Role of the ENTH Domain in Phosphatidylinositol-4,5-Bisphosphate Binding and Endocytosis

See allHide authors and affiliations

Science  09 Feb 2001:
Vol. 291, Issue 5506, pp. 1047-1051
DOI: 10.1126/science.291.5506.1047


Endocytic proteins such as epsin, AP180, and Hip1R (Sla2p) share a conserved modular region termed the epsin NH2-terminal homology (ENTH) domain, which plays a crucial role in clathrin-mediated endocytosis through an unknown target. Here, we demonstrate a strong affinity of the ENTH domain for phosphatidylinositol-4,5-bisphosphate [PtdIns(4,5)P2]. With nuclear magnetic resonance analysis of the epsin ENTH domain, we determined that a cleft formed with positively charged residues contributed to phosphoinositide binding. Overexpression of a mutant, epsin Lys76 → Ala76, with an ENTH domain defective in phosphoinositide binding, blocked epidermal growth factor internalization in COS-7 cells. Thus, interaction between the ENTH domain and PtdIns(4,5)P2 is essential for endocytosis mediated by clathrin-coated pits.

ENTH domains are structural modules of ∼140 amino acids found in mammalian epsin 1 and 2, AP180, and Hip1R, as well as in their yeast homologs, Ent1p through Ent4p, yAP180, and Sla2p (1–4). Mammalian epsin plays a crucial role in clathrin-mediated endocytosis (2). Yeast Ent1p and Ent2p are essential for actin function and for endocytosis. Disruption of both genes in yeast is lethal, and the ENTH domain is required to inhibit lethality. Almost all temperature-sensitive alleles of the ENT1 gene are found within the ENTH domain, supporting its importance (3). The essential function of the conserved ENTH domain from yeast to mammal prompted us to identify its downstream target. Using an ENTH affinity chromatography column, we were not able to detect any protein from bovine brain extract bound to the epsin ENTH domain. Because clathrin-mediated endocytosis is mediated by a specific interaction between endocytic proteins and the lipid bilayer to form invaginated buds and coated vesicles (5,6), and because many biochemical and physiological studies suggest important roles for phosphoinositides in endocytosis and vesicular trafficking (7–9), we examined the possibility that the ENTH domain binds to phosphoinositides.

To determine whether the ENTH domain could bind phosphoinositides, we subjected a glutathione S-transferase (GST) fusion protein of the epsin ENTH domain to liposome binding assay. Although epsin ENTH did not co-sediment with phosphatidylethanolamine (PE)– and phosphatidylcholine (PC)–based liposomes, increasing concentrations of phosphatidylinositol-4,5-bisphosphate [PtdIns(4,5)P2] in the liposomes resulted in co-sedimentation of the ENTH domain (Fig. 1A). Co-sedimentation was not observed in the presence of increased concentrations of PtdIns in the liposomes, demonstrating a high specificity for the interaction with PtdIns(4,5)P2. Co-sedimentation was clearly observed at 0.2% PtdIns(4,5)P2, and the dissociation constantK d for the interaction was estimated at 0.37 μM. The strong interaction between the ENTH domain and PtdIns(4,5)P2 was confirmed by other methods, including overlay assays with protein probe against phospholipids (Fig. 1B) and lipid probe against the ENTH domain blotted onto nitrocellulose membrane (Fig. 1C). The specificity of the binding was then studied with all known mammalian phosphoinositides. PtdIns(3,4,5)P3 also showed substantial binding, whereas PtdIns, PtdIns3P, PtdIns4P, PtdIns5P, PtdIns(3,4)P2, and PtdIns(3,5)P2 exhibited far lower affinities (Fig. 1D). No binding was observed of other acidic phospholipids, such as phosphatidic acid and phosphatidylserine (Fig. 1D). We also carried out liposome binding assays for the AP180 ENTH domain. AP180 ENTH bound to PtdIns(4,5)P2 strongly and also showed a lower affinity for PtdIns(3,4,5)P3 (Fig. 1E). Thus, the ENTH domain is an evolutionally conserved unit designed for strong binding to PtdIns(4,5)P2.

Figure 1

Specific binding of the ENTH domain to PtdIns(4,5)P2. (A) Co-sedimentation assay with GST-epsin ENTH domain. Liposomes (PE/PC = 4/1; total, 100 μg) containing the indicated ratio (weight %) of phosphoinositides were mixed with 5 μg of GST-ENTH (24) and co-sedimented. Proteins in supernatant (s) and precipitate (p) were visualized by SDS–polyacrylamide gel electrophoresis and stained with Coomassie brilliant blue (25). (B) Overlay assay using protein probe against lipids. GST fusion proteins were labeled with 32P by protein kinase A and overlaid (0.5 μg/ml) against phospholipids (4 μg) blotted onto nitrocellulose membrane (26). (C) Overlay assay using lipid probe against proteins. [32P]PtdIns(4,5)P2 was produced by phosphorylation of PtdIns4P with PtdIns4P 5-kinase α and [γ-32P]adenosine 5′-triphosphate as described previously (27) and was used as a probe (1 × 106 cpm/ml) against GST fusion proteins (0.1 μg) on nitrocellulose membrane (28). Quantitative representations of co-sedimentation assay data of GST-ENTH domains from (D) epsin and (E) AP180 (24) with 0.5% of the indicated phospholipids in the liposomes. Results from three independent experiments are represented as mean values ± SEM (error bars).

