Determinants of Rab5 Interaction with the N Terminus of Early Endosome Antigen 1*

The Rab5 effector early endosome antigen 1 (EEA1) is a parallel coiled coil homodimer with an N-terminal C 2 H 2 Zn 2 (cid:1) finger and a C-terminal FYVE domain. Rab5 binds to independent sites at the N and C terminus of EEA1. To gain further insight into the structural determinants for endosome tethering and fusion, we have characterized the interaction of Rab5C with truncation and site-specific mutants of EEA1 using quantitative binding measurements. The results demonstrate that the C 2 H 2 Zn 2 (cid:1) finger is both essential and sufficient for the N-terminal interaction with Rab5. Although the heptad repeat C-terminal to the C 2 H 2 Zn 2 (cid:1) finger provides the driving force for stable homodimerization, it does not influence either the affinity or stoichiometry of Rab5 binding. Hydrophobic residues predicted to cluster on a common face of the C 2 H 2 Zn 2 (cid:1) finger play a critical role in the interaction with Rab5. Although the homologous C 2 H 2 Zn 2 (cid:1) finger of the Rab5 effector Rabenosyn binds to Rab5 with comparable affinity, the analogous C 2 H 2 Zn 2 (cid:1) finger of the yeast homologue Vac1 shows no detectable interaction with Rab5, reflecting non-conservative substitutions of critical residues. Large changes in the intrinsic tryptophan fluorescence of Rab5 accompany binding to the C 2 H 2 Zn 2 (cid:1) finger of l injection of 200 n M GppNHp. Binding and dissociation were monitored following 20- (cid:2) l injections of increasing concentrations of His 6 -EEA1. Following curve alignment, the reference sensogram, which reflects bulk refractive index changes and/or reversible nonspecific binding, was subtracted from the sample sensogram. The SPR signal at equilibrium ( R eq ) was extracted from the fit with a simple 1:1 Langmuir binding model and plotted as a function of His 6 -EEA1 concentration. Dissociation constants ( K d ) were obtained from a fit to the hyperbolic binding function R eq (cid:6) R max (His 6 EEA1)/( K d (cid:1) (His 6 EEA1)), where R max corresponds to the SPR signal at saturation and is treated as an adjustable parameter. Mean values and standard deviations ( (cid:4) n (cid:7) 1 ) were calculated from 2 to 4 independent measurements. Control exper- iments verify that the fitted K d values are independent of flow rate (5–50 (cid:2) l/min) and surface coverage (10-fold range), indicating that the equilibrium data are not limited by mass transfer or rebinding. C 2 H 2 Zn 2 (cid:1) finger shares 29% identity with that of Adr1, which repre- sents the closest homologue of known structure. Non-conserved residues were substituted with the corresponding residues in EEA1, which were modeled in the most frequently observed rotomer conformation compatible with the structure. The resulting homology model repre- sents a rough, working approximation to the actual structure, with an overall fold consistent with the common topology of C 2 H 2 Zn 2 (cid:1) fingers.

The Rab5 effector early endosome antigen 1 (EEA1) is a parallel coiled coil homodimer with an N-terminal C 2 H 2 Zn 2؉ finger and a C-terminal FYVE domain. Rab5 binds to independent sites at the N and C terminus of EEA1. To gain further insight into the structural determinants for endosome tethering and fusion, we have characterized the interaction of Rab5C with truncation and site-specific mutants of EEA1 using quantitative binding measurements. The results demonstrate that the C 2 H 2 Zn 2؉ finger is both essential and sufficient for the N-terminal interaction with Rab5. Although the heptad repeat C-terminal to the C 2 H 2 Zn 2؉ finger provides the driving force for stable homodimerization, it does not influence either the affinity or stoichiometry of Rab5 binding. Hydrophobic residues predicted to cluster on a common face of the C 2 H 2 Zn 2؉ finger play a critical role in the interaction with Rab5. Although the homologous C 2 H 2 Zn 2؉ finger of the Rab5 effector Rabenosyn binds to Rab5 with comparable affinity, the analogous C 2 H 2 Zn 2؉ finger of the yeast homologue Vac1 shows no detectable interaction with Rab5, reflecting non-conservative substitutions of critical residues. Large changes in the intrinsic tryptophan fluorescence of Rab5 accompany binding to the C 2 H 2 Zn 2؉ finger of EEA1. These observations can be explained by a mode of interaction in which a partially exposed tryptophan residue located at the interface between the switch I and II regions of Rab5 lies within a hydrophobic interface with a cluster of non-polar residues in the C 2 H 2 Zn 2؉ finger of EEA1.
