Structural determinants controlling 14-3-3 recruitment to the endocytic adaptor Numb and dissociation of the Numb·α-adaptin complex

Traffic of cargo across membranes helps establish, maintain, and reorganize distinct cellular compartments and is fundamental to many metabolic processes. The cargo-selective endocytic adaptor Numb participates in clathrin-dependent endocytosis by attaching cargoes to the clathrin adaptor α-adaptin. The phosphorylation of Numb at Ser265 and Ser284 recruits the regulatory protein 14-3-3, accompanied by the dissociation of Numb from α-adaptin and Numb's translocation from the cortical membrane to the cytosol. However, the molecular mechanisms underlying the Numb–α-adaptin interaction and its regulation by Numb phosphorylation and 14-3-3 recruitment remain poorly understood. Here, biochemical and structural analyses of the Numb·14-3-3 complex revealed that Numb phosphorylation at both Ser265 and Ser284 is required for Numb's efficient interaction with 14-3-3. We also discovered that an RQFRF motif surrounding Ser265 in Numb functions together with the canonical C-terminal DPF motif, required for Numb's interaction with α-adaptin, to form a stable complex with α-adaptin. Of note, we provide evidence that the phosphorylation-induced binding of 14-3-3 to Numb directly competes with the binding of α-adaptin to Numb. Our findings suggest a potential mechanism governing the dynamic assembly of Numb with α-adaptin or 14-3-3. This dual-site recognition of Numb by α-adaptin may have implications for other α-adaptin targets. We propose that the newly identified α-adaptin–binding site surrounding Ser265 in Numb functions as a triggering mechanism for the dynamic dissociation of the Numb·α-adaptin complex.

Membrane traffic establishes, maintains, and reorganizes distinct compartments of a cell while retaining the compositional (proteins, lipids, etc.) and functional heterogeneity of the donor and acceptor membranes. Clathrin-coated pits/vesicles (CCPs/CCVs), 2 the major membrane traffic carriers, mediate many post-Golgi trafficking routes including internalization and recycling via clathrin-mediated endocytosis, which is fundamental to cell nutrition, neurotransmission, and cellular signaling (1)(2)(3). The adaptor protein 2 (AP2) complex is the most abundant clathrin adaptor. It forms a large globular heterotetrameric (␣-␤2-2-2) core structure with two appendage domains (from the C-terminal domains of the ␣ and ␤2 subunits) and plays important roles in many vesicle trafficking pathways within the cell. The ␣ (also called ␣-adaptin) and ␤2 subunits mediate the binding to the target membrane and the recruitment of clathrin, respectively, whereas the 2 and 2 subunits recognize the YXX (where X represents any amino acid, and indicates a hydrophobic residue hereafter) and dileucine motifs on cargo proteins, respectively (4,5). The appendage domain of ␣-adaptin (referred to as ␣-appendage hereafter) is also responsible for recruiting a large number of accessory/regulatory proteins, including Eps15, by binding to DPF motifs within these otherwise very different proteins (6 -8). Two conserved target-binding sites have been identified on ␣-appendage, namely, the "top" site that recognizes the DPF/FXDXF/FXXFXXL motifs and the side site that binds to the WVX(F/W) motif (6,7,9).
The evolutionarily conserved cell fate determinant Numb displays a complex pattern of functions in asymmetric cell division, cell adhesion, cell migration, endocytosis, and ubiquitination of specific substrates within a number of signaling pathways, and it acts as a tumor suppressor in certain cancers (10 -17). As a cargo-selective endocytic adaptor, Numb regulates both clathrin-dependent and independent intracellular trafficking of multiple molecules including Notch, integrin, and Rab7 (18 -20). Additionally, Numb has been demonstrated to regulate the homotypic fusion of early endosomes during intracellular trafficking (21). From the N to C termini, Numb contains a phosphotyrosine-binding (PTB) domain, a proline-rich region, two DPF motifs, and an NPF motif. The PTB domain of Numb can bind to the intracellular domains of transmembrane proteins and thus promotes their recruitment into CCPs/CCVs and their subsequent internalization through AP2 (18,22). The NPF motif is important for binding to Eps15 homology domain-containing proteins, e.g. Eps15 (23). The C-terminal DPF motif of Numb mediates a conserved interaction with ␣-adaptin of the AP2 complex (22)(23)(24)(25)(26). In HeLa cells, Numb was found to be colocalized with AP2 at substratum plasma membrane punctate and cortical membrane-associated vesicles (26,27). It was recently shown that two conserved Ser residues (Ser 265 and Ser 284 in mouse Numb) in the central region of Numb could be phosphorylated by Ca 2ϩ /calmodulin-dependent protein kinase I (CaM-KI) (25) and atypical protein kinase C (aPKC) (19,26,27) both in vivo and in vitro. Moreover, aPKCdependent phosphorylation of Numb leads to the translocation of Numb from the cortical membrane to the cytosol (26,27). A potential mechanism is the phosphorylation-induced recruitment of 14-3-3, which coincides with the disassociation of AP2 from Numb, suggesting that 14-3-3 binding to Numb may regulate its association with AP2 (28).
Here we solved the crystal structure of 14-3-3 in complex with diphosphorylated Numb (Ser(P) 265 -Ser(P) 284 , referred to as pSS hereafter; see Fig. 1A) peptide, providing an atomic-level picture for the site-specific interaction between Numb and 14-3-3. Next, we discovered another Numb-binding site on ␣-appendage distinct from the well known DPF motif binding top site. We further found that these two weak Numb-binding sites function cooperatively in binding to Numb with high affinity. Structural analysis of the Numb-14-3-3 and Numb-␣-appendage interactions suggested that phosphorylation-induced 14-3-3 binding would impair the dual-site Numb-␣-appendage interaction, resulting in dissociation. In addition to the potential role of 14-3-3 in regulating Numb-␣-adaptin-mediated endocytosis, our study also provides a new dual-site target binding mode for ␣-adaptin.

