Dimeric Switch of Hakai-truncated Monomers during Substrate Recognition

Background: The novel phosphotyrosine-binding domain (HYB) of Hakai forms an atypical, zinc-coordinated homodimer.

Results: C-terminal truncation of the HYB domain causes dramatic structural changes and it becomes a monomer, but in the presence of substrate, it becomes a dimer to recognize the substrate.

Conclusion: HYB dimerization is a unique prerequisite for substrate-binding activity of Hakai.

Significance: The HYB domain may orchestrate the function of Hakai in cancer and cell-cell contacts.

Abstract

Hakai, an E3 ubiquitin ligase, disrupts cell-cell contacts in epithelial cells and is up-regulated in human colon and gastric adenocarcinomas. Hakai acts through its phosphotyrosine-binding (HYB) domain, which bears a dimeric fold that recognizes the phosphotyrosine motifs of E-cadherin, cortactin, DOK1, and other Src substrates. Unlike the monomeric nature of the SH2 and phosphotyrosine-binding domains, the architecture of the HYB domain consists of an atypical, zinc-coordinated tight homodimer. Here, we report a C-terminal truncation mutant of the HYB domain (HYB^ΔC), comprising amino acids 106–194, which exists as a monomer in solution. The NMR structure revealed that this deletion mutant undergoes a dramatic structural change caused by a rearrangement of the atypical zinc-coordinated unit in the C terminus of the HYB domain to a C₂H₂-like zinc finger in HYB^ΔC. Moreover, using isothermal titration calorimetry, we show that dimerization of HYB^ΔC can be induced using a phosphotyrosine substrate peptide. This ligand-induced dimerization of HYB^ΔC is further validated using analytical ultracentrifugation, size-exclusion chromatography, NMR relaxation studies, dynamic light scattering, and circular dichroism experiments. Overall, these observations suggest that the dimeric architecture of the HYB domain is essential for the phosphotyrosine-binding property of Hakai.

Introduction

Phosphotyrosine (Tyr(P))³-binding domains are key determinants of specificity and selectivity in many signal transduction pathways and act by integrating Tyr(P) signals from upstream kinases to downstream effectors that regulate the complex physiology of eukaryotic cells (1–3). The SH2 domain was the first Tyr(P)-binding domain to be discovered (4, 5) and has since been extensively characterized (1, 6–10). The second domain to be identified with a capacity to bind tyrosine residues was the phosphotyrosine-binding domain (11). Subsequently, idiosyncratic Tyr(P)-binding domains have been observed in the C2 domain of protein kinase Cδ (12) and pyruvate kinase M2 (13), and we recently reported the existence of a third Tyr(P)-binding domain in Hakai, coined the HYB domain for Hakai phosphotyrosine (Tyr(P)) binding (14).

Hakai is an E3-ubiquitin ligase first noted for its role in regulating E-cadherin expression and disrupting cell-cell contacts in epithelial tissues (15, 16). Subsequent work has identified the elevated expression of Hakai in human colon and gastric adenocarcinomas (17, 18). The HYB domain interacts with other Src substrates, such as Cortactin, a structural protein involved in coordinating actin rearrangement during cell movement (19), and DOK1, a scaffolding protein that assists in the assembly of signaling complexes (20). Both of these proteins offer important functional contributions in the progression of cancer (21, 22), and both are regulated by the activity of Hakai (14). The novel features of the HYB domain and its interaction with various key molecules may indicate a physiologically important role for Hakai in cancer.

The structure of the HYB domain was determined by x-ray crystallography (14). It comprises amino acids 106–206, and forms an atypical, zinc-coordinated homodimer in an antiparallel, intertwined configuration, utilizing residues from the Tyr(P)-binding region of two Hakai monomers. The dimeric nature of the domain configures the formation of a Tyr(P)-binding pocket that recognizes specific phosphorylated tyrosine residues and flanking acidic amino acids of its substrates (14). The C-terminal region, which harbors the atypical zinc-coordination motif and key residues involved in the Tyr(P) interaction, plays an important role in the dimerization observed in the HYB domain. In the present study, we investigated the consequences of deleting residues from the C terminus of Hakai, whereas maintaining the essential residues involved in the Tyr(P) interaction and the atypical zinc-coordination of the E3 ligase. Through this structural truncation, we identified a C-terminal deletion mutant of Hakai (aa 106–194), herein referred to as HYB^ΔC, which exists as a monomer in solution and flips to a dimeric arrangement in the presence of a Tyr(P) substrate, with the incoming substrate inducing the conformational changes required to initiate dimerization of HYB^ΔC monomers. This monomeric to dimeric switch of HYB^ΔC in the presence of the phosphorylated substrate was further validated by biophysical studies. Taken together, our results suggest that the dimerization of Hakai Tyr(P)-binding domain is a unique prerequisite for substrate binding.

