The Cytoplasmic Tail of CD1d Contains Two Overlapping Basolateral Sorting Signals*

a classic internalization (YQGV) in tail. in cells, the cytoplasmic tail of CD1d for basolateral sorting. a mutation of the critical tyrosine residue of the endosomal sorting signal did not result in the loss of basolateral targeting of the mu-tant CD1d. To search for a basolateral sorting signal we constructed a full set of alanine mutants, but no single alanine substitution inactivated the signal. How-ever, deletions or mutations of either the C-terminal valine/leucine pair or the critical tyrosine residue from the internalization signal and either residue from the C-terminal valine/leucine pair inactivated basolateral sorting. Our data suggest that the cytoplasmic tail contains two overlapping basolateral signals, one tyrosine- and the other leucine-based, each being sufficient to direct CD1d to the basolateral membrane of polarized canine kidney cells. m M CdCl 2 overnight to induce expression of the protein of interest. Clones expressing constructs of interest were identified by screening with the D5 antibody (10). Iodination of Antibodies— D5 antibody was labeled with Na 125 I using IODO-BEADs (Sigma) as described previously (10). Briefly, the antibody (100 m g) was incubated with 1 mCi of Na 125 I and IODO-BEADs for 10 min on ice. Iodinated antibody was then separated from free Na 125 I on Sephadex G-25M columns (Amersham Pharmacia Biotech). The specific activity of the labeled antibody was determined by trichloroace- tic acid precipitation. The amount of soluble radioactivity was generally less than 5% of total radioactivity. Surface Labeling of Polarized Cells— Stably transfected MDCK cells were grown on Transwell polycarbonate filter units (Costar; 10 6 cells/ filter) for 4 days prior to experiments. Expression of the constructs of interest was induced by overnight incubation with 0–25 m M CdCl 2 . To avoid possible artifacts because of the selection of a single clone, different clones were chosen and tested at varying concentrations of CdCl 2 . Care was taken not to overexpress CD1d constructs to levels that might induce missorting. Positive controls (cells that expressed basolaterally sorted wild-type CD1d at same or higher levels as constructs in ques-tion) were included in each set of experiments. Cells were then cooled on ice and incubated with 125 I-D5 (2–10 m g/ml) added apically or basolaterally. The integrity of tight junctions was followed by monitoring the leakage of radioactivity from the basolateral to the apical side. Un-bound antibodies were removed by extensive washing with PBS 11 and filters were excised and counted with a Cobra TM Auto-Gamma counter. Nonspecific binding was determined in parallel experiments with non- transfected MDCK cells and corrected for. Using the same cell line and growth condition we have shown earlier that the apical:basolateral surface membrane is 1:1 (14) and a molecule is distributed polarized when it is significantly different from this. Materials— DMEM and fetal calf serum were obtained from BioWhit-taker. Oligonucleotides for PCR mutagenesis were synthesized by Med- probe (Norway). Materials for PCR amplification, restriction, ligation, and sequencing were from New England Biolabs. All other materials (unless specified otherwise) were purchased from Sigma.

CD1 polypeptides, evolutionarily related to the major histocompatibility complex class I molecules, represent a new class of antigen-presenting molecules that bind and present lipids and glycolipids rather than peptide antigens and are implicated in host defense against invading pathogens (for review, see Refs. [1][2][3][4]. CD1d has been reported to present glycolipid antigens such as ␣-galactosylceramide to the CD1d-restricted natural killer T cells (5)(6)(7)(8), but recent data demonstrate that CD1d may interact with a broader array of T cells (9). We have previously studied the mechanisms of intracellular trafficking of CD1d in MDCK 1 cells and found that the short cytoplasmic tail of CD1d was important for its internalization and basolat-eral sorting (10). Our results showed that CD1d contained a classical tyrosine-based internalization signal in its cytoplasmic tail. Replacing either the tyrosine or the hydrophobic valine residue in the ϩ3 position from the tyrosine residue with alanine resulted in a loss of active internalization. However, alanine substitution of neither the critical tyrosine nor the valine disrupted the basolateral distribution of CD1d. Nonetheless, basolateral sorting information in the cytoplasmic tail of CD1d was sufficient to redirect the otherwise apically distributed protein CD8 to the basolateral surface, indicating that the tail contained sufficient information for basolateral sorting (10).
