Self-Directed Assembly and Clustering of the Cytoplasmic Domains of Inwardly Rectifying Kir2.1 Potassium Channels on Association with PSD-95

Biochimica et Biophysica Acta 1808 (2011) 2374–2389

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

journal homepage: www.elsevier.com/locate/bbamem

Self-directed assembly and clustering of the cytoplasmic domains of inwardly rectifying Kir2.1 potassium channels on association with PSD-95

Svetlana Fomina a,1, Tina D. Howard a,1, Olivia K. Sleator a, Marina Golovanova a, Liam O'Ryan a, Mark L. Leyland b, J. Günter Grossmann c,2, Richard F. Collins a, Stephen M. Prince a,⁎ a Faculty of Life Sciences, University of Manchester, Manchester Interdisciplinary Biocentre, 131 Princess Street, Manchester M1 7DN, UK b Department of Biochemistry, University of Leicester, Henry Wellcome Building, Lancaster Road, Leicester LE1 9HN, UK c Science and Technologies Facilities Council Daresbury Laboratory, Daresbury, Warrington, Cheshire WA4 4AD, UK article info abstract

Article history: The interaction of the extra-membranous domain of tetrameric inwardly rectifying Kir2.1 ion channels Received 9 March 2011 (Kir2.1NC4) with the membrane associated guanylate kinase protein PSD-95 has been studied using Received in revised form 22 June 2011 Transmission Electron Microscopy in negative stain. Three types of complexes were observed in electron Accepted 28 June 2011 micrographs corresponding to a 1:1 complex, a large self-enclosed tetrad complex and extended chains of Available online 5 July 2011 linked channel domains. Using models derived from small angle X-ray scattering experiments in which high

Keywords: resolution structures from X-ray crystallographic and Nuclear Magnetic Resonance studies are positioned, the Potassium channel envelopes from single particle analysis can be resolved as a Kir2.1NC4:PSD-95 complex and a tetrad of this MAGUK unit (Kir2.1NC4:PSD-95)4. The tetrad complex shows the close association of the Kir2.1 cytoplasmic domains Cluster and the inﬂuence of PSD-95 mediated self-assembly on the clustering of these channels. Electron microscopy © 2011 Elsevier B.V. All rights reserved. Small angle X-ray scattering

1. Introduction the intermolecular interactions mediated by scaffold proteins are beyond the resolution of the techniques deployed so far, however Potassium ion channels form groups or clusters in the membrane tomographic techniques allied with immunolabelling have begun to via their interaction with scaffold proteins [1]. The role of scaffold resolve the location of speciﬁc scaffold proteins within cellular proteins is to bring together membrane channels and receptors via fractions [9,10]. bridging interactions mediated by multiple domains within the PSD-95 is a multi-domain scaffold protein with an apparent scaffold protein [2]. Through such interactions inwardly rectifying molecular weight of 95 kDa enriched in the post-synaptic density (Kir) and voltage gated (Kv) Potassium channels have both been (PSD) of neuronal cells [11]. PSD-95 is anchored to the membrane by shown individually to interact or cluster [3–8]. The precise details of palmitoylation at two cysteine residues at the N-terminus of the protein [12]. PSD-95 is the canonical member of a family of proteins termed membrane associated guanylate kinases (MAGUKs) and Abbreviations: Kir, inwardly rectifying potassium channel; Kir2.1, inwardly rectifying ﬁ potassium channel subtype 2.1; Kir2.1NC, (fused amino- and carboxy-terminal) contains the core set of domains which de nes this family, namely cytoplasmic domains of Kir2.1; Kir2.1NC4, homotetramer of Kir2.1NC; Kir2.14,homo- three PSD-95/Drosophila discs large tumor suppressor DlgA/Zona tetramer of Kir2.1; Kv, voltage gated potassium channel; Kv1.2, voltage gated potassium occludens ZO-1 (PDZ) domains, an Src Homology 3 (SH3) domain and β β channel subtype 1.2; Kv1.24, homotetramer of Kv1.2; Kv , potassium channel -subunit; a guanylate kinase (GK) like domain [11]. The MAGUK family proteins Kvβ , homotetramer of Kvβ; MAGUK, membrane associated guanylate kinase; PSD, post- 4 are homologues of the disks large protein and include PSD-93, SAP-97 synaptic density; PSD-95, 95 kDa protein of the post-synaptic density; PDZ, PSD-95/ Drosophila discs large tumor suppressor DlgA/Zona occludens ZO-1; SH3, Src homology 3; and SAP-102. PSD-95 has been found to be important in synaptic GK, guanylate kinase; SAXS, small angle X-ray scattering; EM, electron microscopy; nsTEM, plasticity, PSD-95 expression levels have been implicated in the Transmission Electron Microscopy in negative stain; LCMS, liquid chromatography mass distinction between excitatory and inhibitory synapses [13] and spectrometry; SEC, size exclusion chromatography; SRS, Synchrotron Radiation Source; reported to be altered in schizophrenic brains compared to controls CTF, contrast transfer function; FSC, Fourier shell correlation; SDS-PAGE, sodium- [14]. PSD-95 gene disruption has also been implicated to correlate dodecylsulphate polyacrylamide gel-electrophoresis; E. coli, Escherichia coli;Rg,radiusof gyration with autism spectrum disorders and Williams syndrome in studies ⁎ Corresponding author. Tel.: +44 161 306 8919; fax: +44 161 306 5201. using mouse models [15]. E-mail address: [email protected] (S.M. Prince). PDZ domains consist of ~90 amino acids and class I PDZ domains 1 These researchers made equal contributions to the work reported. (those present in PSD-95) interact with C-terminal target sequences 2 Present address: Molecular Biophysics Group, Institute of Integrative Biology, Faculty of Health and Life Sciences, University of Liverpool, Crown Street, Liverpool L69 of the form -xS/TxI/V [2,16] where x denotes any amino acid. The ~75 7ZB. amino acid SH3 and ~175 amino acid GK domains of PSD-95 have