As master regulators of membrane trafficking, Rab GTPases cycle between active (GTP-bound) and inactive (GDP-bound) conformations (1)(2)(3). In the active conformation, Rab GTPases interact with diverse effectors implicated in vesicle budding, cargo sorting, motor-dependent transport, tethering, docking, and fusion. Guanine nucleotide exchange factors, GTPase-activating proteins, and other accessory factors, including Rab GDP dissociation inhibitor, provide multiple points of regulation throughout the GTPase cycle by modulating nucleotide binding, GTP hydrolysis, and membrane association (4). Protein kinases and phosphatases have also been implicated in the regulation of Rab function, either directly or by phosphorylation of effectors and regulatory factors (5)(6)(7)(8). Thus, through regulated interactions with effectors, Rab GTPases couple signal transduction networks to the membrane trafficking machinery.
The Rab5 effector early endosome antigen 1 (EEA1) was identified as a lupus autoantigen that localizes to early endosomes (24). EEA1 has a modular architecture with an N-terminal C 2 H 2 Zn 2ϩ finger, four consecutive heptad repeats, and a C-terminal region containing a calmodulin-binding (IQ) motif, a Rab5-binding site, and a FYVE domain that binds specifically to PtdIns(3)P (15)(16)(17)25). In cell free reconstitution assays, EEA1 is essential for fusion of early endosomes (11, 26 -28). Endosomal localization requires an intact FYVE domain and is sensitive to inhibitors of phosphoinositide 3-kinase activity as well as mutants of conserved residues in the FYVE domain that disrupt PtdIns(3)P binding (15,25,29,30). Localization also requires a region of ϳ40 residues proximal to the FYVE domain but is not influenced by mutations that disrupt the interaction with Rab5 (25,31,32). EEA1 forms a parallel coiled coil homodimer, and we have shown that the C terminus of EEA1 has an organized quaternary structure that supports a multivalent interaction with membranes containing PtdIns(3)P, explaining the requirement for the proximal 40 residues (33,34). Finally, Rab5 also binds to an independent site at the N terminus of EEA1, which has been shown recently (11,35) to bind Rab22 as well.
The C 2 H 2 Zn 2ϩ finger of EEA1 shares significant homology with the C 2 H 2 Zn 2ϩ finger in the Rab5 effector Rabenosyn as well as the corresponding C 2 H 2 Zn 2ϩ finger in Vac1p, an effector of the yeast Rab5 homologue Ypt51. Like EEA1, Rabenosyn and Vac1p contain a FYVE domain involved in endosome targeting (14,15). Temperature-sensitive mutants implicate Vac1p in intervacuolar trafficking and vacuolar protein sorting (36,37). Immunodepletion of Rabenosyn blocks homotypic * This work was supported by National Institutes of Health Grant GM56324. 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  early endosome fusion as well as heterotypic fusion of endocytic vesicles with early endosomes, suggesting that Rabenosyn plays a critical role distinct from that of EEA1 (14). The GTPbound forms of Rab4 and Rab5 bind to sites in the central and C-terminal regions of Rabenosyn, respectively (38). It is not known whether Rab5 binds directly to the C 2 H 2 Zn 2ϩ finger of EEA1, Rabenosyn, or Vac1p. Two-hybrid data indicate that the integrity of the C 2 H 2 Zn 2ϩ finger is essential for Rab5 binding to the N terminus of EEA1 (11,39). The requirement for an intact C 2 H 2 Zn 2ϩ finger may reflect a direct interaction with Rab5 or an indirect structural role. For example, the double Zn 2ϩ finger of Rabphilin3A is essential for interaction with Rab3A; however, in the crystal structure of the Rab3A-Rabphilin3A complex, the double Zn 2ϩ finger does not contact Rab3A but instead supports interactions with flanking regions (40).
To gain further insight into the structural basis underlying the function of EEA1 in the tethering and fusion of early endosomes and endocytic vesicles, we have characterized the interaction of Rab5C with truncation and site-specific mutants of EEA1 using quantitative binding measurements. The results demonstrate that the C 2 H 2 Zn 2ϩ finger is sufficient for the N-terminal interaction with Rab5C and support a mode of interaction in which an invariant tryptophan residue, which is partially exposed at the interface between the switch I and II regions of Rab5, lies in or near an interface that involves a cluster of hydrophobic residues in the C 2 H 2 Zn 2ϩ finger.