Overall structure of the phospho-Numb/14-3-3 complex
To elucidate the structural basis of the cooperative binding between the diphosphorylated Numb peptide and 14-3-3, we determined the crystal structure of 14-3-3 in complex with the synthetic 30-amino acid pSS peptide at 2.1 Å resolution (Fig. 2, A and B, and Table 1). The electron density map allowed the building of 21 residues of the peptide, clearly illustrating both primary interaction sites. In the Ser(P) 265 and Ser(P) 284 sites, residues 262-267 and 271-287 were visible, respectively (Fig.  2C). The other amino acids in the peptide could not be traced, presumably because of its intrinsic disorder and lack of contacts with 14-3-3. In the complex structure, one Numb pSS peptide asymmetrically interacts with a 14-3-3 homodimer to form a trimeric architecture (Fig. 2, A and B). In line with the crystal structure, ITC data revealed that this peptide bound to a 14-3-3 dimer with 1:1 stoichiometry (n ϭ 1; Fig. 1B). Each monomeric 14-3-3 subunit consisted of nine antiparallel ␣-helices, and residues within the first four ␣-helices formed the dimer interface as well as the floor of the channel ( Fig. 2A). Although neither the Ser(P) 265 nor Ser(P) 284 fragment of Numb was the canonical RSXpSXP or RXXpSXP motif, they resided along the same amphipathic groove between the sidewall and floor of the channel of 14-3-3 in an extended conformation, just as the canonical 14-3-3 binding motifs do (34) (Fig. 2, A and B). Polar contacts were found between main chain amides from Numb pSS and side chains from residues in 14-3-3. O ␦ of Asn 178 (14-3-3) in each monomeric 14-3-3 subunit formed a hydrogen bond with the backbone amide nitrogen of Phe 266(pNumb) or Leu 285(pNumb) , respectively. N ␦ of Asn 229(14-3-3) from each Structural basis of Numb binding to ␣-adaptin and 14-3-3 monomeric 14-3-3 subunit formed a hydrogen bond with the backbone carbonyl oxygen of Gly 264(pNumb) or Leu 283(pNumb) , respectively. At the phospho-site, the phosphate was engaged in a positively charged pocket formed by residues Arg 57 and Arg 132 of 14-3-3, and a hydrogen bond formed between the phosphate and the side chain of Tyr 133(14-3-3) further stabilized the interaction (Fig. 2D). Note that the above contacts between the phosphate and 14-3-3 are completely conserved in the canonical mode 1 and mode 2 14-3-3-target interactions (34) (Fig. 2E). However, because of the lack of further contacts, e.g. the hydrogen bonds formed between S (Ϫ2) of the phosphopeptide and Glu 180 and Trp 228 of 14-3-3 observed in mode 1 interaction or the salt bridges formed between R (Ϫ4) of the phospho-peptide and Glu 180 of 14-3-3 observed in mode 2 interaction (Fig. 2E), single phospho-Numb peptides (Ser(P) 265 or Ser(P) 284 ) could not bind to 14-3-3 robustly. In line with the structure, alanine mutations of Arg 57 and/or Arg 132 in 14-3-3 completely disrupted its interaction with Numb pSS (Fig. S1A). Together, the noncanonical Numb pSS peptide bound to the 14-3-3 dimer in a bidentate fashion by occupying the canonical phospho-sites.