Although the mechanisms for apical sorting remain largely undefined, a number of basolateral sorting signals have been identified. Basolateral sorting signals are currently subdivided into two major classes: signals that are either co-linear or not co-linear with the signals for coated pit localization. Signals that are co-linear with the signals for coated pit localization can be further subdivided into the tyrosine-based basolateral sorting signals, such as those of lysosomal associated membrane protein-1 (11), lysosomal acid phosphatase (12), and TGN38 (13) and leucine-based signals found, for example, in the invariant chain (14), Fc receptor II (15), and furin (16). Signals that are not co-linear with the signals for the coated pit internalization may be either tyrosine-dependent, such as signals in the vesicular stomatitis virus G protein (17) and the low density lipoprotein receptor proximal signal (18), or tyrosine-independent such as in polyimmunoglobulin receptor (19) and the transferrin receptor (20). Tyrosine-and leucine-based sorting signals are believed to interact with one or more of the adaptor complexes, AP1 at the TGN, AP2 at the plasma membrane, and AP3 intracellularly (reviewed in Refs. [21][22][23][24], and it has recently been reported that interaction with AP1B may be a part of the polarization machinery (25,26).
In this study, we sought to identify the basolateral sorting signal in the cytoplasmic tail of CD1d. Our data indicate that there are two overlapping signals within the very last five C-terminal amino acids of CD1d, one tyrosine-and one leucinebased, each being sufficient for its basolateral sorting.

EXPERIMENTAL PROCEDURES
DNA Constructs-Amino acid composition of the cytoplasmic tails of constructs used in this study is given in Fig. 1. Mutations in the cytoplasmic tail of CD1d were created by PCR using the wild-type CD1d cDNA as a template. All constructs were subcloned into the pMEP4 vector (Invitrogen) and sequenced.
Cell Growth-MDCK (strain II) cells were grown in full growth medium (DMEM supplemented with 10% fetal calf serum, 2 mM glutamine, 25 units/ml penicillin, and 25 g/ml streptomycin) in 5% CO 2 in a 37°C incubator.
Stable Transfection of MDCK Cells and Clonal Selection-MDCK cells were stably transfected by the calcium phosphate procedure as described elsewhere (27). Clones expressing DNA constructs under control of the inducible metallothionein promoter in the pMEP4 vector were selected in the presence of hygromycin B (0.3 mg/ml). Resistant clones were isolated and incubated with 25 mM CdCl 2 overnight to induce expression of the protein of interest. Clones expressing constructs of interest were identified by screening with the D5 antibody (10).
Iodination of Antibodies-D5 antibody was labeled with Na 125 I using IODO-BEADs (Sigma) as described previously (10). Briefly, the antibody (100 g) was incubated with 1 mCi of Na 125 I and IODO-BEADs for 10 min on ice. Iodinated antibody was then separated from free Na 125 I on Sephadex G-25M columns (Amersham Pharmacia Biotech). The specific activity of the labeled antibody was determined by trichloroacetic acid precipitation. The amount of soluble radioactivity was generally less than 5% of total radioactivity.
Surface Labeling of Polarized Cells-Stably transfected MDCK cells were grown on Transwell polycarbonate filter units (Costar; 10 6 cells/ filter) for 4 days prior to experiments. Expression of the constructs of interest was induced by overnight incubation with 0 -25 M CdCl 2 . To avoid possible artifacts because of the selection of a single clone, different clones were chosen and tested at varying concentrations of CdCl 2 . Care was taken not to overexpress CD1d constructs to levels that might induce missorting. Positive controls (cells that expressed basolaterally sorted wild-type CD1d at same or higher levels as constructs in question) were included in each set of experiments. Cells were then cooled on ice and incubated with 125 I-D5 (2-10 g/ml) added apically or basolaterally. The integrity of tight junctions was followed by monitoring the leakage of radioactivity from the basolateral to the apical side. Unbound antibodies were removed by extensive washing with PBSϩϩ and filters were excised and counted with a Cobra TM Auto-Gamma counter. Nonspecific binding was determined in parallel experiments with nontransfected MDCK cells and corrected for.
