The Leucine-based Sorting Motifs in the Cytoplasmic Domain of the Invariant Chain Are Recognized by the Clathrin Adaptors AP1 and AP2 and their Medium Chains*

Recognition of sorting signals within the cytoplasmic tail of membrane proteins by adaptor protein complexes is a crucial step in membrane protein sorting. The three known adaptor complexes, AP1, AP2, and AP3, have all been shown to recognize tyrosine- and leucine-based sorting signals, which are the most common sorting signals within membrane protein cytoplasmic tails. Although tyrosine-based signals are recognized by the μ-chains of adaptor complexes, the subunit recognizing leucine-based sorting signals is less clear. In this report we show by surface plasmon resonance that the two leucine-based sorting signals within the cytoplasmic tail of the invariant chain bind independently from each other to AP1 and AP2 but not to AP3. We also show that both motifs can be recognized by the μ-chains of AP1 and AP2. Moreover, by using monomeric as well as trimeric invariant chain constructs, we show that adaptor binding does not require trimerization of the invariant chain.

The major histocompatibility complex (MHC) 1 class II molecules are expressed on antigen-presenting cells and present primarily antigenic peptides from exogenous proteins to T helper cells. The invariant chain (Ii), a type II transmembrane protein, associates with MHC class II molecules in the endoplasmic reticulum and prevents binding of immunogenic peptides to these molecules while escorting them to endosomal compartment where antigen-loading may take place (for review, see Refs. [1][2][3]. Newly synthesized Ii is transport-competent as a trimeric complex (4), which is formed by its luminal domains (5,6). Deletion analysis has shown that information contained in the cytoplasmic domain of Ii is necessary and sufficient for target-ing of the protein to endosomal/lysosomal compartments (7,8). Site-directed mutagenesis localized two signals in the cytoplasmic tail that independently are able to sort Ii to endocytic compartments; one signal contains a leucine and isoleucine (LI) residues at positions 7 and 8 (9), and the second contains a methionine and leucine (ML) residues at positions 16 and 17 of the tail (10, 11). Additionally, the signals are part of an ␣-helix/ turn, and acidic residues N-terminal of both sorting motifs are required for efficient sorting (12,13).
Whether Ii reaches the endocytic pathway directly from the trans-Golgi network or indirectly via the plasma membrane (PM) by fast internalization is still a matter of debate. A dynamin mutant led to accumulation of MHCII/Ii at the plasma membrane (14), and each of the sorting motifs allowed a fast internalization from the PM (10, 15), suggesting an indirect transport route. On the other hand a direct pathway is supported by the finding that only 20% of a fusion protein with the Ii tail were ever exposed on the cell surface (11) and by the observation that the transport of Ii-MHC-II complexes to late endosomes is not inhibited by concanamycin B, which blocks trafficking between early and late endosomes (16). In conclusion the present data support a dual route to endosomes involving both direct transport and routing via the plasma membrane.
Both leucine-and tyrosine-based sorting motifs mediate internalization of membrane proteins from the PM and may sort membrane proteins to endosomes/lysosomes and to the compartment for peptide loading (for review see Refs. 3, 17, and 18). Different experimental approaches have shown that the medium chains of the three known adaptor complexes, AP1, AP2, and AP3, recognize tyrosine-based sorting motifs (19,20). Recognition of leucine-based sorting signals by adaptors is less well characterized. It was recently proposed that these signals bind to the ␤-subunits of AP1 and AP2 (21). When peptides containing leucine-based signals and a photoactivable crosslinker were incubated with adaptors, they were found to crosslink to the ␤-subunits of AP1 and AP2. In addition, also the -chains of AP1 and AP2 have been reported to bind to leucinebased signals. This was found using a random phage display library and by incubating recombinant adaptor -chains with immobilized peptides that were derived from the Ii cytoplasmic tail (22,23). Apart from AP1 and AP2, leucine-based sorting motifs also bind to AP3, e.g. the leucine-based sorting motifs of the lysosomal membrane protein LIMP-II and of the melanosomal membrane protein tyrosinase bind to AP3 but not or only poorly to AP1 and AP2 (24). However, it is not known which of the AP3 subunits mediates binding.
