Abstract
Most transmembrane proteins are selected as transport-vesicle cargo through the recognition of short, linear amino-acid motifs in their cytoplasmic portions by vesicle coat proteins. For clathrin-coated vesicles, the motifs are recognized by clathrin adaptors. The AP2 adaptor complex (subunits , 2, 2 and 2) recognizes both major endocytic motifs: Yxx motifs1 (where can be F, I, L, M or V) and [ED]xxxL[LI] acidic dileucine motifs. Here we describe the binding of AP2 to the endocytic dileucine motif from CD4 (ref. 2). The major recognition events are the two leucine residues binding in hydrophobic pockets on 2. The hydrophilic residue four residues upstream from the first leucine sits on a positively charged patch made from residues on the 2 and subunits. Mutations in key residues inhibit the binding of AP2 to 'acidic dileucine' motifs displayed in liposomes containing phosphatidylinositol-4,5-bisphosphate, but do not affect binding to Yxx motifs through 2. In the 'inactive' AP2 core structure3 both motif-binding sites are blocked by different parts of the 2 subunit. To allow a dileucine motif to bind, the 2 amino terminus is displaced and becomes disordered; however, in this structure the Yxx-binding site on 2 remains blocked.
In clathrin-coated vesicles (CCVs), cargo is selected mainly through the recognition of one of three types of signal by clathrin adaptors: the short, linear, transplantable peptide motifs that are widely found on general cargo; covalently attached ubiquitin molecules (reviewed in ref. 4); and the folded determinants on SNAP receptor (SNARE) proteins5, 6. The two major classes of transplantable motif used in CCVs are the Yxx and acidic dileucine motifs. The latter can be divided into two further classes: the [ED]xxxL[LI] motifs, which, along with Yxx motifs, are recognized by the heterotetrameric AP clathrin adaptor complexes on numerous different pathways (reviewed in ref. 7), and the DxxLL motifs, recognized by GGAs (Golgi-localizing, -adaptin ear homology domain, ARF-interacting proteins) for transport from the trans-Golgi network to endosomes8, 9. AP2 has a pivotal function in regulating the formation of endocytic CCVs destined for early endosomes. Although the mechanism of Yxx motif binding by APs has been characterized at the molecular level, there is no mechanistic insight into [ED]xxxL[LI] binding, although several recent studies have indicated that the binding site resides on the –2 heterodimer10, 11. Eight different 'acidic dileucine' peptides were tried in co-crystallization experiments with the 200-kDa AP2 core, and several produced poorly diffracting crystals. However, co-crystallization with the peptide RM(phosphoS)QIKRLLSE (Q peptide) from the T-cell cell-surface antigen protein CD4, and a variant of this peptide (RM(phosphoS)EIKRLLSE) (E peptide), resulted in crystals that diffracted to resolutions of 3.0 and 3.4 Å respectively. The structure of the Q peptide complex was solved by molecular replacement by using models derived from the closed structure3 (Fig. 1). Rebuilding of the and 2 helical solenoids was required because of alterations in their overall curvature (see below), but even in the initial electron density maps the 'dileucine peptide' could be seen clearly. The identity and orientation of the peptide was confirmed as described in Methods.
Figure 1: Structure of the AP2 adaptor core in complex with the dileucine peptide from CD4.
Orthogonal views of the dileucine motif liganded AP2 complex in ribbon (a, b) and in molecular surface (c, d) representations. In a and c the membrane is parallel to the upper face of the complex, and in b and d the complex is viewed through the membrane, with the membrane interacting surface facing upwards. The subunit is coloured dark blue, 2 green, 2 pale blue, N-2 purple and C-2 mauve. Atoms in the the dileucine motif peptide are shown as spheres, with carbons coloured gold. The two sulphate groups bound to the subunit in the PtdIns(4,5)P2 site are shown.
High resolution image and legend (325K)Download Power Point slide (1,042K)Slides may be downloaded for educational use, according to the terms described in Nature Publishing Group's licensing policy.
