Structure and Asn-Pro-Phe Binding Pocket of the Eps15 Homology Domain

See allHide authors and affiliations

Science  28 Aug 1998:
Vol. 281, Issue 5381, pp. 1357-1360
DOI: 10.1126/science.281.5381.1357


Eps15 homology (EH) domains are eukaryotic signaling modules that recognize proteins containing Asn-Pro-Phe (NPF) sequences. The structure of the central EH domain of Eps15 has been solved by heteronuclear magnetic resonance spectroscopy. The fold consists of a pair of EF hand motifs, the second of which binds tightly to calcium. The NPF peptide is bound in a hydrophobic pocket between two α helices, and binding is mediated by a critical aromatic interaction as revealed by structure-based mutagenesis. The fold is predicted to be highly conserved among 30 identified EH domains and provides a structural basis for defining EH-mediated events in protein trafficking and growth factor signaling.

Protein interaction domains such as Src homology domains 2 and 3 are devoted to the recruitment of ligands into multiprotein complexes (1). The recently discovered EH domain (2) is an interaction module that targets NPF-containing proteins such as RAB, NUMB (3,4), clathrin assembly proteins (5), and synaptojanin (6). Proteins containing these EH domains mediate critical events in endocytosis (7, 8) and actin cytoskeletal organization (8), and they participate in signaling in conjunction with tyrosine kinases (2, 9) and Src homology domains (4, 10). Individual EH repeats consist of about 95 amino acids and have been found in at least 20 proteins in organisms ranging from yeast to mammals (11). The prototypic member of this family, Eps15, is made up of an epidermal growth factor receptor (EGFR) phosphorylation substrate that contains three tandem EH domains, binding sites for Crk's Src homology 3 domain (10) and clathrin adaptor proteins (12), and a coiled-coil oligomerization motif (13). Although Eps15 is clearly an essential player in receptor-mediated endocytosis (7), it also appears to intersect signaling pathways that regulate cellular growth (2,10), synaptic transmission (6), and neuronal development (14). The importance of Eps15 is underscored by the discovery of chromosomal translocations involvingeps15 in leukemia (15) and of lethal mutations of a yeast eps15 homolog (8). We selected the second EH domain of human Eps15 (EH2) for structural studies (16) because of its putative interactions with NPF-containing proteins (3), calcium, and EGFR (2) as well as its high degree of sequence conservation.

The structure of EH2 contains two intimately associated helix-loop-helix motifs connected by a short antiparallel β sheet (Fig. 1). The NH2-terminal helix αA lies roughly perpendicular to a bundle of the other three α helices (αB, αC, and αD) and primarily contacts the centrally located αD. The helix αC lies diagonally across the parallel αB and αD helices. The tight packing of the helices is reflected in the large number of distance restraints (1795), yielding well-defined solution structures with a root-mean-square deviation (rmsd) of 0.29 ± 0.05 Å for the backbone atoms in secondary structure elements. A proline-rich element following αD zigzags over αC and αD and juxtaposes the NH2- and COOH-termini. The proximity of these termini would allow the three EH domains of Eps15 to directly abut each other, facilitating cooperative and multivalent binding to ligand proteins.

Figure 1

Structure of EH2. The four α helices αA (amino acid residues 126 through 136), αB (148 through 156), αC (162 through 172), and αD (182 through 197) are depicted in orange, purple, blue, and red, respectively. The mini–β sheet involving residues 145 through 147 (βA) and 179 through 181 (βB) is shown in light blue, the COOH-terminal proline-rich element is shown in light green, and the calcium ion is shown as a yellow sphere. The NH2- and COOH-termini are labeled N and C, respectively. Figure 1, A and B, Fig. 2B, and Fig. 3, B and C, were generated with InsightII software. (A) Best-fit superposition of the backbone atoms (N, Cα, and C′) in the secondary structure elements of the 20 structures with the lowest nuclear Overhauser effect energies (21, 22). (B) Ribbon diagram of the structure closest to the average of the 20 structures, shown in the same orientation as (A). (C) Amino acid sequence alignment for the three EH domains (residues 9 through 103, 121 through 215, and 217 through 313, respectively) of human Eps15 (11, 17). The secondary structure and solvent exposure are shown above the sequence and were determined with Procheck-NMR (20). The coloring and nomenclature of key amino acids are indicated in the inset. Phosphorylation of Eps15 at Tyr132 (shown in green) by EGFR (2) is unlikely because this residue is buried between helices αA and αD.