In the present study, we used the trimeric invariant chain containing the two leucine-based sorting motifs to analyze binding to adaptor complexes. We provide biochemical evidence that the two leucine-based signals can interact independently of each other with AP1 and AP2 but not with AP3. Furthermore, our experiments support the concept that the leucinebased signals of Ii are recognized by the adaptor -chains.

EXPERIMENTAL PROCEDURES
Construction of Plasmids-Plasmid pET3a.Ii⌬TM-His (25) encodes a soluble C-terminal histidine-tagged Ii molecule lacking the transmembrane domain. A mutant Ii⌬TM molecule lacking the cytoplasmic domain (Ii⌬TM⌬CT) was generated by cleavage of pET3a.Ii⌬TM-His with NdeI and PstI. A double-stranded oligonucleotide (5Ј-TATGGGTC-GACTGGACAAACTGACAGTCACCTCCCAGAACCTGCA-3Ј) was inserted to generate an ATG start codon in front of the complete Ii luminal domain.
The two leucine-based motifs in the Ii cytoplasmic tail were changed by overlap extension using the polymerase chain reaction (26) on pET3a.Ii⌬TM-His. Leucine and Isoleucine in positions 7 and 8 and methionine and leucine in positions 16 and 17 of Ii⌬TM were replaced by alanines generating LI 3 AA and ML 3 AA, respectively. Similarly a double mutant termed LI,ML 3 AA was created. The resulting plasmids were used to produce soluble Ii⌬TM-molecules in Escherichia coli BL21(DE3) (Novagen, Madison/WI).
Expression and Purification of Soluble Ii⌬TM and Its Mutant Forms-Expression of Ii⌬TM was induced by culturing cells in 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 1.5 h at 37°C. Bacteria were collected by sedimentation, resuspended in 50 mM Tris/HCl, pH 7.5, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and disrupted by freeze/thawing in liquid nitrogen and lysozyme treatment (200 g/ml, 15 min on ice). The bacterial lysates were first incubated with 10 units/ml DNase I for 15 min at room temperature, then adjusted to 500 mM NaCl and incubated for another 15 min on ice. Cell debris were removed by centrifugation at 20,000 ϫ g for 30 min at 4°C. The supernatant contained most of the Ii⌬TM molecules.
For affinity purification the extracts were adjusted to 2 mM MgCl 2 , 20 mM imidazole and loaded onto a Ni 2ϩ /NTA-agarose column (Qiagen, Hilden, Germany). After extensive washing with 10 column volumes of 50 mM Tris/HCl, pH 7.5, 20 mM imidazole, and 0.5 M NaCl then 1 M NaCl then 20 mM Hepes/KOH, pH 7.3, 90 mM KCl, and 2 mM MgCl 2 , the His-tagged Ii⌬TM molecules were eluted with 250 mM imidazole in the same buffer.
The eluate was concentrated to half the volume by ultrafiltration with a 30-kDa cut-off filter (Centriprep-30 by Amicon, Witten, Germany), and proteins were separated by gel filtration on a HiLoad 26/60 Superdex 200 prep grade column (Amersham Pharmacia Biotech). The flow rate was adjusted to 2.6 ml/min. The sample was eluted in 20 mM Hepes/KOH, pH 7.3, 90 mM KCl, and 2 mM MgCl 2 , and 4-ml fractions were collected. Ii⌬TM⌬CT eluted with a relative molecular mass of about 100 kDa, and Ii⌬TM and its mutants, with masses of about 120 kDa. Because Ii⌬TM and Ii⌬TM⌬CT have basic pIs, they were further purified by ion exchange chromatography on a Mono S HR 5/5 column (Amersham Pharmacia Biotech). Ii⌬TM molecules eluted from the column between 200 and 400 mM KCl. Using ultrafiltration in Centriprep-30, the protein was concentrated to 1-2 mg/ml, and KCl concentration was adjusted to 90 mM. Concentrated protein was shock-frozen and stored at Ϫ20°C.