EXPERIMENTAL PROCEDURES

Protein Expression and Purification

Sequences corresponding to aa 106–194 of Hakai were cloned into pGEX6P-1 vector (GE Healthcare) using BamHI and SalI restriction sites, and expressed as a GST fusion protein in Escherichia coli BL21 (DE3) cells. Cells were cultured in LB medium at 37 °C until the A_{600 nm} reached 0.6–0.7. Cells were then incubated with 0.15 mm isopropyl 1-thio-β-d-galactopyranoside and 50 μm zinc sulfate, and then grown for an additional 20 h at 15 °C. Cells were collected by centrifugation (8000 × g, 10 min, 4 °C), and the pellets were resuspended in lysis buffer (50 mm BisTris, pH 6.5, 300 mm NaCl, 10 μm zinc chloride, 5% glycerol, 0.5% Triton X-100, 2 mm DTT, 1 mm phenylmethylsulfonyl fluoride) and homogenized using a French Press Cell Disrupter (Thermo Scientific, Wilmington, DE). Cell lysates were then centrifuged at 18,000 × g for 30 min at 4 °C (JA-25.50 fixed angle rotor centrifuge, Beckman Coulter, Fullerton, CA) and the supernatant was passed through glutathione-Sepharose resin (GE Healthcare) for 2–4 h at 4 °C. The resin was subsequently washed with a buffer containing 50 mm BisTris, pH 6.5, 300 mm NaCl, 5% glycerol, and 2 mm DTT, and the bound supernatant was then subjected to an overnight on-column cleavage at 4 °C with GST-PreScission Protease (GE Healthcare) to remove the GST tag. A major portion of GST and GST-PreScission Protease remained bound to the glutathione-Sepharose resin and the flow-through containing the partially purified, untagged proteins were further purified using a Superdex 75 size exclusion column (GE Healthcare) equilibrated with a buffer containing 50 mm sodium phosphate buffer, pH 6.5, and 5 mm DTT.

Circular Dichroism Spectrometry

Far UV spectra (260–190 nm) of HYB (aa 106–206) and HYB^ΔC (aa 106–194) in the absence and presence phosphorylated peptide of E-cadherin (residues 747–759) were measured using a Jasco J-810 spectropolarimeter in phosphate buffer, pH 6.5, at room temperature with a 0.1-cm path length and stoppered cuvettes. The protein concentration was maintained at 20 μm in all cases. Six scans were recorded, averaged, and the baseline subtracted. The scale on the CD spectra was normalized to mean residue ellipticity (MRE), which is measured in degrees cm² dmol⁻¹ residue⁻¹ using the following equation (23),

where θ is raw ellipticity measured in machine units (millidegrees), MRW is mean residue weight (MRW = protein mean weight (in daltons)/(number of residues) for the protein, P is path length in cm, and CONC is protein concentration in mg/ml. The MRE is plotted against the corresponding wavelength.

Dynamic Light Scattering

Dynamic light scattering studies were carried out on a DynaPro Light Scattering instrument (Wyatt Technology Europe GmbH, Dernbach, Germany) with protein concentrations at A_{280 nm} of 1.0, in buffer containing 50 mm sodium phosphate, pH 6.5, and 5 mm DTT.

NMR Spectroscopy

All NMR experiments were carried out at 25 °C on a Bruker Avance 800 MHz spectrometer equipped with a TXI cryogenic probe using 1 mm ¹³C,¹⁵N-labeled HYB^ΔC (aa 106–194) sample prepared in the buffer containing 50 mm sodium phosphate, pH 6.5, and 5 mm DTT. ¹H, ¹³C, and ¹⁵N resonance assignments were achieved by measuring the three-dimensional HNCACB, three-dimensional CBCA(CO)NH (24), and three-dimensional CCH-TOCSY (25) spectra. Inter-proton distance restraints for structural calculation were obtained from three-dimensional ¹³C-edited NOESY-HSQC, three-dimensional ¹⁵N-edited NOESY-HSQC, and two-dimensional NOESY spectra using a 100-ms mixing time. A weakly aligned ¹⁵N-labeled sample (0.8 mm) was prepared using the same buffer by the addition of 6 mg/ml of filamentous phage Pf1 (from ASLA Biotech Ltd, Latvia) for residue dipolar coupling (RDC) measurements. ¹D_NH RDCs were measured using the In-Phase and Anti-Phase method (26). The RDC values were obtained by subtracting the reference value in isotropic solution. NMR spectra (two- and three-dimensional) were processed using the NMRPipe program (27), and data analysis was performed with the help of the Sparky program (28).

Structure Determination

The solution structure for HYB^ΔC (aa 106–194) was calculated using the Xplor-NIH 2.24 software package (29). A total of 1087 NOE distance restraints, 21 hydrogen bonds, and 110 dihedral angle restraints were predicted by the TALOS program (30). Initial structures were calculated using torsion angle molecular dynamics protocol. Structure refinement was performed using a simulated annealing protocol, and 55 RDCs restraints were used in the cooling stage. The position of three zinc ions were determined by the distance restraints derived from the x-ray structure of HYB (PDB code 3VK6). The three zinc ions are bound to the HYB^ΔC through nine cysteines and three histidines exhibiting the topology Cys¹⁰⁹-X₂-Cys¹¹²-Zn₁-Cys¹³⁰-X₂-Cys¹³³, Cys¹²⁵-X-His¹²⁷-Zn₂-Cys¹⁴⁵-X₂-Cys¹⁴⁸, and Cys¹⁶⁶-X₅-Cys¹⁷²-Zn₃-His¹⁸⁵-X₄-His¹⁹⁰, respectively. These zinc ions were included in the structural calculations with an approximately tetrahedral geometry. The distance constraints include four fixed distances for each zinc ion, where the zinc-sulfur (cysteine S^γ) bond distance was set to 2.3 Å and zinc-nitrogen (histidine N^ϵ2) bond lengths was set to 2.0 Å, respectively. A final ensemble of the 20 lowest energy structures from 100 calculated structures was selected for the figure preparation using Chimera program.

