H+ exchanger regulatory factor provides insights into the mechanism of carboxyl-terminal leucine recognition by class I PDZ domains1

doi:10.1006/jmbi.2001.4634 available online at http://www.idealibrary.com on

J. Mol. Biol. (2001) 308, 963±973

Crystal Structure of the PDZ1 Domain of Human Na‡/H‡ Exchanger Regulatory Factor Provides Insights into the Mechanism of Carboxyl-terminal Leucine Recognition by Class I PDZ Domains Subramanian Karthikeyan, Teli Leung, Gabriel Birrane, Gordon Webster and John A. A. Ladias* Molecular Medicine Laboratory and Macromolecular Crystallography Unit, Division of Experimental Medicine, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston MA 02115, USA

The Na‡/H‡ exchanger regulatory factor (NHERF; also known as EBP50) contains two PDZ domains that mediate the assembly of transmembrane and cytosolic proteins into functional signal transduction complexes. The NHERF PDZ1 domain interacts speci®cally with the motifs DSLL, DSFL, and DTRL present at the carboxyl termini of the b2 adrenergic receptor (b2AR), the platelet-derived growth factor receptor (PDGFR), and the cystic ®brosis transmembrane conductance regulator (CFTR), respectively, and plays a central role in the physiological regulation of these proteins. The crystal structure of the human NHERF PDZ1 has Ê resolution using multiwavelength anomalous been determined at 1.5 A diffraction phasing. The overall structure is similar to known PDZ structures, with notable differences in the NHERF PDZ1 carboxylate-binding loop that contains the GYGF motif, and the variable loop between the b2 and b3 strands. In the crystalline state, the carboxyl-terminal sequence DEQL of PDZ1 occupies the peptide-binding pocket of a neighboring PDZ1 molecule related by 2-fold crystallographic symmetry. This structure reveals the molecular mechanism of carboxyl-terminal leucine recognition by class I PDZ domains, and provides insights into the speci®city of NHERF interaction with the carboxyl termini of several membrane receptors and ion channels, including the b2AR, PDGFR, and CFTR. # 2001 Academic Press

*Corresponding author

Keywords: NHERF; EBP50; PDZ; crystal structure; signal transduction

Introduction PDZ (PSD-95/discs-large/ZO-1 homology) domains are structurally conserved protein modules that mediate speci®c interactions between proteins.1 ± 3 A large number of PDZ-containing proteins function as scaffolds for assembling membrane receptors, ion channels, and other signaling molecules in the vicinity of their substrates. By Abbreviations used: b2AR, b2 adrenergic receptor; CFTR, cystic ®brosis transmembrane conductance regulator; EBP50, ezrin-radixin-moesin-binding phosphoprotein-50; GST, glutathione S-transferase; NHERF, Na‡/H‡ exchanger regulatory factor; MAD, multiwavelength anomalous diffraction; PDGFR, platelet-derived growth factor receptor; PDZ, PSD-95/ discs-large/ZO-1 homology. E-mail address of the corresponding author: [email protected] 0022-2836/01/080963±11 $35.00/0

organizing such signal transduction pathways at speci®c intracellular locations, PDZ proteins play fundamental roles in the speci®city and ef®ciency of signal transduction.1 ± 3 Previous structural studies have demonstrated that PDZ domains share a common fold, consisting of a six-stranded antiparallel b-barrel capped by two a-helices.4 ± 10 These protein modules bind to short carboxyl-terminal peptides and have been categorized into two classes on the basis of target sequence speci®city. Class I domains bind to peptides with the consensus sequence (S/T)X(V/I/L) (X denoting any amino acid), whereas class II domains recognize the motif (F/Y)X(F/V/A).11 In addition, PDZ domains can interact with internal protein sequences that adopt b-hairpin structures.10 Peptidic ligands interact with PDZ domains by a process of b-sheet augmentation, in which the peptide forms an additional, antiparallel b-strand in the PDZ b-sheet.4,6,7 ± 10 The loop between the ®rst # 2001 Academic Press

964 two b-strands plays an important role in recognizing and binding the terminal carboxylate group of the target peptide and is referred to as the carboxylate-binding loop. The speci®city of the PDZpeptide interaction is achieved by the residues at positions ÿ3, ÿ2, and 0 of the peptide ligand (position 0 referring to the carboxyl-terminal residue), whereas the residue at position ÿ1 does not play an important role. The mechanism for selecting a valine residue at position 0 by class I PDZ domains has been determined.4,7 ± 9 However, the structural basis for carboxyl-terminal leucine recognition has not been elucidated to date. The Na‡/H‡ exchanger regulatory factor (NHERF) was originally cloned as an essential cofactor for the inhibition of the Na‡/H‡ exchanger NHE-3 by the cAMP-dependent protein kinase A in the renal brush-border membrane.12 The human NHERF was also identi®ed independently as EBP50 (ezrin-radixin-moesin-binding phosphoprotein-50), a membrane-cytoskeleton linking protein that binds to the carboxyl termini of integral membrane proteins through its two PDZ domains and to the cortical actin cytoskeleton through its carboxyl-terminal domain.13 The emerging theme from recent studies of NHERF and its close relative NHERF214 is that these proteins play central roles in the membrane targeting, traf®cking, and sorting of several ion channels, transmembrane receptors, and signaling proteins in many other tissues.15,16 The PDZ domains of human NHERF17 and NHERF2 share a high degree of sequence similarity (Figure 1) and similar ligand-binding speci®cities.15,16 Interestingly, PDZ1 homodimerization leads to NHERF self-association in solution;18,19 however, the PDZ1-PDZ1 interaction interface(s) has not been mapped. The NHERF PDZ1 is a class I domain that interacts speci®cally with carboxyl-terminal sequences present in many membrane proteins, including the b2 adrenergic receptor (b2AR), the platelet-derived growth factor receptor (PDGFR), and the cystic ®brosis transmembrane conductance regulator (CFTR).14,20 ± 23 Through high-af®nity binding of PDZ1 to the b2AR carboxyl-terminal motif DSLL, NHERF plays an important role in b2AR-mediated regulation of Na‡/H‡ exchange,20 and mediates the sorting of internalized b2AR between degrada-