0005-2736/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.bbamem.2011.06.021 S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 2375 been shown to be fused in an association that leaves the SH3 binding respectively. The GST-PSD-95 fusion protein was expressed in E. coli cleft available for interactions [17,18]. The SH3 domain binds the BL21(DE3)pLysS cells, in Luria Broth (400 ml cultures in 2 l flasks at target sequence PxxP [19], PSD-95 itself contains 2 such motifs in the 37 °C with 225 rpm shaking). Expression was induced by the addition linker region between PDZ2 and PDZ3. PSD-95 has been shown to of IPTG to a final concentration of 1 mM when the culture Optical interact both with inwardly rectifying potassium channels (Kir) and Density (OD) at 600 nm was measured in the range of 0.6–1cm−1. voltage gated potassium channels (Kv) which possess the appropriate Cells were harvested by centrifugation 2.5 h after induction and consensus sequence at their C-termini [5,20]. stored frozen at −70 °C until required. Kir channels [21] are either homo- or hetero-tetrameric oligomers. Cells were resuspended in phosphate buffered saline, 0.5 M NaCl, The strong inwardly rectifying channel Kir2.1 is found in a variety of 10 mM dithiothreitol, pH 7.5, protease inhibitor cocktail (Complete tissues [22–24] and the formation of heteromeric channels with the Protease Inhibitor Cocktail, 1 tablet per 50 ml solution: Roche homologous Kir2.3 protein has been proposed. Recent studies of Diagnostics, Charles Avenue, Burgess Hill, West Sussex, RH15 9RY, Kir2.1/2 domains expressed as concatamers in HEK293 cells and UK.), and DNAse1 100 mg/l. Cells were disrupted using a French Press, characterized by electrophysiological measurements have demonstrated insoluble material was removed by high-speed centrifugation functional heteromers of these channels [25]. Gating of Kir channels has (14,500 rpm, Sorvall SS34 rotor, 15 min, 4 °C). The supernatant was been shown to involve the binding of divalent cations and polyamines to applied to a GSTPrep FF 16/10 column and eluted by the application of the cytoplasmic channel domain [26,27]. Each monomer subunit of buffer containing 10 mM reduced glutathione. The eluant was Kir2.1 has a C-terminal sequence containing both SH3 and PDZ desalted into a buffer of 20 mM Tris/HCl, 5 mM DTT, 1 mM EDTA, consensus binding motifs interspersed with a double arginine motif pH 7.5 using a HiPrep 26/10 column. The glutathione-S transferase (−svPlePrplRReSeI). Kir2.1 is known to play a role in the establishment (GST) tag was removed by incubation with PreScission protease (GE of membrane resting potential [28] and mutations in Kir2.1 are Healthcare) at 10 units/mg for 2 h at 4 °C. The protein solution was responsible for Andersen syndrome [29–34]. then filtered, concentrated and applied to a 5 ml Resource Q column. Atomic models of the tetrameric chicken Kir2.2, a prokaryotic Purified PSD-95 was eluted by the application of a 0–1 M NaCl homologue (Kirbac1.1), and the extra-membranous domains of gradient over 20 column volumes. The combined fractions containing (murine) Kir2.1 are available from protein crystallography purified PSD-95 were concentrated and applied to a SEC column [26,35,36]. Structures of each of the PSD-95 interaction domains (Superdex-200 GL 10/300). Fractions under the single resolved peak have been solved using X-ray crystallography or solution Nuclear corresponding to PSD-95 (elution volume ~13.5 ml) were combined. Magnetic Resonance spectroscopy [17,37–40]. A number of models for EM grids were prepared from a sample of PSD-95 at a concentration of the interaction of Kir channels with MAGUK proteins have been 0.4 mg/ml (5 μM) as described below. For the light-scattering analysis proposed [38,41,42]. PSD-95 has been shown to interact with Kir purified PSD-95 was fractionated on a Superdex-200 GL 10/300 column channels in whole neuronal cell extracts [43] in contrast to (P2) and subsequently passed through a modified Wyatt EOS 18-angle laser synaptic fractions where the association of PSD-95 and Kir is not photometer fitted with a Wyatt quasi-elastic light scattering detector for observed [44]: this suggests that MAGUK/Kir complexes are present in the measurement of hydrodynamic radius. The resulting peak was neurones but not in the synapse. analyzed using ASTRA 4.90.08 software (Figs. A.1(a,b)). In this paper we describe models for PSD-95 derived from small angle X-ray scattering (SAXS) and confirmed by Transmission 2.2. Expression and purification of Kir2.1NC4 Electron Microscopy in negative stain (nsTEM); low-resolution structures of the tetrameric cytoplasmic domain of Kir2.1 Design of the Kir2.1NC expression construct — a homology model

(Kir2.1NC4) determined from both SAXS and nsTEM; a Kir2.1NC4: of Kir2.14 was constructed from the tetrameric structures of the extra- PSD-95 complex from nsTEM; and a (Kir2.1NC4:PSD-95)4 complex membranous domains of Kir3.1 [45] and full-length Kirbac1.1 [35]. from nsTEM. Using these data in combination, the high-resolution Brieﬂy, the amino acid sequences corresponding to both structures domain structures for the components of the complexes are were aligned with that of murine Kir2.1 using CLUSTALW [46];a positioned leading to a model for the clustering of Kir2.1 channels chimeric model was made by positioning the Kir3.1 extramembra- by PSD-95. nous domains through a least-squares alignment of residues of high identity; 194–233 and 254–331 of Kir3.1 to 156–195 and 212–289 of 2. Materials and methods Kirbac1.1 respectively (using the program LSQKAB [47,48]); the sequence of Kir2.1 was threaded onto this framework, residues All chemicals were obtained from Sigma-Aldrich (Fancy Road, without a structural representative from either model were omitted Poole, Dorset BH12 4QH, UK) unless stated. Escherichia coli bacterial and sidechain clashes were manually resolved. The resulting model strains were obtained from Novagen/Merck Chemicals (Boulevard was subjected to stereochemical minimization (REFMAC5 [49]) with