For purification of wild type and mutant proteins, cells were suspended in lysis buffer (50 mM Tris, pH 8.0, 0.1 M NaCl, 0.1% mercaptoethanol, 0.1 mM phenylmethylsulfonyl fluoride, 1 mg/ml lysozyme) and disrupted by sonication. Triton X-100 was added to a final concentration of 0.5%, and the cell lysates were centrifuged at 35,000 ϫ g for 40 min. For His 6 fusion proteins, clarified supernatants were loaded onto a nickel-nitrilotriacetic acid-agarose column (Qiagen). After washing with 10 column volumes of buffer (50 mM Tris, pH 8.0, 500 mM NaCl, 10 mM imidazole, 0.1% mercaptoethanol), His 6 fusion proteins were eluted with a gradient of 10 -150 mM imidazole. For GST fusions, the supernatants were loaded onto a glutathione-Sepharose column (Amersham Biosciences) equilibrated with 50 mM Tris, pH 8.0, 0.1 M NaCl, 0.1% 2-mercaptoethanol. After washing with 10 column volumes of the same buffer, GST fusion proteins were eluted with 10 mM reduced glutathione. Subsequent ion exchange chromatography using Source Q or Source S (Amersham Biosciences) followed by gel filtration chromatography over Superdex-75 (Amersham Biosciences) resulted in preparations that were Ͼ99% pure as judged by SDS-PAGE. To generate the untagged form of Rab5C constructs, GST fusion proteins at a concentration of 2-4 mg/ml were incubated with 2 g/ml human ␣-thrombin (Hematologic Technologies) overnight at 4°C in 50 mM Tris, pH 8.0, 2 mM CaCl 2 , and 0.1% mercaptoethanol. Following incubation with glutathione-agarose to remove residual fusion protein, the cleaved Rab5C constructs were further purified by ion exchange and gel filtration chromatography. Typical yields of purified proteins range from 10 to 100 mg/liter of bacterial culture. For Rab GTPases, all buffers are supplemented with 2 mM MgCl 2 .
Co-precipitation-GST-Rab5-(18 -185) was exchanged at 37°C for 30 min with a 25-fold molar excess of GppNHp in 50 mM Tris, pH 8.5, containing 5 mM EDTA, 100 mM NaCl, and 2 units of agarose-immobilized alkaline phosphatase per mg of protein. The exchange reaction was quenched by addition of 10 mM MgCl 2 , and excess nucleotide was removed by gel filtration on Superdex-75. His 6 -EEA1 constructs were incubated in a 1:1 molar ratio with GDP-bound or GppNHp-bound GST-Rab5-(18 -185) at a concentration of 20 M for 30 min at 4°C in buffer A (50 mM Tris, pH 8.5, 100 mM NaCl, 2 mM MgCl 2 , 0.1 mg/ml bovine serum albumin, and 0.1% Tween 20). 50 l of equilibrated glutathione-Sepharose beads (Amersham Biosciences) were added to 100 l of the protein mixture and incubated for 1 h. Following centrifugation, the supernatant was collected and the pellet washed three times with 100 l of buffer A. After washing, the beads were incubated with buffer A containing 10 mM glutathione for 15 min and the fractions analyzed by SDS-PAGE with Coomassie Blue staining.
Surface Plasmon Resonance-SPR sensograms were collected with a Biacore X instrument (Amersham Biosciences AB) using a carboxymethylated (CM5) sensor chip to which a GST antibody was covalently coupled using reagents and protocols supplied by the manufacturer. All proteins were dialyzed into flow buffer (10 mM Tris pH 7.5, 150 mM NaCl, 2 mM MgCl 2 , 0.005% Tween 20) prior to injection. Tandem flow cells were utilized; one was loaded with the 500 nM GST-Rab5C (sample channel), and the other was loaded with an equivalent molar quantity of GST (reference channel) expressed and purified as described above for the GST-Rab5 constructs. GST and GST-Rab5C were injected at a flow rate of 5 l/min, and subsequent injections were conducted at a flow rate of 20 l/min. Conversion to the active conformation was achieved by injecting 50 l of 3 M Rabex-5 followed immediately by a 10-l injection of 200 nM GppNHp. Binding and dissociation were monitored following 20-l injections of increasing concentrations of His 6 -EEA1. Following curve alignment, the reference sensogram, which reflects bulk refractive index changes and/or reversible nonspecific binding, was subtracted from the sample sensogram. The SPR signal at equilibrium (R eq ) was extracted from the fit with a simple 1:1 Langmuir binding model and plotted as a function of His 6 -EEA1 concentration. Dissociation constants (K d ) were obtained from a fit to the hyperbolic binding function R eq ϭ R max (His 6 EEA1)/(K d ϩ (His 6 EEA1)), where R max corresponds to the SPR signal at saturation and is treated as an adjustable parameter. Mean values and standard deviations ( nϪ1 ) were calculated from 2 to 4 independent measurements. Control experiments verify that the fitted K d values are independent of flow rate (5-50 l/min) and surface coverage (10-fold range), indicating that the equilibrium data are not limited by mass transfer or rebinding.