Structure of the NHERF PDZ1 Domain

tive endocytic pathways and plasma membrane recycling.24 Likewise, NHERF potentiates PDGFR activity through PDZ1 interaction with the carboxyl-terminal sequence DSFL of the receptor.23 Remarkably, a single mutation of the carboxylterminal leucine to alanine residue in both b2AR and PDGFR abolishes their interaction with NHERF PDZ1 in vitro and markedly impairs the function of these receptors in vivo.20,23 Furthermore, the NHERF PDZ1-binding carboxyl-terminal motif DTRL of CFTR is essential for the functional expression of CFTR in the apical plasma membrane and mutations that delete or destroy this motif result in abnormal apical polarization of CFTR and defective vectorial chloride transport.25,26 In order to elucidate the structural determinants of the human NHERF PDZ1 ligand-binding speci®city, we solved the crystal structure of this Ê resolution. The structure of domain at 1.5 A NHERF PDZ1 was determined in complex with a ligand provided by the carboxyl terminus of a neighboring PDZ1 molecule related by 2-fold crystallographic symmetry. The structure reveals the molecular mechanism of carboxyl-terminal leucine recognition by class I PDZ domains and provides a starting point for understanding the structural basis of the NHERF interaction with several membrane receptors and ion channels, including the b2AR, PDGFR, and CFTR.

Results and Discussion Structure determination The human NHERF PDZ1 (amino acid residues 11-99) was expressed as a glutathione S-transferase fusion protein in Escherichia coli, cleaved with thrombin, puri®ed, and crystallized using the vapor diffusion method. Numerous attempts to crystallize this domain in complex with high-af®nity peptidic ligands corresponding to the carboxyl termini of CFTR and b2AR, using both co-crystallization and peptide-soaking approaches, were unsuccessful. The NHERF PDZ1 crystals diffracted Ê resolution using CuKa radiation. Howto 1.8 A ever, it was not possible to solve the structure by molecular replacement using other PDZ structures as search models (data not shown). An extensive

Figure 1. Sequence comparison of selected class I PDZ domains. The amino acid sequences of the indicated PDZ domains from human NHERF,17 human NHERF2,14 and rat PSD-95 PDZ3,44 were aligned using the program MACAW45 and visual inspection. Hyphens represent gaps inserted for optimum alignment. Absolutely conserved amino acid residues are shown as white letters on black background. Identical residues in four PDZ domains are shaded. The secondary structure of NHERF PDZ1 is indicated at the top of the sequences.

965

Structure of the NHERF PDZ1 Domain

search to generate heavy-atom isomorphous derivatives of PDZ1 crystals yielded a single HgCl2 derivative, which was used to collect synchrotron diffraction data at three X-ray wavelengths near the mercury LIII absorption edge (Table 1). The NHERF PDZ1 crystal structure was determined to Ê resolution by MAD phasing methods27 using 1.5 A the program SOLVE.28 The model was re®ned to a crystallographic R-factor of 18.1 % and a free R-factor of 20.3 %. Evaluation of the stereochemistry of the re®ned model using PROCHECK29 showed that all the residues are in the allowed region (91.9 % in the most favored and 8.1 % in the additional allowed regions) and no residues are in the disallowed regions. Overview of the structure The overall folding topology of NHERF PDZ1 is similar to previously determined PDZ structures,4 ± 10 consisting of six b-strands (b1-b6) and two a-helices (a1 and a2) (Figure 2). The strands com-

prise an antiparallel b-sandwich with one b-sheet formed by b1, b6, b4, and b5, and the second b-sheet formed by b2, b3, and b4 strands. Hydrophobic residues Leu17, Phe26, Leu28, Ile39, Leu53, Leu59, Val76, Ile79, Val86, and Leu88 form the core of the molecule. The structure is well-ordered Ê 2, except for the with an average B-factor of 16.7 A variable b2-b3 loop, which is slightly disordered Ê 2, and the sidewith an average B-factor of 25 A chain of Lys69 in the b5-a2 loop, which is partially disordered and not visible in the electron density map. In addition, the side-chains of Ile39, Glu49, and Leu53 are modeled in two different conformations. Superposition of the NHERF PDZ1 and the peptide-bound PSD-95 PDZ34 crystal structures shows very similar backbone structures (Figure 3). The root-mean-square deviation (rmsd) in Ca positions Ê between the two models. However, is 1.1 A notable differences are observed in the NHERF PDZ1 b1-b2 loop (residues Cys19-Gly23) which is displaced away from the a2-b6 loop as compared

