Structural Studies of the Nedd4 WW Domains and Their Selectivity for the Connexin43 (Cx43) Carboxyl Terminus*

Neuronal precursor cell-expressed developmentally down-regulated 4 (Nedd4) was the first ubiquitin protein ligase identified to interact with connexin43 (Cx43), and its suppressed expression results in accumulation of gap junction plaques at the plasma membrane. Nedd4-mediated ubiquitination of Cx43 is required to recruit Eps15 and target Cx43 to the endocytic pathway. Although the Cx43 residues that undergo ubiquitination are still unknown, in this study we address other unresolved questions pertaining to the molecular mechanisms mediating the direct interaction between Nedd4 (WW1–3 domains) and Cx43 (carboxyl terminus (CT)). All three WW domains display a similar three antiparallel β-strand structure and interact with the same Cx43CT 283PPXY286 sequence. Although Tyr286 is essential for the interaction, MAPK phosphorylation of the preceding serine residues (Ser(P)279 and Ser(P)282) increases the binding affinity by 2-fold for the WW domains (WW2 > WW3 ≫ WW1). The structure of the WW2·Cx43CT276–289(Ser(P)279, Ser(P)282) complex reveals that coordination of Ser(P)282 with the end of β-strand 3 enables Ser(P)279 to interact with the back face of β-strand 3 (Tyr286 is on the front face) and loop 2, forming a horseshoe-shaped arrangement. The close sequence identity of WW2 with WW1 and WW3 residues that interact with the Cx43CT PPXY motif and Ser(P)279/Ser(P)282 strongly suggests that the significantly lower binding affinity of WW1 is the result of a more rigid structure. This study presents the first structure illustrating how phosphorylation of the Cx43CT domain helps mediate the interaction with a molecular partner involved in gap junction regulation.

Gap junction channels serve to directly interconnect the cytoplasm of neighboring cells, allowing the passage of ions, metabolites, and signaling molecules. Gap junction intercellular communication is an important process that mediates electrical impulse propagation, regulation of cell growth, and whole organ development. Gap junction channels are formed by the apposition of connexons from adjacent cells where each connexon is formed by six connexin proteins. Connexins share a similar topology of four transmembrane domains, two extracellular loops, and three cytoplasmic domains (N terminus, cytoplasmic loop, and C terminus (CT) 3 ). Of the 21 human connexins, connexin43 (Cx43) is the most completely characterized isoform in terms of channel gating properties (1), identified phosphorylation sites (2), known protein partners (3), connexon assembly (4), and channel degradation (5)(6)(7).
Unlike most membrane proteins, connexins exhibit exceptionally high turnover rates with half-lives ranging from 1.5 to 5 h (8). The initial step leading to degradation, internalization from the plasma membrane, is a complex process because docked connexons are unable to be separated under physiological conditions (9). Consequently, removal from the plasma membrane is directionally regulated as internalization of the channels within one of the coupled cells forms a double membrane macrostructure called an annular gap junction or connexosome (10); clathrin and other endocytic adaptors have been shown to be involved in this process (11)(12)(13)(14). Degradation of connexins involves both the lysosomal (endosome or autophagy) and proteasomal pathways (15)(16)(17). Implicated in the regulation of these processes are post-translational modifications (18). For example, studies using cultured cells have shown that activation of protein kinase C (PKC) or mitogenactivated protein kinases (MAPK) induces hyperphosphorylation and subsequent degradation of Cx43 (19 -21).
Another type of post-translational modification universally implicated in protein degradation is ubiquitination. Ubiquitin is a small molecule of 8.5 kDa that covalently attaches to lysine residues of target proteins to alter their cellular location, affect their activity, modulate protein interactions, or signal for their degradation (22). Laing and Beyer (23) first demonstrated that ubiquitin was involved in connexin degradation. Using a cell line containing a thermolabile E1 ubiquitin-activating enzyme, they showed stable Cx43 protein levels upon heat treatment and in the presence of a proteasomal inhibitor. Subsequent studies have determined that the C-terminal domain of Cx43 (Cx43CT) is ubiquitinated (12,(23)(24)(25); however, the lysine residues involved are still unknown (Cx43 contains 23 cytoplasmic lysine residues: three in the N terminus, 11 in the cytoplasmic loop, and nine in the CT domain). Of note, however, a proteome-wide survey by mass spectroscopy recently detected in vivo ubiquitination of Cx43 at lysine residues 9 and 303 (26), but the significance remains unclear.
