Crystal Structure of GRIP1 PDZ6-Peptide Complex Reveals the Structural Basis for Class II PDZ Target Recognition and PDZ Domain-mediated Multimerization

PDZ domains bind to short segments within target proteins in a sequence-specific fashion. Glutamate receptor-interacting protein (GRIP)/ABP family proteins contain six to seven PDZ domains and interact via the sixth PDZ domain (class II) with the C termini of various proteins including liprin-α. In addition the PDZ456 domain mediates the formation of homo- and heteromultimers of GRIP proteins. To better understand the structural basis of peptide recognition by a class II PDZ domain and PDZ-mediated multimerization, we determined the crystal structures of the GRIP1 PDZ6 domain alone and in complex with a synthetic C-terminal octapeptide of human liprin-α at resolutions of 1.5 and 1.8 Å, respectively. Remarkably, unlike other class II PDZ domains, Ile-736 at αB5 rather than conserved Leu-732 at αB1 makes a direct hydrophobic contact with the side chain of the Tyr at the −2 position of the ligand. Moreover, the peptide-bound structure of PDZ6 shows a slight reorientation of helix αB, indicating that the second hydrophobic pocket undergoes a conformational adaptation to accommodate the bulkiness of the Tyr side chain, and forms an antiparallel dimer through an interface located at a site distal to the peptide-binding groove. This configuration may enable formation of GRIP multimers and efficient clustering of GRIP-binding proteins. PDZ domains bind to short segments within target proteins in a sequence-specific fashion. Glutamate receptor-interacting protein (GRIP)/ABP family proteins contain six to seven PDZ domains and interact via the sixth PDZ domain (class II) with the C termini of various proteins including liprin-α. In addition the PDZ456 domain mediates the formation of homo- and heteromultimers of GRIP proteins. To better understand the structural basis of peptide recognition by a class II PDZ domain and PDZ-mediated multimerization, we determined the crystal structures of the GRIP1 PDZ6 domain alone and in complex with a synthetic C-terminal octapeptide of human liprin-α at resolutions of 1.5 and 1.8 Å, respectively. Remarkably, unlike other class II PDZ domains, Ile-736 at αB5 rather than conserved Leu-732 at αB1 makes a direct hydrophobic contact with the side chain of the Tyr at the −2 position of the ligand. Moreover, the peptide-bound structure of PDZ6 shows a slight reorientation of helix αB, indicating that the second hydrophobic pocket undergoes a conformational adaptation to accommodate the bulkiness of the Tyr side chain, and forms an antiparallel dimer through an interface located at a site distal to the peptide-binding groove. This configuration may enable formation of GRIP multimers and efficient clustering of GRIP-binding proteins. glutamate receptor-interacting protein AMPA receptor-binding protein α-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid multiwavelength anomalous dispersion root-mean square deviation Synaptic localization and clustering of ion channels and receptors is often mediated by scaffolding molecules containing the protein-protein interaction motifs called PDZ (Postsynaptic density-95/Discs large/Zona occludens-1) domains (1Sheng M. Sala C. Annu. Rev. Neurosci. 2001; 24: 1-29Google Scholar). One of the most abundant molecular recognition elements, these globular domains each contain two α-helices and six β-strands. They usually bind selectively to the C terminus or a short internal segment of interacting proteins (1Sheng M. Sala C. Annu. Rev. Neurosci. 2001; 24: 1-29Google Scholar) and are categorized into four classes according to their specificity for the C-terminal target sequences (2Songyang 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. Science. 