Table 1. Statistics of NHERF PDZ1 structure determination and re®nement A. Data collectiona Ê) Wavelength (A Ê) Resolution range (A Total reflections Unique reflections Average (I/s(I)) Completeness (%) Redundancy Rsym (%) B. MAD phasing Ê) Resolution range (A Scattering factor f 0 Scattering factor f 00 Phasing power, acentrics (centrics) Rcullis Overall mean figure of merit Mean figure of merit after solvent flattening C. Refinement Ê) Resolution range (A Number of working reflections Number of test reflections Rcryst (%) Rfree (%) Number of non-hydrogen atoms Number of water molecules Ê) Bond lengthsb (A Bond anglesb (deg.) Ê 2) Overall isotropic B-factor (A Ê 2) Main-chain B-factor (A Ê 2) Side-chain B-factor (A Ê 2) Solvent B-factor (A

Native

Hg remote

Hg peak

Hg edge

0.98401 36-1.5 246,628 14,916 35.0 (10.3) 99.7 (98.9) 7.5 (4.6) 3.9 (15.2)

0.99315 44-1.9 193,730 6926 38.3 (18.7) 92.4 (60.0) 14.3 (10.7) 6.4 (12.8)

1.00784 44-1.9 181,058 6820 39.2 (16.9) 90.9 (56.1) 17.3 (12.6) 6.5(14.2)

1.00392 44-1.9 183,331 6780 37.4 (16.8) 89.9 (53.4) 14.3 (10.6) 6.6 (13.8)

44-2.0 ÿ11.0317 9.9044 0.74 (0.62) 0.77 (0.82)

44-2.0 ÿ15.9779 10.1606 -

44-2.0 ÿ23.0652 3.9053 1.22 (0.86) 0.69 (0.76)

0.67 0.86 25-1.5 13,316 1500 18.1 20.3 697 144 0.007 1.5 16.7 13.20 15.18 29.09

Rsym ˆ j(I ÿ hIi)j/(I), where I is the observed integrated intensity, hIi is the average integrated intensity obtained from multiple measurements, and the summation is over all observed re¯ections. Phasing power ˆ hFHi/E, where hFHi is the mean heavy-atom structure factor and E is the residual lack of closure error. Rcullis is the mean residual lack of closure error divided by the isomorphous difference. Figure of merit is jP(a) eia/P(a)j, where P(a) is the phase probability distribution. Rcryst ˆ jjFobsj ÿ kjFcalcjj/(jFobsj, where Fobs and Fcalc are the observed and calculated structure factors, respectively. Rfree is calculated as Rcryst using 10 % of the re¯ection data chosen randomly and omitted from the re®nement calculations. a Ê for native and 1.97-1.90 A Ê for HgCl2-derivative data). Data in parentheses refer to the last resolution bin (1.55-1.50 A b Bond lengths and angles are root-mean-square deviations from ideal values.

966

Structure of the NHERF PDZ1 Domain

termini and neighboring non-crystallographically related molecules.6 PDZ1-peptide ligand interaction without b -sheet augmentation

Figure 2. Ribbon diagram of the NHERF PDZ1 structure. The b-strands are labeled b1-b6 and the a-helices are labeled a1 and a2. The Figure was generated using MOLSCRIPT46 and Raster3D.47

Ê ). Simito the PSD-95 PDZ3 structure (rmsd 3.3 A larly, differences exist in the b2-b3 loops and the carboxyl termini of the two domains (Figure 3).

PDZ1 self-association in the crystalline state The NHERF PDZ1 crystal structure produces a dimeric arrangement of PDZ1 molecules related by 2-fold crystallographic symmetry (Figure 4). This intermolecular association is mediated exclusively by the PDZ1 carboxyl-terminal sequence DEQL which inserts into the peptide-binding pocket of a neighboring PDZ1 molecule. Although this sequence is similar to the D(S/T)XL motif that is a high-af®nity ligand for NHERF PDZ1,14,21 this interaction was unexpected because it was previously shown that aspartate at position ÿ2 of the peptidic ligand is detrimental for binding to this domain.14 Nevertheless, since NHERF PDZ1 selfassociates in solution,18,19 it is conceivable that the region containing the DEQL motif may mediate the PDZ1 self-association in solution by a mechanism similar to that observed in the crystalline state. In this context, the occupation of the peptidebinding pocket by the carboxyl terminus of a neighboring PDZ1 molecule provides an explanation for the failed attempts to co-crystallize the PDZ1 complexed with synthetic peptides. Notably, PDZ self-association was previously observed in crystals of the hCASK PDZ domain, resulting from intermolecular interactions between carboxyl