The neuronal precursor cell-expressed developmentally down-regulated 4 (Nedd4) was the first ubiquitin E3 ligase identified to directly interact with Cx43, and its suppression by siRNA results in an accumulation of Cx43 at the membrane (12,27). Nedd4 was discovered while screening genes down-regulated during development of the central nervous system in embryonic mice but was later found to be ubiquitously expressed across cell types and species (28,29). Nedd4 contains a catalytic HECT (homologous to the E6-AP carboxyl terminus) domain, three to four WW domains, and a C2 domain. The HECT domain binds to a ubiquitin-conjugated E2 and transfers the ubiquitin moiety from the E2 to the substrate, and the C2 domain mediates the localization of Nedd4 to the plasma membrane. The WW recognition domains, named after the presence of two conserved tryptophan residues, regulate the substrate selection. Rat Nedd4 contains three WW domains, which have been shown to interact with the Cx43CT (27). Found in a wide variety of signaling proteins (30), WW domains are well folded compact polypeptides containing 38 -40 amino acids and are divided into five groups according to their binding motif preference. Class I is the largest group and recognizes PPXY motifs where X represents any amino acid. Classes II, III, and V bind PPLP, PXXGMXPP, and PR-rich motifs, respectively. Class IV binds phosphorylated (S/T)P motifs. The Cx43CT binding site for WW2 was identified and defined as a class I motif (PPXY; Cx43 residues 283-286 (27)). This finding was supported by Girão et al. (12) who showed that the Cx43 mutation P283L (LPXY) reduced the binding affinity between Nedd4 and Cx43. Moreover, a subsequent study using a Cx43ubiquitin chimera with a Y286A mutation was shown to have increased internalization (31). Conversely, the absence of ubiquitin combined with the Y286A mutation led to the accumulation of Cx43 at the membrane and a decreased level of ubiquitination, suggesting the PPXY motif to be critical for Cx43 ubiquitination and internalization (31). However, upon examination of the Cx43CT sequence, two MAPK phosphorylation sites, Ser 279 and Ser 282 , are adjacent to the PPXY motif ( 279 SPMSPPGY 286 ) and if phosphorylated would form a class IV motif (pSPXpSP where pS is phosphoserine). Therefore, Cx43CT presents two different but overlapping WW binding motifs (PPXY and pSPXpSP) that could work cooperatively to increase the binding affinity with Nedd4 or independently to dictate which WW domain binds the Cx43CT. The possibility of Nedd4 also interacting with the class IV motif is supported by the observation that Cx43 MAPK phosphorylation after EGF stimulation increased binding to WW2 and WW3 (27). Additional unresolved questions from these studies were the location and nature of the binding motif on the Cx43CT for the WW1 and WW3 domains as well as the mechanism of differential WW domain binding caused by phosphorylation of the Cx43CT. Therefore, in this study, we used a combination of different biophysical techniques to assess these questions to provide a better understanding of the recognition between Nedd4 and Cx43 and help define the interplay between phosphorylation and ubiquitination in degradation of connexins.
Circular Dichroism (CD)-CD experiments were performed on a Jasco J-815 CD spectrometer at 25°C in the far UV (260 -190 nm) or near UV (340 -250 nm) with a 0.1-mm-path length quartz cell using a bandwidth of 1 nm, an integration time of 1 s, and a scan rate of 50 nm/min. Each spectrum is the average of five scans. All spectra were corrected by subtracting the solvent spectrum acquired under identical conditions. The protein concentration for each sample was between 0.55 and 2.2 mM in the far UV and between 3 and 9 mM in the near UV. All CD data were processed using the Spectra Analysis function of Jasco Manager Version 2. Thermal unfolding data were collected for each WW domain in the far UV regions (at 227.5 nm for WW1, 229.5 nm for WW2, and 228.0 nm for WW3) at the same concentrations as for the CD experiments (in a 1-cm cell). The temperature was regulated by a Peltier temperature controller. CD data were recorded every 1°C from 20 to 100°C after the temperature equilibrated for 5 s at Ϯ0.1°C from the target temperature. The thermal unfolding temperature was determined by fitting the curve to a Boltzmann sigmoidal equation using a nonlinear square fitting algorithm (GraphPad Prism software).