The DPF motif is not sufficient for Numb to form a stable complex with ␣-appendage
An interesting phenomenon is that the aPKC/CaM-KImediated phosphorylation of Numb (at Ser 265 and Ser 284 ) and the simultaneous recruitment of 14-3-3 dissociated the stable Numb⅐AP2 complex in vitro (28) and led to the diffusion of Numb from the AP2-marked CCPs in living cells (19). It seems that the interaction between Numb and ␣-appendage in the AP2 complex is independent of Numb phosphorylation (24,28), and thus, the Numb/AP2 dissociation may not be a direct consequence of Numb phosphorylation. In line with this hypothesis, FLAG-tagged full-length Numb WT interacted with GST-␣appendage equally well when preincubated with or without phosphatase inhibitor (Fig. 3A), implying that the phosphorylation state of Numb does not affect its interaction with ␣-appendage. Compared with Numb WT, the S265A or S265A,S284A mutation, which disrupted the binding of Numb to 14-3-3, had no observable effect on the binding to ␣-appendage (Fig. 3A). Whereas in the competition assay, the amount of FLAG-Numb pulled down by GST-␣-appendage was gradually reduced in the presence of increasing amounts of 14-3-3 (Fig. 3B), indicating that 14-3-3 directly competed with ␣-appendage for Numb binding.
Because the well known 556 DPF motif responsible for ␣-appendage binding is located on the very C terminus of Numb (Fig. 3C), it is not clear how 14-3-3 binding to phospho-Numb (around Ser 265 and Ser 284 ) would interfere with its interaction with ␣-appendage. An attractive hypothesis is that there is another important ␣-appendage binding site located near that region. We thus mapped the specific ␣-appendage binding site(s) within Numb by GST pulldown assay. We first con- binding between Numb phospho-peptides and 14-3-3. C, cell lysate GST pulldown assay showing that full-length WT Numb but not the S265A, S284A, or S265A,S284A mutants could be pulled down by GST-14-3-3. D, both 14-3-3 sites in Numb are required for stable interaction with 14-3-3. In this experiment, Trx-His-tagged Numb 241-593 WT or mutants were pretreated with affinity-purified GFP-PKC and then mixed with GST-14-3-3 (see "Materials and methods" for details). The Numb 241-593 WT or mutant proteins were detected by Western blotting analysis using anti-His antibody. IB, immunoblotting.