Intrinsic Tryptophan Fluorescence-Rab5C at concentrations of 1 or 20 M in 10 mM Tris, pH 7.5, 150 mM NaCl, 2 mM MgCl 2 was titrated with His 6 -EEA1-(36 -91), His 6 -EEA1-(36 -218), or the W104A mutant of His 6 -EEA1-(36 -218). Samples were excited at 290 nm (1 M Rab5C) or 300 nm (20 M Rab5C) with a 2 nm bandpass and emission spectra recorded from 300 -400 nm (1 nm bandpass) using an ISS spectrofluorimeter. The magnitude of fluorescence quenching (⌬I) at 340 nm was calculated as ⌬I ϭ I Ϫ I 0 , where I and I 0 are the emission intensities in the presence and absence of EEA1, respectively. Values for the dissociation constant (K d ) and the number of binding sites (n) were obtained by a non-linear least squares fit to a simple two-state binding model t are the total concentrations of EEA1 and Rab5C, respectively, and ⌬I max corresponds to the emission intensity at saturation. Concentrations were determined from the absorbance at 280 nm using calculated extinction coefficients ⑀ 280 (M Ϫ1 cm Ϫ1 ) ϭ number of Trp ϫ 5200 ϩ number of Tyr ϫ 1200 ϩ number of Cys ϫ 120.
Sedimentation Equilibrium-His 6 EEA1 constructs were dialyzed against 50 mM Tris, pH 7.5, 150 mM NaCl and centrifuged to equilibrium in an Optima XLI analytical ultracentrifuge (Beckman Instruments). The absorbance at 230 (A 230 ) or 280 nm (A 280 ) was measured as a function of the radial distance (r) from the axis of rotation. The x values of the data were transformed as m ⅐(r 0 2 Ϫ r 2 /2), where r 0 was taken as the last point in each data set and m was calculated with SEDINTERP (41) using the monomer molecular mass for each construct. Data were compared with the function A(r) ϭ C 0 ϩ C 1 ⅐exp(Ϫn⅐ m ⅐(r 0 2 Ϫ r 2 )/2), where C 0 and C 1 are constants, and n represents the order of the oligomeric state.
Homology Modeling-The sequence of the EEA1 C 2 H 2 Zn 2ϩ finger was threaded against a protein structure data base using the threedimensional PSSM fold recognition server to identify a suitable structure for homology modeling. The NMR structure of a C 2 H 2 Zn 2ϩ finger from the yeast transcription factor Adr1 (Protein Data Bank code 1paa) was selected for further homology modeling on the basis of its low scoring E value and the absence of gaps in the alignment. The EEA1 C 2 H 2 Zn 2ϩ finger shares 29% identity with that of Adr1, which represents the closest homologue of known structure. Non-conserved residues were substituted with the corresponding residues in EEA1, which were modeled in the most frequently observed rotomer conformation compatible with the structure. The resulting homology model represents a rough, working approximation to the actual structure, with an overall fold consistent with the common topology of C 2 H 2 Zn 2ϩ fingers.