In all known PDZ-peptide structures, the peptide ligand is oriented antiparallel to the b2 strand and extends the b-sheet of the PDZ domain.4,6,7 ± 10 By contrast, the NHERF PDZ1 self-association observed in the crystal does not involve a b-sheet augmentation process. The side-chain and carboxylate group of Leu0 enter into the deep hydrophobic cavity of PDZ1 at a steep angle, whereas the sidechain of Gln ÿ1 is oriented towards the solution and does not participate in intermolecular interactions (Figure 5), consistent with previous ®ndings on the limited role of the residue at position ÿ1 for the PDZ-peptide binding speci®city.4,6,7-10 Surprisingly, the Oe1 and Oe2 atoms of Glu ÿ2 hydrogen bond with the Ne2 atom of His72. As a result of this interaction, the large side-chain of Glu ÿ2 dislodges the amino-terminal region of the peptide ligand to a remote position that does not allow interactions with the b2 strand of PDZ1 and consequently, b-sheet augmentation (Figure 5). This interaction was unexpected because the histidine residue at the beginning of the a2 helix is known to play a critical role in the speci®city of class I PDZ domain-peptide interaction through formation of a hydrogen bond with the hydroxyl group of a serine or threonine residue at position ÿ2 of the peptide ligand.4,6,7-10 The backbone of the peptide ligand bends sharply at Asp ÿ3, and the side-chain of this amino acid is oriented towards the solution and does not interact with PDZ1. In this arrangement, the invading peptide is highly ordered and very well de®ned, as indicated by the high-quality electron density map (Figure 5) and low temperature factors (average value for the last Ê 2). The lack of interactions augfour residues 19.4 A menting the PDZ1 b-sheet suggests that the high af®nity of the PDZ1 peptide-binding pocket for the carboxyl-terminal leucine residue is the primary force stabilizing this complex formation. Although it is not presently known whether this mode of interaction occurs in solution, we expect that NHERF PDZ1 will likely interact with high-af®nity peptide ligands containing the shorter side-chains of Ser/Thr at position ÿ2 through b-sheet augmentation, as described for other PDZ domains. Structural basis for carboxyl-terminal leucine recognition The side-chain and carboxylate group of Leu 0 enter into a deep cavity formed by Tyr24, Gly25, Phe26, Leu28, Val76, and Ile79 residues (Figure 6). The Cd2 atom of Leu0 makes a hydrophobic interaction with Cg1 of Val76 and an intricate network of hydrogen bonds stabilizes the carboxylate group within the hydrophobic pocket (Figures 6 and 7). The carboxyl-terminal oxygen atom of Leu0

Structure of the NHERF PDZ1 Domain

967

Figure 3. Stereo drawing showing the superposition of the human NHERF PDZ1 and rat PSD-95 PDZ3 structures. The Ca backbone traces of NHERF PDZ1 shown in red and the peptide-bound PSD-95 PDZ3 (PDB 1be9) shown in blue, were superimposed with the peptide ligands removed from the models using the program O. The Figure was made using MOLSCRIPT.

hydrogen bonds directly with the amide nitrogen atoms of Tyr24 and Gly25, and indirectly with the carbonyl oxygen atom of Phe26 through an ordered water molecule (Wat20). In addition, the carbonyl oxygen atom of Leu0 hydrogen bonds directly with the amide nitrogen atom of Tyr24 and indirectly with Nz of Lys19 through two ordered water molecules (Wat10 and Wat5). In this

Figure 4. Ribbon diagram showing the self-association of two NHERF PDZ1 domains in the crystalline state. Each carboxyl terminus serves as a ligand for a neighboring PDZ1 molecule related by 2-fold crystallographic symmetry (the 2-fold crystallographic axis is perpendicular to the page plane).

respect, the NHERF PDZ1 differs from the PSD-95 PDZ3 structure,4 where the carbonyl oxygen atom of Val0 hydrogen bonds with the NZ1 atom of Arg318 through one ordered water molecule (corresponding to Wat10 of NHERF PDZ1). How do class I PDZ domains discriminate between the side-chains of carboxyl-terminal leucine and valine residues? Comparison of the NHERF PDZ1 and PSD-95 PDZ3 hydrophobic cavities bound to Leu0 and Val0, respectively, reveals that they have different sizes and shapes (Figure 8). The NHERF PDZ1 cavity is large (calculated surÊ 2) and the isobutyl group of face area of 183.43 A Leu0 ®ts snugly in this pocket. It is conceivable that the smaller side-chains of valine and alanine would leave vacated spaces within the hydrophobic cavity that would be energetically unfavorable.30 Thus, the tight ®t of the leucine side-chain in the hydrophobic cavity provides an explanation for the strict requirement for carboxylterminal leucine in all the high-af®nity ligands of NHERF PDZ1, and the poor af®nity of this domain for carboxyl-terminal valine and alanine residues.14,20,21 On the other hand, the smaller cavity of PSD-95 PDZ3 (calculated surface area Ê 2) has evolved to interact tightly with the 110.69 A isopropyl group of valine, and accommodation of the larger isobutyl group of leucine would be stereochemically challenging. It therefore appears that sequence variation among different PDZ domains generates hydrophobic cavities with distinct volumes and shapes, thus providing a selectivity mechanism for peptide ligand interaction based on the stereochemical complementarity of

968

Structure of the NHERF PDZ1 Domain

Figure 5. PDZ1-peptide ligand interaction without b-sheet augmentation. Stereo view ball-and-stick representation of the PDZ1 carboxyl-terminal region (residues 94-99) bound to a crystallographically related PDZ1 molecule. Ê resolution and contoured at 0.8s is superimposed on the pepA 2Fobs ÿ Fcalc electron density map calculated at 1.5 A tide ligand. Carbon atoms are shown in black, oxygen atoms in red, nitrogen atoms in blue, and hydrogen bonds as broken lines. The Figure was made using MOLSCRIPT, BOBSCRIPT,48 and Raster3D.