Expression and Purification of Histidine-tagged Adaptor Medium Chains-DNA coding for mouse 1 and rat 2 was kindly provided by Dr. T. Kirchhausen (Harvard Medical School). The full-length medium chains were cloned in-frame into the type IV pQE30 vector (Qiagen) to express constructs containing a histidine tag at their N termini. The oligonucleotides used for polymerase chain reaction amplification were TCCCGGGGATCCATGTCCGCCAGCGCCGTCTAC and TCTAGGCA-AAGCTTTCACTGGGTCCGGACCTGATA for 1 and GAGCTCGGTA-CCATGATCGGAGGCTTATTCATC and TCTAGGCAAAGCTTCTAGC-AGCGGGTTTCGTAAAT for 2. Amplified constructs were purified with QIAEX II kit (Qiagen) and cloned into BamHI and HindIII (for 1) or KpnI and HindIII (for 2) sites of pQE30. Proteins were expressed in the bacterial strain M15 (Qiagen) according to the manufacturer's protocol. Both proteins formed inclusion bodies, which were solubilized in 6 M guanidinium hydrochloride containing 10 mM ␤-mercaptoethanol. Proteins were purified in one step under denaturing conditions on Ni 2ϩ -NTA resin (Qiagen) according to specifications of the manufacturer. Purified proteins were diluted to a concentration of 10 -20 g/ml and refolded in the binding buffer (0.1 M Tris, 5 mM EDTA, 0.1% Triton X-100, pH 7.5). Before the binding assay, proteins were centrifuged for 1 h at 100,000 ϫ g (Airfuge) to remove the insoluble material. His-tagged dihydrofolate reductase was expressed from the control plasmid pQE16 supplied with the kit and purified according to manufacturer's recommendations. Protein concentration was determined from Coomassie-stained gels by comparison with protein standards.
Expression and Purification of the GST Fusion Proteins-Cytoplasmic tails of the wild type Ii (Met-1 to Arg-30) and its L7A,L17A mutant were fused in-frame to the C terminus of the GST protein. The cytoplasmic tails of the Ii and the L7A,L17A mutant were amplified by the polymerase chain reaction using the full-length Ii and the double alanine mutant (11) as a template. The primers in both cases were TC-CCGGGGATCCATGGATGACCAGCGCGAC (5Ј primer, BamHI site) and CACGATGAATTCGCGGCTGCACTTGCTCTC (3Ј primer, EcoRI site). The amplified constructs were cloned into pGEX-2t vector (Amersham Pharmacia Biotech) using the BamHI-EcoRI sites, and the frame was verified by sequencing. Fusion proteins were expressed and purified as recommended in the Amersham Pharmacia Biotech manual. Briefly, BL21 cells carrying the constructs of interest were induced with 0.25 M isopropyl-1-thio-␤-D-galactopyranoside for 3 h and collected by centrifugation. The fusion proteins were released by a series of 15-s sonication bursts, purified on the GST-Sepharose (Amersham Pharmacia Biotech), and dialyzed overnight against phosphate-buffered saline, pH 7.4. A GST protein (Amersham Pharmacia Biotech) was also purified for use as a negative control in the subsequent experiments.
Cross-linking with 3,3Ј-Dithiobis(sulfosuccinimidylpropionate) (DTSSP)-To determine the oligomeric state of Ii molecules, the Ii oligomers were stabilized by chemical cross-linking using the dithiothreitol-cleavable homobifunctional cross-linking reagent DTSSP (Pierce). 0.25 mg/ml Ii polypeptides in 20 mM Hepes/KOH, pH 7.3, 90 mM KCl, and 2 mM MgCl 2 were incubated with increasing concentrations of DTSSP for 30 min at room temperature. After stopping the reaction with 50 mM glycine, the samples were divided into halves, trichloroacetic acid-precipitated, and analyzed by SDS-polyacrylamide gel electrophoresis in the presence or absence of dithiothreitol. After Western blotting, the Ii polypeptides were visualized by using the QIAexpress detection system (Quiagen Ab, Hilden, Germany).
