Mechanism of Interaction between Leucine-based Sorting Signals from the Invariant Chain and Clathrin-associated Adaptor Protein Complexes AP1 and AP2*

The cytoplasmic tail of the invariant chain contains two leucine-based sorting signals, and each of those seems sufficient to route the invariant chain to its intracellular destination in either normal or polarized cells. It is believed that the intracellular routing of the invariant chain is mediated by its interactions with the clathrin-associated adaptor protein complexes AP1 and AP2. We (1) have previously demonstrated the in vitrointeractions between the cytoplasmic tail of the invariant chain and AP1/AP2 complexes. These interactions were specific and depended on the critical leucine residues in the invariant chain's sorting signals. In the present study, we decided to investigate the molecular mechanism of these interactions. To this end, we constructed a set of glutathioneS-transferase fusion proteins that contained the intact cytoplasmic tail of the invariant chain and its various mutants to define residues important for its interactions with AP1 and AP-2. Our results demonstrated the importance of several residues other than the critical leucine residues for such interactions. A strong correlation between in vitro binding of AP2 to the invariant chain and in vivo internalization of the invariant chain was observed, confirming the primary role of AP2 in recognition of endocytic signals. In addition, we demonstrated different requirements for AP1 and AP2 binding to cytoplasmic tail of the invariant chain, which may reflect that the different sorting pathways mediated by AP1 and AP2 involve their recognition of the primary structure of the sorting signal.

The major histocompatibility complex (MHC) 1 class II and its associated invariant chain (Ii) molecule play an important role in the immune response. The heterodimeric (␣ and ␤ chain) MHC class II molecule is expressed on antigen-presenting cells, and its function is primarily to present exogenous antigenes to helper T-cells. Ii is a type II transmembrane protein associated with MHC class II molecules in the endoplasmic reticulum. It is believed that Ii has a dual role of preventing endogenous polypeptides from binding to the MHC class II groove and mediating the sorting of the MHC class II to the endosomes either directly or via the plasma membrane in clathrin-coated vesicles (CCVs) (2)(3)(4).
The cytoplasmic tail of Ii consists of 30 amino acids. Deletion analysis has shown that this cytoplasmic tail was necessary and sufficient for targeting of the nonameric ␣␤Ii complex to peptide loading compartments (5,6). Within the cytoplasmic tail of Ii two leucine-based signals have been identified: the membrane-distal sorting signal (Leu-Ile residues at positions 7 and 8) and the membrane-proximal signal (Met-Leu residues at positions 16 and 17). Either signal is independently sufficient for endosomal localization of Ii (7,8). The two leucine-based endosomal sorting signals within the cytoplasmic tail of Ii were both individually sufficient for basolateral targeting and internalization of Ii in MDCK cells (9,10).
There is increasing evidence that leucine sorting signals (as well as related tyrosine sorting signals) can interact directly with the clathrin-associated adaptor protein complexes AP1, AP2, AP3, and AP4 (11)(12)(13). It has to be noted that the precise amino acid requirements for the in vivo basolateral targeting and internalization are not always identical for either leucine or tyrosine sorting signals (14 -17). This indicates that amino acids in the vicinity of the critical residues determine whether the motif functions as a signal for endocytosis, basolateral targeting, or both.
The clathrin-associated adaptor protein complexes can promote clathrin cage assembly, link clathrin to the membrane, and interact with membrane proteins that contain leucine and tyrosine signals for sorting into CCV (12,18). Each AP contains four polypeptides, called adaptins: two large chains of ϳ100 kDa, a medium chain of ϳ50 kDa, and a small chain of ϳ25 kDa. AP1 complexes are associated with CCVs derived from the trans-Golgi network (TGN), AP-2 complexes are associated with the endocytic CCVs, and the functional localization of AP-3 is less clear; it seems to be involved in direct transport between TGN and melanosomes/vacuoles and mediating sorting on endosomes, whereas functions of AP-4 is less well characterized, but seems to participate in basolateral sorting (13, 19 -24). All APs seem to recognize tyrosine sorting signals via their respective -subunit and are associated with specific populations of transport vesicles and confer distinct sorting properties onto these vesicles (13,18). It is less clear how AP complexes recognize leucine signals; one laboratory has reported interactions between several leucine signals and the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ Supported by a grant from the European Union (XCT 960058) and by a personal grant from the German Science Foundation.