Nuclear Magnetic Resonance (NMR) Data Acquisition, Assignment, and Structure Calculation-NMR data were acquired at 7 or 25°C using a 600-MHz Varian INOVA NMR spectrometer outfitted with a cryoprobe. NMR experiments to determine the WW1, WW2, WW3, and WW2⅐Cx43CT 276 -289 (Ser(P) 279 , Ser(P) 282 ) peptide backbone sequential and side chain assignments were collected as described previously (33). NMR spectra were processed using NMRPipe (35) and analyzed using NMRView (36). NOE distance constraints for WW1, WW2, and WW2⅐Cx43CT 276 -289 (Ser(P) 279 , Ser(P) 282 ) were obtained from 15 N NOESY-HSQC and 13 C NOESY-HSQC spectra. Hydrogen bonds were identified from temperature dependence experiments and introduced as a pair of distance restraints. Model structures were calculated by simulated annealing using torsion angle dynamics as implemented in the program Crystallography & NMR System (37) and refined with the ARIA 1.2 (Ambiguous Restraints for Iterative Assignment) software (38,39). In the first ARIA round, manually assigned NOEs (cross-peak volumes) were included for the generation of Cx43CT/Nedd4 WW Domain Interaction initial folded structures. The unambiguous and ambiguous NOE restraints derived from ARIA outputs were further analyzed and used as inputs for the next round of calculations. The resulting list of NOE distance restraints was subjected to a final run of structure calculations where 20 structures were calculated using the simulated annealing protocol of ARIA. The 10 best energy-minimized structures were evaluated using PROCHECK-NMR (40) and selected for display and structural analysis. MOLMOL (41) and PyMOL programs were used for structure visualization. The 10 NMR ensemble structures have been deposited in Protein Data Bank under codes 2n8s, 2n8u, and 2n8t for WW1, WW2, and WW2⅐Cx43CT 276 -289 (Ser(P) 279 , Ser(P) 282 ), respectively.
Calculating Binding Affinity by NMR-Gradient-enhanced two-dimensional 15  Isothermal Titration Calorimetry (ITC)-Heat produced by the binding of Cx43CT 276 -289 peptide phosphorylated (Ser(P) 279 , Ser(P) 282 ) or non-phosphorylated with each of the Nedd4 WW domains was measured by ITC using the MicroCal iTC200 isothermal titration calorimeter from Malvern (Worcestershire, UK). All proteins were equilibrated in 1ϫ PBS at pH 7.4 by overnight dialysis. ITC binding isotherms were collected at 25°C by injecting 20 ϫ 2 l of peptide (1.7-7.9 mM) into a solution of each of the WW domains (90 -790 M), representing a 1:10 -1:30 molar ratio. The heat from each injection was measured by integrating the area of the injection peak, corrected for the background heat produced by the dilution of the peptide into the buffer, and plotted as a function of the WW domain/peptide molar ratio. K D values were calculated by fitting the titration curves according to a single binding site model with Origin 7 software with ITC add-ons supplied by Malvern.

Structural Characterization of the Nedd4 WW Domains-
Leykauf et al. (27) observed differential binding between the three Nedd4 WW domains (Fig. 1A) and the Cx43CT regardless of the Cx43CT phosphorylation state. Possible explanations for these observations include differences between the WW domains in their structure, dynamics, and/or residues at the binding interface. Therefore, CD spectra of the WW1-3 domains in the far UV (190 -250 nm) were collected to detect any difference in their overall secondary structure (Fig. 1B). The far UV spectrum for each WW domain is consistent with that of a protein containing ␤-sheet structure. The minimum usually observed around 217 nm for pure ␤-sheet is shifted to ϳ205 nm due to a significant content of random coil structure (42% for each domain) as calculated by DICHROWEB (42,43). Consist-ent with other members of the WW domain family, the two conserved tryptophan residues cause a positive ellipticity around 228 nm (44,45). Interestingly, WW1 also has a negative peak around 196 nm, which is normally positive in ␤-sheet structures, as observed for the WW2 and WW3 domains. The 196 nm peak is the result of the ⌸ 3 ⌸* transition (i.e. light absorbed), which is dominated by the backbone carbonyl ⌸ bond and is also affected by the involvement of the nitrogen in the ⌸ orbitals. All peaks observed in CD rely on the intensity and energy of transitions that depend on the and angles; therefore, the data suggest that the WW1 structure could be slightly different from the WW2 and WW3 structures. Moreover, the negative peak around 196 nm for WW1 could be caused by a number of amino acid side chains that have transitions in the peptide region (i.e. sequence-specific differences between the WW domains).