1997; 275: 73-77Google Scholar). Class I PDZ domains bind to a C-terminal motif with the sequence X-Ser/Thr-X-Val/Leu-COOH, whereX represents any residue, while class II PDZ domains preferX-Φ-X-Φ-COOH, where Φ is usually a large hydrophobic residue. Both class I and II domains have a preference for a hydrophobic residue at the 0 position of the ligand. Class III PDZ domains prefer the sequenceX-Asp-X-Val-COOH in which a negatively charged amino acid is at the −2 position (3Stricker N.L. Christopherson K.S. Yi B.A. Schatz P.J. Raab R.W. Dawes G. Bassett Jr., D.E. Bredt D.S. Li M. Nat. Biotechnol. 1997; 15: 336-342Google Scholar), while class IV domains prefer the sequence X-Ψ-Asp/Glu-COOH in which an acidic residue is at the C-terminal position and where Ψ represents an aromatic residue (4Vaccaro P. Brannetti B. Montecchi-Palazzi L. Philipp S. Citterrich M.H. Cesareni G. Dente L. J. Biol. Chem. 2001; 276: 42122-42130Google Scholar). In addition, there are other classes of PDZ domains that do not fall into any of the aforementioned classes (5Maximov A. Sudhof T.C. Bezprozvanny I. J. Biol. Chem. 1999; 274: 24453-24456Google Scholar, 6Borrell-Pages M. Fernandez-Larrea J. Borroto A. Rojo F. Baselga J. Arribas J. Mol. Biol. Cell. 2000; 11: 4217-4225Google Scholar), and there are minor discrepancies in the proposed classifications of PDZ domains (7Bezprozvanny I. Maximov A. FEBS Lett. 2001; 509: 457-462Google Scholar, 8Harris B.Z. Lim W.A. J. Cell Sci. 2001; 114: 3219-3231Google Scholar). Members of the GRIP1 family proteins (GRIP1 and GRIP2/ABP) contain six to seven PDZ domains (9Srivastava S. Osten P. Vilim F.S. Khatri L. Inman G. States B. Daly C. DeSouza S. Abagyan R. Valtschanoff J.G. Weinberg R.J. Ziff E.B. Neuron. 1998; 21: 581-591Google Scholar, 10Wyszynski M. Valtschanoff J.G. Naisbitt S. Dunah A.W. Kim E. Standaert D.G. Weinberg R. Sheng M. J. Neurosci. 1999; 19: 6528-6537Google Scholar,11Dong H. Zhang P. Song I. Petralia R.S. Liao D. Huganir R.L. J. Neurosci. 1999; 19: 6930-6941Google Scholar). GRIP PDZ45, which is classified as a class II PDZ domain (1Sheng M. Sala C. Annu. Rev. Neurosci. 2001; 24: 1-29Google Scholar), binds to the C terminus of the GluR2/3 subunit of AMPA glutamate receptors (9Srivastava S. Osten P. Vilim F.S. Khatri L. Inman G. States B. Daly C. DeSouza S. Abagyan R. Valtschanoff J.G. Weinberg R.J. Ziff E.B. Neuron. 1998; 21: 581-591Google Scholar, 10Wyszynski M. Valtschanoff J.G. Naisbitt S. Dunah A.W. Kim E. Standaert D.G. Weinberg R. Sheng M. J. Neurosci. 1999; 19: 6528-6537Google Scholar, 12Dong H. O'Brien R.J. Fung E.T. Lanahan A.A. Worley P.F. Huganir R.L. Nature. 1997; 386: 279-284Google Scholar), while GRIP PDZ6, also a class II PDZ domain, interacts with the C terminus of ephrin-B1 ligand and EphB2/EphA7 receptor tyrosine kinases (13Bruckner K. Pablo Labrador J. Scheiffele P. Herb A. Seeburg P.H. Klein R. Neuron. 1999; 22: 511-524Google Scholar, 14Lin D. Gish G.D. Songyang Z. Pawson T. J. Biol. Chem. 1999; 274: 3726-3733Google Scholar, 15Torres R. Firestein B.L. Dong H. Staudinger J. Olson E.N. Huganir R.L. Bredt D.S. Gale N.W. Yancopoulos G.D. Neuron. 1998; 21: 1453-1463Google Scholar). GRIP PDZ6 also interacts with the C terminus of the liprin-α family of multidomain proteins (16Wyszynski M. Kim E. Dunah A.W. Passafaro M. Valtschanoff J.G. Serra-Pages C. Streuli M. Weinberg R.J. Sheng M. Neuron. 2002; 34: 39-52Google Scholar), which interact with the leukocyte antigen-related protein family of receptor tyrosine phosphatases (17Serra-Pages C. Kedersha N.L. Fazikas L. Medley Q. Debant A. Streuli M. EMBO J. 1995; 14: 2827-2838Google Scholar, 18Serra-Pages C. Medley Q.G. Tang M. Hart A. Streuli M. J. Biol. Chem. 1998; 273: 15611-15620Google Scholar). Interestingly, the PDZ456 region also reportedly mediates homo- and heteromultimerization of GRIPs (11Dong H. Zhang P. Song I. Petralia R.S. Liao D. Huganir R.L. J. Neurosci. 