RESULTS
The active forms of Rab5A, Rab5B, and Rab22 have been shown to interact directly with EEA1-(1-209) (11,35,39). This region of EEA1 encompasses a hydrophilic sequence of ϳ35 residues, the C 2 H 2 Zn 2ϩ finger, and two consecutive heptad repeats. As shown in Fig. 1A, His 6 EEA1-(1-218) co-precipitates with GST-Rab5C loaded with the non-hydrolyzable GTP analogue, GppNHp, but does not co-precipitate with the GDPbound form or in the presence of the Zn 2ϩ -chelating agent TPEN. These results are consistent with two-hybrid experiments in which mutation of a cysteine residue involved in Zn 2ϩ coordination disrupts the interaction with constitutively active Rab5A and Rab5B mutants (11,39). Although these observations demonstrate a requirement for an intact C 2 H 2 Zn 2ϩ finger, it is not clear whether this reflects a direct interaction between the C 2 H 2 Zn 2ϩ finger and Rab5 or whether the C 2 H 2 Zn 2ϩ finger plays an indirect structural role by supporting interactions with flanking regions as is the case for the double Zn 2ϩ finger of the Rab3A effector Rabphilin3A (40).
As shown in Fig. 1, B-D, the interaction of GST-Rab5C-GppNHp with the N terminus of EEA1 can also be detected and quantitatively analyzed by SPR in a BIAcore instrument using a monoclonal GST antibody coupled to a CM5 sensor chip. When injected at concentrations in the low micromolar range, His 6 EEA1-(1-218) exhibits reversible binding to the GppNHpbound form of GST-Rab5-(18 -185) as judged by the amplitude of the SPR signal compared with the GST reference channel (Fig. 1B). The signal in the reference channel rises and decays within the response time of the instrument, scales linearly with the concentration of His 6 EEA1-(1-218), and therefore represents either a bulk refractive index change or weak, reversible nonspecific binding indistinguishable from a bulk refractive index change. Under the conditions of these experiments, the association of N-terminal EEA1 constructs with GST-Rab5C approaches equilibrium on the time scale of the injection (Fig.  1C). The quantity bound at equilibrium (R eq ) saturates at low micromolar concentrations of His 6 EEA1-(1-218) (Fig. 1D). The data are well fit by a simple Langmuir binding isotherm, yielding a dissociation constant (K d ) of 3.3 M. In contrast, GST-Rab5-(18 -185) loaded with GDP shows no detectable binding to His 6 EEA1-(1-218), as expected for a bona fide GTPaseeffector interaction. An equivalent affinity (K d ϭ 2.3 M) is observed for the binding of His 6 EEA1-(36 -218) to full-length GST-Rab5-GppNHp, which includes the hypervariable N-and C-terminal extensions, indicating that the interaction determinants reside within the GTPase domain.
To map the minimal interaction site at the N terminus of EEA1 and determine whether the C 2 H 2 Zn 2ϩ finger is sufficient for Rab5 binding, SPR experiments were used to analyze quantitatively the binding of GST-Rab5-(18 -185) loaded with GppNHp to a panel of His 6 EEA1 truncation constructs ( Fig.  2A). Elimination of the first 35 residues, corresponding to the hydrophilic N terminus, has no significant effect on the interaction. Likewise, C-terminal truncations eliminating part or all of the heptad repeats show relatively small differences in affinity, which likely reflect systematic variations in the physical properties of the constructs. Indeed, His 6 EEA1-(36 -91), which lacks the N-terminal hydrophilic region and both heptad repeats, binds in a nucleotide-dependent manner to GST-Rab5-(18 -185) with an affinity comparable with that of His 6 EEA1- (1-218) (compare Fig. 2, B and C, with Fig. 1, C and D). Consistent with this observation, EEA1-(91-218), which lacks the N-terminal hydrophilic region and C 2 H 2 Zn 2ϩ finger, shows no detectable binding to GppNHp-bound GST-Rab5- (18 -185) at concentrations up to 150 M (the highest concentration tested). A shorter construct corresponding to the minimal C 2 H 2 Zn 2ϩ finger defined by homology (EEA1-(36 -74)) expressed poorly in bacteria and could not be fully purified. Although not suitable for quantitative analysis by SPR, this construct coprecipitates with GST-Rab5-(18 -185) loaded with GppNHp but not GDP and, by this measure, does not differ significantly from EEA1-(36 -91) (data not shown). We therefore conclude that the C 2 H 2 Zn 2ϩ finger is both necessary and sufficient for the interaction of Rab5C with the N terminus of EEA1.