Industrial Park, Padge Road, Beeston, Nottingham, NG9 2JR, UK). All the imposition of C4 molecular symmetry. Examination of the resulting columns were purchased from GE Healthcare (The Grove Centre, model showed that residues 67 and 189 were in close proximity. White Lion Road, Amersham, Bucks, HP7 9LL, UK); chromatography A pET15b plasmid representative of residues 1–67 and 189–427 of was undertaken using a GE Healthcare AKTA Fast Protein Liquid Kir2.1 (UniProt ID: P35561) connected by two glycine residues was Chromatography (FPLC) rig. Murine Kir2.1 cDNA was a kind gift of Ian constructed. Briefly, the DNA sequence representative of the N-terminus Ashmole of Warwick University. of Kir2.1 plus the linker abutted by NdeIandBstXI restriction endonuclease sites was amplified by PCR and subsequent ligation into 2.1. Expression and purification of PSD-95 an pGEM-T Easy vector (Promega, Delta House, Enterprise Rd. Chilworth ResearchCentre,SO167NS,UK);similarlytheDNAsequenceforthe An expression plasmid for PSD-95 was made by cloning full-length C-terminus of Kir2.1 was prepared with BstXI and BamHI restriction rat PSD-95 into a pGEX-6P plasmid between BamHI and XhoI sites; a 3-way ligation was used to join the restricted and purified restriction endonuclease sites. The translated PSD-95 protein from components. The resulting plasmid was analyzed by DNA sequencing to this plasmid differs from the published sequence (UniProt ID: confirm the orientation and correct insertion of the gene. The P31016) as follows T9→A, E23→K, E51-Q53 omitted, S216→N, N-terminally tagged protein was expressed overnight in E. coli BL21 Q594→R. The first two point mutations and the 3 residue excision are (DE3)pLysS at 37 °C using Studier autoinduction media [50] (400 ml in the (flexible) N-terminal tether; the remaining two point cultures in 2 l flasks with 225 rpm shaking). Cells were harvested by mutations are to surface residues in the PDZ2 and GK domain centrifugation and stored frozen at −70 °C until required. 2376 S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389

E. coli cells were resuspended in 20 mM phosphate buffer, pH 8.0, smaller domain of the pair in all models was computed. This distance 0.5 M NaCl, 20 mM imidazole, 1 mM reduced GSH, 50 mM L—glutamic was used as a restraint on the maximum displacement between any acid, 50 mM L—arginine [51], protease inhibitor cocktail (Roche Cα coordinate of the two domains. The SH3 and GK domains of PSD- Complete) and DNAse1 100 mg/l and lysed by French press. Insoluble 95 were treated separately for the purpose of restraint calculation, material was removed by high-speed centrifugation and the super- although the fused domain was submitted for refinement. As the natant was applied to a 5 ml HisTrap nickel affinity column. Kir2.1NC individual PDZ domains of PSD-95 would be indistinguishable at low was eluted by the application of 150 mM of imidazole and was resolution the final restraint set was expressed in a manner to account effectively pure. The eluant was desalted into a buffer of 20 mM Tris/ for this. A series of 20 refinements were carried out using these HCl pH 7.5, 150 mM NaCl, 1 mM reduced GSH, 1 mM EDTA, 50 mM L— restraints and the program BUNCH [58]. The final X (qb0.5 Å−1) glutamic acid, 50 mM L—arginine, using a HiPrep 26/10 column. The values for these refinements were in the range of 3.8–3.3 (Fig. 1(a,b)). hexahistine affinity tag was removed by incubation with thrombin Both SASREF and BUNCH gave models consistent with the initial (10 units/mg) for a minimum of 2 h at 4 °C. The protein was ab-initio envelopes. The program EOM [59] was employed to assess subsequently concentrated and applied to a SEC column (Superdex- the level of compactness of the PSD-95 domain based model; random 200 GL 10/300). Fractions under the peak corresponding to the coil models for the non-domain sequences of PSD-95 gave marginally −1 Kir2.1NC4 tetramer (retention volume 11.5 ml, Fig. A.1(c)) were better results (Χ=3.5, qb0.5 Å ) compared to native-like chain combined. Preparations showing significant aggregate peaks below models (Χ=3.6). The best ensemble, selected from a pool of 10,000 this retention volume were discarded. EM grids were prepared from a models covering all accessible space, had a median Rg of 36.6 Å 20 mg/ml sample of Kir2.1NC4 with a 10 fold dilution (final concentra- compared to the pool median of 49.5 Å (Fig. 1(c)). tion 14 μM) prior to gridding as described below (Section 2.5). For the 4.25 m camera length the concentration of Kir2.1NC4 was 0.5 mg/ml (3.6 μM) for the 1 m camera a concentration of 4.8 mg/ml