¹⁵N Relaxation Measurements

The ¹⁵N single-labeled free HYB^ΔC sample (0.8 mm) and complex sample (∼1 mm peptide added) were used to measure the ¹⁵N relaxation time, T₁ and T₂, respectively. The ¹⁵N relaxation times, T₁ and T_1ρ, were measured at 25 °C using inversely detected two-dimensional NMR methods (31, 32) on a Bruker 800 Avance machine equipped with a TXI cryogenic probe. For the free HYB^ΔC sample, relaxation times, T₁, were determined by collecting eight points with delays of 10, 250 (×3), 400, 500, 650, 750, and 850 ms using a recycle delay of 3 s and 8 scans. Relaxation times, T_1ρ, were measured by collecting eight points with delays of 1, 30 (×2), 45, 60, 75, 90, and 110 ms using a spin-lock power of 1.6 kHz, a 2.0-s recycle delay, and 8 scans. For the complex sample, relaxation times, T₁, were determined by collecting eight points with delays of 10, 250 (×2), 400, 500, 650, 750, and 1000 ms using a recycle delay of 3 s and 16 scans. Relaxation times, T_1ρ, were measured by collecting eight points with delays of 1, 15, 30 (×2), 45, 60, 75, and 87 ms using a spin-lock power of 1.6 kHz, a 2.0-s recycle delay, and 16 scans. All data were recorded using 220 and 1280 complex points in t₁ and t₂ dimensions, respectively, and with spectral widths of 1946 Hz (¹⁵N) and 11160 Hz (¹H). Relaxation times were fitted as single exponential decays to peak height data. The spin-spin relaxation time, T₂, was calculated from T_1ρ and T₁ according to the following equation (32),

where θ = arctan(Δω/ω₁), and Δω and ω₁ are the resonance offset and spin-lock field strength, respectively. The experiment error was analyzed by repeating the T₁ measurement (250 ms) three times. The relative standard deviation for HSQC peak intensities was about 0.99%. The relaxation constants were also given with the fitting errors. The differences in relaxation parameters observed were significantly higher than the standard deviation; thus, it was outside the error range.

Isothermal Titration Calorimetry

The phosphorylated peptide of E-cadherin (residues 747–759) was titrated at a molar concentration of 575 μm against 75 μm HYB^ΔC in a VP-ITC microcalorimeter (Microcal, Northhampton, UK) at 293 K. The titrations were carried out using 30 10-μl injections of the peptide into the sample cell containing HYB^ΔC and the data were analyzed with a one-site binding model using the Origin software package version 7.0 supplied by Microcal. All measurements were repeated three times.

Analytical Ultracentrifugation (AUC)

The oligomeric states of the HYB^ΔC in the absence and presence of the Tyr(P) peptide were investigated by monitoring their sedimentation properties in AUC sedimentation velocity experiments. Samples (400 μl) in 50 mm sodium phosphate, pH 6.5, and 5 mm DTT with an absorbance of 1.0 at 280 nm were used. Sedimentation velocity profiles were collected by monitoring the absorbance at 280 nm. The samples were centrifuged at 40,000 × g at 25 °C in a Beckman ProteomeLab XL-I centrifuge fitted with a four-hole AN-60 Ti rotor and double-sector aluminum centerpieces and equipped with absorbance optics. The scans were analyzed using the Sedfit program (33).

Chemical Shift Perturbation Analysis

Chemical shift perturbations of backbone amides by the binding of phospho-E-cadherin-(747–759) peptide to HYB^ΔC (peptide to protein ratio is 1.2:1) were calculated. The chemical shift perturbations were defined as Δδ = [(Δδ_HN² + Δδ_N²/25)/2]^0.5 for amide NH, where Δδ_HN and Δδ_N are the chemical shift differences of amide ¹H, amide ¹⁵N, between the samples in the presence of phospho-E-cadherin-(747–759) peptide and in the absence of phospho-E-cadherin-(747–759) peptide. The NH HSQC assignments for HYB^ΔC complex were confirmed by three-dimensional ¹⁵N-edited NOESY-HSQC experiment.