the peptidic carboxyl-terminal residue and the volume/shape of the cavity. Perspective The present structure provides insights into the molecular mechanism by which class I PDZ domains recognize a carboxyl-terminal leucine residue, and serves as a starting point for understanding the structural basis of the NHERF interaction with several membrane receptors and ion channels, including b2AR, PDGFR, and CFTR. Upon binding of NHERF PDZ1 to the carboxyl termini of these proteins, Leu0 will likely occupy the peptide-binding pocket in a manner similar to that observed in our structure. However, use of this model to derive the structural determinants underlying the speci®city of the NHERF PDZ1 interaction with physiological ligands, such as the carboxyl termini of b2AR, PDGFR, and CFTR, is limited because of the unusual mode of NHERF PDZ1 self-association observed in the crystal. In fact, we expect that the carboxyl termini of b2AR, PDGFR, and CFTR will interact with NHERF PDZ1 through a b-sheet augmentation process, because these proteins contain a serine or threonine residue at position ÿ2. Consequently, the hydrophobic interactions and the hydrogen bonding networks that stabilize the physiologic ligands into the NHERF PDZ1 pocket will likely differ from those described in the present structure. Elucidation of the structural determinants of the NHERF PDZ1 interaction with b2AR, PDGFR, and

CFTR may have important clinical applications. For example, the recent ®nding that NHERF PDZ1 binds with high af®nity to PDGFR carboxyl terminus and potentiates the mitogenic activity of this receptor,23 in combination with the observation that the NHERF gene is rapidly activated by estrogen in estrogen receptor positive breast cancer cells,31 raise the intriguing possibility that NHERF may play an important role in breast cancer development. Therefore, the analysis of the NHERF PDZ1-PDGFR interface at the atomic level will provide the structural framework for the design of inhibitors that will disrupt the interaction of these proteins and decrease the PDGFR proliferative activity, with potential applications in cancer treatment. Furthermore, since many of the CFTR mutations causing cystic ®brosis affect the traf®cking of CFTR from the endoplasmic reticulum to the apical plasma membrane,32 elucidation of the molecular mechanisms underlying the NHERF PDZ1mediated functional localization of CFTR in the plasma membrane may lead to the development of new strategies for improving the clinical phenotype of this disease.

Materials and Methods Protein expression and purification A PCR-ampli®ed DNA fragment encoding the human NHERF PDZ1 (residues 11-99) was cloned into a modi®ed pGEX-2T vector (Pharmacia). Recombinant PDZ1 was expressed in E. coli BL21 (DE3) cells as a glutathione

969

Structure of the NHERF PDZ1 Domain

Figure 6. Stereo view representation of the NHERF PDZ1 peptide-binding pocket bound to the carboxyl-terminal peptide ligand (gray). The side-chains of His27 and Leu28 are not shown for clarity. Atoms are colored as described in the legend to Figure 5. Water molecules are shown as green spheres and hydrogen bonds as orange broken lines. A hydrophobic interaction between Leu0 and Val76 is depicted as a black broken line.

S-transferase (GST) fusion protein by growing the cells at 37  C until they reached an A600 of 1.0, followed by induction with 0.1 mM isopropyl-b-D-thiogalactopyranoside for three hours. The cells were collected, resuspended in 1  PBS containing 5 mM DTT and protease inhibitors (Boehringer Mannheim), and lysed using a pressure gradient cell disrupter (Energy Service Co.). Triton X-100 was added to the lysate to a ®nal concentration of 1 % (v/v) and the sample was rocked at room temperature for 30 minutes, followed by centrifugation at 31,000 g for ten minutes. The GST-PDZ1 fusion protein was puri®ed from the supernatant using glutathione-Sepharose 4B resin following standard methods (Pharmacia). The PDZ1 protein was released from the resin-bound fusion protein by digestion with thrombin (Haematologic Technologies; 40 units per liter of cell culture, two hours incubation at room temperature), and puri®ed to homogeneity on a Q12 anion exchange column (BioRad) using a linear gradient of 0-1 M NaCl in a buffer of 25 mM Tris-HCl (pH 8.0). The puri®ed PDZ1 protein was dialyzed against 10 mM NaCl, 10 mM Hepes, 0.5 mM DTT (pH 7.5), and was concentrated using a Centriprep concentrator to 20 mg/ml.

Crystallization and data collection Recombinant PDZ1 protein at a concentration of 20 mg/ml was crystallized using the sitting drop vapor diffusion method in 0.05 M potassium sodium tartrate, 0.06 M sodium citrate, 2 M ammonium sulfate (pH 5.9), and 20 mM MnCl2, at 20  C. Crystals with maximum dimensions of 0.6 mm  0.3 mm  0.2 mm grew in 24 hours. Mercury derivatives of PDZ1 crystals were obtained by soaking them in mother liquor supplemented with 1.4 mM HgCl2 for 12 hours, and the structure was solved by the multiwavelength anomalous diffraction (MAD) method using mercury as the anomalous scatterer. For synchrotron data collection, native and HgCl2-derivatized crystals were cryoprotected in 40 % (v/v) 2-methyl-2,4-pentanediol (MPD) for ten minutes and ¯ash-frozen in a stream of liquid nitrogen. MAD data of a HgCl2-derivatized crystal were collected at three wavelengths near the LIII absorption edge of mercury, using a Quantum-4 CCD detector (Area Detector Systems Corp.) on the F2 beamline at the Cornell High Energy Synchrotron Source (CHESS), Ithaca, NY. Monochromatic data of the native crystals diffracting to Ê were also collected at the CHESS F2 beamline. The 1.5 A data were indexed and the intensities integrated, scaled,