The structure shows that the peptide binds in an extended conformation near the -subunit binding site for phosphatidylinositol-4,5-bisphosphate (PtdIns(4,5)P2), which is replaced here by two sulphate ions (Fig. 1 and Supplementary Fig. 3). The LL moiety at positions L0 and L+1 (for nomenclature see ref. 7 and Supplementary Fig. 1b) binds in two adjacent hydrophobic pockets on the small 2 subunit that are lined by several hydrophobic residues (Fig. 2), most of which are conserved in all subunits from APs 1–4 in species from yeast to mammals (Fig. 3 and Supplementary Information). Mutation of several of these to hydrophilic residues (2L65S, 2V88D, 2V98S or 2L103S), or filling in the pocket by replacement of 2A63 or 2N92 with tryptophan, strongly inhibited the binding of recombinant AP2 core complexes to different dileucine motifs displayed on PtdIns(4,5)P2-containing liposomes (Supplementary Table 2, Methods and ref. 12). The Kd for the binding of wild-type AP2 to the CD4 dileucine motif was 0.85 M, whereas the 2V88 and 2L103 mutations decreased binding to below detectable levels. A double mutant in 2R15 and R21 decreased binding to the dileucine motif by around an order of magnitude, with the Kd for 2R15S/2R21E double mutant decreasing to 7.8 M. In contrast, the binding of wild-type and mutant AP2 cores to PtdIns(4,5)P2 and to Yxx was unaffected, with Kd values for binding to the TGN38 Yxx motif between 0.31 and 0.38 M, indicating that all recombinant core variants were correctly folded and functional (Fig. 3).
Figure 2: Details of binding of the CD4 dileucine signal by the 2 and subunits of AP2.
a, The dileucine peptide is bound mainly by the pale blue 2 subunit near its interface with the subunit (dark blue). The peptide is shown in its final 2mFo - DFc electron density (cropped around the peptide and contoured at 0.11e Å-3). b, Schematic representation of the dileucine peptide with the principal side chains involved in its binding. Mutants in the boxed residues were analysed kinetically. c, Details of the polar Q(L-4) binding pocket. d, As c, with a semi-transparent electrostatic surface representation (coloured from red at -0.5 V to blue at +0.5 V), showing the positive charge principally due to R21, 2K13 and 2R15. e, Details of the deep and shallow pockets on 2 involved in recognizing the leucine residues at positions L0 and L+1. f, As e, with a molecular surface. Labelled side chains are in the 2 subunit unless otherwise indicated.
High resolution image and legend (371K)Download Power Point slide (754K)Slides may be downloaded for educational use, according to the terms described in Nature Publishing Group's licensing policy.
Figure 3: Confirmation of location and conservation among different subunits of the dileucine-motif-binding site.
a, The dileucine-peptide-binding site with residues whose mutation strongly inhibit dileucine peptide binding while not affecting Yxx motif binding are coloured red (see Supplementary Table 1 and Supplementary Fig. 8). b, c, Sensorgrams and Kd values for binding of wild-type AP2 core and three mutants thereof that strongly inhibit the binding of AP2 to PtdIns(4,5)P2-containing liposomes displaying the CD4 Q peptide motif (b) but do not affect binding to PtdIns(4,5)P2-containing liposomes displaying the TGN38 Yxx motif (c). WT, wild-type; RU, resonance units. d, Mammalian 2, 1a, 3 and 4 were aligned by using ClustalW (Supplementary Fig. 7) and the residue conservation was plotted from dark purple (absolute conservation) to white (no conservation) onto the surface of mammalian 2 in two views related by a rotation of 180°. The binding site for the dileucine-motif peptide is the outstanding feature of surface residue conservation.
High resolution image and legend (394K)Download Power Point slide (780K)Slides may be downloaded for educational use, according to the terms described in Nature Publishing Group's licensing policy.