Preparation of Adaptors-Clathrin-coated vesicles were prepared from bovine brain essentially as described by (27) except using Hepes instead of Mes buffer. The coat components were released from the membranes with 0.5 M Tris/HCl, pH 7, 2 mM EDTA (28) and separated from the vesicles by centrifugation at 240,000 ϫ g for 45 min at 4°C. The extract was adjusted to 20 mM Hepes/KOH, pH 7.3, 100 mM potassium acetate, 2 mM MgCl 2 and concentrated to 1-2 mg/ml protein using ultrafiltration in the stirring cell with a 10-kDa cut-off filter (Amicon, Witten, Germany). This clathrin-coated vesicle extract was frozen in liquid nitrogen and stored at Ϫ80°C.
For surface plasmon resonance studies, the clathrin adaptor complexes AP1 and AP2 were purified according to standard procedures from bovine or porcine brain (29). Briefly, clathrin-coated vesicles were purified from brain after homogenization and differential centrifugation. Adaptor complexes were released from clathrin-coated vesicles with 0.5M Tris, pH 7.8, plus 2 mM EDTA for 30 min at 4°C. After centrifugation at 100,000 ϫ g for 30 min, the material was applied to a Superose-6 column (2.5 ϫ 75 cm) connected to a fast protein liquid chromatography system at a flow rate of 0.5 ml/min. The column was equilibrated in 0.5M Tris-HCl, pH 7.8. Fractions containing adaptor complexes were identified by SDS-polyacrylamide gel electrophoresis. AP1 was separated from AP2 by hydroxylapatite chromatography exactly as described by Manfredi and Bazari (30). AP1 and AP2 were concentrated to 0.2 mg/ml and dialyzed against 20 mM Hepes, pH 7.0, 150 mM NaCl, 2 mM MgCl 2 , 50 mM KCl, 3 mM EDTA. This buffer was used for all experiments using surface plasmon resonance.
For analyzing AP3 binding, cytosolic fractions of pig brain derived from gel filtration were used. These AP3-containing fractions were shown to be devoid of AP1 and AP2 (for details, see Ref. 24) Surface Plasmon Resonance Interaction Analysis-The interaction between the different Ii constructs and adaptors was analyzed in real time by surface plasmon resonance (31) using a BIAcore 2000 Biosensor (BIAcore AB). Ii constructs were immobilized via their hexahistidine tag to the surface of a NTA sensor chip according to the manufacturer's instruction. Buffers that were used include the running buffer A (20 mM Hepes-NaOH, pH 7, 150 mM NaCl, 10 mM KCl, 2 mM MgCl 2 , 50 M EDTA, 0.005% Surfactant P20), the dispenser buffer (buffer A at 3 mM EDTA), the NiCl 2 solution (500 M NiCl 2 in running buffer), and the regeneration solution (running buffer A at 0.35 M EDTA).
In brief, the flow cell was first washed with 20 l of the regeneration solution at a flow rate of 20 l/min to remove contaminating metal ions. Subsequently the NTA surface was saturated with nickel by injecting the NiCl 2 solution for 2 min at a flow rate of 20 l/min. Ii constructs (at 100 nM in running buffer) were then immobilized at a flow rate of 5 l/min until the base-line shift was around 2000 response units, corresponding to 2 ng/mm 2 .
GST fusion proteins of the Ii tail were immobilized to a CM5 sensor surface that was first coated with an anti-GST antibody (BIAcore AB) to a density of 5000 response units. GST-Ii fusion proteins were immobilized at a density of 1000 response units.