The peptide motif NPF was recently established as the essential target for the EH domains of Eps15. Binding is enhanced when Thr or Ser occupy the two positions preceding NPF and when a hydrophobic or basic residue follows (3, 4). The peptide sequence PTGSSSTNPFL (17) contains all of these consensus binding residues and corresponds to the COOH-terminus of RAB, which is a cofactor of the human immunodeficiency virus REV protein. Moreover, the association of RAB and a fusion protein containing EH2 was previously demonstrated (3).

The presence of an exposed binding pocket for this peptide (NPFRAB) is apparent from progressive changes in the1H, 13C, and 15N resonances of EH2 upon titration with NPFRAB (Fig. 2, A and B). Helices αB and αC flank this hydrophobic pocket, which can readily accommodate the NPF residues. The center of the binding site is located approximately 10 Å from the calcium ion and on a face opposite the termini. The bottom of the binding pocket is occupied by Leu155, Leu165, and Trp169 (Fig. 2, C and D). The high conservation of the latter two residues and of a hydrophobic residue at the position corresponding to Leu155 suggests that they contact the conserved elements in EH domain ligands. The largest chemical shift changes were observed for Leu156 and Leu165 (Fig. 2B), which implies that they directly contact the Phe side chain of the NPF ligand. The Gly148, Lys152, Val162, and Gly166 residues form the edge of the binding pocket, where they may contribute to EH domain specificity.

Figure 2

NPF binding site of EH2. (A) Superimposed regions of four1H-13C HSQC spectra of 1 mM EH2with the following NPFRAB concentrations: 0 mM, dark blue; 0.25 mM, light blue; 0.5 mM, orange; and 2 mM, red. The inset shows the Trp169 side chain nomenclature. Observation of fast exchange on the NMR time scale is consistent with theK D of the EH2:NPFRABinteraction estimated by surface plasmon resonance (18). (B) Space-filling model showing the NPF binding site. Atoms are colored on the basis of 1H, 13C, and15N chemical shift differences (Δδ) induced by NPFRAB addition. Red, orange, yellow, and light blue indicate very large, large, medium, and small differences, respectively, as listed in the inset. Dark blue indicates that the difference was not measured. (C) Ribbon diagram displaying side chains in the NPF binding site. The orientation is identical to that in (B). Leu155 (dark blue), Leu165(yellow), and Trp169 (red) make up the base of the binding site, whereas residues Gly148 (dark green), Lys152 (magenta), Leu156 (orange), Val162 (light blue), and Gly166 (light green) form the walls of the pocket. Calcium is indicated as a yellow sphere. (D) Molecular surface of EH2 with the same orientation and color coding as in (C) [(C)and (D) were generated with GRASP (23)].

Because Trp169 lies prominently in the NPF binding pocket of EH2, two mutants were generated in which this amino acid was substituted with either an Ala or a Tyr residue. Both mutants possess a native-like fold, as evident from their1H nuclear magnetic resonance (NMR) spectra, which indicates that any change in binding kinetics is directly related to the Trp169 mutations. A dissociation constant (K D) of 560 ± 40 μM was measured for the interaction of wild-type EH2 with NPFRAB, using surface plasmon resonance detection (18). This binding is likely reinforced through multivalent interaction of EH domain clusters with target proteins, which commonly contain multiple NPF motifs (3–6). The Trp169 → Ala169 substitution reduced NPF binding beyond detection, demonstrating the critical role of Trp169 in NPF binding. In contrast, the Trp169 → Tyr169 substitution reduced the affinity only by a factor of 2.8. The high degree of conservation of this Trp (2, 11) and its replacement only by Tyr in a few other EH domains emphasize the importance of the aromatic character of this residue for NPF recognition. Protein interactions involving any EH domain may now be selectively eliminated by mutation of this aromatic residue, allowing the dissection of EH-mediated pathways.