2.3. SAXS analysis of PSD-95 and Kir2.1NC4 (34 μM) was used. The Rg for Kir2.1NC4 was computed as 45.3 Å. Initial ab-initio envelopes for Kir2.1NC4 were computed using the program SAXS data were collected at Daresbury Laboratory on the Synchro- GASBOR [54] with the condition of C4 symmetry; the agreement with − tron Radiation Source station 2.1. Samples were injected between mica the experimental scattering curve I[0.015bqb0.8055 Å 1] in the range windows into a brass sample cell, the cell was maintained at 4 °C of 1.8–2.0. These envelopes showed a squat structure comprising a core throughout data collection. Multiple exposures of 60 s were collected domain with extended regions. The GASBOR models were aligned and using the multiwire gas detector using a wavelength of 1.54 Å at two averaged (DAMAVER/SUPCOMB) and the composite model was refined camera lengths. using DAMMIN with the condition of C4 symmetry and an oblate Data for the 4.25 m camera length were collected using a PSD-95 starting model to give a final ab initio model for Kir2.1NC4 (Χ=1.7) concentration of 1 mg/ml (13 μM assessed by solution UV absorption (Fig. 1(d,e)). Subsequently a series of 10 refinements were carried out at 280 nm), data for the 1.0 m camera length were collected at a using the individual domain of Kir2.1NC4 obtained from the PDB ID: concentration of 8 mg/ml (100 μM), in both cases data on buffer 1U4F and C4 symmetry using the program BUNCH. Maximum samples were collected for later subtraction. Successive exposures separation restraints of 5 and 7 Å were applied at the base of the were examined for signs of radiation damage (sample aggregation molecule between the adjacent and opposite Cα coordinates forming etc.). Data were processed using the programs OTOKO [52] and GNOM the cytoplasmic pore with an additional restraint of 9 Å at the −1 [53]. The radius of gyration (Rg) for PSD-95 was computed from the cytoplasmic face between adjacent domains. The X (qb0.5 Å ) values Guinier plot as 43.6 Å.16ab-initio envelopes for PSD-95 were for these refinements were in the range of 3.6–2.6. BUNCH refinements computed using the program GASBOR [54] and these showed a gave models consistent with the initial ab-initio envelopes (Fig. 1(d,e)). comma shaped envelope for PSD-95 and agreement (Χ=√Χ2, where 2 2 nΧ =[[δI−σI]/I] and I is the scattering intensity, δI is the difference 2.4. Preparation of complexes of Kir2.1NC4 and PSD-95 between observed and calculated I, σI is the estimated experimental error in I and n is the number of observations) with the experimental For the 1:1 complex Kir2.1NC4 was prepared as described (Section scattering curve (intensity I versus momentum transfer q: I 2.2). The GST-PSD-95 fusion construct was expressed and affinity −1 [0.007bq=2πsb0.766] where s is the nominal Bragg resolution purified as described (Section 2.1). 200 μL of Kir2.1NC4 at 1.4 mg/ml expressed in Å) in the range of 1.6−1.7. The GASBOR models were (10 μM) was added to 2 ml of GST-PSD-95 at 0.11 mg/ml (1 μM) to aligned and averaged (DAMAVER [55]/SUPCOMB [56])and give equimolar concentrations of GST-PSD-95 and Kir2.1NC4. PreScis- the composite model was refined using DAMMIN [57] to give a final sion protease was added to the resulting solution and this was ab-initio model for PSD-95 (Χ=1.6) (Fig. 1(a,b)). Subsequently, incubated for 2 h at 4 °C. The solution was concentrated and individual domains of PSD-95 obtained from the PDB IDs: 1IU0, 1QLC, subsequently fractionated on a Superose-6 GL 10/300 column. EM 1TP5 and 1KJW were refined against the data using the program grids were prepared from fractions under a peak eluting at 11.5 ml. SASREF [58]. 10 models (qb0.5 Å−1; X=5.0–4.9) from this process Individual micrographs imaged from grids of this sample showed 1:1 3 were analyzed in order to obtain a set of loose inter-domain restraints. complexes of Kir2.1NC4:PSD-95 (nN10 ) and a minority of Kir2.1NC4 Briefly: the distances between Cα coordinates of pairs of domains in particles (n≈102, resolved analytically during EM refinement). each model were computed using the program PDBCUR [47]; a mean For higher order Kir2.1NC4/PSD-95 complexes, purified Kir2.1NC4 relative displacement enclosing on average ~50% of residues of the was prepared as described (Section 2.2). The GST-PSD-95 fusion

Fig. 1. SAXS derived models of PSD-95 and Kir2.1NC4: (a) Agreement of SAXS models of PSD-95 with the experimental scattering data. Raw data are represented as triangles ln(I)+/−σI/I where I is the reduced scattering intensity and σI is the error estimate. The DAMMIN model agreement is shown as a dotted line in red. The agreement of the best BUNCH model is shown as a solid line. (b) Ab initio DAMMIN model of PSD-95, shown as gray spheres (3.25 Å radius), with the superimposition of 10 BUNCH domain models of PSD-95 (each colored blue–red according to sequence) the crystal structure of the fused SH3GK domain [17] shows that the C-terminus of PSD-95 emerges from the GK domain to associate with SH3, this element of the structure is partially blocked by the PDZ3 domain in the view shown. Generated using CHIMERA [61]. (c) A histogram showing the distributions of radii of gyration from domain-based models generated by the program EOM (Rg frequency is expressed as a fraction of the total number of models). The histogram from the randomly generated pool of structures is shown with boxed points and a dotted line. The histogram of the best representative domain structures is shown as diamond/solid line. (d) Agreement of SAXS modelsofKir2.1NC4 with the experimental scattering data. Representation as for (a). (e) Ab initio DAMMIN model of Kir2.1NC4 shown as gray spheres (3.25 Å radius), and superimposition of 10 BUNCH (domain) models of Kir2.1NC4 (shown as black Cα trace models). Generated in PyMOL (DeLano Scientiﬁc). S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 2377

construct was expressed as described above (Section 2.1). The lysate at the affinity stage 5 mg (62 nmol) of GST-PSD-95 was estimated to from a 1.6 l culture volume of E. coli expressing GST-PSD-95 was be bound to the column. A total of 28 mg (0.2 μmol) of purified applied to a GST affinity column (Amersham GSTPrep FF 16/10) Kir2.1NC4 was passed slowly over the column giving an exposure of followed by a wash step. Based upon the average yield of GST-PSD-95 GST-PSD-95 to Kir2.1NC4 exceeding a 3-fold molar excess. After 2378 S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389

washing 100 units of PreScission protease was added to the column 2.8. Data processing of Kir2.1NC4:PSD-95 micrographs and the column was sealed and incubated for 1 h at room temperature. The material liberated by the protease digestion was 15 micrographs imaged in a Philips Tecnai 10 electron microscope washed from the column, concentrated and fractionated on a within a defocus range of 1.75 (Fig. A.4(a))–0.6 μm and recorded on Superdex-200 GL 10/300 column (Fig. A.1(e)). EM grids were ﬁlm were digitized with a ﬁnal pixel resolution of 3.67 Å/pixel. The prepared from 200 μl fractions close to the void volume of the column micrographs appeared homogeneous by eye and an automated