We next studied intracellular localization of the epsin ENTH domain using green fluorescent protein (GFP) fusion proteins (10). COS-7 cells transfected with the GFP-pleckstrin homology (PH) domain of phospholipase C–δ1 (PLC-δ1) showed localization at the plasma membrane (11, 12), whereas GFP alone showed no specific localization. Transfected GFP-epsin ENTH domain was also localized to the plasma membrane, even after the treatment with 300 nM wortmannin [Web fig. 1 (10)], indicating an interaction with the plasma membrane, presumably via PtdIns(4,5)P2 but not through PtdIns(3,4,5)P3 in vivo.

To identify the phosphoinositide binding site of the epsin ENTH domain, we investigated the interaction of the epsin ENTH domain with inositol-1,4,5-trisphosphate [Ins(1,4,5)P3], the head group of PtdIns(4,5)P2, using nuclear magnetic resonance (NMR) spectroscopy. Changes in the backbone amide 1H and15N chemical shifts of the epsin ENTH domain were measured as a function of Ins(1,4,5)P3 concentration (Fig. 2A). Mapping of the changes in the chemical shift onto the three-dimensional structure showed that the most perturbed residues are localized at three sites: the NH2-terminal unstructured region (site 1), the solvent-exposed surface of the first helix (site 2), and the region around loop 1 (loop connecting helices 1 and 2) and loop 3 (loop connecting helices 3 and 4) and helix 4 (site 3) (Fig. 2, A and B). Calculation of the electrostatic surface potential of the epsin ENTH domain showed that sites 1 and 3 are positively charged, whereas site 2 is negatively charged (Fig. 2C). In addition, highly conserved residues in ENTH domains were concentrated at site 3 (Arg63, Trp71, Arg72, and Lys76) (Fig. 2D). To evaluate the importance of these sites, several residues were mutated, and the binding affinities for PtdIns(4,5)P2 were studied (Fig. 2, E and F).

Figure 2

NMR analysis of the interaction between epsin ENTH domain and Ins(1,4,5)P3(29). (A) Chemical shift differences observed in the two-dimensional 15N-HSQC spectra for backbone NH groups in the presence of 2.0 mM Ins(1,4,5)P3. The deviations (in Hz) were quantified with the formula Δ = √[̅(̅Δ̅H̅N̅)̅2̅+̅ ̅(̅Δ̅N̅)̅2̅]̅, where ΔHN and ΔN are the chemical shift differences (in Hz) of the amide proton and nitrogen, respectively. Chemical shift differences in the side-chain NH groups of three tryptophan residues (Trp33, Trp61, and Trp71) are displayed at the right of the graph in the order of residue number. (B) On a representation of the human epsin ENTH domain structure (30), residues are colored on the basis of the 1H and 15N chemical shift differences shown in (A) (31). Red, orange, and yellow indicate large, medium, and small differences, respectively. The figure was created with the MIDASPlus program (32). (C) The electrostatic potential surface (31). Positive, negative, and neutral electrostatic potentials are shown in blue, red, and white, respectively. The potential was calculated with Delphi (MSI) and was drawn with the program Insight 98 (MSI). (D) Space-filling model (31). Highly conserved residues are shown in green. This figure was made with the Insight98 program (MSI). PtdIns(4,5)P2 co-sedimentation assay data of alanine-substituted mutants for (E) site 3 or (F) site 1 (33). The co-sedimentation assay was carried out at 0.5% PtdIns(4,5)P2 in the liposomes, and the binding activities are shown relative to wild type (100%). ΔN18 represents a mutant in which NH2-terminal 18 residues are deleted. Results from three independent experiments are represented by mean values ± SEM (error bars). (G) Model for interaction between Ins(1,4,5)P3 and residues in the ENTH domain (31, 34). According to the degree of inhibition of PtdIns(4,5)P2 binding upon substitution to alanine (E and F), residues are colored as follows: yellow (little effect), orange (medium effect), and red [abolition of PtdIns(4,5)P2binding].

Substitution of alanine for Arg63 and Lys76 almost completely abolished PtdIns(4,5)P2binding of the epsin ENTH domain, whereas substitution for Arg72 only slightly decreased the affinity. Substitution for Lys86, which is located at the end of helix 4 but is distant from site 3, did not affect phosphoinositide binding (Fig. 2E). Thus, specific residues in site 3 appeared to be involved in the direct binding of PtdIns(4,5)P2.