RESULTS

HYB^ΔC Is a Monomer in Solution

We previously demonstrated that the HYB domain consists of an atypical zinc-coordinated, intertwined homodimer formed by an anti-parallel arrangement of two Hakai (aa 106–206) monomers (14). The dimerization mainly occurs through the C-terminal region of Hakai (aa 106–206), which harbors key Tyr(P)-interacting residues as well as the atypical zinc-coordinated motif formed by two histidine residues (His¹⁸⁵ and His¹⁹⁰) and one cysteine residue (Cys¹⁷²) from one monomer and a second cysteine residue (Cys¹⁶⁶) from the adjacent monomer. In this study, we expressed and purified a C-terminal truncation mutant comprising residues 106–194, herein referred to as HYB^ΔC, which retains all of the residues involved in the atypical zinc-coordinated motif and the Tyr(P) interaction. The gel-filtration elution profile showed that HYB^ΔC had an apparent molecular weight equivalent to a monomeric subunit (Fig. 1A). Furthermore, the monomeric nature of HYB^ΔC was verified using AUC analysis, which showed a peak corresponding to an apparent molecular mass of monomeric HYB^ΔC (∼10,200 Da) (Fig. 1B). The dynamic light scattering experiments also revealed an apparent molecular weight of monomeric HYB^ΔC.

FIGURE 1.

Gel filtration, analytical ultracentrifugation, and circular dichroism of the HYB^ΔC (aa 106–194). A, gel filtration profile of HYB^ΔC. HYB^ΔC was loaded onto a calibrated Superdex-75 gel filtration chromatography column. The elution profile suggests that HYB^ΔC exists as a monomer. The SDS-PAGE analysis of HYB^ΔC depicts the purity and the molecular mass of the peak fraction (∼10 kDa). B, AUC analysis of HYB^ΔC. The monomeric nature of HYB^ΔC was studied using sedimentation velocity analysis. The molecular mass profile of HYB^ΔC indicates that the protein exists as a monomer in solution, with an apparent molecular mass of 10 kDa. C, circular dichroism analysis shows a significant difference in the secondary structure composition between HYB^ΔC and the untruncated HYB domain. Furthermore, in the presence of the phosphotyrosine peptide ligand (E-cadherin-(747–759)), the CD spectrum of HYB^ΔC shows marked resemblance with that of the HYB domain, suggesting a conformational change in HYB^ΔC in the presence of the phosphorylated ligand.

Circular Dichroism Analysis

Circular dichroism (CD) analysis revealed significant differences in the secondary structure composition between the monomeric HYB^ΔC and the dimeric HYB domain (Fig. 1C). In the presence of the phosphorylated peptide of E-cadherin (corresponding to residues 747–759), the CD spectra demonstrated conformational changes to the HYB^ΔC. These significant differences in the conformation of HYB^ΔC in the CD spectral analysis motivated us to determine the NMR structure of HYB^ΔC to understand the structural basis of this ligand-associated conformational change.

NMR Structure of HYB^ΔC

To gain structural insights into the tertiary fold of HYB^ΔC, we solved the solution NMR structure. The NMR structure of HYB^ΔC was refined to a final r.m.s. deviation of 0.63 ± 0.21 Å (backbone) (Table 1) for the 20 best structures (Fig. 2, A and B). The NMR structure of HYB^ΔC revealed a monomeric fold, comprising an N-terminal RING domain and a C-terminal zinc-coordinated domain bearing a C₂H₂-type zinc finger (ZnF) configuration (Fig. 2, B and C). As expected from knowledge of the HYB structure, the solution structure of HYB^ΔC contained three zinc-coordination sites. The RING domain (residues 106–148) was shown to adopt a typical RING architecture containing two β-strands (β1 and β2) and one α-helix (α1), which engage two zinc ions. The C-terminal domain (residues 148–194) lay in close association with the N-terminal RING domain by means of the third β-strand, β3, which formed an anti-parallel β-sheet arrangement with β1 and β2 of the RING domain. The C-terminal domain contained the third zinc-coordination site of HYB^ΔC, formed by the residues Cys¹⁶⁶, Cys¹⁷², His¹⁸⁵, and His¹⁹⁰, resembling a typical C₂H₂ ZnF motif (Fig. 2C); the two cysteines (Cys¹⁶⁶ and Cys¹⁷²) formed a hairpin loop and the two histidines (His¹⁸⁵ and His¹⁹⁰) were found juxtaposed to the α-helix (α2).

TABLE 1.

NMR data and structure determination details for HYB^ΔC

NMR data
All NOE distance restraintsa	1087
Intra-residue	224
Sequential (|i–j| = 1)	338
Medium range (1 <|i–j|<5)	193
Long range (|i–j|≥5)	344
Hydrogen bond restraints	21
Dihedral angle restraints (ϕ, ψ)b	110
Residual dipolar coupling restraints	55
Energy statistics X-PLOR energy (kcal mol−1)
Enoe	76.1 ± 4.8
Ecdih	1.26 ± 0.26
Erdc	20.2 ± 3.7
Deviations from idealized covalent geometryc
R.m.s. deviations of bond lengths (Å)	0.0031 ± 0.0031
R.m.s. deviations of bond angles (°)	0.430 ± 0.015
R.m.s. deviations of improper angles (°)	0.358 ± 0.011
Deviations from experimental restraints
R.m.s. deviations of distance restraints (Å)	0.0369 ± 0.0012
R.m.s. deviations of dihedral angle restraints (°)	0.265 ± 0.074
Ramachandran plot analysis (%)d
Residues in allowed region	96.9%
Residues in generally allowed regions	2.3%
Residues in disallowed regions	0.8%
Average R.m.s. deviations from mean structure (Å)e
Heavy atoms	1.29 ± 0.21
Backbone atoms (N, CA, C′, O)	0.63 ± 0.21

^a The distance restraints were obtained by classifying the NOE cross peaks into three categories: strong (1.8–2.9 Å), medium (1.8–3.5 Å), and weak (1.8–5.0 Å).