All interaction experiments were performed with buffer A (see above) at a flow rate of 20 l/min. When using isolated adaptor -chains, buffer A was adjusted to 0.5% Triton X-100. Association for 2 min was followed by dissociation for 2 min, during which buffer A was perfused. A short pulse injection (15 s) of 20 mM NaOH, 0.5% SDS was used to regenerate the sensor chip surface after each experimental cycle. The peptidederivatized sensor chips remained stable and retained their specific binding capacity for more than 100 experimental cycles of association/ dissociation and regeneration. AP1 and AP2 were used at 200 nM unless otherwise stated.
Determination of Kinetic Constants-The rate constants (k a for association and k d for dissociation) of the interaction between Ii construct and purified AP1 or AP2 were calculated by using the evaluation software of the BIAcore 2000. Association was determined 15-20 s after switching from buffer flow to adaptor solution to avoid distortions due to injection and mixing. The dissociation rate constants were determined 5-10 s after switching to buffer flow. After a first dissociation phase, for around 30 s further dissociation of adaptors was very low.
The association constant k a , the dissociation constant k d , and the calculation of the equilibrium constant K D ϭ k d / k a were determined by using the BIAevaluation software version 1.2, assuming a first order kinetic A ϩ B ϭ AB. The model calculates the association rate constant k a and the steady state response level Req by fitting data to the equation, where t is the time in s, Req is the steady state response level, and C is the molar concentration of adaptors in the injection solution. The steric interference factor N that describes the valency of the interaction between the adaptors and the Ii constructs was set to 1. The dissociation rate constant k d was determined by fitting data to the equation, where R 0 is the response level at the beginning time t 0 of the dissociation phase. This model, which has recently been applied to describe adaptor tail interaction (32), is described in more detail elsewhere (33,34). It should be noted that the above-described models allow the determination of rate constants without reaching equilibrium during the experimental cycle.

Isolation of Soluble Trimeric
Ii-To study the interaction between the cytoplasmic tail of the Ii and adaptor complexes, we generated soluble Ii molecules devoid of their membranespanning region (Ii⌬TM). Furthermore we produced mutant Ii molecules that lacked the entire cytoplasmic tail (Ii⌬TM⌬CT) or contained alanine residues instead of either one (LI 3 AA; ML 3 AA) or both leucine-based motifs (LI,ML 3 AA) (Fig. 1). To facilitate purification, we added a hexahistidine tag to the C termini. The different Ii constructs were expressed in E. coli and purified using a Ni 2ϩ -NTA-agarose matrix. All Ii constructs expressed in E. coli were found to be largely soluble and not aggregated. After separation by gel filtration, they showed estimated molecular masses of about 120 kDa for Ii⌬TM and 100 kDa for Ii⌬TM⌬CT ( Fig. 2A). Because the monomeric Ii⌬TM⌬CT and Ii⌬TM molecules have molecular weights of about 26 and 28 kDa, respectively, we conclude that the Ii⌬TM molecules, like the authentic Ii molecules, form trimers (5,6,35). To more directly investigate the oligomeric assembly of Ii⌬TM and Ii⌬TM⌬CT, we cross-linked the subunits with the cleavable cross-linker DTSSP and separated the molecules under reducing and nonreducing conditions (Fig. 2, B and C).
Ii⌬TM forms under nonreducing conditions mainly dimers, even in the absence of the cross-linker. In contrast, Ii⌬TM⌬CT forms monomers. After cross-linking with 0.1 mM DTSSP, both Ii⌬TM and Ii⌬TM⌬CT migrated as dimers and trimers.