To further characterize the structural differences between the WW domains, NMR was used to solve the structures of the WW1 and WW2 domains. The NMR solution structure of WW3 has already been solved albeit in complex with a small peptide (46). All 1 H, 15 N, and 13 C backbone resonances and most side chain and aromatic groups of the WW1 and WW2 domains (97 and 95% of the expected resonances, respectively) were assigned in 1ϫ PBS (pH 7.4) at 25°C (Fig. 1C). The chemical shift data were deposited into the BioMagResBank database under accession numbers 25864 (WW1) and 25866 (WW2). Of note, most backbone and side chain resonances were assigned for the WW3 domain alone ( Fig. 1C; chemical shift data were deposited into the BioMagResBank database under accession number 26698) to calculate the binding affinity with different Cx43CT peptides. The solution structures of the WW1 and WW2 ( Fig. 2A, left and center, respectively) domains display a bundle of three antiparallel ␤-strands, resembling that of WW3 ( Fig. 2A, right; from Ref. 46). The coordinates were deposited in the Protein Data Bank under accession codes 2n8s (WW1) and 2n8u (WW2). The template modeling score that focuses on evaluating the topology fold similarity was above 0.5 for all WW pairings ( Fig. 2B) in agreement with a very comparable fold (47). Moreover, the root mean square deviation (r.m.s.d.) value observed between the WW domains is ϳ1 Å, which is indicative of a good structural superposition. WW1 has a larger number of hydrogen bonds and NOE restraints compared with WW2 (Table 1), suggesting a more rigid structure. The NMR data are supported by the near UV CD spectra of the WW domains (250 -340 nm), which generate a fingerprint of the chiral environment of aromatic residues, thus providing a tool to monitor their local conformation (Fig. 3A). The three WW domains have a similar tertiary structure based on the overall similarity in shape of the spectra between 250 and 290 nm. One difference is that the WW1 signal is significantly higher in intensity (caused by closer aromatic side chain packing) as compared with the WW2 and WW3 domains, suggesting a more rigid structure (48).
To confirm this difference in dynamics, CD was also used to evaluate the thermal stability of the three WW domains (Fig.  3B) higher temperature to unfold). Finally, protein flexibility was predicted using the random coil index method, which quantitatively estimates backbone root mean square fluctuations of structural ensembles and order parameters using only chemical shifts (49). In Fig. 3C, the model-free order parameter S2 is provided as a function of residue number. The order parameter S2 describes the amount of local mobility with S2 ϭ 1 for no local motion and S2 ϭ 0 for completely unrestricted local motion of the NH vectors (50). Although the local motion in ␤-strand 1 is similar between the WW1 and WW2 domains, the WW1 domain is more rigid in ␤-strands 2 and 3. Altogether, the structural data identified differences in the dynamic properties and thermostability between the Nedd4 WW domains. Although this may account for the different binding affinities with the Cx43CT (27), another possibility is that the sequence-specific residues in the binding pocket differ. To address this possibility, we next determined the residues involved in the Cx43CT interaction with the WW1 and WW3 domains. On the consensus line, stars (*), colons (:), and periods (.) represent fully conserved residues, residues with strongly similar properties, and semiconserved residues among the three domains, respectively. B, CD spectra of each Nedd4 WW domain in the far UV in 1ϫ PBS (pH 7.5). C, 15 N HSQC spectra of each Nedd4 WW domain in 1ϫ PBS (pH 7.5). Each amide cross-peak is labeled. Location of the Cx43CT Interaction with the Nedd4 WW1 and WW3 Domains-Although all three WW domains of Nedd4 were shown to interact with the Cx43CT (27), only the area of interaction for WW2 has been identified (within residues Cys 271 -Leu 288 ; contains PPXY motif), and a different Cx43CT region and binding motif were suggested for WW1 and WW3. To determine the portion of Cx43CT interacting with WW1 and WW3, 15 N HSQC spectra of Cx43CT 236 -382 were collected in the presence of various concentrations of each WW domain (Fig. 4A). A summary of the residues affected by each WW domain is displayed in Fig. 4B. The three Nedd4 WW domains strongly affect the Cx43CT (peaks broaden beyond detection) over an area encompassing Met 281 -Gly 291 , which includes the PPXY ( 283 PPGY 286 ) motif. A second area between Gly 261 and Gly 270 (peaks shift), which is commonly affected by protein partners binding anywhere along the Cx43CT domain (51), was weakly affected by the interaction with the WW domains. The Met 281 -Gly 291 sequence also overlaps with a potential Class IV WW domain binding motif pSPXpSP ( 279 pSPMpSP 283 ), suggesting that phosphorylation of the Cx43CT could influence the binding affinity with the WW domains.