970

Structure of the NHERF PDZ1 Domain

Figure 7. Two-dimensional schematic representation of the contacts observed in the NHERF PDZ1 binding pocket. PDZ1 residues (orange) making hydrogen bonds and hydrophobic contacts with the peptide ligand (purple) are shown in ball-and-stick representation. Carbon atoms are shown in white, oxygen atoms in red, and nitrogen atoms in blue. Water molecules are shown as green spheres. Hydrogen bonds are depicted as broken lines and numbers Ê . Val76 involved in hydrophobic interaction with the Cd2 atom of Leu0 is shown as an arc with indicate distances in A radial spokes. The Figure was generated using the program LIGPLOT.49

and merged using the programs DENZO and SCALEPACK in the package HKL 1.96.33 The crystals belong to space group P3221, with unit cell dimensions Ê , c ˆ 58.9 A Ê , and contain one molecule in a ˆ b ˆ 51.6 A the asymmetric unit. The calculated Matthews coef®cient Ê 3 Daÿ1, corresponding to 48.3 % (v/v) solVM is 2.38 A vent content.34 The data collection statistics are shown in Table 1. Structure determination and refinement Two mercury atoms were located by analyzing MAD data using the automated structure determination pack-

age SOLVE. Similar solutions for the mercury positions were also obtained with the programs SHARP,35 SHELXL-97,36 and SnB.37 These sites were re®ned, and initial phases were generated for the data in the resolÊ using the program MLPHARE as ution range 44-2.0 A implemented in the CCP4 software suite.38 Because the handedness of the calculated phases was not known, assignment of the space group as P3221 or P3121 was not possible. Therefore, both the original positions of the mercury sites and their opposite handedness were re®ned using MLPHARE. The obtained phases were further improved by solvent ¯attening and histogram matching with a solvent content of 48 %, using the pro-

Structure of the NHERF PDZ1 Domain

971

Figure 8. Carboxyl-terminal leucine and valine recognition by PDZ domains. Surface representation of the hydrophobic pockets of (a) human NHERF PDZ1 and (b) rat PSD-95 PDZ3 bound to carboxylterminal leucine and valine residues, respectively. The limits of the hydrophobic cavities are denoted with jagged edges. The Figure was made with the program GRASP.50

gram DM in the CCP4 package.38 After 11 cycles of phase extension with DM, the Rfree for space group P3221 converged from 54.4 % to 39.2 %, whereas the Rfree for the enantiomorphic space group P3121 changed only from 57.9 % to 46.1 %, indicating that the correct space group was P3221. The calculated phases were re®ned Ê for the native data set using DM. and extended to 1.5 A The ®nal map after phase re®nement and solvent ¯attening was of excellent quality and readily interpretable. An initial model was built automatically using the program ARP/wARP39 by placing free atoms into the electron density peaks. The protein model was built based on the position of these free atoms, and the structure factors were improved by combining the initial experimental phases with those calculated from the model. In total, 503 protein atoms (out of 697 of the ®nal model) were built into the map with ARP/wARP. Crystallographic re®nement was carried out with molecular dynamics using the program CNS.40 A total of 10 % of the re¯ections were reserved throughout re®nement for the calculation of the free R-factor.41 Several cycles of simulated

annealing at 3000 K, including minimization and individual B-factor re®nement were performed. After each cycle of re®nement, manual model building was carried out using sA-weighted 2Fobs ÿ Fcalc and Fobs ÿ Fcalc maps,42 and the complete model was built into the map using the program O.43 After manual assignment of 144 ordered water molecules, the ®nal R and Rfree factors were 18.1 % and 20.3 %, respectively, for 13,316 re¯ecÊ . The phasing and tions in the resolution range 25-1.5 A the re®nement statistics are given in Table 1. The ®nal model consists of the entire NHERF PDZ1 domain (residues 11-99), comprising 697 non-hydrogen atoms and 144 water molecules. The recombinant PDZ1 protein used for crystallization contains at its amino terminus ®ve additional vector-derived residues (GSSRM), from which the methionine residue is well-ordered, the arginine residue is poorly de®ned and modeled as glycine residue, while the remaining three residues are disordered and not included in the ®nal model. The average temperature factors for all main-chain and side-

972 Ê 2 and 15.2 A Ê 2, respectively. The chain residues are 13.2 A temperature factors for the water molecules range from Ê 2, with an average value of 29.1 A Ê 2. 10.6 to 52.9 A Protein Data Bank accession code The atomic coordinates and structure factors for the human NHERF PDZ1 crystal structure have been deposited in the RCSB Protein Data Bank with accession number 1g9o.

Acknowledgments We thank Dr Jerome E. Groopman for his generous support that made this project possible. We also thank Dr Vijaya Ramesh for providing the human NHERF clone, and the staff of the Macromolecular Diffraction Facility at the Cornell High Energy Synchrotron Source, beamline F2, for assistance during data collection. J.A.A.L. is an Established Investigator of the American Heart Association.