(8 ml) and well before the elution volume of Kir2.1NC4 tetramers particle picking routine implemented in the program BOXER was (11.5 ml). Individual micrographs imaged from grids of this sample used to pick 26,947 particles in 64×64 pixel boxes (later reduced to 3 showed a majority (n~10 ) of large tetrad complexes a small number 48×48 during processing, Fig. A.4(a,b)). An initial reference free C4 (n~101) of long chain aggregate particles and a minority (n~102)of symmetric envelope was constructed using the startcysm routine in particles of the size of Kir2.1NC4:PSD-95 1:1 complexes (therefore EMAN. This envelope was refined against the particle set filtered to presumed to have dissociated from larger complexes during EM the resolution of the first zero of the CTF of the furthest from focus gridding). micrograph. The envelope differed from the Kir2.1NC4 alone envelope having continuous lobes of density above the main body of the 2.5. Preparation of grids for electron microscopy particle. As the fractionation procedure employed to isolate the complex would not be able to resolve the Kir2.1NC4:PSD-95 complex fi Samples were adsorbed onto freshly glow discharged carbon film grids and Kir2.1NC4 alone, the multire ne routine in EMAN was employed (400 mesh Copper, Agar Scientific, Unit 7 M11 Business Link, Parsonage to assign two particle sets from the data (Fig. A.4(c)). The starting Lane, Stansted, Essex, CM24 8GF, UK) as follows: 1 μlofsamplesolution envelopes were the initial envelope from the Kir2.1NC4:PSD-95 fi was pipetted into the grid followed by blotting; 1 μl of de-ionized water micrographs here and the starting envelope from the lm data for fi was then applied for 10 s followed by blotting; 1 μlof2%w/vuranyl Kir2.1NC4, particles ltered at the 1st zero of the CTF were used. acetate solution was applied followed by a final blotting step. 16,440 particles were assigned to the Kir2.1NC4:PSD-95 envelope and this data set was further refined relaxing the C4 symmetry to C1 symmetry. In order to obtain a scattering curve for CTF correction the 2.6. Data processing of PSD-95 micrographs best BUNCH models of Kir2.1NC4 and PSD-95 were docked into this resulting envelope using CHIMERA [61]. A scattering curve for the Two micrographs imaged in a Philips Tecnai 10 electron complex was calculated using the program CRYSOL [62] and this curve μ microscope with defocus values of 2.25 (Fig. A.2(a)) and 0.9 m was used to correct the raw particles for the CTF. The multirefine step fi fi and recorded on lm were digitized with a nal pixel resolution of was repeated without CTF correction. The final refinement used this 3.67 Å/pixel. The micrograph appeared homogeneous by eye, particle set, C1 symmetry and a full CTF correction, and proceeded although the particle density was high. An envelope for particle from a reference-free starting envelope (Fig. A.4(d–h)). picking was obtained from the DAMMIN model of PSD-95 using the EMAN [60] routine pdb2mrc and a resolution of 25 Å. This envelope was used with the model based particle picking routine in the EMAN 2.9. Data processing of micrographs containing tetrads and long chain routine BOXER to pick 7854 particles in 48×48 pixel boxes (Fig. A.2(a,b)). aggregates The initial model-based envelope was subsequently refined against the fi particle set ltered at the 1st zero of the contrast transfer function (CTF). These micrographs contained large (tetrad) particles with apparent C fi fi 4 This re nement was followed by further re nement on CTF corrected symmetry along with smaller particles of the approximate dimensions of fi particles using the SAXS pro le of PSD-95 and implemented in the CTFIT Kir2.1NC tetramers and a minority of long chains of particles (Fig. A.5(a), – 4 routine in EMAN (Fig. A.2(c g)). 6(d)).Aseriesof17setsof3micrographseachrecordedfrom17separate fields at tilts of −45°, 0° and +45° were imaged in a Philips Tecnai 10

2.7. Data processing of Kir2.1NC4 micrographs electron microscope. Film micrographs were digitized with a final pixel resolution of 3.67 Å/pixel. The defocus range of the 0° micrographs in the A single micrograph imaged in a Philips Tecnai 10 electron set was 1.6 (Fig. A.5(a))–0.6 μm. Tetrad particles only were picked from a microscope with a defocus value of 0.75 μm and recorded on film was matched pair of −45° and +45° micrographs and a tomographic digitized with a final pixel resolution of 3.67 Å/pixel. The micrograph reconstruction was made from these particles using EMAN (Steven appeared homogeneous by eye and the automated particle picking Ludtke, personal communication). The tomographic reconstruction was routine implemented in the EMAN routine BOXER was used to pick 6469 representative of the gross dimensions of the tetrad particles (Fig. A.5(b)) particles in 48×48 pixel boxes. An initial reference free C4 symmetric and was used in model-based particle picking from all of the 0° tilt envelope was constructed using the startcsym routine in EMAN. This micrographs mentioned above, plus 14 additional micrographs recorded envelope was refined against the particle set filtered at the 1st zero of at 0° tilt under similar conditions (Fig. A.5(a,c)). The resulting set of the CTF. A set of 22 micrographs were subsequently collected on the particles was examined and badly miscentered particles were deleted same grid using a Jeol 1220 microscope equipped with a Gatan Orius manually. Initial rounds of refinement were performed on automatically CCD camera. These micrographs were recorded within the defocus centered particles (implemented in the cenalignint routine in EMAN) in range 1.25 (Fig. A.3(a))–0.6 μm and the final imaging resolution was which 4195 particles were omitted. The centered particles were filtered to 2.93 Å/pixel. The envelope obtained from the earlier data set was used in the resolution of the first zero of the furthest from focus micrograph and a model-based particle picking routine in BOXER to obtain 49,012 refined using the tomographic reconstruction as the starting envelope, particles in 64×64 pixel boxes (Fig. A.3(a,b)). These were split into two with C4 symmetry now imposed (Fig. A.5(d)). In order to obtain a groups of 26,697 and 22,315 and refined independently from the scattering curve for CTF correction the Kir2.1NC4:PSD-95 model obtained starting reference free model obtained from the film data set. Initial earlier was docked into ¼ of the resulting envelope using CHIMERA with a refinement with particles filtered to the resolution of the first zero of the model for the complete tetrad (Kir2.1NC4:PSD-95)4 being generated by CTF of the furthest from focus micrograph was undertaken. This was the application of C4 symmetry using the program PDBSET [47].A followed by further refinement on CTF corrected particles using the scattering curve for the tetrad complex was calculated using the program