Influence of Cx43CT Phosphorylation on the Interaction with Nedd4 WW1-3 Domains-Cx43CT phosphorylation upon MAPK activation (EGF treatment) was shown to increase the amount of Cx43 pulled down by Nedd4 GST-WW domains, notably WW2 and WW3 (27). MAPK has been identified to phosphorylate Cx43CT residues Ser 255 , Ser 262 , Ser 279 , and Ser 282 (52). Cx43 phosphorylation at Ser 279 and Ser 282 showed weak binding to WW2 and no binding to WW1 and WW3, suggesting that phosphorylated residues other than Ser 279 and Ser 282 increased the binding affinity between Cx43CT and the WW domains (27). However, because all three Nedd4 WW domains interacted with the same Cx43CT residues that contain the PPXY motif and overlap with residues of the pSPXpSP motif, Cx43CT peptides were designed to study the impact of the two motifs and the effect of Cx43CT phosphorylation on the binding affinity with the WW domains.
The K D values were calculated for each WW domain with the Cx43CT peptides by holding the concentration of the 15 N-labeled WW domain constant (50 M) and titrating the unlabeled CT peptides from 50 M to 2.5 mM (Fig. 5A). The changes in chemical shift of at least 5 residues affected as a result of the increasing Cx43CT peptide concentrations were then fitted according to the nonlinear least square method to estimate the binding constant (Fig. 5B). For the control peptide (no phosphorylation; PPXY motif), WW2 exhibits the highest binding affinity (K D ϭ 0.73 Ϯ 0.05 mM (mean Ϯ S.D.)) followed by WW3 and WW1 (1.30 Ϯ 0.20 and 1.70 Ϯ 0.10 mM, respectively) (Fig. 5C). Any modification of the critical PPXY motif residue Tyr 286 (mutation, Y286A; phosphorylation, Tyr(P) 286 ) with or without phosphorylation at Ser 279 and Ser 282 significantly reduced the binding affinity with all the WW domains. Next, the peptides (PPXY motif) were differentially phosphorylated to determine the influence of the pSPXpSP (Ser(P) 279 and Ser(P) 282 ) motif on the binding affinity with the WW domains. Ser(P) 279 alone had no influence on binding the WW domains. Ser(P) 282 significantly increased the binding affinity (WW2 ϳ WW3 Ͼ Ͼ WW1); however, phosphorylation at both Ser(P) 282 and Ser(P) 279 further increased the binding affinity for only WW2. These binding affinities were independently confirmed using a different biophysical technique (ITC; Fig. 6A) and a phosphate-independent buffer (HEPES; Fig. 6B). Altogether, the data indicate that the Cx43CT PPXY motif is critical for binding the WW domains, whereas the pSPXpSP motif modulates this interaction.
The importance of the PPXY motif was also evident when solution conditions (ϩMeOH or ϩKCl) were tested to identify the main driving force (hydrophobic or electrostatic) mediating the WW/Cx43CT domain interaction (Fig. 7). [ 15 N]WW2 and the Cx43CT peptide (Ser(P) 279 , Ser(P) 282 ; 1:4 molar ratio) were collected in the presence of 10 or 25% MeOH, 1 M KCl, or both 10% MeOH and 1 M KCl. Although addition of 10% MeOH clearly reduced the shifts observed upon addition of the CT peptide compared with the shifts in 1ϫ PBS (i.e. reduced the binding affinity), the addition of high salt (1 M KCl) only had a minor impact on the interaction. When 1 M KCl was added to a solution already containing 10% MeOH, the interaction further decreased in comparison with 10% MeOH alone. The data suggest that the dominant force mediating the Cx43CT and WW domain interaction is hydrophobic with a contribution from electrostatic interactions, most likely involving Tyr 286 and Ser(P) 279 /Ser(P) 282 , respectively.