AP1 and AP2 Binding to the Cytoplasmic Tail of Immobilized Ii Polypeptides-To analyze the interaction between the Ii cytoplasmic tail and AP1 or AP2, we used a biosensor system monitoring surface plasmon resonance. This method has been used successfully in other studies on the interaction between adaptors and cytoplasmic tails of the epidermal growth factor receptor, hemagglutinin, lysosome-associated membrane protein-1, and mannose 6-phosphate receptor (24,29,32,36,37). The different Ii forms were immobilized via their hexahistidine tag to the surface of a NTA sensor. When binding of purified AP1 and AP2 to Ii⌬TM and Ii⌬TM⌬CT were tested, very low binding was found for the tailless mutant, whereas both adaptors bound with high affinity to Ii⌬TM. An equilibrium constant of 50 nM was determined for AP1 and 200 nM for AP2. The association constant (k a , M Ϫ1 /s Ϫ1 ) was 4.4 ϫ 10 4 for AP1 and 3.1 ϫ 10 4 for AP2; the dissociation constant (k d , s Ϫ1 ) was 2.2 ϫ 10 Ϫ4 for AP1 and 6.2 ϫ 10 Ϫ4 for AP2 (Fig. 3).
The Ii⌬TM leucine mutants showed reduced binding in comparison with Ii⌬TM (see Table IA). When both leucine-based motifs were mutated, binding to AP1 and AP2 was too low to allow calculation of the rate constants. The on-rates for AP1 and AP2 binding to the ML 3 AA or LI 3 AA mutants were 1.3 to 4 times slower as compared with those determined for Ii⌬TM. The off-rates for these mutants were 1.5 times faster for AP2 and 0.8 times slower for AP1 as compared with Ii⌬TM. These experiments demonstrate that either of the leucinebased motifs can mediate high affinity binding of AP1 and AP2 and that their substitution by alanine residues results in complete loss of AP1 and AP2 binding.
Binding of GST-Ii to AP1 and AP2-The experiments described above for adaptor binding to Ii were performed with soluble Ii trimers immobilized to the sensor surface. We also analyzed binding of adaptors with monomeric GST-Ii tail fusion proteins immobilized to the sensor surface. AP-1 bound to a wild-type GST-Ii tail fusion protein with a K D of 114 nM and to AP-2 with a K D of 250 nM. Two controls, a GST-Ii tail fusion protein in which both leucine-based motifs were replaced by alanines and GST alone, did not bind AP-1 nor AP-2 (see Fig. 4 and Table IB). In conclusion, oligomerization of Ii into trimers is not necessary for the interaction with AP1 or AP2; also Ii monomers can bind to AP1 and AP2.
Binding of Ii to the -Chains of AP1 and AP2-Although AP-1 and AP-2 have been shown to bind to the leucine-based sorting motifs of several membrane proteins such as cationindependent mannose 6-phosphate receptor, CD3-␥, CD4, and glucose transporter 4 (21,37,38), it is still a matter of debate which of the adaptor subunits mediates this interaction. We have tested the binding of -chains by passing recombinant 1 and 2 over biosensor surfaces derivatized with Ii. Both 1 and 2 bound to the wild-type Ii tail with kinetics very similar to those obtained for the fully assembled AP-1 and AP-2 complexes (Table I,  The wild-type Ii tail or the L7A,L17A mutant tail were expressed and purified as monomeric GST fusion proteins. For surface plasmon resonance analysis, the purified fusion proteins were diluted into buffer A and captured by an anti-GST monoclonal antibody (BIAcore AB) that was covalently immobilized on a CM5 sensor surface (see "Experimental Procedures" for details). Subsequent to capture of the Ii, AP1 or AP2 binding was monitored. Note that the monomeric GST-wild type (wt) Ii tail bound AP1 and AP2 with high affinity, whereas adaptor binding to the GST-Ii L7A,L17A mutant tail construct was abolished. RU, response units.

FIG. 5.
Binding of AP1/AP2-and AP3-enriched cytosolic fractions to Ii. The tail constructs of Ii and tyrosinase were immobilized on a sensor surface and analyzed for their binding capacity for AP1/AP2and AP3-enriched cytosolic fractions. Note that the sensorgrams shown represent the difference of binding to wild type and mutant forms of the tail constructs (see Fig. 1). AP3 binding was only detectable for tyrosinase, whereas Ii strongly bound AP1/AP2 but not AP3. RU, response units.