ʈ ␤-subunit of AP1 complex using a cross-linking assay (25), whereas we have observed interactions between leucine signals and chains of AP1 and AP2 using either protein-protein interactions assay on magnetic beads (26) or surface plasmon resonance technique (1). It is believed that the interactions between sorting signals and the adaptor protein complexes mediates the sequestration of cargo membrane proteins into a specific type of transport vesicle for delivery to its intermediate or final destination.
We have previously shown that Ii binds in vitro to AP-1 and AP-2, but not to AP-3 (1). To investigate the mechanism of interactions between the leucine signals of Ii and AP-1 and AP-2 adaptor protein complexes, we constructed a set of GST fusion proteins containing the wild-type cytoplasmic tail of Ii and its various mutants. These GST fusion constructs were expressed in Escherichia coli, and AP1/AP2 adaptor binding to these constructs was studied with surface plasmon resonance. This technique has been successfully used previously to detect binding of various adaptor protein complexes to the cytoplasmic tails of a number of proteins containing leucine or tyrosine signals (27)(28)(29)(30)(31), as well as in our previous experiments (1).

MATERIALS AND METHODS
DNA Constructs-The cytoplasmic tails of the wild-type Ii (Met-1 to Arg-30) and its various mutants (see Table I) were fused in-frame to the C terminus of GST using BamHI and EcoRI sites of pGEX-2T vector (Amersham Biosciences), practically as described elsewhere (1). The composition of all constructs was verified by sequencing.
Expression and Purification of GST Fusion Proteins-Fusion proteins were expressed and purified as recommended by the manufacturer (Amersham Biosciences). Briefly, BL21 cells carrying the constructs of interest were induced with 0.25 M isopropyl-1-thio-␤-Dgalacropyranoside for 3 h and collected by centrifugation. The fusion proteins were released by a series of 15-s sonication steps or by a Bugbuster kit (Novagen) and then purified on GST-Sepharose (Amersham Biosciences). The purity and size of the proteins were verified on 12% SDS-PAGE gels and then the proteins were dialyzed overnight against BIA buffer (20 mM, HEPES pH 7, 150 mM NaCL, 10 mM KCl, 2 mM MgCl 2 , 0,2 mM dithiothreitol). The GST protein without a fusion partner was also purified and used as a negative control for the biosensor experiments.
Preparation of AP1 and AP2-AP1 and AP2 were prepared from pig brain essentially as described elsewhere (27,30). Briefly, clathrincoated vesicles were purified from brain after homogenization and differential centrifugation. The adaptor proteins were released from clathrin-coated vesicles with 0.5 M Tris/HCl, pH 7.0, and applied to a Superose-6 column (2.5 ϫ 75 cm, equilibrated in the same buffer) connected to a fast protein liquid chromatography system at a flow rate of 0.3 ml/min. Fractions containing AP1 and AP2 were identified by SDS-PAGE and separated from each other by subsequent hydroxylapatide chromatography as described elsewhere (32). Fractions containing either AP1 or AP2 were dialyzed against 20 mM HEPES, pH 7.0, 150 mM NaCl, 2 mM MgCl 2 , 10 mM KCl, 0.2 mM dithiothreitol (BIA buffer), which was used for all experiments using surface plasmon resonance.
Surface Plasmon Resonance-The interaction between the different Ii constructs and adaptors was analyzed in real time by surface plasmon resonance (33) using a BIAcore 3000 biosensor (BIAcore AB). Ii constructs were immobilized via their GST moiety of the GST-Ii chimera to the surface of a CM5 sensor chip coated with anti-GST antibodies. The peptides were coupled to a CM5 sensor chip via their primary amino groups exactly according to the manufacturer's instructions. The subsequent interaction experiments were performed at a flow rate of 20 l/min. Association was recorded for 2 min during which adaptor proteins at different concentrations were injected and followed by recording dissociation for 2 min during which buffer was perfused. A short pulse injection (15 s) of 20 mM NaOH, 0.5% SDS was used to regenerate the surface after each experimental cycle. The anti-GST surface retained its binding capacity for at least 15 cycles of association, dissociation, and regeneration. AP1 and AP2 were used at concentrations ranging from 20 to 200 nM.