Using the same cell line and growth condition we have shown earlier that the apical:basolateral surface membrane is 1:1 (14) and a molecule is distributed polarized when it is significantly different from this.
Materials-DMEM and fetal calf serum were obtained from BioWhittaker. Oligonucleotides for PCR mutagenesis were synthesized by Medprobe (Norway). Materials for PCR amplification, restriction, ligation, and sequencing were from New England Biolabs. All other materials (unless specified otherwise) were purchased from Sigma.

RESULTS AND DISCUSSION
In a previous study (10), we reported that a deletion of the last six amino acids from the CD1d cytoplasmic tail ( Fig. 1 delta SYQGLV construct) abolished basolateral sorting of CD1d in MDCK cells (Fig. 2). Furthermore, fusing the last eight CD1d cytoplasmic amino acids to the transmembrane and extracellular domains of the CD8 molecule re-routed this otherwise apical protein to the basolateral surface of MDCK cells. Interestingly, mutation in the tyrosine residue critical for internalization did not impair basolateral sorting of CD1d. We therefore undertook an alanine scan of the CD1d cytoplasmic tail (Fig. 1) to identify possible important amino acid residue(s) for its basolateral sorting.
The alanine scan did not reveal any single critical residue important for basolateral sorting of CD1d (Fig. 2). However, all individual alanine substitutions somewhat increased the amount of CD1d routed to the apical side as compared with the wild-type molecule. The largest effect (over 20% sorted apically compared with less then 5% for the wild-type molecule) was registered when either of the two most C-terminal residues, valine and leucine, was mutated to alanine (Fig. 2).
This led us to investigate in more detail the role of the C-terminal valine and leucine residues in basolateral sorting of CD1d. As internalization and endosomal sorting were abolished by mutation of the tyrosine alone, we concluded that these residues were not essential for endosomal sorting of CD1d. A single mutation of the valine or the leucine was not sufficient to abolish basolateral sorting, and we therefore made mutants with a dual substitution or deletion of both residues (Fig. 1, constructs VA, LA, and delta VL). As shown in Fig. 3, these mutants were no longer sorted basolaterally, indicating that valine and leucine residues were in fact involved in basolateral sorting. It is well established that leucine-based endosomal sorting signals are not necessarily recognized for basolateral sorting (11, 28 -32), and our results document the reverse: a basolateral sorting signal that is not active in internalization.
Previous studies have shown that single point mutations within a leucine-based sorting motif were sufficient to abolish basolateral sorting (e.g. Refs. 14 and 15). This was clearly not the case for the VL-based signal of CD1d, as both residues had to be mutated to impair the basolateral sorting. We, however, noticed that the valine was also a part of the tyrosine-based internalization signal. We therefore decided to reinvestigate the role of the cytoplasmic tyrosine residue in basolateral sorting of CD1d. Two constructs were made; either valine or leucine was mutated to alanine in addition to the tyrosine (Fig.  1, constructs YA, VA and YA, LA). Both constructs were no longer sorted basolaterally. From this we conclude that the tyrosine is also involved in targeting CD1d to the basolateral surface.
Curiously the YA, LA, delta VL, and the VA, LA constructs were distributed predominantly to the apical membrane (Fig.  3). This might in principle indicate the presence of apical sorting information in these mutants as a truncation of the last eight cytoplasmic residues of CD1d led to truly non-polarized sorting (50/50 apical/basolateral distribution, Fig. 2). However, we cannot draw any conclusion until more is known.
