The Structure and Dynamics of Tandem WW Domains in a Negative Regulator of Notch Signaling, Suppressor of Deltex*

WW domains mediate protein recognition, usually though binding to proline-rich sequences. In many proteins, WW domains occur in tandem arrays. Whether or how individual domains within such arrays cooperate to recognize biological partners is, as yet, poorly charac-terized. An important question is whether functional diversity of different WW domain proteins is reflected in the structural organization and ligand interaction mechanisms of their multiple domains. We have determined the solution structure and dynamics of a pair of WW domains (WW3–4) from a Drosophila Nedd4 family protein called Suppressor of deltex (Su(dx)), a regulator of Notch receptor signaling. We find that the binding of a type 1 PPPY ligand to WW3 stabilizes the structure with effects propagating to the WW4 domain, a domain that is not active for ligand binding. Both WW domains adopt the characteristic triple-stranded (cid:1) -sheet structure, and significantly, this is the first example of a WW domain structure to include a domain (WW4) lacking the second conserved Trp (replaced by Phe). The domains are connected by a flexible linker, which allows a hinge-like motion of domains that may be important for the recognition of functionally relevant targets. Our results contrast markedly with those of the only previously determined three-dimensional structure of tandem WW domains, that of the rigidly oriented WW domain pair from

WW domains are small protein interaction modules found in a wide range of eukaryotic signaling and structural proteins (1). The domain is a small three-stranded ␤-sheet stabilized by the stacking of several conserved aromatic and proline residues (2). Differences in residue identity at the binding surface result in a variation in ligand specificity that is used as the basis to divide WW domains into groups. For example, in group I WW domains that bind PPXY sequences (3), the Tyr is a key specificity residue and is accommodated by a largely hydrophobic pocket on the concave binding surface consisting of conserved Ile (or Val/Leu), His, and Gln (or Arg/Lys) residues. The Pro residues of the ligand contribute to the binding by stacking against the Trp and Tyr residues that form a second interaction site (4). It is evident that, since a number of proteins are likely to contain WW domain recognition sites, further factors most probably contribute to increasing affinity and specificity of a WW domain for a target. For example, the Pro-rich sequence in ␤-dystroglycan targeted by the dystrophin WW domain requires a composite binding surface provided by the WW domain and an adjacent EF hand (4). The binding of murine Nedd4 to the amiloride-sensitive epithelial sodium channel requires direct involvement of two of its three WW domains (5). Indeed, WW domains often exist in multiple numbers within a protein. Multiple modules may act in concert to achieve greater specificity for a target as observed for the SH2 1 domains in SHP-2 phosphatase (6,7). Alternatively, their presence may indicate a functional diversity borne out by an ability to bind more than one target. The number and spatial arrangement of WW domains are sometimes modified by alternative splicing (8), highly suggestive of a purpose to their tandem arrangement as discussed above. Therefore, it is likely that the length and structure of the interdomain linkers are of crucial significance.
Little is currently known about the functional importance of the linker region in proteins with multiple WW domains. The notable exceptions are as follows. First, the phosphorylation at sites in the linker regions diminishes the human Nedd4-2epithelial sodium channel interaction (9). This effect is biologically very important, enabling the regulation of the human body Na ϩ ion balance and blood pressure. Secondly, the recent solution structure of the WW domain pair of the yeast-splicing factor Prp40 (10) reveals a single rigid structure with the linker assuming an ␣-helical conformation. This linker conformation results in a defined relative orientation of the WW domain binding surfaces in keeping with a bridging function for Prp40 early in the splicing process. In contrast, incomplete structural studies on the second and third tandem WW domains from the rat Nedd4 protein (11) suggest a disordered interdomain linker. Therefore, differences in the structure and dynamics of tandem WW domains are probable and possibly affect their protein recognition function.