Determination of Kinetic Rate Constants-The association and dissociation constants k a and k d for the interactions were calculated by using the evaluation software of the BIAcore 3000. The mathematical models used are described in more detail elsewhere (34,35). In brief, the association was determined after 15-20 s following the switch from buffer solution to adaptor solution to avoid distortions due to injection and mixing. The dissociation rate constants were determined after 5-10 s following the switch to buffer solution. After a rapid (ϳ30 s) dissociation phase of adaptor from Ii-GST, the dissociation kinetics decreased to a low rate. 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 BIA evaluation software version 1.2, assuming a first order kinetic A ϩ B ϭ AB. Relative binding values were then calculated from the K D values.

RESULTS
To investigate the effect of amino acid residues surrounding the critical leucine residues Leu-7 and Leu-17 on AP1 and AP2 binding, we generated a set of GST-Ii constructs (see Table I). Briefly, we studied the effect of residues surrounding the membrane-proximal (containing the Leu-17 residue) signal in the context of L7A mutation to abolish AP1 and AP2 binding to the membrane-distal signal, and the effect of residues surrounding the membrane-distal signal was likewise studied in the context of L17A mutation. Fig. 1A illustrates the validity of this approach for AP1 binding; the GST fusion containing the wildtype bound AP1 specifically (K D ϭ 130 nM), whereas no binding was detected for the fusion protein containing double L7A,L17A mutation. This confirms our early results (1) and demonstrates that a mutation in the critical leucine residue is enough to knock out binding of AP1 to this signal completely. Likewise, no interactions between pure GST protein and AP1 were detected (not shown). The fusion constructs containing single mutations in ether Leu-7 or Leu-17 retained their ability to bind AP1 complex, confirming earlier data obtained with the full-length Ii constructs (1). Interestingly, AP1 binding to the L17A construct was around 60% higher than its binding to the wild-type Ii fusion construct, whereas AP1 binding to the L7A construct was almost 40% lower than binding to wild-type Ii fusion. This may reflect more favorable steric conformation of the membrane-distal leucine signal in the L17A fusion construct for AP1 binding compared with the wild-type fusion construct. It is not clear whether this is of importance to the in vivo interactions. Fig. 1B demonstrates that specific AP2 binding to wild-type Ii fusion construct (K D ϭ 234 nM) is likewise abolished in the double L7A,L17A mutant. Both the L7A and the L17A single mutant fusion constructs retained their ability to bind AP2, although at somewhat reduced rate (51 and 65% of the wildtype, respectively).
We then studied binding of AP1 and AP2 to a number of constructs that contained a mutation in one or more residues other than the critical leucine residues that have been shown to be important for in vivo internalization of Ii. As explained above, these mutant constructs were made in the context of either L7A or L17A mutation. Fig. 2A demonstrates binding of AP2 to such constructs. AP2 is found mostly on the plasma membrane, and there is increasing evidence that it mediates internalization of molecules with leucine or tyrosine sorting signal (13,18). From the earlier in vivo studies, it is known that the acidic residues upstream of the leucine residues form a part of the motif required for internalization from the plasma membrane for a number of molecules (36). In addition, a number of other residues have been shown to be important for internalization of Ii. Motta et al. (37) and Simonsen et al. (10) showed that Asp-2, Asp-3, and Gln-4, in addition to Leu-7 and Ile-8 were required for internalization of the Ii chain construct lacking the membraneproximal signal. However, a study by Pond et al. (38) demonstrated that only Asp-2 and Asp-3 were important for Ii internalization, but not Gln-4. In the case of membrane-proximal signal, available data show that Glu-12 is important for internalization in addition to Met-16 and Leu-17 (37,38). Additionally, Pro-15 was found to be important for internalization of a chimeric construct containing cytoplasmic tail of Ii (39).
As seen in Fig. 2A, mutation in acidic residue Asp-3 upstream of the membrane-proximal signal reduced AP2 binding significantly to less than 60% of the L17A level. Interestingly, mutation in Asp-2 reduced AP2 binding only to around 80% of the L17A level; this correlates with the in vivo internalization data as the construct carrying the D2A mutation internalized significantly slower than the wild type but still noticeably faster than the Asp-3 mutant (37,38). Furthermore, mutations in both Asp-2 and Asp-3 did not decrease AP2 binding compared with mutation in Asp-3 only, and this again correlates with the in vivo internalization data (37,38). This suggests that Asp-3 is more important for AP2 binding in vivo and subsequent internalization from the plasma membrane than Asp-2, possibly due to its proximity to the critical leucine residue in the membrane-distal signal. The membrane-proximal signal contains only one acidic residue Glu-12 upstream of its critical leucine. When this residue was mutated to alanine in the context of L7A mutation, the binding of AP2 to such construct was completely abolished.