^b Dihedral angles of backbone φ and ψ were predicted by TALOS (30) using the chemical shifts of Cα, Cβ, Hα, N, and HN.

^c Twenty lowest-energy conformers with no NOE violations greater than 0.3 Å and no torsion angle violations greater than 3° were selected from 100 conformers to represent the NMR ensembles.

^d Calculated with PROCHECK-NMR (66).

^e Calculated with MOLMOL (67) over secondary structure region α1 (131–140), α2 (179–190), β1 (119–123), β2 (127–130) and β3 (154–158).

FIGURE 2.

NMR structure of HYB^ΔC (aa 106–194). A, superposition of 20 energy-minimized conformers representing the NMR structure of HYB^ΔC. B, ribbon diagram of the lowest-energy conformer representing the three-dimensional NMR structure of HYB^ΔC. Each monomer contains three zinc-coordination sites: Sites 1, 2, and 3. The zinc ions are shown as red spheres. C, metal coordinations in the NMR structure of HYB^ΔC. The coordination of zinc ions in the RING domain and the C-terminal domain of HYB^ΔC are shown.

The overall structure of HYB^ΔC was compared with other proteins present in the Protein Data Bank using the DALI server (34). The DALI search showed several structural homologs containing a similar RING domain, but none bore any resemblance beyond amino acid residue 166 of the C-terminal domain of HYB^ΔC, the C₂H₂ ZnF site. This suggested that combination of the N-terminal RING domain and the C-terminal domain containing the C₂H₂ ZnF adopts a novel fold in HYB^ΔC. To determine whether any protein in the PDB database contained a fold resembling the C₂H₂ ZnF of HYB^ΔC, a second DALI search was performed for the localized fold comprising amino acid residues 159–194 of HYB^ΔC. The results showed only four proteins that contain a fold similar to the C₂H₂ ZnF of HYB^ΔC: zinc finger protein 406 (PDB code 2elx), SPIF2 (PDB code 1sp2), RAS-related protein RAB-5A (PDB code 3mjh), and SWI5 (PDB code 1zfd), bearing a sequence similarity of 30, 27, 28, and 19%, and Z-scores of 2.4, 2.3, 2.1, and 2.1, respectively. These proteins are transcription factors, with the exception of RAB-5A, which is involved in Rab5 binding using the distinctive N-terminal C₂H₂ zinc finger (35).

Comparison of the Structure of HYB^ΔC with HYB Domain

To analyze the structural differences between monomeric HYB^ΔC and HYB domain, the monomeric HYB^ΔC was superimposed onto that of one of the monomeric counterparts of Hakai (aa 106–206). The structural segment corresponding to residues 106–166 of HYB^ΔC and one monomer of the HYB domain were found to be similar, with a r.m.s. deviation of 0.9 Å (Fig. 3). This implies that the RING domain (residues 106–148) and the fold of the linker sequence (residues 149–166) overlap in both HYB^ΔC and the monomer of the HYB domain. Structural differences arose, however, from residue 167 onward, which corresponds to the C-terminal zinc-coordination site and the terminal α-helix (α2). The C-terminal helix α2 of HYB^ΔC was flipped by 180° and is separated from the α2 of HYB monomeric counterpart by a distance of 25 Å (Fig. 3C). Fig. 4, A–F, show comparisons between the monomeric HYB^ΔC and the dimeric fold of the HYB domain. From these comparisons, it became apparent that the differences in the folding arose from changes in the C₂H₂ ZnF in HYB^ΔC, with all of the zinc-coordination residues situated on the same monomeric chain of HYB^ΔC; this is in contrast to the sharing of zinc ions between the two monomers in the atypical zinc-coordination motif of the HYB domain. Thus, the NMR structure provides the basis for the existence of HYB^ΔC as a monomer in solution and bearing a different fold, likely because its lacks the atypical zinc-binding motif, which is critical for the dimerization of the full-length HYB domain. The overall topology of HYB^ΔC and the HYB domain is compared in Fig. 4, E and F.

FIGURE 3.