Suppressor of deltex (Su(dx)) is a Drosophila member of the Nedd4 family of type 3 ubiquitin ligases (12) defined by a common modular architecture; an N-terminal lipid-interacting C2 domain, between two and four WW domains (Su(dx) has four arranged in pairs) and a C-terminal E3 ubiquitin ligase HECT domain. Su(dx) was identified in Drosophila as a negative regulator of Notch receptor signaling (13)(14)(15)(16). The Notch signal transduction cascade is of fundamental importance for pattern formation and correct execution of multiple cell-fate decisions (17)(18)(19). Direct binding partners for Su(dx) that mediate its effect on Notch signaling are yet to be identified but most probably interact via the Su(dx) WW domains. The four WW domains (all group I) are organized in a pairwise fashion with domains one (WW1) and two (WW2) and, likewise, domains three (WW3) and four (WW4) separated by short linkers: 8 residues between WW1 and WW2 and 14 residues between WW3 and WW4. The linker between the two module pairs, WW1-2 and WW3-4, is much longer (40 amino acids). To obtain further insight into the operation of the multiple WW domains of Su(dx), we undertook NMR studies of the tandem pair WW3-4. The latter was of particular interest, because it lacks the second conserved Trp (replaced by Phe) that gives the WW domain its name (Fig. 1). The conservation of this mutation in WW4 from nematode worms to humans ( Fig. 1) indicates a unique role for this particular domain, a role that remains unknown. No structure of this WW module class has previously been determined.

MATERIALS AND METHODS
Sample Preparation-The region encoding WW domains 3 and 4 (amino acids 474 -545 of Su(dx)) was amplified by PCR from Su(dx) cDNA using primers (MWG Biotech) that generate flanking EcoR1 and XhoI sites for insertion into pGEX 4T-1 (Amersham Biosciences) as a fusion with glutathione S-transferase. Expression in Escherichia coli BL21 cells was induced by the addition of 0.1 mM isopropyl-1-thio-␤-Dgalactopyranoside at 37°C. Cells were lysed in phosphate-buffered saline by lysozyme and a freeze-thaw cycle. Protein was purified by affinity binding to glutathione beads in phosphate-buffered saline followed by thrombin cleavage to release the WW domains from the glutathione S-transferase tag and a final gel filtration step (Sephacryl S-100, Amersham Biosciences) to remove higher molecular weight contaminants and to exchange the protein into the buffer used in NMR experiments. After cleavage, the WW3-4 construct retains an N-terminal leader sequence of GSPEFHM arising from vector sequences that link the cloned gene to that for glutathione S-transferase. To prepare the 15 N-labeled sample, cells were grown in M9 minimal medium with 15 NH 4 Cl as a sole source of nitrogen. The yield of purified protein was relatively low in all of the conditions tested and required up to 10 liters of medium/NMR sample, making 13 C labeling fiscally unviable. The individual WW3 and WW4 domains were prepared similarly. Purified GPPPPPGYPG-peptide (Pept1) was purchased from Pepceuticals Ltd. (Leicester, United Kingdom). The buffer used for all of the experiments contained 50 mM NaCl, 2 mM KCl, 5 mM Na 2 HPO 4 , and 1 mM KH 2 PO 4 (pH 7.1). WW3-4 tends to aggregate at concentrations above 0.5 mM; therefore, a 45 mM mixture of Arg and Glu amino acids was added upon advice (20) in order to enhance protein solubility and obtain solutions of higher concentration for NMR experiments.
Fluorescence Equilibrium Binding Studies-Intrinsic tryptophan fluorescence of Su(dx) WW domains (individual and paired), used to derive equilibrium dissociation constants for Pept1 binding, was measured in a Varian Cary Eclipse fluorimeter spectrometer at 20°C. Excitation and emission wavelengths were 297 and 343 nm, respectively. Dissociation constants were determined by fitting the data to a single-site binding model (equation is given in Fig. 2) using WinCurveFit.
NMR Spectroscopy-NMR experiments were carried out at 15°C on a Bruker DMX 600 MHz spectrometer equipped with pulse-field gradients and triple resonance probes. The data were processed and analyzed using NMRPipe (21) and NMRView (22) software. For the assignment and structure calculations, two-dimensional homonuclear NOESY, E-COSY, and TOCSY and three-dimensional heteronuclear 15  chemical shift changes in 1 H-15 N HSQC spectra (25). The combined chemical shift changes were calculated using Equation 1.