Structural basis of Numb binding to ␣-adaptin and 14-3-3
firmed that the C-terminal part of Numb (aa 241-593, referred to as NumbC hereafter) but not the PTB domain (aa 1-240) was responsible for binding to ␣-appendage (Fig. 3, C and D). In line with our cell lysate GST pulldown results (Fig. 3A), the phosphorylation mimic S265E, S284E, or S265E,S284E mutants of NumbC bound to ␣-appendage with a similar affinity as WT did (Fig. 3E). Note that the Numb fragment containing the C-terminal DPF motif alone (aa 541-593) had a very weak binding affinity to ␣-appendage (Fig. 3, C and D), further implying the existence of another ␣-appendagebinding motif within Numb. Unfortunately, NumbC was very unstable, possibly because of its long unstructured random coil region with easy access for proteolysis. To find a relatively stable Numb fragment capable of binding to ␣-appendage as efficiently as NumbC, we made several truncation fragments of NumbC covering both the 14-3-3-binding region and the C-terminal DPF motif, and we finally found a satisfactory Numb fragment (aa 260 -329 fused with aa 541-570, referred to as NumbF hereafter; Fig. S2). Note that although there are two DPF motifs ( 333 DPF and 556 DPF) within Numb (Fig. 3C), 333 DPF plays a minor role in binding to ␣-appendage ( Fig. 3C and Fig. S2), which is consistent with previous reports (22,26). Importantly, either the N-terminal part (aa 260 -329) or the 556 DPF motif showed very weak to moderate binding to ␣-appendage, and the 556 DPF motif could not form a stable complex with ␣-appendage, as NumbF did, when analyzed by analytical gel-filtration assay (Fig. 3, D and F). When ␣-appendage was incubated with both NumbF and the 556 DPF motif in a 1:1:1 molar ratio, only NumbF was found in complex with ␣-appendage (Fig. 3F). Together, these results demonstrated that the 556 DPF motif was essential but not sufficient for Numb binding to ␣-appendage, and the N-and C-terminal fragments of NumbF functioned cooperatively to interact with ␣-appendage with high affinity.
To understand how the two sites within NumbF collaboratively interact with ␣-appendage, we tried to obtain crystals of ␣-appendage in complex with NumbC, NumbF, or other Numb constructs. However, although we could obtain a stoichiometric Numb⅐␣-appendage complex (Figs. 3F and Fig. S2), our extensive attempts to grow crystals have failed.

Another target-binding site on ␣-appendage
Because the aPKC/CaM-KI phosphorylation-induced 14-3-3 recruitment of Numb impaired its binding to ␣-appendage, we were motivated by the idea that the other unknown

Structural basis of Numb binding to ␣-adaptin and 14-3-3
␣-appendagebinding site might be partially overlapped with the 14-3-3 binding site in Numb, e.g. near the Ser 265 and Ser 284 phosphorylation sites. In line with this hypothesis, truncation of amino acids 260 -290 (encompassing both phosphorylation sites) of full-length Numb dramatically weakened its binding ability to ␣-appendage (Fig. 4A). As a positive control, alanine substitution of the 556 DPF motif (referred to as DPF/A hereafter) robustly impaired the binding, and the combined mutation of ⌬260 -290 and DPF/A completely disrupted the interaction (Fig.  4A). We then constructed several N-terminal truncation fragments of NumbF and tested their binding to ␣-appendage. Truncation of the N-terminal 10 amino acids (⌬260 -269) significantly weakened the binding 2-fold, whereas truncation of the N-terminal 20 amino acids (⌬260 -279) resulted in a comparable though slightly weaker binding affinity to ␣-appendage than that of the ⌬260 -269 mutant (Fig. 4B), indicating that amino acids 260 -269 of Numb may play an important role in ␣-appendage binding.
Sequence alignment of Numb revealed two evolutionarily conserved clusters within amino acids 260 -269 of Numb: the bulky hydrophobic cluster (Phe 266 and Phe 269 ) and the positively charged cluster (Arg 262 , Gln 263 , and Arg 267 ; Fig. 4C). Alanine substitution of each cluster (referred to as FF/A and RQR/A, respectively) had negligible influence on the Numb/␣appendage interaction, whereas the combined mutation of both clusters (referred to as RQFRF/A) significantly weakened the interaction to an extent comparable with that of the ⌬260 -279 truncation (Fig. 4, B and D, and Figs. S1B and S3A), indicating that both clusters act synergistically in ␣-appendage binding. As expected, the mutation of both RQFRF and 556 DPF motifs further weakened the Numb-␣-appendage interaction, as indicated by the increased K d (Fig. 4, D and E).
Next, we tried to map out the corresponding binding site of the RQFRF motif on ␣-appendage. Previous studies had identified the top and side target-binding sites on ␣-appendage, which recognize the DPF/FXDXF/FXXFXXL and WVXF motifs, respectively (Fig. 4F) (7, 9, 40). Site-directed mutagenesis in ␣-appendage revealed that the top site (W840A) played an important role in binding Numb (most probably through the 556 DPF motif), whereas the side site (F740E) seemed not to be involved in the interaction (Fig. 4, D and G, and Fig. S1C), implying that other site distinct from these two sites may exist on ␣-appendage for accommodating the RQFRF motif. Because both the hydrophobic and positive-charge properties of the RQFRF motif were required for its interaction with ␣-appendage (Fig. S3A), the corresponding properties should be present on the binding site of ␣-appendage. Surface analysis of ␣-appendage revealed two potential RQFRF-binding sites that are hydrophobic patches surrounded by negatively charged amino acids: the "E" cluster (containing Glu 830 , Glu 863 , Glu 878 , and Glu 879 ) and the "ED" cluster (containing Glu 702 , Asp 703 , Asp 760 , Asp 761 , Glu 797 , Glu 932 , and Glu 936 ; Fig. S3B). Mutation in the E cluster (by substituting the Glu residues with alanine, referred to as E/A) weakened the binding to NumbF by 2-fold, whereas mutation in the ED cluster (by substituting the E/D residues with alanine, referred to as ED/A) had an undetectable influence on the interaction (Fig. 4, D and G). Thus, it is attractive for us to propose that the central RQFRF motif and the C-terminal DPF motif of Numb bind to the E cluster and top site of ␣-appendage, respectively, to cooperatively interact with ␣-appendage with high efficiency.