As it was required to mutate residues both within the tyrosine-based internalization signal in addition to the residues in the putative leucine-based basolateral sorting signal in order to disrupt basolateral sorting of CD1d, it is tempting to conclude that the cytoplasmic tail of CD1d contains two overlapping basolateral sorting signals. Efficient basolateral sorting by the tyrosine-based signal required an intact tyrosine residue and one of the bulky hydrophobic residues at the position ϩ3 or ϩ4. A basolateral tyrosine sorting signal with a hydrophobic residue in position ϩ4 has to our knowledge not been reported before, but in most studies the context of tyrosine signals was not investigated in detail. Our data thus suggest that we have identified a new context for a tyrosine-based basolateral signal. It is noteworthy that endosomal sorting mediated by the same tyrosine-based signal was abrogated by single mutation of the ϩ3 valine only (10). This strongly suggests that the context requirements for internalization and basolateral sorting by the tyrosine-dependent signal are different.
The leucine-based basolateral signal is also special, as this type of sorting signal has so far not been found at the very end of naturally occurring molecules. However, it is reported that internalization of certain molecules was still efficient and dependent on a leucine-based signal when all residues C-terminal to the signal were deleted. This has been described, for example, for the dileucine signal in CD3␥ (33) and in the interleukin-6 signal transducer gp130 (34). It is therefore not surprising that the leucine-based basolateral sorting signal at the very C-terminal end of the CD1d molecule is functional.
We chose to investigate the steady-state distribution of CD1d constructs because our previous studies have shown that newly synthesized CD1d molecules use several hours to reach the cell surface (10), and detectable levels of metabolic labeling were achieved only after 2-3 h of labeling. Therefore, measurements of surface appearance of the newly synthesized CD1d molecules would be difficult to perform and interpret. Internalization and possible recycling of CD1d molecules at the cell surface might also be relatively rapid adding to the problem. We nonetheless believe that CD1d molecules are sorted directly from the TGN to the cell surface like most newly synthesized proteins in the MDCK cells (35). Indirect evidence for this is that CD1dYA and CD1dVA mutants have lost their internalization signal (10) but were still sorted basolaterally (Fig. 2). Had the newly synthesized CD1d molecules been initially delivered to the apical membrane and subsequently internalized and transported to the basolateral membrane, a mutation in the internalization signal should have led to a predominantly apical or non-polarized distribution of the CD1dYA and CD1dVA mutants, which is not the case.
Endosomal tyrosine-and leucine-based sorting signals have been shown to interact with adaptor molecules (e.g. Refs. 36 -42). It is generally accepted that tyrosine-based signals may interact with the medium chain of adaptor complexes (36,37,40) whereas the leucine signals have been reported to bind the ␤ chain of AP2 (41) and/or the medium chains of AP1 and AP2 (42,43). A study by Ohno et al. (26) has identified a novel medium chain (1B) that is only expressed in polarized cells. It was recently shown that this chain (which is able to replace 1A in AP1) may reconstitute polarized sorting in a cell line lacking this molecule (25). Furthermore, the AP4 adaptor may be involved in basolateral sorting of molecules containing both tyrosine-and leucine-based sorting signals. 2 This leads to the suggestion that internalization/endosomal and basolateral sorting signals may be recognized by different adaptors. Our finding that the context of the CD1d tyrosine signal is different for internalization and basolateral sorting and that its leucinebased signal is only functional for polarized sorting lends support to the existence of separate adaptor machineries for internalization and basolateral sorting.
At this point, we can only speculate why there are two basolateral signals within the CD1d molecule. This redundancy is not unique as other molecules also contain more than one polarization signal. For instance, two basolateral signals have been found in the low density lipoprotein receptor (29), and the complex consisting of major histocompatibility complex class II and invariant chain contains no less than four different basolateral signals (14). Separate sorting signals in these molecules may in principle bind more than one adaptor at the same time. In contrast, the two basolateral sorting signals in CD1d are overlapping, and steric hindrance will most likely not allow that they function simultaneously. Because only the tyrosine, but not the leucine-based signal, is involved in internalization of CD1d, it is clear that the signals are able to interact with different components of the intracellular machinery, but the precise mechanisms of such interactions remain to be elucidated.