Rab5C contains two tryptophan residues (Trp-74 and Trp-114), whereas the N terminus of EEA1 contains a single tryptophan residue (Trp-104). When titrated with His 6 EEA1-(36 -91), which lacks tryptophan residues, the intrinsic tryptophan fluorescence of untagged GppNHp-bound Rab5-(18 -185) undergoes significant quenching accompanied by a small shift in the emission maximum (Fig. 3B). Both effects saturate at low micromolar concentrations of His 6 EEA1-(36 -91), indicative of a binding interaction. As shown in Fig. 3C, the change in intrinsic tryptophan fluorescence is well described by a simple hyperbolic binding model, which yields a K d of 1.1 M, in good agreement with the affinity of His 6 EEA1-(36 -91) for GppNHpbound GST-Rab5-(18 -185) measured by SPR. Consistent with these observations, the intrinsic tryptophan fluorescence of untagged Rab5-GDP is not perturbed by addition of His 6 EEA1- (36 -91).
To determine whether homodimerization influences the stoichiometry of Rab5C binding to the N terminus of EEA1, titration experiments employing intrinsic tryptophan fluorescence to monitor binding were conducted under conditions where the concentration of the fixed component (GppNHp-bound Rab5-(18 -185)) was roughly 7-fold greater than the measured K d . The resulting data were analyzed with a titration binding model (see "Experimental Procedures") that relates the change in intrinsic tryptophan fluorescence to the binding stoichiometry (n), K d , and the maximum change in intrinsic tryptophan fluorescence at binding saturation (⌬F max ). Because K d values cannot be determined accurately under titration conditions, it was fixed at the value obtained from the experiment in Fig. 3C. Titration with His 6 EEA1-(36 -91) yields a stoichiometry of 0.89, consistent with 1:1 binding. Titration with His 6 EEA1-(36 -218) is complicated by the presence of a single tryptophan residue in the longer EEA1 construct. Because the magnitude of the change in GppNHp-bound Rab5-(18 -185) intrinsic tryptophan fluorescence is considerably larger than the intrinsic tryptophan fluorescence contributed by EEA1-(36 -218), the observed signal decreases monotonically until the majority of GppNHp-bound Rab5-(18 -185) is bound by EEA1- (36 -218), at which point the signal increases monotonically reflecting the contribution from excess EEA1-(36 -218) (data not shown). However, titration with the W104A mutant of EEA1-(36 -218), which binds with an affinity comparable with the wild type protein in the SPR experiment, yields a stoichiometry of 1.1, consistent with two molecules of GppNHp-bound Rab5C binding to identical independent sites at the N terminus of homodimeric EEA1.
To facilitate further characterization of the structural requirements underlying the interaction of Rab5C with the N terminus of EEA1, a working homology model for the EEA1 C 2 H 2 Zn 2ϩ finger was constructed from the NMR structure of a C 2 H 2 Zn 2ϩ finger from the Adr1 transcription factor (42). The Adr1 structure was identified by threading against a structural data base using the three-dimensional PSSM fold recognition server (43) and was selected from the lowest E value structures on the basis of the sequence identity (29%) and the absence of gaps in the alignment. Non-identical residues in the Adr1 C 2 H 2 Zn 2ϩ finger were replaced with the corresponding residues in EEA1, which were modeled in the most common rotomer conformation. Although the resulting homology model does not represent an accurate representation of the actual structure, it is likely that the overall topology and approximate location of residues are preserved. The latter assertion is supported by extensive structural studies of weakly homologous C 2 H 2 Zn 2ϩ fingers, which share a common ␤␤␣-fold.
The C 2 H 2 Zn 2ϩ fingers of EEA1 and Rabenosyn conserve a number of residues in addition to those required for Zn 2ϩ coordination or stability (Fig. 4A). Most of these residues are partially exposed in the homology model and cluster on a common surface, suggestive of a putative Rab5 interaction epitope (Fig. 4B). To test this hypothesis, seven residues (Glu-39, Phe-  Table I). Consistent with these observations, we are unable to detect the binding of GppNHp-bound Rab5C- (18 -185) to N-terminal EEA1 constructs by isothermal titration microcalorimetry (data not shown), suggesting that the interaction is entropically driven, presumably by burying exposed hydrophobic surfaces. The results of the mutational analysis suggest that Rab5C should be capable of binding to the C 2 H 2 Zn 2ϩ finger of Rabenosyn but not to the C 2 H 2 Zn 2ϩ finger of Vac1, due to non-conservative substitutions involving critical residues (see Fig. 4A). To test this hypothesis, the C 2 H 2 Zn 2ϩ fingers of Rabenosyn and Vac1 were expressed as GST fusions, and the interaction with His 6 Rab5C-(18 -185) was determined by SPR (Fig. 6). Whereas the C 2 H 2 Zn 2ϩ finger of Rabenosyn binds to Rab5C-(18 -185) with a K d of 0.6 M, the C 2 H 2 Zn 2ϩ finger of Vac1 exhibits no detectable binding at concentrations of Rab5C-(18 -185) as high as 150 M (the highest tested).