Mutations in "secondary" residues that form the leucine signal (Ile-8 and Met-16) resulted in greatly decreased binding of AP2 complex, as expected (30 -35% of the respective L17A and L7A mutants). Interestingly, these mutations did not abolish such binding completely like mutations in both L7A and L17A did (compare Figs. 1A and 2A), indicating that the leucine residues may be more important for AP2 recognition of the signal than these residues.
We also studied AP2 binding to the construct containing Q4A mutation in the context of L17A. The decrease in AP2 binding was not significant (around 95% of AP2 binding to the L17A), supporting the report by Pond et al. (38) that stated that Gln-4 was not important for Ii internalization. Then, we determined AP2 binding to constructs containing D6A or R5A single mutations or R5A,D6A double mutation in the context of L17A. Neither Asp-6, nor Arg-5 has been shown to play a role in the internalization (in fact, the double R5A,D6A mutant has been shown to internalize more rapidly than the wild-type Ii (10,37). Therefore, these mutants served as a control to answer what happens to AP2 binding to Ii tail in vitro if any one or two residues are mutated to alanines. These constructs both retained high AP2 binding (around 85-101% of the L17A construct). We made another control construct for an unspecific mutation in the membrane-proximal signal. This construct contained the L14A mutation in the context of L7A, as it is has been shown that residue Leu-14 does not significantly affect internalization (37,38). This construct also retained high AP2 binding ( Fig. 2A). Finally, we tested the relevance of Pro-15 for AP2 binding, and the mutation to alanine (P15A) resulted in reduced AP2 binding to about 30% of L7A levels ( Fig. 2A), which fits well to the recently obtained reduced levels of internalization of this mutant in vivo (39).
In the next set of experiments, we determined binding of AP1 to the same constructs. Since AP1 is mostly found at the TGN and on some vesicular structures, it is believed to play a role in the clathrin-mediated transport from the TGN to endosomes (18,40). It should, however, be noted that recently obtained data suggest an important function of AP1 on endosomes in retrograde transport to the TGN (41). Transport from the TGN to the basolateral or apical domain of the plasma membrane in polarized cells such as MDCK appears to be direct, although in other cell lines it may involve an endosomal intermediate (42). Here we investigated the correlation between AP1 binding to Ii fusion constructs and the fate of previously reported corresponding mutants in polarized MDCK cells. Wild-type Ii is found almost exclusively at the basolateral domain of polarized   FIG. 1. AP1 and AP2 binding to wild-type Ii and L7A and L17A single and double mutant constructs. Ii-GST fusion proteins were immobilized on a sensor surface and monitored for AP1 or AP2 binding as described under "Materials and Methods." From the obtained sensorgrams, the rate constants of AP1 binding were calculated and are plotted as relative values to the AP1 binding to the wild-type Ii construct (A). Rate constants of AP2 binding were likewise calculated and plotted (B).
MDCK cells, and both leucine signals in Ii act as independent basolateral sorting signals in addition to the third basolateral signal, the precise structure of which has yet to be determined (10,43). Previous results have shown that for the membranedistal signal, residues Asp-2, Asp-3, Gln-4, and Ile-8 are important for basolateral sorting, but not Arg-5 or Asp-6 (10). For the membrane-proximal signal, residues Pro-15 and Met-16 were shown to be important for the basolateral sorting (43), whereas the effect of residues Glu-12 and Leu-14 has not been investigated in vivo. Fig. 2B shows binding of AP1 complex to various Ii fusion constructs. Residues Asp-2, Asp-3, Ile-8, Pro-15, and Met-16 were important for AP1 binding in addition to Leu-7 and Leu-17, as well as for the in vivo basolateral sorting. At the same time, residue Gln-4 that is important for basolateral sorting (10) was not important for AP1 binding (90% binding relative to the control), whereas mutant constructs D6A and R5A,D6A that are sorted basolaterally in vivo showed significant decrease in AP1 binding (30 -45% of the control levels). Interestingly, mutation in the acidic residue Glu-12 had no significant effect on AP1 binding (around 80% of the control), whereas the acidic residues Asp-2 and Asp-3 had an important effect (around 50% of the control levels). There is, however, no data on the importance of residue Glu-12 for basolateral sorting. Residue Leu-14 was apparently important for the AP1 binding, but its importance for basolateral sorting has not been, to our knowledge, investigated. and AP2 (filled bars) are plotted as values relative to the AP1 or AP2 binding, respectively, to either L7A (for constructs that had L7A mutation) or L17A (for constructs that had L17A mutation).