Comparison of monomeric HYB^ΔC with monomeric Hakai of the dimeric HYB domain. A and B, schematic representations of HYB^ΔC (aa 106–194; green) and the Hakai (aa 106–206; yellow) monomeric counterpart (PDB code 3vk6) of the HYB domain, respectively, in the same orientation. C, superposition of HYB^ΔC on Hakai reveals an overlap with a r.m.s. deviation of 0.9 Å for a stretch of residues from 106 to 166, which contains the RING domain and part of the C-terminal domain in both structures. Differences in the structure arise after residue Cys¹⁶⁶ (highlighted by a red dotted circle, lower right of the figure). Beyond this, the folds of HYB^ΔC and the HYB monomeric counterpart are completely different, with the helix α2 flipping by 180° and separating by ∼25 Å. The relative rotations that result in the conformational change are indicated by arrows orientated in a clockwise direction. D, schematic representation of the conformational changes in C of the structures of HYB^ΔC and Hakai (aa 106–206) of the HYB domain with the relative rotation resulting in the observed structural changes. The yellow dot represents the original position of the C-terminal zinc ion at Site 3 of the monomeric Hakai counterpart in the dimeric HYB domain, and the green dot represents the new position of the Site 3 zinc ion in HYB^ΔC.

FIGURE 4.

C₂H₂ ZnF in the C-terminal domain of HYB^ΔC is critical for its monomeric fold. Analysis of the structural differences between HYB^ΔC and HYB (A) and the dimeric fold of HYB domain containing paired Hakai (aa 106–206) (PDB code 3vk6) monomers (B). The red dot in A represents the zinc ion in the C-terminal domain of HYB^ΔC, whereas the cyan and yellow dots in B represent the C-terminal zinc ion shared between the respective monomers of the dimeric HYB-fold. C, the monomeric fold of HYB^ΔC (green) contains a C₂H₂ ZnF in the C-terminal domain (highlighted by red dotted circle). D, the HYB domain (PDB code 3vk6) contains a dimeric fold of paired Hakai (aa 106–206) monomers (yellow monomer in ribbon representation; cyan in schematic representation for clarity). In monomeric HYB^ΔC, all four zinc-coordinating residues are situated in the same monomeric chain. Comparatively, the dimeric fold of the HYB domain is formed by the sharing of two zinc ions (red spheres) between the Hakai monomers, such that three of the four zinc-binding residues come from one monomer (His¹⁸⁵, His¹⁹⁰, and Cys¹⁷²) and the fourth from the adjacent monomer (Cys¹⁶⁶). The zinc-coordinating side chains are shown as sticks. The structural rearrangement arises because of a systematic reshuffling among the zinc-coordinating residues in the monomeric and dimeric conformers. Topology of HYB^ΔC (E) is compared with that of the HYB domain (PDB code 3vk6) (F). The zinc ions are shown as spheres (color coded according to the monomer colors) and the three zinc-coordination sites are indicated for each monomer (Sites 1, 2, and 3). In HYB^ΔC, Site 3 is formed by four zinc-coordination residues, all situated from the same monomer. For the HYB domain, Site 3 and Site 3′ share residues from both monomers.

HYB^ΔC Binds Phospho-E-cadherin Ligand as a Dimer

We next questioned whether monomeric HYB^ΔC could still engage with a phosphorylated substrate ligand. Hence, we conducted isothermal titration calorimetry (ITC) experiments to investigate the binding property of HYB^ΔC with the peptide ligand corresponding to our previously established Tyr(P) motif of E-cadherin, comprising residues 747–759 (14). The ITC experiment revealed that HYB^ΔC binds with the Tyr(P) ligand of E-cadherin with a binding stoichiometry of 2:1 (HYB^ΔC:Tyr(P) ligand) and an affinity value of K_d = 3.7 μm (Fig. 5A). This indicates that the phospho-E-cadherin-(747–759) ligand interacts with a dimeric form of HYB^ΔC.

FIGURE 5.

ITC, gel filtration profiles, AUC, and NMR relaxation studies of HYB^ΔC with phospho-E-cadherin-(747–759). A, the tyrosine-phosphorylated E-cadherin-(747–759) was titrated against HYB^ΔC using ITC. The top panel shows the heat release profile after baseline correction and the lower panel indicates the binding isotherm for the interaction. The dissociation constant (K_d) and binding stoichiometry (N) are shown in the table. B, comparison of gel filtration profiles of HYB^ΔC in the presence (blue) and absence (red) of phospho-E-cadherin-(747–759) using a Superdex 75 gel filtration column. The elution profile suggests that HYB^ΔC exists as a dimer in the presence of ligand but as a monomer in the absence of the ligand. C, AUC analysis of HYB^ΔC in the presence of phospho-E-cadherin-(747–759) ligand. The ligand-induced dimerization of HYB^ΔC was studied using sedimentation velocity analysis. The results show that the protein exists as a dimer in the presence of the ligand with an apparent molecular mass of 20,000 Da. D, ¹⁵N relaxation T₁ (a) and T₂ (b) values as well as the error values for each residues in free HYB^ΔC sample (labeled as a red star) and phospho-E-cadherin^747–759 bound complex (labeled as ●). The results show the T₁ and T₂ values change in the presence and absence of the substrate peptide. The secondary structures and domain boundary of HYB^ΔC also are illustrated in the middle of the figure. The missing residues and weak intensity residues (except Val¹²⁸) are located in the region between β3 and α2.