DISCUSSION
The crystal structure of a constitutively active mutant of Rab3A bound to the minimal Rab3A binding domain of Rabphilin3A provides the only available structural data on the interactions of Rab GTPases with effectors (40). The Rab3A binding domain of Rabphilin3A consists of an N-terminal helix, a double Zn 2ϩ finger with a fold similar to that of a FYVE domain, and a C-terminal loop/helix. Although the double Zn 2ϩ finger does not contact Rab3A directly, it serves an indirect structural role by supporting interactions with the N-terminal helix and C-terminal loop/helix. The interaction of Rab5 with the C terminus of EEA1 requires an intact FYVE domain and is disrupted by point mutants in the proximal coiled coil region (25,32). Although the interaction with the C terminus has not been quantitatively analyzed, Rab5 does not co-precipitate   with either the proximal coiled coil or the FYVE domain alone, suggesting that the binding site encompasses both regions (32). By using purified reagents and quantitative measures of binding, we have shown that C 2 H 2 Zn 2ϩ finger (residues 36 -74) is both necessary and sufficient for the interaction of Rab5C with the N terminus of EEA1. Thus, the function of the C 2 H 2 Zn 2ϩ finger at the N terminus of EEA1 differs fundamentally from the indirect structural role of the double C 2 H 2 Zn 2ϩ finger of Rabphilin3A. The mode of interaction also differs from that at the C terminus of EEA1, which requires both the FYVE domain and the proximal coiled coil.
Two variations are common in classic C 2 H 2 Zn 2ϩ fingers, the "consensus" motif (CX 2 CX 3 FX 5 LX 2 HX 3-5 H) and the "swapped" motif (CX 2 CXFX 7 LX 2 HX 3-5 H) (44,45). These motifs differ in the location of the central phenylalanine residue, which in either case packs in the small hydrophobic core with the conserved leucine and one of the conserved histidine residues. Both motifs adopt a similar ␤␤␣-fold stabilized by a tetrahedral Zn 2ϩ ion, which forms a tetrahedral coordination complex with the thiol groups of the conserved cysteine residues contributed by adjacent strands of the ␤-hairpin and the imidazole side chains of the conserved histidine residues at the C terminus of the ␣-helix. The C 2 H 2 Zn 2ϩ finger of EEA1 closely resembles the consensus motif with the exception that the phenylalanine residue following the conserved cysteine residues is replaced by a leucine residue. The corresponding substitution has been studied in the context of the C 2 H 2 Zn 2ϩ finger of ZFY, where it has little effect on the structure or affinity for Zn 2ϩ , although it does increase the overall dynamic mobility (45). Whether the latter effect would be compensated by other substitutions in EEA1 or otherwise contribute to the interaction with Rab5 is not clear. Although it is possible that the observed defects in some mutants reflect an indirect effect on the folding and/or structure of the C 2 H 2 Zn 2ϩ finger, alanine residues occur naturally in C 2 H 2 Zn 2ϩ fingers at each of the positions examined in this study. Consistent with this observation, C 2 H 2 Zn 2ϩ fingers have been shown to be tolerant of alanine substitutions at non-consensus positions (46). Furthermore, each mutated residue occupies an exposed or partially exposed position in the structural homology model. Finally, the mutants with severe defects involve residues that cluster on a common surface. Therefore, we favor the interpretation that the defects arise from altered interactions with Rab5.