Many residues in site 1 showed large chemical shift changes, suggesting a large structural change upon phosphoinositide binding (Fig. 2A). Deletion of the entire site 1 region (NH2-terminal, 18 residues) resulted in a complete loss of PtdIns(4,5)P2binding (Fig. 2F). Of the positively charged residues within this region, only Arg8 was shown to be essential for binding (Fig. 2F). Results suggest that site 1, the NH2-terminal unstructured region, was necessary for binding and adopted a specific conformation upon binding to PtdIns(4,5)P2.

The most probable binding model, based on these data, is as follows: the head moiety of PtdIns(4,5)P2 binds to the cleft composed of loop 1, helix 3, and helix 4. Subsequently, the NH2-terminal unstructured region binds to site 2, and Arg8 is positioned near the binding site. The side chains of three basic residues (Arg8, Arg63, and Lys76) then form salt bridges with the phosphate groups of PtdIns(4,5)P2 (Fig. 2G).

In addition to the ENTH domain, epsin contains several motifs involved in linking clathrin-coat proteins such as AP-2, Eps15, and clathrin heavy chain (2, 13). The interactions with these proteins are mediated by Asp-Pro-Trp (DPW), Asn-Pro-Phe (NPF), and clathrin-binding motifs present in the central to COOH-terminal region of the epsin molecule (Fig. 3A). Overexpression of epsin DPW motifs results in an inhibition of clathrin-mediated endocytosis (2).

Figure 3

Inhibition of EGF internalization by epsin K76A. (A) Schematic representations of epsin constructs (35). (B) Texas Red–EGF (EGF, red) internalized into COS-7 cells was observed within 10 min (WT and R72A, green) (36). Overexpression of epsin ΔENTH or K76A inhibited internalization (ΔENTH and K76A, green; indicated by arrowheads in the lower panels). (C) Quantitative representation of (B). Cells that internalized Texas Red–EGF (defined as >10 spots visible within the cell body) were counted and represented as a percentage relative to all cells observed (total of >50 cells from three independent experiments). Similar results were obtained from an internalization assay with 125I-labeled EGF [Web fig. 2 (10)].

To evaluate the importance of the interaction between the ENTH domain and phosphoinositide in endocytosis in vivo, we studied the effect of mutant epsin overexpression on endocytosis in COS-7 cells. COS-7 cells treated with Texas Red–conjugated epidermal growth factor (EGF) showed internalization within 10 min. EGF internalization was not affected by overexpression of wild-type epsin in COS-7 cells (Fig. 3B). In contrast, overexpression of epsin ΔENTH in which the NH2-terminal ENTH domain was deleted inhibited internalization (Fig. 3, B and C). This inhibition occurred in a competitive manner in which coat proteins such as AP-2 and Eps15 were presumably sequestered by DPW or NPF motifs present in the COOH-terminal half of the mutant epsin molecule overexpressed in the cell. EGF internalization was also inhibited by overexpression of epsin Lys76 → Ala76 (K76A) (Fig. 3, B and C), in which Lys76 of the ENTH domain was substituted with an alanine residue. Overexpression of epsin Arg72 → Ala72 (R72A), which still binds to PtdIns(4,5)P2 (Fig. 2E), did not show any inhibitory effect on EGF internalization (Fig. 3, B and C). Thus, the phosphoinositide-binding ability of the ENTH domain was essential for epsin to induce clathrin-mediated endocytosis, and the lack of this ability was comparable to a loss of the entire ENTH domain.

The epsin ENTH domain associates with a transcription factor, promyelocytic leukemia Zn2+ finger protein (PLZF) (14), and epsin ENTH R72A fails to bind PLZF. In our study, ENTH R72A still bound to PtdIns(4,5)P2 (Fig. 2E), and overexpression of epsin R72A did not inhibit EGF internalization (Fig. 3, B and C), suggesting that clathrin-mediated endocytosis does not require an interaction between the ENTH domain and PLZF.

Thus, interaction between PtdIns(4,5)P2 and the ENTH domain of epsin is indispensable for EGF internalization mediated by clathrin-coated pits. Internalization of insulin into CHO-IR cells is inhibited by overexpression of the epsin ENTH domain (15), which may indicate a competitive inhibition of ENTH-phosphoinositide binding. All known proteins carrying ENTH domains, such as epsin, AP180, and Hip1R (Sla2p), have been reported to bind clathrin and AP-2 directly or to colocalize with clathrin-coated pits (2,4, 16, 17). This family of proteins is thought to play a regulatory role in the formation of coated pits, invagination of the plasma membrane, or the formation of coated vesicles (2, 6, 18). In this process, ENTH domains may regulate the linkage of the clathrin triskelion to the plasma membrane through phosphoinositides [Web fig. 3 (10)]. Phosphoinositide-metabolizing enzymes like synaptojanin may modify these ENTH-membrane interactions (9, 19).

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


View Abstract

Navigate This Article