The L-4 position of AP2-binding dileucine motifs is most commonly occupied by the acidic residues glutamate or aspartate; however, other residues are possible, such as glutamine (as in the human CD4 dileucine motif used here), histidine (in feline CD4) or arginine (in GLUT4) (reviewed in ref. 7). The binding of the residue in the L-4 position occurs on a hydrophilic patch with an overall positive charge. The electron density for its side chain is not as strong as that for the two leucine residues or the peptide backbone, indicating some degree of flexibility in binding. Indeed, simultaneous mutation of two basic residues in the patch (R21 and 2R15) were necessary to decrease dileucine motif binding significantly (Fig. 3). In the E peptide, where the L-4 position is the more commonly found glutamate7, the electron density is similar to that for the corresponding glutamine in the Q peptide complex (see Supplementary Information). The use of general electrostatic complementarity rather than a specific interaction between side chains is supported by the fact that exchange of a motif's L-4 glutamate to a shorter aspartate has little effect on the binding of dileucine motifs to AP2 (ref. 11). This general electrostatic interaction probably assists in correctly orienting the peptide, a proposal supported by the observations that mutation of the L-4 acidic residue to alanine significantly weakens binding but does not abolish it, whereas mutation to a positively charged arginine abrogates binding altogether in the context of the sequence EDEPLL11. Consistent with this was our observation of a slightly weaker binding of AP2 to the Q peptide than to the E peptide (see data in Supplementary Table 3). However, in GLUT4 transporters (RRTPSLL) the L-4 and L-5 positions of the internalization motifs are arginine residues (reviewed in ref. 7); suggesting that AP2 might possess alternative modes of dileucine motif binding that would probably use the same dileucine-moiety binding pocket but could place the remainder of the peptide in a different location. Taken together, these structural and mutagenesis data explain the inhibition of the endocytosis of dileucine-motif-containing cargoes when the key L[LI] and [ED] residues are mutated11, 13, 14.
Point mutation and motif swapping indicates that the identity of the residues in the L-1 to L-3 positions have some effect on motif binding10, 11. In the structure presented here, the arginine in the L-1 position interacts with 2N97. A proline, often found at this position, would increase the strength of binding by favouring the conformation of the peptide backbone observed in the bound peptide14. In this AP2–CD4 peptide complex the amino-acid residue side chains at the -2 and -3 positions point into solvent, away from the AP2 surface. In vivo, however, the conformational context of the motif may be affected by these and other surrounding residues, as has been proposed for the L-5 position phosphorylated serine residue, which has no visible electron density for the phosphate in the structures presented here, despite its presence as confirmed by mass spectrometry (data not shown). The apparent lack of sequence specificity for residues other than the L[LI] moiety—manifested in the increased tolerance of AP2 for alterations in residues adjacent to the L[LI] in comparison with AP1 and AP3 (ref. 11)—is reflected in the lack of strong contacts between non-L[LI] residues of the motif and AP2 shown here. Such comparatively broad specificity, as is demonstrated by the similar binding affinities displayed by AP2 for several different 'acidic dileucine' motifs (Supplementary Table 3), may ensure the efficient internalization of dileucine-motif-containing proteins that arrive at or mis-sort to the cell surface. The ability of the other AP complexes to discriminate between different L[LI] signals may be necessary to 'fine tune' the directing of cargoes to different intracellular compartments.
In the closed and therefore inactive conformation of AP2 (ref. 3), the hydrophobic L[LI] binding pockets are occupied by side chains from residues in the N-terminal extension to the first helix of the 2 subunit. 2F7 occupies the deep LØ pocket, whereas 2Y6 lies in the broader and shallower L+1 L[LI] pocket (Fig. 4 and Supplementary Fig. 8c). Phosphorylation of 2Y6 by EGFR15 would inhibit the binding of 2Y6 in the L+1 pocket and therefore would greatly weaken the interaction between the N terminus of 2 and 2. In the overall structure of the heterotetrameric complex, the helical solenoids of the and 2 chains form a puckered ring around 2 and the N-terminal domain of 2 (N-2), which together make a bowl for the carboxy-terminal domain of 2 (C-2). Comparing this dileucine-motif-liganded, unlocked structure with the 'inactive' conformation, the /2 ring splits approximately as two rigid bodies, (1–400)–2 and (401–end) 2/N–2, with a relative rotation of about 20° around a point near residue 400. This widens the opening of the ring at the N terminus of 2, and the N-terminal extension swings away into solvent, thus exposing the peptide-binding site. C-2 forms a third rigid body, bridging between and 2, and roughly preserving its contacts with both and 2, such that the Yxx-binding site on 2 remains blocked by 2V365 and 2Y405. In AP1, and very probably therefore also in AP2, it has been proposed that the binding of either an acidic dileucine or a Yxx motif drives a conformational change that favours the binding of the other motif16. A further conformational change must therefore occur so as to produce a fully active conformation that is able to bind both types of motif and PtdIns(4,5)P2 simultaneously. The failure to attain a fully active 'open' conformation is driven by a combination of factors including crystal packing, the lack of a Yxx peptide to compete out the autoinhibitory binding of 2 to 2, and the high concentration of ammonium sulphate in the crystallization conditions, increasing the apparent strength of the interaction between 2V365 and the C-2 -residue binding pocket and also helping to stabilize the interaction of 2 residues 350 and 500 with C-2.