Rab5C contains two tryptophan residues, one located in the ␤3-strand at the interface between the switch I and II regions (Trp-74) and the other located in the ␣3-helix adjacent to the switch II region (Trp-114). In the crystal structure of GppNHpbound Rab5C, Trp-74 is partially exposed, whereas Trp-114 is buried (47). Although it is possible that the observed changes in intrinsic tryptophan fluorescence arise indirectly from structural rearrangements in the vicinity of one or both tryptophan residues, a more straightforward explanation is that the partially exposed Trp-74 is located in or adjacent to the epitope for interaction with the C 2 H 2 Zn 2ϩ finger of EEA1. Trp-74 is flanked by two other partially exposed hydrophobic residues, Phe-58 in the switch I/␤2-region and Tyr-90 in the switch II region. This triad of invariant hydrophobic residues constitutes a prominent, exposed hydrophobic patch in the active form of Rab GTPases with known structure and lies at the core of the interface between the switch regions of Rab3A and Raphillin3A (40). We have shown previously (47) that the conformation of the invariant hydrophobic triad varies between Rab GTPases and thereby contributes to the specificity of Rab-effector interactions. Moreover, the hydrophobic triad is located adjacent to variable loop regions also implicated in Rab-effector specificity (40). We therefore propose that the invariant hydrophobic triad in Rab5 lies at the core of a largely non-polar interface with the cluster of conserved hydrophobic residues in the C 2 H 2 Zn 2ϩ finger of EEA1. This hypothesis is generally consistent with known modes of GTPase-effector interaction and would explain the apparent lack of a significant enthalpic component to the free energy of binding inferred from the inability to detect the interaction by isothermal titration microcalorimetry. A more detailed understanding of the structural basis underlying the interaction of Rab5 with the N terminus of EEA1 will require the structure of the C 2 H 2 Zn 2ϩ finger in complex with Rab5.
In an earlier study, we have shown that the C-terminal endosome targeting region of homodimeric EEA1 forms an organized quaternary structure that would support additive, bivalent binding of two molecules of PtdIns(3)P in the context of a lipid bilayer (34). Bivalent binding is expected to enhance significantly the affinity for membranes containing PtdIns(3)P; however, the isolated FYVE domain of EEA1 does not form stable dimers. These observations explain the strict requirement of the proximal heptad repeat for endosome targeting and why it can be bypassed by fusing two FYVE domains in tandem (48). Although the heptad repeat at the N terminus also provides the driving force for stable dimerization, it neither enhances nor interferes with Rab5 binding. Moreover, titration experiments indicate that the N terminus of the EEA1 homodimer supports simultaneous, independent binding of two Rab5 molecules. Whether one or two molecules of Rab5 bind in a cellular context would depend on the effective concentration of active Rab5 on the membranes of endosomes or endocytic vesicles.
EEA1 has been hypothesized to tether early endosomes and endocytic vesicles by interacting with Rab5 on one endosome/ endocytic vesicle and PtdIns(3)P on another. It seems unlikely that the low micromolar K d for Rab5 binding to the N terminus of EEA1 would be sufficient for stable endosome tethering. However, the combined action of multiple EEA1 molecules, each with the potential for bivalent Rab5 binding, could generate a stably tethered intermediate with considerable dynamic flexibility, given the glycine-rich hydrophilic sequence following the C 2 H 2 Zn 2ϩ finger, several predicted hinge regions in the heptad repeat near the N terminus of EEA1, and the glycine/proline rich hydrophilic sequence connecting the GTPase domain of Rab5 to the dual prenylation motif at the C terminus.
EEA1 possesses two distinct Rab5-binding sites, one corresponding to the C 2 H 2 Zn 2ϩ finger and the other overlapping the C-terminal FYVE domain and proximal coiled coil region. We estimate that the solution affinity of unprenylated Rab5 for C-terminal constructs of EEA1 is at least an order of magnitude weaker than that observed for the C 2 H 2 Zn 2ϩ finger. Though weaker, the C-terminal interaction may nevertheless contribute significantly in vivo, as a consequence the restricted dimensionality resulting from co-localization on endosome membranes. Indeed, point mutants that disrupt Rab5 binding to the C-terminal site exhibit a dominant negative phenotype when expressed at high levels in cultured cells (32). Like EEA1, Rabenosyn also possesses two apparently independent Rab5binding sites corresponding to the C 2 H 2 Zn 2ϩ finger (present study) and a second site within the C-terminal third of the protein (38). Although the C 2 H 2 Zn 2ϩ fingers of EEA1 and Rabenosyn bind Rab5 with an affinity and nucleotide specificity characteristic of bona fide GTPase effectors, the functional significance of these interactions has not been tested in vivo. If, as hypothesized, the N-terminal interaction with Rab5 plays a critical role in the tethering and/or fusion of early endosomes, then cells expressing full-length EEA1 constructs carrying the F41A or I42A mutations in the C 2 H 2 Zn 2ϩ finger should exhibit a strong dominant negative phenotype. We anticipate that these mutants and the analogous point mutations in Rabenosyn will provide useful reagents for exploring the in vivo functional role of the homologous C 2 H 2 Zn 2ϩ fingers in the context of the full-length proteins.