Mechanism of Cx43CT Phosphorylation-based Increase in the WW Binding Affinity-The WW1 and WW2 domains of Smurf1 or YAP (Yes kinase-associated protein) can simultaneously bind a peptide from Smad1 that contains both a PPXY and a pSP motif (53). Because the Nedd4 WW2 and WW3 domains are in close proximity (21 amino acids apart), the possibility exists that the PPXY and pSPXpSP motifs are meant to interact simultaneously with different WW domains in a cooperative  ; Fig. 8A). 15 N HSQC spectra of WW2-WW3 in the presence of various concentrations of a Cx43CT peptide containing the PPXY and pSPXpSP (Ser(P) 279 , Ser(P) 282 ) motifs were collected. The amount of Cx43CT peptide needed to reach saturation with WW2-WW3 was doubled compared with the individual WW domains. After calculating the K D for each WW domain, the values were found to be consistent with the binding affinity obtained when the WW domains were alone (Fig. 8B), suggesting a 1:2 binding ratio and no cooperative binding. To understand the structural basis for the WW2 domain to have the highest binding affinity for the Cx43CT and the role of phosphorylation in this interaction, we solved the solution structure of the WW2 domain in the presence of the Cx43CT Ser(P) 279 /Ser(P) 282 peptide (Ala 276 -Val 289 ) by NMR. Structure calculation by torsion angle dynamics followed with refinement by simulated annealing and energy minimization led to the family of structures shown in Fig. 9A. The structures were superimposed on the basis of the backbone coordinates from the WW2 residues Gly 407 -Lys 432 , which excluded the disordered N and C termini (Ser 401 -Gly 407 and Pro 433 -Asp 437 ). The WW2 and Cx43CT peptide backbone r.m.s.d. values for this alignment were 0.87 Ϯ 0.21 and 1.94 Ϯ 0.61 Å, respectively. The structural models fit the NMR data well with no violations of experimental distance restraints greater than 0.29 Å. Structural statistics are presented in Table 2. The overall structure of the  WW2 domain displays a bundle of three antiparallel ␤-strands similar to the WW2 domain alone (r.m.s.d., 1.24 Å; data not shown). The overall structure of the Cx43CT peptide displays a horseshoe-shaped arrangement. The critical Cx43 tyrosine residue (Tyr 286 ) anchors to the side chains of WW2 residues Tyr 419 , Val 421 , His 423 , Lys 426 , Thr 427 , and Thr 428 ; they comprise the front face of the WW2 ␤2-loop-␤3 structure (Fig. 9B, top). Pro 283 and Pro 284 form a short polyproline II helix that recognizes a pocket known as the XP groove formed by the second conserved tryptophan (Trp 430 ) and the highly conserved tyrosine (Tyr 419 ) in the WW domain (54). Phosphorylation of Cx43 residue Ser(P) 282 enables close contacts (Յ5 Å) with the backbone atoms from WW2 residues Thr 428 , Thr 429 , and Trp 430 (end of ␤3). Phosphorylation of Cx43 residue Ser(P) 279 facilitates close contacts (Յ5 Å) with the side chain atoms from WW2 residues Thr 427 and Ser 425 (back face of ␤3). Together, the contacts formed by phosphorylation cause the Cx43CT to form a horseshoe-shaped arrangement that fits into a groove formed by the Nedd4 WW2 domain (Fig. 9B, bottom).

Discussion
The studies of Leyauf et al. (27) and Girão et al. (12) helped define the biological importance of Nedd4 in the regulation of gap junction intercellular communication. Nedd4 was the first ubiquitin protein ligase identified to interact with Cx43, and its suppressed expression results in accumulation of gap junction plaques at the plasma membrane (27). Nedd4-mediated ubiquitination of Cx43 is required to recruit Eps15 and target Cx43 to the endocytic pathway (12). Although the Cx43 residues that undergo ubiquitination are still unknown, in this study, we address other unresolved questions pertaining to the molecular mechanisms mediating the direct interaction between Nedd4 and Cx43.
Where Do the Nedd4 WW1 and WW3 Domains Interact with the Cx43CT?- Leykauf et al. (27) demonstrated that all three Nedd4 WW domains interact with Cx43 in a pulldown assay. However, surface plasmon resonance experiments only detected a direct interaction between WW2 and a PPXY motif containing Cx43CT peptide (Cys 271 -Leu 288 ). The conclusion was that WW1 and WW3 were interacting with a different Cx43CT domain (27). The sequence similarity between all three WW domains is between 60 and 80% (Fig. 1A), suggesting a common overall fold and mechanism of binding with the Cx43CT. The solution structure of the three rat Nedd4 WW domains confirmed a similar fold (r.m.s.d., 1.03-1.24 Å) consisting of three antiparallel ␤-strands that are stabilized by a hydrophobic cluster formed in part by conserved Trp (position ϩ10) and Pro (position ϩ25) residues. Using the full-length Cx43CT domain, NMR titration experiment in the presence of the Nedd4 WW domains identified that WW1 and WW3 bind to the same Cx43CT PPXY (Pro 283 -Tyr 286 ) motif as WW2. The difference between the Nedd4 WW domains is that WW2 has a higher binding affinity (0.73 Ϯ 0.05 mM) than WW3 (1.3 Ϯ 0.02 mM) or WW1 (1.7 Ϯ 0.1 mM). The close proximity of Cx43CT residues Ser 279 and Ser 282 to the PPXY motif suggests that phosphorylation by MAPK would affect the binding affinity of Cx43CT for the Nedd4 WW domains (55,56).