Discussion
It is well documented that the important endocytic protein Numb is incorporated into the clathrin-coated pits/vesicles via the recognition of its C-terminal 556 DPF motif by the ␣-appendage of AP2, where it acts as an adaptor to further recruit cargos into CCPs/CCVs for trafficking (22,24,26). Intriguingly, the aPKC/CaM-KI-mediated phosphorylation of Numb at Ser 265 and Ser 284 results in the dissociation of Numb from AP2, which coincides with the recruitment of 14-3-3 by phospho-Numb (25). In this study, we discovered a conserved RQFRF motif located adjacent to the phosphorylation sites (Ser 265 and Ser 284 ) of Numb, which cooperated with the DPF motif to bind to ␣-appendage with high affinity. Upon phosphorylation at Ser 265 and Ser 284 , Numb recruits 14-3-3, leading to the release of the adjacent RQFRF motif from ␣-appendage through steric hindrance, and then the DPF motif of Numb dissociates from the top site on ␣-appendage because of instability (Fig. 5). This finding defines the RQFRF motif of Numb as a triggering site necessary for the dynamic dissociation of the Numb⅐AP2 complex.
As the key component of the AP2 complex, ␣-appendage acts by recruiting a large number of accessory/regulatory proteins into CCPs/CCVs (6 -8). During the past decades, two conserved target-binding sites have been identified on ␣-appendage, namely, the top site that recognizes the DPF/FXDXF/ FXXFXXL motifs and the side site that binds to the WVXF/W motif (7,9,40). The E cluster discovered here, which could accommodate the RQFRF motif of Numb, provides a new mode for target recognition of ␣-appendage. Multiple sequence alignment showed that both the E cluster and the hydrophobic amino acids around it are evolutionarily highly conserved (Fig.  S4), implying that the identified E cluster may be another general target-binding site on ␣-appendage, and new targets of ␣-appendage could be explored. Importantly, the E cluster could be regarded as a gatekeeper site, necessary for robust binding to certain targets (e.g. Numb) but with almost no intrinsic affinity to

Structural basis of Numb binding to ␣-adaptin and 14-3-3
the target itself, which therefore allows modulated target binding at this site by competitional regulatory factors (e.g. 14-3-3). This dual-site recognition mode seen in Numb might be common in other ␣-appendage cargos, allowing the spatiotemporal fine-tuning of cargo recruitment and the release of the AP2 complex. The structure of the Numb pSS/14-3-3 complex solved here reveals the high diversity of 14-3-3 target recognition. It has been shown that a synthetic peptide containing two chemically linked identical 14-3-3 sites gained a 30-fold higher affinity for 14-3-3 over a single 14-3-3 site (34). Meanwhile, the cooperation of a divergent 14-3-3 site (with extremely weak affinity to 14-3-3) and a consensus 14-3-3 site (with a modest affinity to 14-3-3) could also achieve markedly enhanced target binding (39,41). In the extreme Numb⅐14-3-3 case, neither of the phospho-sites in Numb was the canonical 14-3-3 site, and the mono-phospho-peptide had barely detectable binding to 14-3-3. Nevertheless, both diverse 14-3-3 sites function cooperatively to interact with 14-3-3 with high efficiency. Our finding expands the potential target pool of 14-3-3.