1999; 19: 6930-6941Google Scholar), suggesting that the PDZ domain is a module mediating multimerization as well as peptide recognition. The first reported crystal structure of a class II PDZ domain (from hCASK) revealed the presence of a second hydrophobic pocket not seen in class I PDZ domain (19Daniels D.L. Cohen A.R. Anderson J.M. Brunger A.T. Nat. Struct. Biol. 1998; 5: 317-325Google Scholar). In addition, this study showed that hCASK PDZ forms a homotetramer by binding to the truncated C-terminal tail of a neighboring PDZ domain that partially mimics the specific peptide ligand, though the self-associated structure does not faithfully represent the actual binding mode of class II PDZ domains due to differences in the amino acid sequences of the true target peptides and the truncated C-terminal tail (19Daniels D.L. Cohen A.R. Anderson J.M. Brunger A.T. Nat. Struct. Biol. 1998; 5: 317-325Google Scholar). Another structure of a class II PDZ domain in InaD revealed its association with the −1 position of the ligand via a distinctive intermolecular disulfide bond (20Kimple M.E. Siderovski D.P. Sondek J. EMBO J. 2001; 20: 4414-4422Google Scholar), which is not a canonical non-covalent peptide-PDZ interaction. To better understand the structural basis of peptide recognition by class II PDZ domains and the mechanisms underlying PDZ-mediated GRIP multimerization, we determined the crystal structures of the GRIP1 PDZ6 domain alone and in complex with a synthetic C-terminal octapeptide of liprin-α1 at resolutions of 1.8 and 1.5 Å, respectively. This is the first description of the crystal structure of a class II PDZ domain non-covalently complexed with its specific peptide ligand, showing an additional role of PDZ domains in the multimerization of PDZ-containing proteins. Recombinant GRIP1 PDZ6 (residues 665–761) from Rattus novergicus with a cleavable glutathione S-transferase tag was expressed in BL21(DE3) Escherichia coli, cleaved, purified, and crystallized as previously described (21Park S.H. Im Y.J. Rho S.H. Lee J.H. Yang S. Kim E. Eom S.H. Acta Cryst. D. 2002; 58: 1063-1065Google Scholar). The C-terminal octapeptide (ATVRTYSC) of the human liprin-α1 used for PDZ-peptide cocrystallization was chemically synthesized. The mutants Y671D and R718D were obtained by site-directed mutagenesis of the plasmid carrying theGRIP-PDZ6 gene using the QuikChange mutagenesis kit from Invitrogen. The expression and purification of the mutants were done by same procedures with the wild type protein. A native data set was collected from a peptide-free frozen crystal to 1.5 Å resolution using an ADSC Quantum 4R CCD detector at beamline X8C in the National Synchrotron Light Source. Since bromine is a convenient anomalous scatterer for MAD phasing, a bromine MAD data set was collected from a single crystal soaked for 30 s in cryoprotection solution containing 1m NaBr (22Dauter Z. Dauter M. Rajashankar K.R. Acta Cryst. D. 2000; 56: 232-237Google Scholar). A data set was collected from a PDZ-peptide complex to 1.8 Å Bragg spacing using an ADSC Quantum 4R CCD detector at beamline BL-18B at Photon Factory, Tsukuba, Japan. The structure of the peptide-free PDZ6 domain was determined by MAD using bromine as an anomalous scatterer. The positions of three bromines in the asymmetric unit were located and refined using the program SOLVE (23Terwilliger T.C. Berendzen J. Acta Cryst. D. 1999; 55: 1872-1877Google Scholar). The initial phases (overall figure of merit, 0.51) were improved by solvent flattening using the program DM (24Collaborative Computational Project, Number 4 Acta Cryst. D. 1994; 50: 760-763Google Scholar). The resultant map was readily interpretable, and model building proceeded using the program O (25Jones T.A. Zou J.-Y. Cowan S.W. Kjeldgaard M. Acta Cryst. A. 1991; 47: 110-119Google Scholar), after which the initial model was refined using the program CNS (26Brunger A.T. Adams P.D. Clore G.M. DeLano W.L. Gros P. Grosse-Kunstleve R.W. Jiang J.