FIG. 2. Effect of various mutations
in Ii cytoplasmic tail on its binding to AP2 and AP1. Ii-GST fusion proteins were immobilized on a sensor surface and monitored for AP1 or AP2 binding as described under "Materials and Methods." From the obtained sensorgrams, the rate constants were calculated. In A, the constants for AP2 binding are plotted as values relative to the AP2 binding to either L7A (for constructs that had L7A mutation) or L17A (for constructs that had L17A mutation). In B, the constants for AP1 binding are plotted as values relative to the AP1 binding to either L7A (for constructs that had L7A mutation) or L17A (for constructs that had the L17A mutation). more important for binding to either AP1 or AP2 but not to both of these complexes. Residue Asp-6 was clearly important for AP1 binding (less than 50% of control), but not for AP2 binding (around 90% of control). Oppositely, residue Glu-12 was important for AP2 binding (binding was lost completely when the residue was mutated to alanine), but not for AP1 binding (over 80% of the control). There were also noticeable differences in AP1 and AP2 binding to the secondary residues of the leucine signals, Ile-8 and Met-16. For AP1, no binding was detected to I8A mutant, whereas binding to the M16A mutant was over 60% of the control levels. Binding of AP2 to I8A and M16A was 29 and 37%, respectively. This confirms that AP1 and AP2 have different mechanisms of recognition of Ii primary structure. DISCUSSION Our results demonstrate a very strong correlation between residues important for in vivo endocytosis of Ii and those important for the in vitro binding of Ii to AP2, providing further support to the critical role of or AP2 in endocytosis of proteins containing leucine sorting signals. All constructs harboring mutations in the residues important for endocytosis of the invariant chain were also shown to exhibit significantly decreased binding to AP2 in vitro. At the same time, constructs containing mutations in residues that were not important for the invariant chain endocytosis did not show a decrease in AP2 binding. Taken together, these data have demonstrated the significance of our in vitro approach and illustrated the strong correlation between our data presented here and the in vivo internalization data.
Our results demonstrated no convincing correlation between AP1 binding and basolateral sorting of Ii in polarized MDCK cells, despite such correlation for some residues. This indicates that AP1 may be either not involved in direct delivery on freshly synthesized Ii from the TGN to the basolateral or apical domain of the plasma membrane, or AP1 is not the only determining factor in such delivery. It is noteworthy that a study by Ohno et al. (44) has identified a second AP-1 medium chain (1B) that is only expressed in polarized cells. It was found that this chain reconstituted polarized sorting in a cell line lacking this molecule (45). It is thus conceivable that there are two distinct subpopulations of AP1 complexes, and only AP1B (containing 1B) is specifically involved in the sorting from the TGN to basolateral membrane, whereas AP1 is involved in general sorting from the TGN. To complicate the matter further, the AP4 adaptor complex may also play a role in basolateral sorting of molecules containing both tyrosine-and leucine-based sorting signals (20). This suggests that the interactions between sorting signals of and elements of the intracellular sorting machinery in polarized cells may be complex. Further studies should identify other factors that may participate in the polarized sorting of Ii and potentially other proteins containing leucine signals.
It is also of interest that previous studies have identified a stretch of ten cytoplasmic amino acids in the immediate proximity to the trans-membrane domain of Ii that contained an unknown basolateral signal in addition to the two leucine signals, although the precise structure of this signal has not been reported (9,43). Our in vitro binding data show that AP1 binding is completely abolished when both leucine signals are knocked out (Fig. 1A). Therefore, our results demonstrated no convincing correlation between AP1 binding and basolateral sorting of Ii in polarized MDCK cells, despite such correlation for some residues.
Finally, our results demonstrate differential requirements for AP1 and AP2 binding to Ii at the level of residues around the critical leucine residues, suggesting that there are different mechanisms of recognition of leucine signals by AP1 and AP2 complexes, conceivably at the level of the primary structure of the sorting signal. In this context, it is also interesting to note that recent studies have shown that posttranslational the modification of the AP2 -subunit by phosphorylation can determine the affinity for sorting signals and is critical for the rate of endocytosis in vivo. Since such modifications might occur that may be adaptor-specific and may be restricted to certain cell types offers another level of complexity in the control of adaptor binding to sorting signals (46,47).