References 1. Craven, S. E. & Bredt, D. S. (1998). PDZ proteins organize synaptic signaling pathways. Cell, 93, 495498. 2. Ponting, C. P., Phillips, C., Davis, K. E. & Blake, D. J. (1997). PDZ domains: targeting signalling molecules to sub-membranous sites. BioEssays, 19, 469-479. 3. Garner, C. C., Nash, J. & Huganir, R. L. (2000). PDZ domains in synapse assembly and signalling. Trends Cell Biol. 10, 274-280. 4. Doyle, D. A., Lee, A., Lewis, J., Kim, E., Sheng, M. & MacKinnon, R. (1996). Crystal structures of a complexed and peptide-free membrane protein-binding domain: molecular basis of peptide recognition by PDZ. Cell, 85, 1067-1076. 5. Morais Cabral, J. H., Petosa, C., Sutcliffe, M. J., Raza, S., Byron, O., Poy, F., Marfatia, S. M., Chishti, A. H. & Liddington, R. C. (1996). Crystal structure of a PDZ domain. Nature, 382, 649-652. 6. Daniels, D. L., Cohen, A. R., Anderson, J. M. & BruÈnger, A. T. (1998). Crystal structure of the hCASK PDZ domain reveals the structural basis of class II PDZ domain target recognition. Nature Struct. Biol. 5, 317-325. 7. Schultz, J., Hoffmuller, U., Krause, G., Ashurst, J., Macias, M. J., Schmieder, P., Schneider-Mergener, J. & Oschkinat, H. (1998). Speci®c interactions between the syntrophin PDZ domain and voltage-gated sodium channels. Nature Struct. Biol. 5, 19-24. 8. Tochio, H., Zhang, Q., Mandal, P., Li, M. & Zhang, M. (1999). Solution structure of the extended neuronal nitric oxide synthase PDZ domain complexed with an associated peptide. Nature Struct. Biol. 6, 417-421. 9. Kozlov, G., Gehring, K. & Ekiel, I. (2000). Solution structure of the PDZ2 domain from human phosphatase hPTP1E and its interactions with C-terminal peptides from the Fas receptor. Biochemistry, 39, 2572-2580. 10. Hillier, B. J., Christopherson, K. S., Prehoda, K. E., Bredt, D. S. & Lim, W. A. (1999). Unexpected modes of PDZ domain scaffolding revealed by structure of nNOS-syntrophin complex. Science, 284, 812-815.

Structure of the NHERF PDZ1 Domain 11. Songyang, Z., Fanning, A. S., Fu, C., Xu, J., Marfatia, S. M., Chishti, A. H., Crompton, A., Chan, A. C., Anderson, J. M. & Cantley, L. C. (1997). Recognition of unique carboxyl-terminal motifs by distinct PDZ domains. Science, 275, 73-77. 12. Weinman, E. J., Steplock, D., Wang, Y. & Shenolikar, S. (1995). Characterization of a protein cofactor that mediates protein kinase A regulation of the renal brush border membrane Na‡-H‡ exchanger. J. Clin. Invest. 95, 2143-2149. 13. Reczek, D., Berryman, M. & Bretscher, A. (1997). Identi®cation of EBP50: a PDZ-containing phosphoprotein that associates with members of the ezrinradixin-moesin family. J. Cell Biol. 139, 169-179. 14. Hall, R. A., Ostedgaard, L. S., Premont, R. T., Blitzer, J. T., Rahman, N., Welsh, M. J. & Lefkowitz, R. J. (1998). A C-terminal motif found in the b2-adrenergic receptor, P2Y1 receptor and cystic ®brosis transmembrane conductance regulator determines binding to the Na ‡/H ‡ exchanger regulatory factor family of PDZ proteins. Proc. Natl Acad. Sci. USA, 95, 8496-8501. 15. Minkoff, C., Shenolikar, S. & Weinman, E. J. (1999). Assembly of signaling complexes by the sodiumhydrogen exchanger regulatory factor family of PDZ-containing proteins. Curr. Opin. Nephrol. Hypertens. 8, 603-608. 16. Weinman, E. J., Minkoff, C. & Shenolikar, S. (2000). Signal complex regulation of renal transport proteins: NHERF and regulation of NHE3 by PKA. Am. J. Physiol. Renal Physiol. 279, F393-F399. 17. Murthy, A., Gonzalez-Agosti, C., Cordero, E., Pinney, D., Candia, C., Solomon, F., Gusella, J. & Ramesh, V. (1998). NHE-RF, a regulatory cofactor for Na‡-H‡ exchange, is a common interactor for merlin and ERM (MERM) proteins. J. Biol. Chem. 273, 1273-1276. 18. Fouassier, L., Yun, C. C., Fitz, J. G. & Doctor, R. B. (2000). Evidence for ezrin-radixin-moesin-binding phosphoprotein 50 (EBP50) self-association through PDZ-PDZ interactions. J. Biol. Chem. 275, 2503925045. 19. Shenolikar, S., Minkoff, C. M., Steplock, D. A., Evangelista, C., Liu, M. & Weinman, E. J. (2001). N-terminal PDZ domain is required for NHERF dimerization. FEBS Letters, 489, 233-236. 20. Hall, R. A., Premont, R. T., Chow, C. W., Blitzer, J. T., Pitcher, J. A., Claing, A., Stoffel, R. H., Barak, L. S., Shenolikar, S., Weinman, E. J., Grinstein, S. & Lefkowitz, R. J. (1998). The b2-adrenergic receptor interacts with the Na‡/H‡-exchanger regulatory factor to control Na‡/H‡ exchange. Nature, 392, 626630. 21. Wang, S., Raab, R. W., Schatz, P. J., Guggino, W. B. & Li, M. (1998). Peptide binding consensus of the NHE-RF-PDZ1 domain matches the C-terminal sequence of cystic ®brosis transmembrane conductance regulator (CFTR). FEBS Letters, 427, 103-108. 22. Short, D. B., Trotter, K. W., Reczek, D., Kreda, S. M., Bretscher, A., Boucher, R. C., Stutts, M. J. & Milgram, S. L. (1998). An apical PDZ protein anchors the cystic ®brosis transmembrane conductance regulator to the cytoskeleton. J. Biol. Chem. 273, 19797-19801. 23. Maudsley, S., Zamah, A. M., Rahman, N., Blitzer, J. T., Luttrell, L. M., Lefkowitz, R. J. & Hall, R. A. (2000). Platelet-derived growth factor receptor association with Na‡/H‡ exchanger regulatory fac-

Structure of the NHERF PDZ1 Domain

24.