To further understand the dimeric nature of this interaction, HYB^ΔC was separated on a calibrated gel-filtration column in the presence and absence of the ligand. A comparison of these two elution profiles showed that HYB^ΔC elutes as a dimer in the presence of the ligand and as a monomer in its apo form (Fig. 5B). To further verify this presumed ligand-induced dimerization of HYB^ΔC, sedimentation velocity AUC experiments were conducted with the samples containing HYB^ΔC in the presence of the phospho-E-cadherin-(747–759) peptide. The analysis of the AUC data using Sedfit (33) showed that the apparent molecular mass of HYB^ΔC in the presence of ligand is 20,000 Da (Fig. 5C), which is equivalent to twice the monomeric molecular weight of HYB^ΔC. Interestingly, the ¹⁵N relaxation data clearly showed a change in the T₁ and T₂ values in the presence and absence of the substrate peptide (Fig. 5D). The average T₁ and T₂ values for the free HYB^ΔC were 0.88 s and 120 ms, respectively. However, upon binding to the peptide, the average T₁ and T₂ values for the complex were 1.02 s and 100 ms, respectively. These results indicate that a high molecular weight complex is formed upon binding. Dynamic light scattering results further confirmed these findings, with an apparent molecular weight corresponding to a dimer in the presence of the substrate peptide. The combined results suggest that HYB^ΔC switches from a monomeric conformation to a dimeric form in the presence of a Tyr(P)-containing ligand.

Proposed Model of Tyr(P) Recognition by HYB^ΔC

To understand the mode of Tyr(P) recognition by HYB^ΔC, we undertook chemical shift perturbation analyses of HYB^ΔC in the presence of phospho-E-cadherin-(747–759). The results showed a significant shift for α2 helix in the presence of phospho-E-cadherin-(747–759) (Fig. 6, A and B). This chemical shift perturbation upon ligand binding with HYB^ΔC bears marked similarity to the one observed in our previous study wherein the ¹⁵N dimeric HYB domain was titrated with the phospho-E-cadherin-(747–759) peptide (14). Given that the presence of a Tyr(P) ligand changes the oligomerization, we analyzed the proposed dimeric configuration formed by the HYB^ΔC monomers in the presence of Tyr(P) ligand. We compared the HYB^ΔC monomer structure with the corresponding monomer of the untruncated HYB domain and observed a major conformational change in the α2 helix of the HYB^ΔC (Figs. 3 and 4). Our observations from the solution experiments, structural analysis, chemical shift perturbation data, and previous mutational data (14) suggest that the Tyr(P) ligand induces the dimerization to create a binding pocket in HYB^ΔC for substrate recognition (Fig. 6C). The electrostatic surface potential representation of HYB^ΔC in the phospho-E-cadherin-(747–759) peptide complex model suggests that the Tyr(P) along with the surrounding acidic residues on the phospho-E-cadherin-(747–759) peptide potentially interact with the positively charged HYB^ΔC binding pocket (Fig. 6D).

FIGURE 6.

Proposed model of Tyr(P) recognition by HYB^ΔC. A, an overlay of the HN-HSQC spectra of HYB^ΔC in the absence (green) or presence (red) of phospho-E-cadherin-(747–759) peptide. B, chemical shift perturbations of backbone amides by the binding of phospho-E-cadherin-(747–759) peptide to HYB^ΔC. The chemical shift perturbations were defined as Δδ = [(Δδ_HN² + Δδ_N²/25)/2]^0.5 for amide NH, where Δδ_HN and Δδ_N are the chemical shift differences of amide ¹H, amide ¹⁵N, between the samples in the presence and absence of phospho-E-cadherin-(747–759) peptide. The NH HSQC assignments for HYB^ΔC complex were confirmed by three-dimensional ¹⁵N-edited NOESY-HSQC experiment. The empty regions represent no information available for residues in those regions because the NH HSQC peaks of those residues were invisible or unassigned. The secondary structures of HYB^ΔC also are illustrated in the middle of the figure. C, a proposed model of Tyr(P) recognition by the ligand-induced dimeric configuration of the HYB^ΔC monomers. The respective monomers are shown in green and magenta, with the peptide ligand in orange. The zinc ions are shown as red spheres. The HYB^ΔC·Tyr(P) peptide complex model was generated based on the chemical shift data as well as our previous mutational and binding studies (14), which showed the interaction of His¹²⁷, Tyr¹⁷⁶, His¹⁸⁵, and Arg¹⁸⁹ of HYB with the E-cadherin. Although the interacting side chains from the Tyr(P) peptide are not yet clearly established, our previous mutational studies showed that Asp⁷⁵⁰, Val⁷⁵², Tyr(P)⁷⁵⁴, Asp⁷⁵⁶, and Glu⁷⁵⁷ are the interacting sites from the phospho-E-cadherin-(747–759) peptide (14). D, an electrostatic surface potential representation of the model presented in C suggests that the Tyr(P) along with its surrounding acidic residues of the phospho-E-cadherin-(747–759) interacts with the positively charged HYB^ΔC binding pocket. The Tyr(P) peptide ligand is shown in orange.