Does MAPK Phosphorylation of the Cx43CT Affect the Interaction with the Nedd4 WW1-3 Domains?-NMR titration experiments using the three Nedd4 WW domains and different Cx43CT phosphopeptides revealed that Ser(P) 282 had the most dramatic effect, increasing the Cx43CT binding affinity ϳ2-fold for each of the WW domains (K D , WW2 ϳ WW3 Ͼ Ͼ WW1). Conversely, phosphorylation of Cx43CT residue Ser 279 alone had no effect on the binding affinity. Only WW2 utilized phosphorylation of both Cx43CT residues Ser 279 and Ser 282 , which also had the highest binding affinity (K D , WW2 Ͼ WW3 Ͼ Ͼ WW1). The close sequence identity between the WW1-3 residues that interact with the Cx43CT PPXY motif and Ser(P) (Fig. 1A) strongly suggest that the significant difference in binding affinity of WW1 is the result of a more rigid structure. The NMR data identified that the increased rigidity for WW1 is localized to ␤-strands 2 and 3, which are the same ␤-strands involved in the direct interaction with the Cx43CT. Consistent with previous studies, Cx43CT residue Tyr 286 is necessary for the direct interaction with the Nedd4 WW domains (31). NMR titration data also revealed that phosphorylation of Tyr 286 would inhibit the interaction with the Nedd4 WW domains. We are unclear to as why the WW1 and WW3 domains did not interact with biotinylated Cx43CT peptides attached to the streptavidin surface of the sensor chip in the Leykauf et al. (27)

Cx43CT/Nedd4 WW Domain Interaction
study. Additionally, the binding affinities for both the Cx43CT control (1.1 versus 730 M) and the Ser(P) 279 /Ser(P) 282 phosphopeptide (0.6 versus 290 M) with WW2 were different between our studies. This large difference led us to 1) ensure this was not a buffer-dependent effect (HEPES versus PBS) and 2) use ITC as a third, independent measurement of the binding affinity. ITC data for the three Nedd4 WW domains with the Cx43CT control (e.g. WW2, 630 Ϯ 140 M) and the Ser(P) 279 / Ser(P) 282 phosphopeptide (e.g. WW2, 430 Ϯ 30 M) were similar to the NMR titration data. One possible explanation for these differences is that surface plasmon resonance requires immobilization of one of the binding partners onto a surface, whereas NMR and ITC do not. Another potential explanation is that biosensor experiments have the potential of artifacts caused by mass transport, nonspecific binding, and avidity effects that could alter the apparent binding constants (57,58). Of note, when using surface plasmon resonance, our laboratory has experienced nonspecific binding caused by the intrinsically disordered Cx43CT as well as higher binding affinities for protein partners when compared with well established values in the literature.