1 D NH residual dipolar couplings (RDCs) were measured in the 5% solution of 8-alkyl-poly(ethylene glycol/octanol) (26) using the IPAP (in-phase/anti-phase) method of spectra recording (27). The RDC values were obtained by subtracting the reference value in isotropic solution. Alignment tensors were calculated from RDC values using the PALES program (28). Heteronuclear 1 H-15 N NOEs and longitudinal (R 1 ) and transverse (R 2 ) 15 N relaxation rates were measured using the standard two-dimensional methods (29,30). The relaxation delays were set to 10 (run twice), 20, 50, 100, 200, 352, and 500 ms, 1.512 and 2.520 s for R 1 measurements, and 18 (run twice), 36,55,110,147,202,294,496, and 993 ms for R 2 measurements. Correlation times were calculated for the WW domains from these experimental measurements (31). The 1 H-15 N NOE experiments were run twice in an interleaved fashion with and without (reference experiment) proton saturation during the recovery delay. Errors in the peak intensities were estimated from the average base-line noise.
Structure Calculation-Structure calculation was performed using the ARIA program (32). Intraresidue, sequential, and backbone NOEs across ␤-sheets were assigned manually. The remaining NOESY crosspeaks were assigned automatically during ARIA cycles and then visually inspected for the assignment of peaks near the water signal. All of the fully assigned peaks in NOESY spectra were integrated using NMRView. A floating chirality approach was used for methylene and isopropyl groups and Phe and Tyr aromatic protons (33). Hydrogen bonds between ␤-sheets were applied as indicated by H N , H ␣ , and N chemical shifts and characteristic interstrand NOEs. The bounds for hydrogen bond restraints were 1.8 -2.3 Å for H N -O distances and 2.8 -3.2 Å for N-O distances. Restraints for the backbone angles were derived from 3 J HNHA coupling constant and H N and N chemical shifts (34). The structures were calculated using a simulated annealing protocol that employs torsion angle dynamics during the first six iterations and Cartesian dynamics during the final two iterations (32). RDCs were used in the final annealing stages. The final structures were analyzed using the PROCHECK-NMR program (35). Insight II (Molecular Simulations Inc.) was used for visualization and figure preparation.

Fluorescence Analysis of Ligand Binding to the WW3-4 Tandem Pair-
The direct interaction of the second Trp residue within WW domains with ligand allows the monitoring of ligand binding by Trp fluorescence. To determine whether the recombinant WW3-4 pair was active in ligand binding, we monitored fluorescence changes upon the addition of a synthetic peptide. A peptide sequence, GPPPPPYPG (Pept1), was selected based on its previously identified high affinity binding to the group I WW domains (36). Titration of the WW3-4 pair with increasing concentrations of Pept1 gave a K d of 15.3 M. A similar K d of 17.1 M was obtained for the individual WW3 domain (Fig. 2). Such similar values would indicate that binding is occurring to WW3 and that there is a negligible effect of the second domain on its binding affinity. Furthermore, the dissociation constant obtained upon the titration of an equimolar mixture of the individual domains with Pept1 (15.8 M, data not shown) is within experimental error of the apparent dissociation constants for the binding of this peptide to WW3 alone and to WW3-4. The binding of Pept1 to WW3 only is consistent with previous studies showing that group I WW domains with a Trp3 Phe/Tyr mutation do not bind to peptides with the PPPY motif (10,37). However, as WW4 lacks the Trp residue used to monitor fluorescence changes, we cannot exclude the possibility of some Pept1 binding to WW4 by this method, particularly if it does so with a near identical affinity.
Resonance Assignments and Structure Determination of the WW3-4 Pair-Direct assignment of the unligated WW3-4 pair was not possible as the 1 H-15 N HSQC spectrum of free WW3-4 displayed a paucity of amide 1 H-15 N peaks (Fig. 3A) and a very narrow dispersion of resonances. The addition of Pept1 resulted in significantly improved spectra (Fig. 3B), allowing sequence-specific 1 H and 15 N NMR resonance assignments based on three-dimensional 1 H, 15 N HSQC-TOCSY, three-dimensional 1 H, 15 N HSQC-NOESY, and homonuclear NOESY, TOCSY, and E-COSY. All of the backbone and most side-chain resonances with the exception of the first four residues have been assigned. The assignment of methylene protons was non-stereospecific.