-S. Kuszewski J. Nilges N. Pannu N.S. Read R.J. Rice L.M. Simonson T. Warren G.L. Acta Cryst. D. 1998; 54: 905-921Google Scholar). When the crystallographic R value for the model was 28.3%, the coordinates were used in the refinement procedures against the 1.5-Å native data set. The final crystallographic R value for the peptide-free PDZ6 model using data from 15 to 1.5 Å was 25.4% (R free = 27.8%). The structure of the PDZ6-octapeptide complex was determined using standard molecular replacement methods using the structure of peptide-free PDZ6 as a starting model. There were two PDZ6 domains related by NCS in the asymmetric unit. After applying a simulated annealing procedure using data to 1.8 Å, the location of the two bound peptides was determined from a fo − fc difference electron density map. Densities of all eight residues of the peptide were obvious, indicating they were well ordered within the structure. The octapeptide was modeled using the program O, and the peptide-bound PDZ6 domain was refined to a final crystallographic R value of 20.0% and a free R value of 22.2%. The refined model of the peptide-bound PDZ6 domain consisted of a PDZ domain dimer (Ala-668-Gln-753) related by 2-fold NCS, two octapeptides, and 240 water molecules. Because of disorder, the N-terminal three and C-terminal eight residues were not modeled. The stereochemistry of the model was analyzed with PROCHECK (27Laskowski R.A. MacArthur M.W. Moss D.S. Thornton J.M. J. Appl. Crystallogr. 1993; 26: 283-291Google Scholar); no residues were found in the disallowed regions of the Ramachandran plot. Data collection and refinement statistics are summarized in TableI. The coordinates and structure factors of PDZ6 and peptide-PDZ6 complex have been deposited with the Protein Data Bank accession numbers 1N7E and 1N7F, respectively.Table IData collection and refinement statisticsData setPeptide-free by in the solution containing NaBr for 30 = − I figure of = − free with of all from refinement using resolution bond bond by in the solution containing NaBr for 30 R = − I R = − R free with of all from refinement using resolution in a GRIP1 PDZ6 is a globular domain containing eight segments of six that an antiparallel and two α-helices the crystal structure of the all eight amino acid residues of the peptide ligand were well as by the difference electron density map of the peptide and the electron density that the octapeptide was is of most PDZ the ligand was in the the and the to as an additional In addition, the a sequence (residues often to as the GRIP1 PDZ6 binds to the liprin-α C-terminal sequence via a class II hydrophobic PDZ two hydrophobic that accommodate hydrophobic residues at the 0 and −2 positions of the ligand. In the peptide-bound PDZ6 domain forms an antiparallel dimer in an asymmetric unit of the crystallized peptide-free PDZ6 forms a dimer through a crystallographic 2-fold in its was seen in the crystal structures of the hCASK and PDZ domains, the peptide binding pocket of peptide-free PDZ6 is by the C-terminal of a neighboring the recognition of the peptide ligand (19Daniels D.L. Cohen A.R. Anderson J.M. Brunger A.T. Nat. Struct. Biol. 1998; 5: 317-325Google Scholar, S. T. G. G. J. Mol. Biol. 2001; Scholar). of the crystal structures of the peptide-free and peptide-bound PDZ6 domains shows a slight in the which the peptide binding pocket and of the side chain of tyrosine from the ligand. The free and peptide-bound structures of the PDZ6 domain showed an root-mean square deviation of Å the six conserved were used for the an of Å for the residues positions helix showed the structural Because the amino acid sequence of the GRIP1 PDZ6 domain used in this study is to human GRIP1 PDZ6, we used a synthetic octapeptide (ATVRTYSC) that mimics the C terminus of human liprin-α1 as a ligand. The residues of