Protein expression and purification
Various mouse Numb fragments (Fig. 3), mouse ␣-adaptin appendage domain (aa 693-938), and mouse 14-3-3 (aa 1-246) were individually cloned into pGEX-6P-1 or a modified version of the pET32a vector (42), with the resulting protein containing a GST or Trx-His tag in its N terminus. All of the mutations were Figure 3. Discovery of another ␣-appendage binding site within Numb. A, the interaction between Numb and ␣-appendage was phosphorylation-independent. HEK293T cells were transfected with full-length FLAG-Numb WT or various mutants. Lysates were incubated with GST-␣-appendage with or without phosphatase inhibitor. B, competition assay. 14-3-3 captured phosphorylated Numb from ␣-appendage in a dose-dependent manner. The amounts of 14-3-3 were 0.1, 0.4, and 0.8 mg. Lysates were preincubated with phosphatase inhibitor. C, mapping the ␣-appendage binding region of Numb. A summary of the binding affinities between various Numb constructs and ␣-appendage, as shown in D, is given on the right. D, GST pulldown experiments showing that NumbC (aa 241-593) and NumbF (aa 260 -329 ϩ 541-570) bound to ␣-appendage with high affinity, whereas the C-terminal DPF motif (aa 541-593) showed weak binding to ␣-appendage. E, phosphorylation mimic S/E mutations on NumbC had undetectable impact on Numb/␣-appendage interaction. F, an analytical gel filtration assay showed that NumbF but not the C-terminal DPF motif (aa 541-593) could form a stable complex with ␣-appendage. The elution volumes of the molecular size markers are indicated by arrowheads at the top. IB, immunoblotting. ␣-adaptin and 14-3-3 created by a standard PCR-based mutagenesis method and confirmed by DNA sequencing. Recombinant proteins were expressed in Escherichia coli BL21 (DE3) host cells at 16°C and were purified by using GST or nickel-nitrilotriacetic acid-agarose affinity chromatography followed by size-exclusion chromatography.

GST pulldown assay
For the GST pulldown assay, various GST-tagged ␣-appendage proteins (4 nmol) were first loaded onto GSH-Sepharose 4B slurry beads and then incubated with 12 nmol of potential binding partner proteins in 500 l of assay buffer (containing 50 mM

Structural basis of Numb binding to ␣-adaptin and 14-3-3
Tris, pH 8.0, 100 mM NaCl, 1 mM DTT, and 1 mM EDTA) at 4°C for 1 h. After being washed three times, proteins captured by the affinity beads were eluted by boiling, resolved by 12% SDS-PAGE, and detected by Coomassie Blue staining.

Cell lysate GST pulldown assay
Human HEK293T cells were transiently transfected with 6 g of full-length FLAG-Numb WT or various mutants using polyethylenimine transfection reagent. The cells were collected 24 h post-transfection and lysed in a buffer containing 50 mM Tris (pH 7.4), 150 mM sodium chlorate, 0.5% Nonidet P-40, 10 mM sodium fluoride, 1 mM sodium metavanadate, 1 mM DTT, and 10 mM phenylmethylsulfonyl fluoride with or without phosphatase inhibitors. Each lysate was incubated with GSTfusion proteins at 4°C for 2 h. For the competition assay, the lysate was incubated with GST-␣-appendage protein in the presence or absence of 0.1, 0.4, or 0.8 mg of purified His-14-3-3.

In vitro kinase assay
Approximately 1 mg/ml purified Trx-tagged WT or mutant mouse NumbC (aa 241-593) was dialyzed in the kinase buffer containing 20 mM MOPS (pH 7.4), 15 mM MgCl 2 , and 2 mM DTT. GFP-tagged PKC was transfected in HEK293T cells and harvested 24 h after transfection. The cell lysate was incubated with GFP antibody-coupled protein G beads. After being washed with the lysis buffer and the kinase assay buffer, respectively, PKC-bound beads were mixed with 50 l of NumbC sample and incubated at 25°C for 1.5 h in the kinase assay buffer containing 0.1 mM ATP.