DISCUSSION

Phosphotyrosine-binding domains are critical, modular components that bind to phosphorylated tyrosine residues in acceptor proteins to create multiprotein complexes and regulate several intracellular signaling pathways (1, 8, 36). Dysregulation of these pathways is often associated with oncogenic transformation (37), rendering Tyr(P)-binding domains as attractive targets for directed therapies (38–40). So far, all of the major Tyr(P)-binding domains, including SH2 and phosphotyrosine-binding domains, tend to be monomeric (7, 41), barring the few exceptions where they function as homodimers. These exceptions include the SH2 domains of Grb10, Grb14, Grb7, APS, and SH2-B (7, 41–43). In addition, in STAT proteins, tyrosine phosphorylation mediated dimerization has been reported to occur via SH2 domains (44–46). The recently identified Tyr(P)-binding fold in Hakai, HYB, bears a novel dimeric fold consisting of two atypical ZnFs shared between the paired Hakai (aa 106–206) monomers that exist in an intertwined configuration across the flexible C-terminal region (14). Previously we showed that the atypical, intermolecular zinc coordination is necessary for the dimerization of Hakai and hence its ability to interact with the substrate (14). Mutations of these residues involved in intermolecular zinc coordination results in dimer to monomer transition; and also abrogate the ability of full-length Hakai to bind its targets like E-cadherin and Cortactin, as revealed in the mammalian cell based studies (14). In this study, we showed that deletion of the flexible C-terminal region of HYB also results a monomeric fold in solution. A similar observation was recently reported in the C-terminal domain of SARS-CoV main protease (M^pro-C), where the truncation of the disordered C-terminal helix results in its transition from a dimer to a monomer (47). In yet another study, a dimer/tetramer equilibrium observed in Escherichia coli DNA mismatch repair protein MutS is converted into a monomer/dimer equilibrium upon deletion of the C-terminal 53 amino acids (48).

To understand the structural basis for this monomeric conformation of HYB^ΔC, we determined its solution NMR structure. The NMR structure revealed that the characteristic atypical ZnF, which plays a key role in the dimeric fold of the HYB domain, is replaced by a C₂H₂-type ZnF in HYB^ΔC, formed by two cysteines and two histidines, all situated on the same monomer. This represents a unique monomeric to dimeric switching mechanism.

One of the best studied physiological functions of the HYB domain is its interaction with and regulation of tyrosine-phosphorylated E-cadherin at cell-cell junctions (14, 15, 49, 50). As such, we selected a tyrosine-phosphorylated E-cadherin-(747–759) peptide as a model substrate to further assess this monomeric conformation of HYB^ΔC and to determine whether it is still functional. The ITC interaction studies showed that the ligand binds with the protein only in its dimeric form. This result is consistent with our previous studies that a dimeric HYB fold is necessary to create the Tyr(P) binding pocket; this was also validated with full-length Hakai in cell-based experiments (14). A series of solution studies further validated this switch to a dimeric conformation of HYB^ΔC in the presence of the Tyr(P) E-cadherin-(747–759) ligand. Furthermore, all of these experiments were carried out at different concentrations of HYB^ΔC and suggest that the ligand-induced dimerization of HYB^ΔC is independent of protein concentration. Previous studies have demonstrated ligand-mediated dimerization as a novel mechanism for protein-carbohydrate recognition (51, 52). The present study extends upon this paradigm to Tyr(P) signaling and regulation in Hakai.

Protein dimerization is an important regulatory mechanism in signal transduction (53–55). Numerous RING finger ubiquitin ligase family members require dimerization for their function; for example, cIAP, XIAP, RNF4, SIAH, and TRAF2 act as homodimers (56–60), whereas heterodimers form between MDM2 and MDMX, BRCA1 and BARD1, and RING1b and BMI1, respectively (61). Previous studies have shown that dimerization is often necessary for the interaction of the E2 ubiquitin conjugate with the RING domain (56). However, the dimerization of Hakai monomers reported in this study likely represents a unique mechanism, as the incoming substrate, rather than association with the RING domain, mediates the dimerization. This might represent a critical mechanism to facilitate the conformational switch Hakai requires for substrate binding and thus for the ubiquitination of the substrates such as E-cadherin.

In conclusion, we have identified a novel monomeric fold of HYB^ΔC, which impairs its Tyr(P) binding property. However, in the presence of the Tyr(P) ligand, the monomeric HYB^ΔC becomes a dimer to create the Tyr(P) binding pocket to engage the substrate. The findings from this study suggest that the dimeric architecture of the HYB domain is necessary to engage the Tyr(P) ligand, which is in sharp contrast to all the other known Tyr(P)-binding domains that predominantly function as monomers. Selectively targeting the dimeric interface of therapeutically important enzymes has emerged as an attractive method of allosteric inhibition (62–65). The importance of the dimeric architecture of HYB in Tyr(P) substrate binding demonstrated by the present study makes it an ideal target for the design of selective allosteric inhibitors that abrogate HYB dimerization and potentially act as novel therapeutic interventions against cancer.