A role for calcium in the activity of EH proteins has been inferred on the basis of the presence of canonical EF hand sequence motifs in several EH repeats (2). However, calcium binding has not yet been demonstrated, and calcium does not appear to influence Eps15's protein interactions (2). Our NMR data pinpoint a binding site for a single calcium ion in the second helix-loop-helix motif of EH2 (Fig. 3A). The backbone conformation (Fig. 3, B and C) and the amino acid sequence (Fig. 1C) of this loop segment match those of canonical EF hands (19), which indicates that calcium is directly ligated by the carboxyl groups of Asp173, Asp175, Asp177, Glu184, and the carbonyl oxygen of Met179. The first helix-loop-helix motif does not bind calcium, but its perpendicular helices and its pairing with the second helix-loop-helix define it as an EF hand variant (Fig. 3, B and C). The loop connecting αA and βA differs most substantially from classical EF hands: it is one residue shorter, contains a Pro residue, and consists of two turns. In light of the EH2 structure, a calcium ion could be readily accommodated in the first helix-loop-helix in EH domains that bear a canonical EF hand sequence at this position (Fig. 1C). The calcium-bound and free states of EH2 exhibit slow exchange on the NMR time scale (Fig. 3A). Such tight binding of calcium indicates full occupancy at intracellular calcium concentrations and implies a structural role for calcium rather than a regulatory switch as found in some EF hand proteins (19).

Figure 3

Calcium binding by EH2. (A) 1H-15N HSQC spectrum of 1 mM EH2 after EDTA treatment. Pairs of1H-15N cross peaks belonging to the same amino acid in calcium-bound and free states are colored green and red, respectively, and are connected by blue lines. An asterisk marks1H-15N cross peaks where only one of the pair could be assigned. Cross peaks of amino acids that did not shift upon calcium removal are drawn in black. Amino acids labeled in blue hold calcium in place through chemical interactions, whereas those in magenta are located in the EF hand loops. (B) Ribbon diagram of EH2 showing the location of the calcium binding site. The ribbon is colored on the basis of 15N, HN, and Hα chemical shift differences (Δδ) between calcium-bound and free states. Red, orange, green, and blue indicate large, medium, small, and negligible chemical shift differences, respectively, as listed in the inset, with the maximum Δδ indicated on top. The yellow sphere represents calcium; the arrow points to the NPF binding site. (C) Comparison of the canonical EF hands and paired loops of calmodulin (sites II and IV), troponin C (site II), parvalbumin, calbindin (blue) (19), and EH2 (red). The backbone atoms of the first six residues in the second EF hand loop are superimposed. The helices of the calcium-loaded EF hand of EH2 are considerably closer than the corresponding helices in the other structures.

Comparison of EH sequences indicates that the structurally critical residues involved in the packing of the hydrophobic core (Fig. 1C) are highly conserved throughout the family. EH domains share the double EF hand fold with other domains within the EF-hand protein superfamily (19), but a few notable differences exist. The location of the peptide binding pocket is unprecedented. The αD helix is completely buried within the EH2 structure and is followed by a proline-rich COOH-terminal element. The modular arrangement and sequence of EH domains further emphasize the fact that these domains constitute a separate class of EF hand proteins. The structure of EH2 provides a foundation for defining the mechanisms through which this family of domains mediates a host of biological processes, including protein internalization and signaling.

  • * To whom correspondence should be addressed. E-mail: MichaelOverduin{at}


View Abstract

Navigate This Article