SAXS profile of Kir2.1NC4 and implemented in the CTFIT program in CRYSOLandthiscurvewasusedtocorrecttherawparticlesfortheCTF. EMAN. All 49,012 particles were used in the final rounds of refinement The final refinement used all 14,477 uncentered CTF corrected particles

(Fig. A.3(c–g)). and C4 symmetry (Fig. A.5(e–i)). S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 2379

Fig. 2. Multi-particle averaging of nsTEM envelopes: (a) Top: construction/selection of envelopes representing the 1:1 complex after individual refinements; envelopes for PSD-95 shown in magenta; Kir2.1NC4 (cyan); 1:1 complex (yellow) and ¼ of the tetrad (blue). Middle: the averaged 1:1 complex. Bottom: envelopes consistent with the averaged 1:1 complex for the next round of refinement, colors as for the top representation. (b) FSC curves are shown for the comparison of envelopes at each stage of refinement averaging: The dotted gray line shows the comparison of the independent 1:1 complex envelope with the first averaged envelope, the 5 curves colored red–blue compare successive steps of averaged 1:1 complex envelopes. Convergence is reached after the 5th cycle.

2.10. Docking of SAXS models into nsTEM maps the case of the (Kir2.1NC4:PSD-95)4 envelope the Kir2.1NC4:PSD-95 model envelope was initially docked into ¼ of the map by eye, and C4 The best PSD-95 and Kir2.1NC4 models obtained from rigid body symmetry was used to generate the other three subunits. In a second analysis with BUNCH were used to calculate model envelopes using round of fitting a model envelope was generated from the three the pdb2mrc routine in EMAN. The resolution used in these model subunits and subtracted from the map; the Kir2.1NC4:PSD-95 model envelope calculations matched that of corresponding nsTEM map to envelope was then automatically docked into the remaining density. be fitted (either the filter resolution or in the later stages of The final model was again generated by the application of C4 refinement the resolution determined by a Fourier shell correlation symmetry. Care was taken at every stage to maintain consistent (FSC=1/2) calculation [63]). The calculated envelopes were docked chirality for all of the maps and models. into the nsTEM maps in an automated fashion using CHIMERA. The The resolution of the nsTEM envelopes was estimated as follows: resulting re-positioned calculated model envelope was then used to each set of particles was split into two groups and the last 2 relocate the rigid body SAXS model, again by auto-fitting in CHIMERA. refinement cycles were repeated for each envelope independently This procedure was performed directly to dock the PSD-95 and using the two sets of (unmasked) particles. The FSC was computed by

Kir2.1NC4 SAXS models into the corresponding nsTEM envelopes. In comparing the envelopes from these two reﬁnements using the 2380 S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389

Fig. 3. nsTEM derived envelope of PSD-95 and docked SAXS models: (a–c) Three orthogonal views of the final 22.9 Å resolution nsTEM envelope for PSD-95 refined with particles corrected for the CTF function, two BUNCH models are fitted (colored blue–red according to sequence), both models are included in the coordinate deposition PDB ID: 2XKX, one model (chain B in the deposition) is shown in the remaining figures in this paper. The scalebar represents 25 Å. The outer contour encloses 78 kDa, the inner contour approximately that volume occupied by the ordered domains. Generated using CHIMERA. (d) Equivalent views of the fitted PSD-95 envelope with a simplified representation of the models fitted, colored balls (PDZ radius 13.8 Å; SH3GK, 21.6 Å) representing the domains (PDZ1, blue; PDZ2, cyan; PDZ3, green; SH3GK orange) (e) Fourier shell correlation and structure factor comparison: FSC for the PSD-95 nsTEM envelope (red solid line) with fitted curve (red dotted line); red axis; structure factors for the nsTEM envelope (blue line), a normalized SAXS curve calculated from the DAMMIN model (yellow line), normalized SAXS curve after the application of a Debye–Waller B-factor of 1360 (green line); black axis. S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 2381