The refinement procedure for the liganded WW3-4 utilized NOE distance constraints, hydrogen bonds, torsional angles, and RDC (see Table I). Structure calculations were performed for 20 structures. The statistics are given in Table I for an ensemble of the 10 lowest energy structures of the 20 calculated (Protein Data Bank code 1TK7). None of the converged structures has any distance constraint violation greater than 0.4 Å or any violation of the dihedral angle constraints. The inclusion of RDC constraints in the calculations does not result in a specific orientation of the two domains, and their mobility remains high. The inclusion of RDC also has a minimal effect on calculations of root mean square deviation (r.m.s.d.) from the mean structure for the 10 converged structures. Taking each domain individually, the r.m.s.d. values close to 1.0 Å are calculated for backbone heavy atoms (see Table I   ligand complexes, they are implicated in the interaction with ligand. Thus, this line broadening probably reflects ligand binding and an exchange rate in an intermediate time regime, even at the large peptide excess. Although Pept1 is binding in the usual site for group I WW domains, the weak NOEs from these residues and our inability to assign the resonances of sequential prolines in the GPPPPPYPG-peptide in the absence of isotopic labeling precludes the structure determination of the whole complex and we restricted the study to the structure of WW3-4 only in a liganded state. The solution structures (Fig. 4)  The structure of the interdomain linker (residues 44 -56) is poorly defined by the experimental data suggesting that it may be highly flexible (see below). No interdomain NOEs have been identified, further suggesting that there is no single preferred relative orientation of the two domains. However, it is apparent that the linker region is not totally unstructured because a number of medium-range d NN (i,iϩ2), d ␣N (i,iϩ2), and d ␣N (i,iϩ3) NOEs were assigned through this segment (data not shown).
Protein Dynamics of the WW3-4 Structure-The WW3-4 structure, as complexed with a peptide ligand (Fig. 4), implies considerable flexibility of the possible orientations of WW3 with respect to WW4, in contrast to the only previous solved structure of a WW domain pair (10). To determine whether this reflected true conformational flexibility or simply an experimental consequence of a lack of observed NOEs, we further investigated the dynamics of this lack of fixed orientation using 15 N relaxation measurements (29,30). For this purpose, we recorded R 1 , R 2 , and 1 H- 15 5A). In contrast, the interdomain linker has smaller values indicating restricted internal motion and the lack of a well ordered structure. The same conclusion can be drawn from the R 2 /R 1 ratios, which follow the same trend (Fig. 5B).
Further information regarding interdomain orientation and motion is obtained by the measurement of RDCs (40,41). RDCs proved to be a valuable tool in identifying the relative orientation of WW domains in Prp40 (10). The measured 1 H-15 N RDCs are markedly different for domains WW3 and WW4 (Fig. 6). The RDCs are mostly negative for the residues in the ␤-strands of WW4 but positive for the corresponding residues in WW3. The opposite sign of RDCs immediately points to non-colinearity of the two domains. At the same time, the axial components (D a NH ) and the rhombicity () of the alignment tensors are practically identical if calculated using RDCs from WW3 only (D a NH ϭ 8.1 Hz, ϭ 0.44), WW4 only (D a NH ϭ 8.4 Hz, ϭ 0.42), or the whole molecule (D a NH ϭ 8.2 Hz, ϭ 0.44). These values would suggest some restriction to interdomain motion, because independent or semi-independent orientation usually results in markedly different parameters (40). Thus, we conclude that, although linker flexibility probably leads to some orientation independence between domains 3 and 4, there remains to be limitations to this flexibility and that the two domains, on average, tumble as a unit. Consistent with this conclusion is the presence of the aforementioned medium-range NOEs within the linker region together with additional NOEs between the WW4 C-terminal residue, Arg 87 (side chain), and the linker residues, Asn 53 and Glu 54 (backbone atoms). Furthermore, the calculation of the correlation times from experimental spin-lattice and spin-spin relaxation times (T 1 and T 2 ) gives a value of 7.8 ns for each domain. The expected correlation time estimated (at temperature ϭ 288 K) using an empirical method that utilizes the dependence of correlation time on number of protein residues (42) is 9 ns for an 88-residue protein, 3.5 ns for a single 32-residue domain, and 7.8 ns for 75 residues (two domains plus bound peptide excluding the flexible linker). Correlation time was also estimated from the Protein Data Bank coordinates (43) of WW3-4, giving a value of 8.2 ns. Thus, the experimental correlation times for individual domains correspond to those expected for the full protein rather than the smaller individual domains, demonstrating unambiguously that the domains tumble together and that there is interaction between them at a structural level.
