integral membrane proteins but not with the dileucine motif suggesting that the latter signal may depend on different structural elements (53).
Taken together, these different lines of evidence suggest that basolateral sorting of proteins may be mediated by several different mechanisms relying on different secondary structural elements. The lack of previously described sequence motifs suggests that the basolateral signal in N-CAM may define a novel class of basolateral sorting signal. However, structural analysis will be required to discern the precise nature and conformational requirements directing basolateral transport of N-CAM.
Our data indicate that other downstream membrane trafficking events may also be modulated by the N-CAM basolateral signal. It is well demonstrated that basolateral signals can overlap with signals directing a variety of different sorting events. For example, tyrosine-based signals mediate targeting to the endosomal/lysosomal network from the TGN (lysosomal acid phosphatase, lamp 2), TGN localization (TGN 38/41 (54)) as well as clathrin-dependent internalization of the low density lipoprotein receptor (55), and the transferrin receptor (31). Furthermore, dileucine-based signals also mediate diverse sorting processes, including internalization from the cell surface (IgG Fc receptor (52)), targeting of both CD-M6PR and CI-M6PR to endosomes (56), and sorting to the lysosome from the TGN (CD36/LIMP II (57,58)). Preliminary experiments in which the stability of the internalized pool of p75t/1-40 and p75t is analyzed indicates that internalized p75t/1-40 is more labile than p75t, suggesting that the N-CAM-140 basolateral signal contains information directing rapid degradation. Although this aspect of the N-CAM basolateral signal requires further characterization, it is interesting to speculate on how such a signal could play a role in the regulation of N-CAM surface expression. For instance, in processes such as synaptic remodeling, undefined intracellular signals could activate the internalization of surface N-CAM, which would then be diverted into a degradative pathway hence ensuring the constitutive degradation of internalized N-CAM and preventing inappropriate recycling of the molecule back to the surface.
Apical Information in the Ectodomain of N-CAM Is Unable to Direct 712t to the Apical Surface-Although many cytoplasmic tail deletion mutants of basolateral proteins are delivered apically (27-29, 33, 59), some truncated basolateral proteins are not, suggesting that they lack a positive apical signal. For example, both vesicular stomatitis virus G protein (60) and transferrin receptor (61) deleted of their cytoplasmic tails are missorted to both surfaces. Ectodomain apical sorting signals have been proposed to be contained in post-translational modifications such as N-linked carbohydrates (62). However, as many exceptions to this rule exist (63,64), the possibility cannot be discarded that the apical information is present in threedimensional patches formed by nonlinear protein segments (21) sensitive to conformational changes induced by carbohydrate residues. Surprisingly, we found that progressive truncations of N-CAM switch the pattern of protein expression from basolateral (788t) to nonpolar (770t), then to apical (749t, 725t, and 719t), and finally back to a nonpolar pattern (712t). Although it is possible that the first nine cytoplasmic amino acids contain apical sorting information, the apical release of soluble forms of N-CAM suggests that the ectodomain of N-CAM carries sufficient sorting information to specify apical localization. However, we cannot determine from our results why this ectodomain signal is insufficient to drive apical sorting of the cytoplasmic deletion mutant 712t. The lack of effect upon fusion of two lysine residues to 712t suggests that the lack of positive charge in the cytoplasmic tail, a frequent feature of transmembrane proteins, is not the reason for the loss of polarized expression of 712t (data not shown). However, the short length of the cytoplasmic tail of 712t may affect the conformation of the ectodomain, weakening the ability of the apical signal to interact with putative apical sorting receptors in the TGN. Evidence for a role of cytoplasmic residues in altering ectodomain interactions has been recently described in the normally basolaterally expressed asialoglycoprotein receptor. Deletion within the tail of this receptor results in its TGN retention, a localization determined by the length of the cytoplasmic tail rather than specific sequence and mediated through the protein ectodomain (61). Increasing tail length may inhibit interactions between the ectodomain and TGN components presumably by steric hindrance in the ectodomain or by altering the position of the protein in the membrane, thereby allowing transport to the cell surface.
Although apical delivery of soluble mutants of N-CAM suggests that the ectodomain contains apical sorting information, we cannot exclude the possibility that additional apical sorting information is contained within the transmembrane spanning region of membrane anchored N-CAM. Indeed, influenza virus neuraminidase has been shown to contain two independent apical signals, one in the ectodomain and another in the transmembrane and/or cytoplasmic domain (63). Conformational changes in the context of 712t may affect both signals, a possibility that can be addressed by further mutational analysis.
Regardless of the precise mechanism responsible for apical sorting, the strong basolateral localization of 712t has important implications for the sorting of proteins to the basolateral surface. It has been suggested that sorting to the basolateral surface is an active process by which only proteins containing basolateral determinants are included in a basolateral vesicle. Apical proteins, alternatively, lacking such a signal would fail to be incorporated into these vesicles and would thereby be removed via apically targeted vesicles by default (65). However, the ability of 712t to be directly delivered to the basolateral surface implies that proteins can be incorporated into basolateral vesicles via a non-cytoplasmic signal-mediated mechanism. If proteins can enter a basolateral vesicle without a signal, it is therefore necessary that apically directed traffic also be signal mediated to maintain asymmetry. Alternatively, different populations of basolateral vesicles may exist: one that specifically incorporates signal-containing basolateral proteins and one directing non-signal mediated delivery to the basolateral surface. Finally, it cannot be discounted that the transmembrane domain of N-CAM may contain some basolateral targeting information. Indeed, if the position of 712t in the lipid bilayer is altered, this may allow the incorporation of the protein into basolateral vesicles via a non-cytoplasmic mediated event in a manner similar to that by which a mutant form of hemagglutinin is incorporated into coated pits, a feature presumably mediated through the mutant HA transmembrane domain (66).