Fig. 4. nsTEM derived envelope of Kir2.1NC4 anddockedSAXSmodel:(a–d). Views of the final 17.2 Å resolution nsTEM envelope for the Kir2.1NC4, the best BUNCH model is superimposed. The scale bar represents 25 Å. The outer contour encloses 140 kDa, the inner contour approximately that volume occupied by the ordered domains. Generated using CHIMERA. Top side and bottom views (a, b, c) and the side view rotated by 45° (d) are shown. (e) FSC and structure factor comparison: representation as for Fig. 3(e) with a Debye–Waller B-factor of 1581 applied. 2382 S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 proc3d routine in EMAN. In an effort to use the complete FSC curve in polyacrylamide gel-electrophoresis (SDS-PAGE) gel (Fig. A.1(d)). The the determination of resolution an equivalent filter function of the purified protein was found to be monomeric and homogeneous by size form FSC=(1+exp[as+b])−1 was applied to the FSC plot; where a exclusion chromatography (SEC) and light scattering (Fig. A.1(a,b)). The and b are determined from the gradient and y-axis intercept of a identity of the protein was confirmed by tryptic digest and liquid straight line fit of ln((1−FSC)/FSC) versus s for points where FSCN1/2. chromatography mass spectrometry (LCMS) from an excised gel band. The resolution of the nsTEM envelope was determined from these Fig. 1(b) shows the final ab-initio DAMMIN envelope and 10 SAXS rigid parameters: for FSC=1/2; s−1 =a/−b(thefit is shown as a dotted line body domain models along with the agreement of the best of these in the FSC plots in Figs. 3(e), 4(e), 5(e), 6(f) and A.2(d), A.3(d) A.4(e) along with the SAXS intensity profile (Fig. 1(a)). 76% of the sequence of and A.5(f)). the PSD-95 molecule is represented by high-resolution domain structures. The initial nsTEM envelope of PSD-95 (Fig. A.2(e–g))isnot 2.11. Averaging of nsTEM maps truly independent of the SAXS model since the ab-initio SAXS model for PSD-95 was used in model-based particle picking. However the final An averaging procedure was undertaken to ensure the consistency refined nsTEM envelope draws on contributions from all but the of the nsTEM maps: The process is represented schematically in Fig. 2. Kir2.1NC4 particle set and remains consistent with the SAXS data. The Starting from the individually refined envelopes for PSD-95, Fig. A.2 overall form of the final nsTEM envelope suggests that there is some

(e–g); Kir2.1NC4, Fig. A.3(e–g); PSD95:Kir2.1NC4 Fig. A.4(f–h) and variation in the conformation of PSD-95 (Fig. 3(a–c)). The models (PSD-95:Kir2.1NC4)4 Fig. A.5(g–i); three envelopes were averaged obtained with rigid-body modeling using BUNCH also show some together: The unmodiﬁed PSD95:Kir2.1NC4 envelope; an envelope degree of variation especially for the N-terminal and linker regions representative of the 1:1 complex constructed from the individual (between PDZ2 and PDZ3, and between PDZ3 and SH3). The array of

PSD-95 and Kir2.1NC4 envelopes; and an envelope representative of PSD-95 rigid body models reveal two populations, one in which PDZ2 is the 1:1 complex consisting of ¼ of the (PSD-95:Kir2.1NC4)4 envelope closest to PDZ3 and another in which PDZ1 is closer to PDZ3. Two (Fig. 2(a)). The relative translations for overlaying the individual models representative of each of these pools are shown docked into the maps were obtained from the fit of the identical BUNCH SAXS models nsTEM envelope. A more compact conformation of PSD-95 is also to each map, centered maps were then aligned using the align3d supported by the ensemble analysis conducted (Fig. 1(c)) and this is in routine in EMAN. Envelopes for each of the four particle sets were contrast with the truncated PDZ1-3 construct of SAP-97 which was then derived from this averaged 1:1 complex: In the case of PSD-95 observed to have a large degree of flexibility [64]. The shape and domain and Kir2.1NC4 by masking out one component; for the (PSD-95: arrangements for PSD-95 as determined by nsTEM and SAXS are Kir2.1NC4)4 via the application of C4 symmetry, while for the PSD95: consistent with earlier studies [65]. The proximity of the SH3 domain to Kir2.1NC4 envelope the averaged envelope was submitted for PDZ3 is consistent with an intra-molecular association of PtsP or PreP refinement. Four cycles of refinement from these envelopes using sequences close to the core fold of PDZ3 with SH3 [66]. the individual particle sets were then carried out before the next cycle The Kir2.1NC protein was successfully expressed in E. coli and purified of averaging. The convergence of this process was measured by to a single band on a Coomassie stained SDS-PAGE gel (Fig. A.1(d)). The comparing the FSC of each successive averaged 1:1 complex between identity of the protein was confirmed by tryptic digest and LCMS. The averaging cycles (shown in Fig. 2(b)). Convergence was achieved after protein was found to migrate in a resolvable peak consistent with a

5 cycles of averaging/refinement. Fig. 2(a) shows envelopes from the Kir2.1NC4 complex by SEC (Fig. A.1(c)). Fig. 1(e) shows the final ab-initio last round of averaging. SAXSmodelwith10SAXSdomainmodelssuperimposed,alongwiththe A comparison of the structure factors from the final averaged maps agreement of the best of these (Fig. 1(d)) with the SAXS intensity profile. and the structure factors obtained from the DAMMIN SAXS models is For SAXS model analysis C4 molecular symmetry was imposed for all also shown in Figs. 3–6. In each case a Debye–Waller B-factor was residues in all models along with restraints to maintain the relative calculated by comparing the SAXS and EM derived structure factors. orientations of the four folded Kir2.1NC domains. The crystallographic

The relative B-factor in each case is in the range of 1200–1600 and structure of Kir2.1NC4 has C2 molecular symmetry but the core domains arises primarily from the effect of the negative stain. The application are within 0.6 Å of C4 symmetry (the root mean square deviation on main of the B-factor scaled SAXS amplitude to the refined envelopes was chain coordinates on superposition of a 90° rotated copy). Exploratory found to be essentially equivalent to filtering at the resolution rigid body refinements imposing C4 molecular symmetry but where the determined by FSC=1/2 for each envelope. inter-domain restraints were relaxed produced worse agreement with the SAXS intensity profile and were characterized by rotations of the folded 2.12. Accession numbers unique Kir2.1NC domain giving unreasonable inter-domain contacts. In

the ﬁnal Kir2.1NC4 SAXS model set 68% of the sequence is represented by Maps have been deposited in the European Bioinformatics high-resolution domain structures and the models are consistent with a Institute Electron Microscopy Data Bank (EBI-EMDB), and coordinates core tetramer and disordered far N and C-termini. It should be noted that deposited at the Protein Data Bank (PDB), accession codes EBI-EMDB the far N and C-termini are unlikely to be constrained to C4 molecular ID: EMD-1761, PDB ID: 2XKX, PSD-95; ID: EBI-EMDB EMD-1764, symmetry in reality. Single particle nsTEM studies gave an envelope for

PDB ID: 2XKY, Kir2.1NC4; EBI-EMDB ID: EMD-1765, Kir2.1NC4:PSD- Kir2.1NC4 (Fig. A.3(e–g)). The initial Kir2.1NC4 envelope was reﬁned from 95, and EBI-EMDB ID: EMD-1766, (Kir2.1NC4:PSD-95)4. The coordi- a reference free envelope for which particles were picked automatically nate transformations for positioning copies of 2XKX and 2XKY in using a small subset of examples present on each micrograph. Fig. 4(a–d) EMD-1765 and EMD-1766 are included in each entry. shows the ﬁnal envelope obtained with the FSC (Fig. 4(e)). The envelope

obtained shows a high degree of fidelity with the Kir2.1NC4 SAXS model as 3. Results exemplified by the docking of this model into the envelope shown in the figure.