FIG. 2. Soluble Ii polypeptides form trimers.
Bacterially synthesized Ii polypeptides were purified via Ni 2ϩ -NTA-agarose affinity chromatography followed by gel filtration to separate Ii trimeric complexes. A, scheme of elution profile from gel filtration column. When Ii⌬TM and Ii⌬TM⌬CT were fractionated on a Hiload 26/60 Superdex 200 prep grade column, high molecular weight aggregates (gray box) always eluted before the relevant protein peaks, which are indicated by the curves and which where used for binding studies. The column was calibrated with dextran blue (M r 2,000,000), ferritine (M r 440,000), aldolase (M r 158,000), albumin (M r 67,000), ovalbumin (M r 43,000), chymotrypsin (M r 25,000), and ribonuclease A (M r 13,700). Ii⌬TM (B) and Ii⌬TM⌬CT (C) were incubated with increasing concentrations of the cross-linking reagent DTSSP. The reaction products were analyzed by SDS-polyacrylamide gel electrophoresis (12.5%) in the absence (Ϫ) and presence (ϩ) of dithiothreitol (DTT, 3 mM) followed by a Western blot and staining of Ii polypeptides with Ni 2ϩ -NTA conjugate. The Ii trimer (Ii ϫ 3), dimer (Ii ϫ 2), monomer (Ii ϫ 1) are indicated.
FIG. 3. Surface plasmon resonance analysis of AP1 and AP2 binding to trimeric Ii. Trimeric Ii tail constructs were immobilized on a NTA sensor surface and subsequently probed for binding of purified AP1 and AP2 (200 nM). High affinity binding of AP1 and AP2 required the wild type Ii tail. Binding of AP1 and AP2 to a tailless Ii mutant (Ii⌬TM⌬CT) was negligible. See Table IA for determination of the rate constants of AP1 and AP2 binding. RU, response units.
leucine-based sorting signals to AP1 and AP2, the leucinebased motifs in the cytoplasmic tails of the lysosomal membrane protein LIMP-II and the melanosomal enzyme tyrosinase are known to bind AP3 but not AP1 or AP2 (24). We tested the possibility of AP-3 binding to Ii by incubating the immobilized Ii tail constructs with cytosolic fractions enriched in either AP-1/AP-2 or AP-3. When Ii⌬TM was incubated with a cytosolic fraction enriched in AP-1 and AP-2, we observed strong binding. In contrast, no binding was observed when Ii⌬TM was incubated with an AP-3-enriched fraction (Fig. 5). As a control for the binding activity of the AP3-enriched fraction, we show that tyrosinase strongly binds AP-3. We therefore conclude that Ii, although it contains two leucine-based sorting motifs, does not bind AP-3.

DISCUSSION
In this study we show that the Ii is able to bind the cytosolic adaptors AP1 and AP2 with high affinity. The binding of AP1 is about 4-fold stronger as compared with AP2, mainly due to a lower dissociation rate. The rate constants were determined using a Biosensor surface to which soluble forms of the Ii lacking the transmembrane domain were immobilized. Deletion of the cytoplasmic tail abrogated the binding of adaptors, clearly showing that the cytoplasmic tail of Ii mediated binding.
Ii is transported as a homotrimer and, together with MHC class II ␣and ␤-chains, as a heterononameric complex to endosomal compartments (4). The soluble forms of the Ii used in this study behave in solution also as trimers. However, trimerization of Ii is not required for adaptor binding. Monomeric fusion proteins of the cytoplasmic tail of Ii with GST bound AP1 and AP2 almost as strong as the trimeric forms. Thus, in vivo more than one adaptor molecule may bind to the oligomeric Ii complexes. The binding curves obtained by surface plasmon resonance fitted best when assuming binding of a single adaptor complex per Ii trimer, suggesting a 1:1 stoichiometry for adaptor-Ii complexes.