Pulldown experiments were performed by Leykauf et al. (27) using GST-tagged WW domains along with control and EGFtreated (leads to phosphorylation of Cx43CT residues Ser 279 and Ser 282 ) cell lysate. Analysis with a polyclonal and phosphospecific antibody (Ser(P) 279 and Ser(P) 282 ) to the Cx43CT led to the following conclusions. 1) The WW2 domain mainly interacted with the non-phosphorylated form of Cx43, 2) phosphorylation of Cx43CT residues Ser 279 and Ser 282 led to the inhibition of the interaction between Cx43 and Nedd4 domain WW3, and 3) phosphorylation of Cx43CT residues other than Ser 279 and Ser 282 appears to be significant for the Cx43 interaction with WW3. Of note, also from the Leykauf et al. (27) study, the immunoprecipitation experiment showed that Nedd4 interacted only with the Cx43 P1 and P2 isoforms in untreated WB-F344 cells, whereas in the pulldown experiment with the same cells, Nedd4 interacted predominately with the Cx43 P0 isoform. These findings are not what we would predict based upon the three Nedd4 WW domains interacting with the same Cx43CT residues, the order of binding affinity (WW2 Ͼ WW3 Ͼ Ͼ WW1), Cx43CT phosphorylation enhancing the binding affinity in a similar manner (WW2 Ͼ WW3 Ͼ Ͼ WW1), and the WW2⅐Cx43CT (Ser(P) 279 , Ser(P) 282 ) structure. One possibility is that our studies involve the use of protein fragments in solution, conditions likely different from the microenvironment of the CT when integrated in a gap junction plaque. However, several studies have shown that, even when not fused with the pore-forming region, the Cx43CT domain retains the biochemical and functional properties consistent with those found in a gap junction plaque (e.g. Refs. 59 -63). These conclusions were formulated based upon the belief that the P0 Cx43 isoform is completely non-phosphorylated. Our laboratory and others have identified that the P0 isoform can also contain phosphorylated Cx43 (64,65). Interestingly, we have identified that MAPK phosphorylation of Cx43 at residue Ser 282 alone results in Cx43 migration at P0. This finding, along with the higher binding affinity for Cx43 when Ser 282 is phosphorylated as compared with no phosphorylation, could explain why the Cx43 P0 isoform was the dominate isoform pulled down by the WW domains in the Leykauf et al. (27) study. This would also explain why the MAPK phosphospecific antibody (Ser(P) 279 and Ser(P) 282 -specific) did not recognize WW1 and WW3 and barely WW2 if the predominant Cx43 isoform in the pulldown experiment contained phosphorylation at Ser 282 only. This illustrates the difficulty with interpreting absolute Cx43 phosphorylation levels with the different migrating isoforms.
Why Does MAPK Phosphorylation of Cx43 Increase the Binding Affinity for the Nedd4 WW Domains?-To the best of our knowledge, the Cx43CT sequence is unique in that a critical Class I motif (PPXY) is aided by multiple, preceding Ser(P) residues to achieve a higher binding affinity with a WW2 domain. A previous study has identified that threonine residues in the immediate vicinity (one before and one after) of the PPXY motif increase the binding affinity for Nedd4 2-3-fold (66). The closest structure comparisons of Cx43CT and Nedd4 can be made with the Nedd4L (66% identity) and Pin1 (38% identity) WW  alone has no effect on the Cx43CT/Nedd4 WW2 domain binding affinity indicates that Ser(P) 282 is necessary to coordinate the Ser(P) 279 interaction. Of note, none of the WW2 residues within 5 Å of Ser(P) 279 and Ser(P) 282 are positively charged; thus the salt bridges formed by the phosphate of Smad3 Thr(P) 179 with the Nedd4L or Pin1 WW domains would explain the approximately 2 orders of magnitude higher binding affinity than the Cx43CT/Nedd4 WW domain interaction.
Conclusion-This study presents the first structure illustrating how phosphorylation of the Cx43CT domain helps to mediate the interaction with a molecular partner. The similarity in structure between the three Nedd4 WW domains highlights how small changes in the primary sequence can alter the dynamics and binding interface to impact the binding affinity with the Cx43CT. Moreover, previous work from our laboratory suggested that different mechanisms of regulation exist between connexin isoforms even when involving the same molecular partners (67,68). When examining the amino acid sequence of the CT of Cx40 (TPPPDF) and Cx45 (TPSAPPGY), two other cardiac connexins that co-express and co-oligomerize with Cx43, a similar PPX⌽ motif can be observed. However, the position of the surrounding potential Ser(P)/Thr(P) residues is not directly adjacent to the PPXY motif, suggesting a distinct mechanism of interaction with and potential mode of regulation by Nedd4. Additionally, future studies need to address the Nedd4 (as well as other multidomain molecular partners) interaction with Cx43 in the context of a connexon. Can Nedd4 bind multiple Cx43CT domains? Can Nedd4 interact with the Cx43CT and a second protein partner at the same time? Does binding of Cx43 to each of the different WW domains of Nedd4 lead to a different response in the cell? Or is it just that increasing binding to any of them facilitates the same process (ubiquitination and endocytosis)?
Author Contributions-G. S. and F. K. conducted most of the experiments, analyzed the results, and edited the manuscript. J. L. K., H. L., S. Z., K. L. S., and R. G. helped G. S. and F. K. conduct the experiments and edit the manuscript. P. L. S. conceived the idea for the project, analyzed the results, and wrote the manuscript.