Immunoblotting
After extensive washing of the beads with the lysis buffer, the captured proteins were boiled in SDS-PAGE loading buffer and subjected to SDS-PAGE. The proteins were transferred to a 0.45-mm nitrocellulose membrane (Millipore), and the nitrocellulose membrane was blocked with 3% BSA in TBST (20 mM Tris-HCl, pH 7.4, 137 mM NaCl and 0.1% Tween 20) buffer at room temperature for 1 h, followed by incubation with the anti-GFP (ABclonal, AE012), anti-FLAG (ABclonal, AE005), or anti-His (Abci, ABT505) antibody at a 1/3,000 dilution at 4°C overnight. The membranes were washed three times with TBST buffer, incubated with horseradish peroxidase-conjugated goat anti-mouse antibody (ABclonal, AS003) and visualized on a LAS3000 chemiluminescent imaging system.

Analytical gel filtration assay
Analytical gel filtration experiments were carried out on an AKTA FPLC system (GE Healthcare). Proteins (20 M, 100 l) were loaded on a Superose 12 10/300 GL column 20 (GE Healthcare) equilibrated with buffer containing 50 mM Tris (pH 8.0), 100 mM NaCl, 1 mM DTT, and 1 mM EDTA. Protein elution was detected by the absorbance at 280 nm.

Crystallography
Freshly purified 14-3-3 was concentrated to 40 mg/ml, and crystals were grown by the hanging drop vapor diffusion method at 16°C in a reservoir solution containing 100 mM trisodium citrate dehydrate (pH 5.6), 20% isopropanol, and 20% polyethylene glycol 4000. The diffraction data of the crystals were collected at the Shanghai Synchrotron Radiation Facility in China Beamline BL17U1 at a wavelength of 0.9792 Å. The data were processed and scaled using HKL2000. The phasing problem of 14-3-3 and Numb peptide was solved by molecular replacement using the 14-3-3 structure (Protein data Bank code 4HKC) as the search model against the 2.1 Å resolution data set. The initial model was further rebuilt, adjusted manually with COOT (43), and refined by the phenix.refine program of PHENIX (44). The final model had 98.4% of the residues in the favored region of the Ramachandran plot with no outliers. The final refinement statistics are summarized in Table 1.

ITC measurements
ITC measurements were performed on an ITC200 Micro calorimeter (MicroCal) at 25°C. All of the protein samples were dissolved in a buffer containing 50 mM Tris (pH 8.0), 1 mM EDTA, and 100 mM NaCl. The titrations were carried out by Figure 5. Model of the 14-3-3-regulated dissociation of the Numb/AP2 complex during the endocytic process. The central RQFRF motif and the C-terminal DPF motif of Numb cooperatively interact with ␣-appendage with high efficiency, by binding to the E cluster and the top site, respectively. The aPKC/CaM-KI-mediated phosphorylation of Numb at Ser 265 and Ser 284 recruits 14-3-3 and disrupts the binding between the RQFRF motif and the E cluster. The interaction between the DPF motif and the top site is not sufficient to maintain Numb/␣-appendage as a stable complex, and Numb then dissociates from AP2. For clarity, ␤2-appendage is omitted in the figure. ␣-adaptin and 14-3-3 injecting 40-l aliquots of various Numb fragments (0.5-0.7 mM) or phosphorylated Numb peptides into solutions of ␣-appendage or 14-3-3 (0.03-0.05 mM) at time intervals of 2 min to ensure that the titration peak returned to the baseline. The titration data were analyzed using the program Origin 7.0 and fitted by the one-site binding model.

Structural basis of Numb binding to
Author contributions-X. C. data curation; X. C. and W. W. formal analysis; X. C., Z. L., Z. S., W. Y., and A. G. investigation; X. C. writing-original draft; W. W. conceptualization; W. W. supervision; W. W. funding acquisition; W. W. project administration; W. W. writing-review and editing.