25.

26.

27.

28. 29.

30.

31.

32. 33. 34. 35.

36.

tor potentiates receptor activity. Mol. Cell. Biol. 20, 8352-8363. Cao, T. T., Deacon, H. W., Reczek, D., Bretscher, A. & von Zastrow, M. (1999). A kinase-regulated PDZdomain interaction controls endocytic sorting of the b2-adrenergic receptor. Nature, 401, 286-290. Moyer, B. D., Denton, J., Karlson, K. H., Reynolds, D., Wang, S., Mickle, J. E., Milewski, M., Cutting, G. R., Guggino, W. B., Li, M. & Stanton, B. A. (1999). A PDZ-interacting domain in CFTR is an apical membrane polarization signal. J. Clin. Invest. 104, 1353-1361. Moyer, B. D., Duhaime, M., Shaw, C., Denton, J., Reynolds, D., Karlson, K. H., Pfeiffer, J., Wang, S., Mickle, J. E., Milewski, M., Cutting, G. R., Guggino, W. B., Li, M. & Stanton, B. A. (2000). The PDZ-interacting domain of cystic ®brosis transmembrane conductance regulator is required for functional expression in the apical plasma membrane. J. Biol. Chem. 275, 27069-27074. Hendrickson, W. A. & Ogata, C. M. (1997). Phase determination from multiwavelength anomalous diffraction measurements. Methods Enzymol. 276, 494-523. Terwilliger, T. C. & Berendzen, J. (1999). Automated MAD and MIR structure solution. Acta Crystallog. sect. D, 55, 849-861. Laskowski, R. J., MacArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283-291. Eriksson, A. E., Baase, W. A., Zhang, X. J., Heinz, D. W., Blaber, M., Baldwin, E. P. & Matthews, B. W. (1992). Response of a protein structure to cavitycreating mutations and its relation to the hydrophobic effect. Science, 255, 178-183. Ediger, T. R., Kraus, W. L., Weinman, E. J. & Katzenellenbogen, B. S. (1999). Estrogen receptor regulation of the Na‡/H‡ exchange regulatory factor. Endocrinology, 140, 2976-2982. Kopito, R. R. (1999). Biosynthesis and degradation of CFTR. Physiol. Rev. 79, S167-S173. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307-326. Matthews, B. W. (1968). Solvent content of protein crystals. J. Mol. Biol. 33, 491-497. de la Fortelle, E. & Bricogne, G. (1997). Maximumlikelihood heavy-atom parameter re®nement for multiple isomorphous replacement and multiwavelength anomalous diffraction methods. Methods Enzymol. 276, 472-494. Sheldrick, G. M. & Schneider, T. R. (1997). SHELXL: high-resolution re®nement. Methods Enzymol. 277, 319-343.

973 37. Weeks, C. M. & Miller, R. (1999). The design and implementation of SnB version 2.0. J. Appl. Crystallog. 32, 120-124. 38. Collaborative Computational Project Number 4 (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, 760-763. 39. Perrakis, A., Morris, R. & Lamzin, V. S. (1999). Automated protein model building combined with iterative structure re®nement. Nature Struct. Biol. 6, 458-463. 40. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W., Jiang, J. S., Kuszewski, J., Nilges, M., Pannu, N. S., Read, R. J., Rice, L. M., Simonson, T. & Warren, G. L. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallog. sect. D, 54, 905-921. 41. Brunger, A. T. (1992). Free R-value: a novel statistical quantity for assessing the accuracy of crystal structures. Nature, 355, 472-475. 42. Read, R. J. (1986). Improved Fourier coef®cients for maps using phases from partial structures with errors. Acta Crystallog. sect. A, 42, 140-149. 43. Jones, T. A., Zou, J. Y., Cowan, S. & Kjeldgaard, M. (1991). Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallog. sect. A, 47, 110-119. 44. Cho, K. O., Hunt, C. A. & Kennedy, M. B. (1992). The rat brain postsynaptic density fraction contains a homolog of the Drosophila discs-large tumor suppressor protein. Neuron, 9, 929-942. 45. Schuler, G. D., Altschul, S. F. & Lipman, D. J. (1991). A workbench for multiple alignment construction and analysis. Proteins: Struct. Funct. Genet. 9, 180190. 46. Kraulis, P. J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, 946-950. 47. Merritt, E. A. & Murphy, M. E. P. (1994). Raster3d version-2.0: a program for photorealistic molecular graphics. Acta Crystallog. sect. D, 50, 869-873. 48. Esnouf, R. M. (1997). An extensively modi®ed version of MolScript that includes greatly enhanced coloring capabilities. J. Mol. Graph. Model. 15, 132134. 49. Wallace, A. C., Laskowski, R. A. & Thornton, J. M. (1995). LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions. Protein Eng. 8, 127-134. 50. Nicholls, A., Sharp, K. A. & Honig, B. (1991). Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins: Struct. Funct. Genet. 11, 281-296.

Edited by D. Rees (Received 30 January 2001; received in revised form 16 March 2001; accepted 18 March 2001)