The membrane distal LI (position 7 and 8) and membrane proximal ML (position 16 and 17) signal have been shown in in vivo studies to be critical for the sorting of the Ii to the endosomal pathway (9 -11). The two signals were therefore likely candidates for mediating AP1 and AP2 binding. Indeed, substitution of the LI and ML signals by alanine residues abolished binding of adaptors to the soluble trimeric form of the Ii as well as to the monomeric GST-Ii tail fusion protein. When either one of the two signals was substituted, the affinity to AP1 and AP2 decreased only moderately. This is in line with the observation that both signals function independently. In vivo studies had further indicated that the LI signal is more efficient than the ML signal in mediating rapid internalization from the plasma membrane (12). This is reflected in the rate constants for binding of AP2 (and AP1) to trimeric Ii, such that substitution of the LI signal had a greater effect than mutation of the ML signal.
Leucine-based signals in the lysosomal membrane protein LIMP-II and the melanosomal enzyme tyrosinase have recently been found to interact with AP3 (24), an adaptor complex found in association with trans-Golgi network membranes as well as with more peripheral membranes, including part of the endosomal system (39,40). Neither of the two leucine-based signals in the Ii bound AP3. This is in agreement with in vivo data showing that down-regulation of AP3 by microinjection of antisense DNA did not alter the localization of Ii and MHC class II. 2 The structural features that specify the selective affinity of leucine-based signals for AP3 (as in LIMP-II and tyrosinase (24)) or for AP1 and/or AP2 (as in Ii, CD4, CD3-␥, glucose transporter 4 or the mannose-6-phosphate receptors (21,37,41)) remain to be defined.
It is still a matter of debate by which subunit(s) of the adaptor complexes the leucine-based signals are recognized. The leucine-based signals of CD3␥ was shown to interact with the ␤ 2 -chain of AP2 using a photocross-linking approach (21), whereas the reports on the binding to the 1and 2-chains of AP1 and AP2 are conflicting. Ohno et al. (42) failed to see an interaction between 1 and 2 and a leucine-based signal in the yeast two-hybrid system, whereas Rodionov and Bakke (23) demonstrated binding of 1 and 2 to immobilized peptides containing the leucine-based signals of Ii. Here we show that isolated 1and 2-chains bind to soluble trimeric forms of the Ii with affinities strikingly similar to those found for the fully assembled heterotetrameric AP1 and AP2. The binding is strictly dependent on the two leucine-based signals. Loss of a single signal has differential effects on 1 and 2 binding, which were also seen for the binding of AP1 and AP2 and for Ii sorting in vivo. Thus, binding of leucine-based signals to medium chains can fully account for their binding to AP1 and AP2. Our data show that binding of Ii to AP1 and AP2 was dependent on the two leucine-based signals that are also the major determinants for sorting of Ii in vivo. In this context it has to be noted that in vivo studies on the role of AP1 for sorting of Ii have provided conflicting data. Although overexpression of dominant negative forms of the clathrin heavy chain failed to affect Ii localization (43), overexpression of Ii increased recruitment of AP1 to Golgi membranes (44). The high affinity binding of Ii for AP1 supports the view that multimeric complexes of Ii with MHC class II ␣and ␤-chains are sorted at the trans-Golgi network into AP1/ clathrin-coated vesicles, allowing direct transfer from the Golgi to endosomal membranes. The leucine-based sorting signals of Ii are also responsible for polarized sorting to the basolateral PM and basolateral endosomes, and it is the same signal context that is recognized both for polarized sorting and PM internalization (45,46). Our results, showing that both AP1 and AP2 recognize the same signals, is thus in accordance with the polarized studies if AP1 mediates basolateral sorting, but this has not been proven. The ability of the leucine signals to bind to AP2 may have a role in the efficient internalization of Ii alone and Ii-MHC class II complexes from the plasma membrane (10, 15). The details of the regulation of the same signals with two adaptor complexes is, however, not answered, but it is tempting to speculate that this is due to differential binding or local factors at the different sorting stations where these adaptors are present.