Structural Changes to WW3-4 on Addition of Pept1-To investigate possible structural changes involved in ligand binding, we recorded the 1 H-15 N HSQC spectra of free WW3-4 and WW3-4/Pept1 complexes at different concentrations of the added peptide. The addition of Pept1 produces substantial changes in the 1 H-15 N HSQC spectrum. At a 1:4 WW3-4/Pept1 ratio, the spectrum exhibits a full set of 1 H-15 N peaks (Fig. 3B) as confirmed by an assignment of the protein backbone resonances. The examination of spectra at different titration points led us to divide WW3-4 resonances into two classes: 1) resonances that exhibit only the chemical shift drift upon titration with little or no changes in the line widths and 2) resonances that have very broad line widths in the free WW3-4 form and cannot be detected below 1:1 WW3-4/Pept1 ratio (Fig. 7). Based on the successful assignment of the WW3-4/Pept1 complex, the assignment to the uncomplexed WW3-4 was traced wherever possible (Figs. 7 and 8A). The resonances that exhibit differential line broadening are localized in all three ␤-strands of WW3 and the third ␤-strand of the domain WW4 (Fig. 8A, residues with negative bars). The observed differential line  (11). However, the situation is more complex in the present case. Many WW3-4 resonances have exactly the same line widths and intensity in both free and bound forms (Fig. 7). The gradual chemical shift changes without any effect on the line width are indicative of a fast exchange regime for these residues, inconsistent with aggregation. Furthermore, in support of ligand-induced structural changes, the Trp 17 and Trp 61 residues, which form part of the core of WW3 and WW4, respectively, undergo large chemical shifts on Pept1 binding (Fig.  8, A and C). In fact, Trp 61 has the largest chemical shift change among the residues for which it can be reliably measured (Fig.  8C). The analysis of chemical shift changes thus supports the conclusion that ligand binding induces a conformational change to both domains from an at least partially unfolded state to one that is fully folded and stable. This is in contrast to NMR studies of the YAP65 (38), Prp40 (10), and Pin1 WW domains (44,45) where shifts of the equivalent residues to Trp 17 and Trp 61 are not significantly affected by ligand binding and the domains are stably folded in absence of ligand.
Chemical shift mapping has previously proved useful in identifying residues directly involved in ligand binding by WW domains (10,37,38,44,45). Based on the sequence homology with other group I domains, these residues are Tyr 28 , Val 30 , His 32 , Arg 35 , Thr 37 , and Trp 39 in domain WW3 and Phe 72 , Val 74 , His 76 , Arg 79 , Thr 81 , and Phe 83 in domain WW4. Chemical shift mapping in the case of WW3-4 is complicated by the aforementioned differential line broadening for certain residues (particularly in WW3) and the structural changes to both domains induced by ligand binding. However, a direct comparison between chemical shifts of ligand binding residues for free and liganded forms is possible for the Tyr 28 -Phe 72 and His 32 -His 76 pairs (representing equivalent positions in the two domains). The Tyr 28 and His 32 residues in WW3 domain display chemical shift changes (Fig. 8B) of a magnitude consistent with the direct involvement of WW3 in ligand binding, whereas the topologically equivalent residues in WW4 show smaller shifts consistent with an indirect effect of Pept1 on the global conformation of WW4. Notably, the chemical shift change for Phe 83 is very small in contrast to the large chemical shift change for Trp61 (Fig. 8C), which lies in the core of the domain. Phe 83 is in the position of the normally conserved second tryptophan, and a chemical shift change would undoubtedly be observed for this residue in the event of direct ligand binding to this domain. Taken together, the experimental data (NMR and fluorescence) are consistent with ligand binding only to WW3, which induces a global conformational change transmitted to WW4.