Figure 4: A conformational change in AP2 is required for dileucine peptide binding.
a, In the IP6-liganded inactive conformation the N terminus of 2 is held in place by 2F7 and 2Y6 sitting in the L0 and L+1 pockets. b, For dileucine motifs to bind, the N terminus of 2 is displaced from the surface of 2. c, Close-up of the dileucine-binding site: the dileucine peptide (gold) runs in the opposite direction to the N terminus of 2 (green). d, Schematic representation of the conformational change: a 20° hinge movement in moves the N terminus of 2 out of the dileucine-binding site, allowing the motif to bind. The 2 subunit has been omitted for clarity. e, Front and back views of the conformational change: the IP6-liganded inactive conformation is shown in pale colours, the unlocked dileucine-peptide-bound conformation in dark colours. The Yxx site on 2 remains blocked, and is remote from the dileucine-binding site.
High resolution image and legend (418K)Download Power Point slide (799K)Slides may be downloaded for educational use, according to the terms described in Nature Publishing Group's licensing policy.
The binding site on 2 used by AP2 to bind acidic dileucine motifs is not the same as that used by other proteins with -like or 'longin domain' folds to bind their cognate ligands: these include the SNAREs Sec22b17 and Vamp7 (ref. 5) and the signal recognition particle SR subunit18. However, the hydrophobic L[LI]-binding pocket, but not the acidic-residue-binding patch at the -4 position, is largely conserved in the COP subunit of the subcomplex of COPI that is homologous in structure to AP complexes (PDB 2HF6 and Supplementary Information), pointing to a possible conservation of function in the binding of a di-hydrophobic motif, candidates for which would include the FF motif of the COPI cargoes, the p24 family and ERGIC-53 (ref. 19) and the N terminus of -COP.
The mechanism of the binding of a dileucine motif by the AP2 clathrin adaptor complex conforms to the established model of a CCV–cargo interaction. The peptide is bound in an extended conformation by a folded adaptor domain, with specificity arising from a few residue side chains fitting into compatible pockets. The mode of binding is therefore superficially similar to the binding of DxxLL motifs by the VHS domains of GGAs8, 9, but the design of the LL and acidic-residue recognition pockets, their spacing and the structures underlying them are completely different (Supplementary Fig. 9). The interaction between the two isolated components, a dileucine-motif-containing peptide molecule and a single AP2 complex, is dynamic, with a Kd in the micromolar range (1–3 M)12, and buries about 1,100 Å2 of solvent-accessible surface area. These values are both similar to those shown for other cargo–clathrin adaptor interactions6. When the signal is presented in, and so simultaneously recognized with, a PtdIns(4,5)P2 -containing membrane, the apparent strength of the interaction with AP2 is increased: the Kd decreases to the high nanomolar range, as shown here, and it is this coincident detection of PtdIns(4,5)P2 and acidic dileucine motif that allows the signal to outcompete the N terminus of 2 for binding to 2. The need for sorting signals to be presented in a PtdIns(4,5)P2-containing membrane for efficient AP2 binding also prevents inappropriate recognition of similar sequences in cytoplasmic proteins.
Methods Summary
Recombinant AP2 cores were expressed and purified as in ref. 3. Crystals of the CD4–AP2-core complexes were grown over a period of three weeks by hanging-drop vapour diffusion against a reservoir containing 1.7–2.2 M ammonium sulphate, sodium citrate pH 6.5 and 5 mM dithiothreitol from a mixture of 10 mg ml-1 AP2 core and 7 mg ml-1 peptide. Crystals were of space group P43212, with two molecules in the asymmetric unit, and diffracted to a best resolution of 3.0 Å. After cryoprotection, data were collected at 100 K on Diamond beamline I03 and the structure was solved by molecular replacement with PHASER20. A complete explanation of the structure determination and of work performed to confirm the identity and orientation of the peptide is given in Methods. Mutant versions of AP2 core were sequenced throughout and the surface plasmon resonance (SPR)-based liposome binding assay was performed as in ref. 12, with the peptides described in Methods.
Full methods accompany this paper.