liprin-α are and the residues at the 0 and −2 positions are well conserved as with from other PDZ6 binding receptor The C terminus of the ligand binds as an additional to the of the PDZ domain and makes hydrophobic with helix αB, while the peptide of the C-terminal four residues is to by in the sequences of PDZ domains, their are conserved in of structure and the to the ligand The C-terminal of octapeptide with the of and in the and was within of two water molecules The amino acid residues at ligand positions 0 and −2 are to the specific recognition by the PDZ The side chain of is at the of the first hydrophobic which is of and and is of to accommodate various hydrophobic side from of to not The second hydrophobic which the Tyr residue at ligand position is of from the and Ile-736 and from helix the structures of previously described class I and class II PDZ domains, the residue at ligand position −2 usually binds to the side chain of a residue located at the N-terminal of helix of the PDZ (7Bezprozvanny I. Maximov A. FEBS Lett. 2001; 509: 457-462Google Scholar, S. T. J. Biol. Chem. 2001; 276: Scholar). In class I PDZ domains, or at the −2 position forms with the side chain of a residue at the of the In class II PDZ domains, the residue at the αB1 position is conserved and to for the specificity of ligand In hCASK for at αB1 the residue at ligand position −2 (19Daniels D.L. Cohen A.R. Anderson J.M. Brunger A.T. Nat. Struct. Biol. 1998; 5: 317-325Google Scholar). Another class II PDZ domain, the crystal structure of showed that the and of at αB1 in the hydrophobic with at the ligand −2 position (20Kimple M.E. Siderovski D.P. Sondek J. EMBO J. 2001; 20: 4414-4422Google Scholar). the structure of the GRIP1 by the side chain of not a direct contact with the conserved Leu-732 at the Tyr aromatic was the first hydrophobic and with the side chain of Ile-736 located at second of the helix Leu-732 to to the ligand binding by the hydrophobic accommodate a hydrophobic residue at ligand position This a mode of peptide recognition in which a hydrophobic residue at the αB5 position the role of the conserved hydrophobic residue at the αB1 position in class II PDZ Interestingly, of peptide-free and peptide-bound structures showed that a slight reorientation of the helix to accommodate the side chain of The of the structures showed helix to have the the structure in the peptide-free and peptide-bound which is located in the to in the C-terminal of helix αB, the peptide The resultant in the C-terminal of helix the second hydrophobic suggesting that a role in the conformational adaptation of helix by to the In addition, a of Å in the of at which interacts with ligand due to the bulkiness of the Tyr side chain this the four peptide positions of the ligand in the self-associated structure and in the PDZ-peptide complex were in there was a difference of Å in the of which interacts with due to the difference of the bulkiness of the side chain at that position the of helix is determined by the difference in the of the residue at ligand position This structural may the of GRIP1 PDZ6 to bind various target peptides with hydrophobic amino at the −2 position the structure of the ligand residues at the −1 and positions also to to its recognition. with in the in of the while with in the other The residues at the position of PDZ6 are conserved and all contain direct to the specificity and for the GRIP1 PDZ6 The side chain of at ligand position also in with at the terminus of the there is no conserved of interaction at the and positions PDZ domains, the specificity of recognition. PDZ multimerization is a for a of PDZ domain proteins. multimerization of which is mediated by its or domain, does not the binding of target proteins A. Li C. J. Cell Biol. 1998; Scholar) does of which is mediated by its two PDZ domains R.A. 2001; Scholar). these no structural for the of multimerization or does not ligand the structures of InaD or and the crystal structures of the domains no S. T. G. G. J. Mol. Biol. 2001; Scholar, S. T. J. Biol. Chem. 2001; 276: Scholar). proteins reportedly homo- or heteromultimers through their PDZ456 domains (11Dong H. Zhang P. Song I. Petralia R.S. Liao D. Huganir R.L. J. Neurosci. 