Differential Binding of Su(dx) WW Domains to a Peptide
Ligand-Drosophila Su(dx) acts as a negative regulator of the developmentally important Notch signaling receptor through controlling the availability of the receptor protein at the apical adherens junction. 2 However, little is currently known regarding the directly interacting biochemical targets of Su(dx) in vivo. This study of the WW3-4 pair represents the first attempt to define the structure-function relationship of this protein. Analysis of the Su(dx) WW3 sequence predicts that it is a group I WW domain, a group that recognizes peptides with the core sequence PPXY. This was confirmed in our study when the PPPY containing Pept1 sequence was found to bind to the isolated WW3 domain. Interestingly, our data indicate that WW4 does not contribute to the binding affinity for Pept1. WW4 is a member of an unusual class of group I WW domain that possess a Phe (as in this case) or Tyr in place of the highly conserved second Trp. The specificity of this subset of WW domains is poorly understood, but the lack of binding of Pept1 to WW4 is consistent with the disruption of ligand binding by the Trp3 Phe/Tyr mutation at this second site in the YAP65 WW domain (46). Interestingly, we have noted that domains of this type are, in all of the cases to date (47), located in the C-terminal position of an array prior to a catalytic domain, suggestive of a unique role in protein function or regulation of activity.
Structure and Domain Orientation of the WW3-4 Pair-As expected from sequence alignments to other WW domains, we found that, in saturating ligand conditions, both WW3 and WW4 adopted a global fold based around a triple stranded ␤-sheet, similar to the conformation of all of the other WW domains determined to date. Importantly, WW4 is the first example of a structure determined for a member of the group I WW domain subset that lacks the second Trp.
In proteins possessing arrays of multiple WW domains such as Su(dx), it is likely that the number of domains and the spatial organization between them adds further specificity to their activities in vivo, although this is poorly understood. Although the structures of a number of isolated WW domains have been determined (4, 38, 39, 48 -50), only one structure of a pair of WW modules, that of Prp40, has previously been solved (10). In that study, a rather rigid orientation of the two WW domains was observed. Both WW domains were aligned in the same direction with the two respective specificity pockets on opposite faces of the domain pair, enabling Prp40 to interact simultaneously and independently with different partners in a fixed spatial arrangement. In contrast, we have shown that the WW3-4 domains of Su(dx) lack a fixed orientation with respect to each other, although RDC alignment tensor parameters and rotational dynamics calculations indicate that the molecule does tumble on average as a single unit, implying that there is some restriction to the interdomain dynamics.
In the absence of extensive interdomain contacts, the linker structure may play the key role in defining the orientation of adjacent WW domains. In Prp40, the linker forms a well defined helix and conserved leucine residues at either end make close contacts with the cores of the respective WW domains, ensuring a rigid orientation (10). In contrast, in Su(dx) WW3-4, the linker does not adopt a helical conformation and the data on backbone dynamics indicate that the linker is in a more flexible dynamic state than the flanking WW domains, resulting in a range of possible domain-domain orientations. However, it is clear that the flexibility of the linker is not unrestricted, because it displays some medium range interresidue NOEs and terminal residues of the linker make contacts with the WW4 C-terminal region. Therefore, it is possible that the partially restricted linker motion consists of a hinge bending action, although this cannot be confirmed from the current data.
Response of WW3-4 to Ligand Binding-Another important difference between the WW domain pairs of Prp40 and Su(dx) concerns the mechanism of ligand binding. In Prp40, a comparison of the NMR spectra of the ligand-bound and ligand-free WW domain pair indicated that each WW domain has a well defined structure in both states and that ligand-induced chemical shift changes were restricted to residues in the specificity pocket. Therefore, it was concluded that the ligand has a negligible affect on the global conformation of both WW domains and that there is little communication between each binding site (10). Ligand recognition by Prp40 thus appears to be of a lock and key type. In contrast, in the absence of ligand, the Su(dx) WW3-4 pair has a relatively poorly defined structure as shown by a poorly dispersed spectrum. Differential line broadening of a substantial number of residues together with ligandinduced changes in the chemical shift of structural residues in both domains remote from the binding surface supports the conclusion that both domains undergo a structural transition from a partially folded state to the fully stable WW domain fold. Thus, large conformational changes in WW3 on direct contact with peptide propagate to WW4, indicating that unlike with Prp40 there is some coupling between the domains. This structural coupling cannot be fully explained without better quality spectra of the apoprotein. However, a possible scenario is one where WW3 requires the bound ligand for full stability, and a partially unstructured and possibly aggregating apodomain may interact with WW4, thus affecting the structure and stability of WW4. The differential line broadening observed for residues in all three ␤ strands of WW3, and the third ␤ strand of WW4 (Fig. 8A, negative bars) could be construed as an indication that such an interaction does indeed occur. The structural effects of ligand binding to WW3 may synergistically enhance ligand binding to WW4. The lack of a known peptide ligand for WW4 has thus far precluded such binding studies, although they form an essential component of our future plans. signals can be detected below 1:1 the protein/ligand ratio (such differential line broadening is indicative of structural change as described under "Results") to distinguish from residues for which chemical shift changes could not be measured accurately, simply because of chemical shift overlap (or Pro residues). B, chemical shift change upon titration of WW3-4 with Pept1 as a function of protein/peptide stoichiometry for the residues Tyr 28  Undoubtedly, our study reveals that, even with similar peptide binding specificities, the mechanism of ligand binding to WW domains and the relative domain positioning and dynamics can be very different in different proteins, which may have important biological consequences.
Functional Implications of the WW3-4 Structure-It is interesting to speculate on how the conformational flexibility and possible cooperativity between the WW domains is relevant to the biological activity of Su(dx). The current model of Prp40 function is that it bridges between target sites within the splicing machinery components branch-point-binding protein and Prp8, holding them in a precise orientation (10). In contrast, individual members of the Nedd4 family are known to interact with many proteins through their WW domains (12) not only proteins targeted for endocytosis but also, in some cases, components of the endocytic machinery itself (51)(52)(53). Therefore, being adaptable to a number of spatial orientations imposed by different substrates and accessory molecules might be facilitated by flexible domain orientation. Recent evidence suggests that Su(dx) may indeed regulate different target proteins associated with separable activities of Notch down-regulation and imaginal disc epithelial cell proliferation (54). The dynamics of WW3-4 (and for that matter higher order conformation of the entire WW domain array in Su(dx)) may also be important for positioning of the target protein(s) for ubiquitination by the C-terminal catalytic HECT domain or for steric up-or down-regulation of HECT domain activity as observed for the SH3 and SH2 domains of the Src family kinases (55). Therefore, the interaction requirements of Su(dx) may differ considerably from those of Prp40.
In vivo, Su(dx) colocalizes at the apical adherens junctions with Notch and triggers the depletion of the receptor from the cell surface. 2 Interestingly, we have shown that in vivo Su(dx) interacts only transiently with Notch at the cell surface and does not accompany Notch into the early endosome, thus the regulatory complex, which includes Notch and Su(dx), must be dynamically assembled and disassembled during the progression of Notch through its trafficking pathway. Therefore, it is interesting to speculate on whether the structural changes on target binding that are transmitted between the adjacent domains may result in cooperativity of interactions among different ligand binding sites. This would facilitate the rapid assembly and disassembly of complexes as appropriate in such a dynamic process as endocytic trafficking.

CONCLUSIONS
This study provides only the second example of a threedimensional structure of a tandem pair of WW domains and highlights the variety by which conserved motifs can interact with their targets and each other with probable implications for their target specificity and biological function. It will be interesting to further compare the WW3-4 pair with the WW1-2 pair from Su(dx) in which individual domains are separated by a shorter linker and may have a different interdomain relationship. Further work is also in progress to determine the directly interacting ligands for Su(dx) WW domains. The ability to combine structural investigation with studies of protein function in the genetically tractable Drosophila model organism will facilitate understanding of the factors determining biological specificity of WW domain proteins in an in vivo context.