1999; 19: 6930-6941Google Scholar). is not these three PDZ domains from and that GRIP1 PDZ6 forms a dimer free or which that GRIP1 PDZ6, the to dimer with that the crystal structure of peptide-bound GRIP1 PDZ6 showed formation of a PDZ6 dimer related by a 2-fold in the asymmetric unit 4 The interface the two PDZ domains a and an from each the the of the 2-fold with the and C termini of each PDZ domain in The peptide binding were located at the distal of the in which binding of target This interaction was by six the two and hydrophobic in the The of dimer formation is or of the of each target binding by PDZ multimers was also with InaD and proteins using A. Li C. J. Cell Biol. 1998; Scholar, R.A. 2001; Scholar). In the peptide-free crystal structure with the there was in the asymmetric unit. the same interaction described was via a crystallographic 2-fold is that the dimer in the solution into a crystallographic in a crystallographic shows difference of the for all the two in the asymmetric unit of a peptide-bound Another intermolecular interaction with a in the peptide-free crystal structure is the association through C-terminal into the ligand binding two In this C termini as for neighboring PDZ which is of the crystal structures of and hCASK PDZ domains (19Daniels D.L. Cohen A.R. Anderson J.M. Brunger A.T. Nat. Struct. Biol. 1998; 5: 317-325Google Scholar, S. T. G. G. J. Mol. Biol. 2001; Scholar). the C-terminal amino acid sequence of the PDZ6 from the class II PDZ target peptides The of the C-terminal four residues of the neighboring located in the shows difference from the location C-terminal octapeptide of liprin-α1 complexed with the hydrophobic of GRIP1 PDZ6 hydrophobic residues or the residues that a hydrophobic the in the peptide-free crystal structure is not interaction molecular interaction within the crystal with this a PDZ6 in which the seven residues (residues were a dimer in solution by not To the presence of PDZ6 dimer in which is in 4 we the residues at the interface and the molecular of the mutants and wild type PDZ domains in solution using The Y671D which was to the hydrophobic interaction in the was as a while the wild type PDZ6 was a dimer in solution 4 that is located at the interface and an role in hydrophobic interaction of the the of by the R718D the residue in of the domains in the not the of the PDZ with the of peptide-bound PDZ6 in the and in solution that the structure of PDZ6 in A is of PDZ6 in that the GRIP PDZ6 domain a role in heteromultimerization of a of amino acid sequence with and the two in the region containing the PDZ456 domains and in the PDZ6 the residues in the region of the interface and are for conserved in GRIP1 and in 4 these that GRIP PDZ6 the same interface for homo- and In we a mode of peptide recognition by the class II GRIP PDZ6 Ile-736 at the αB5 position was in specific recognition of the ligand, and the conformational adaptation of the helix by ligand binding the hydrophobic at ligand position In addition, the PDZ6 structure described in this study is the first of a role of PDZ domains in of antiparallel PDZ6 domain is mediated by an interface located at a site from the peptide-binding in target binding by the PDZ This enable efficient clustering of various target proteins though multimerization of GRIP proteins. N. and M. and N. for in data collection at beamline BL-18B of Photon Factory, Tsukuba, Japan. J. Berendzen and L. for data collection at beamline X8C of National Synchrotron Light at National also H. S. Lee and G. H. Kim at the of