The PSD-95 protein was successfully expressed in E. coli and puriﬁed The SEC fraction of the combined Kir2.1NC4/PSD-95 samples at an to a single band on a Coomassie stained sodium-dodecylsulphate 11.5 ml retention volume contains two populations of particles, those

Fig. 5. The ﬁnal nsTEM envelope of the Kir2.1NC4:PSD-95 1:1 complex: (a–d) Orthogonal views of the 20.5 Å resolution nsTEM envelope for Kir2.1NC4:PSD-95, the outer contour encloses 218 kDa, the inner contour approximately that volume occupied by the ordered domains; the BUNCH models shown in Figs. 3 and 4 for Kir2.1NC4 and PSD-95 are ﬁtted

(Kir2.1NC4 cyan; PSD-95 magenta). The scalebar represents 25 Å. Generated using CHIMERA. (e) FSC and structure factor comparison: representation as for Fig. 3(e) comparing transformed DAMMIN models for PSD-95 and Kir2.1NC4 and with a Debye–Waller B-factor of 1211 applied. S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 2383 2384 S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389

of 1:1 complexes (218 kDa) and unbound Kir2.1NC4 (140 kDa). The tissues and so general conclusions relevant to the interactions of Kir other possible solution components, unbound PSD-95 (78 kDa), ion channels and scaffold proteins may be drawn from this study. glutathione-S transferase (52 kDa dimer/26 kDa monomer) and Both Kir2.1NC and PSD-95 proteins have been heterologously monomeric Kir2.1NC (35 kDa) are each resolved by the column at expressed in E. coli and therefore post-translational modifications higher retention volumes, any higher order complexes are found such as palmitoylation and phosphorylation are not present. The closer to the void volume (8 ml). Two envelopes were obtained from binding of Kir2.3 channels to PSD-95 is known to be influenced by the nsTEM analysis of the SEC fractionated Kir2.1NC4/PSD-95 protein kinase A phosphorylation [20]. As the eSeI and PtsP motifs of complexes (Fig. A.4(c)): A Kir2.1NC4 envelope, consistent with the Kir2.1 and PSD-95 respectively are predicted phosphorylation targets Kir2.1NC4 envelope determined above (Fig. A.3(e–g)); and an [68], the expression of the unmodified proteins in E. coli may have been envelope representative of a Kir2.1NC4:PSD-95 complex. The latter critical in observing these clustering phenomena. The use of the largest envelope was refined with C1 molecular symmetry. Refinements representative constructs, which could be studied in the solution phase starting from a C1 envelope or a reference free envelope developed (without the inclusion of detergent) has allowed the elucidation of with C1 symmetry directly give consistent results, (Fig. A.4(f–h) models for the clustering of these domains. The use of specificcleavable shows the particle refined from the reference free envelope). Single tagging systems for PSD-95 (glutathione-S-transferase tag) and particle analysis without the benefit of molecular symmetry is Kir2.1NC (hexa-histidine tag) was important in isolating the individual notoriously difficult and the form of the resulting envelope was proteins and in the observation of large complexes via the co- poorer than that which would be expected of a particle with higher purification method employed. However it should be noted that symmetry and the concomitant internal standard. Nevertheless it was under the solution conditions applied the majority of the proteins straightforward to position the SAXS models for Kir2.1NC4 and PSD- formed Kir2.1NC4:PSD-95 complexes. The formation of the (Kir2.1NC4: 95 within the initial envelope, and the process of multi-particle PSD-95)4 complex and extended aggregates required that the binding averaging produced a readily interpretable envelope for the 1:1 partners were introduced at an earlier stage of the PSD-95 preparation complex. Fig. 5(a–d) shows the Kir2.1NC4:PSD-95 envelope along with where the PSD-95 protein is more abundant and that saturating the docked model of the complex. The docked model was consistent quantities of pure Kir2.1NC4 are applied to the GST-tagged PSD-95 with a 1:1 multi-domain association of PSD-95 with Kir2.1NC4. This protein when bound to the affinity column. Nevertheless as ion channel envelope suggests a greater degree of interaction between the folded and scaffold proteins are effectively concentrated by membrane co- domains of PSD-95 and those of Kir2.1NC4 than might be expected as the localization in vivo, and the Kir2.1 protein is likely to be trafficked to the interaction motifs on Kir2.1NC4 are at the C-terminus of the extended plasma membrane in a complex containing MAGUK proteins [43,69] we chains. maintain that the clusters seen here are highly relevant to ion channels Fig. 6(d) shows a segment of a micrograph recorded from a grid in the membrane. containing higher order complexes. Tetrad particles can be seen (shown boxed) along with a long, extended aggregate. Both the self- enclosed tetrad complexes and longer aggregates are assumed to be 4.2. Precision of the structural models formed from a unit consisting of the Kir2.1NC4:PSD-95 complex as they are only formed after the co-elution of Kir2.1NC4:PSD-95 The process of determining the SAXS models was undertaken complexes from a GSH affinity column during preparation. Relatively using a careful approach in order to avoid any bias in the restraints few extended aggregates were observed on each micrograph and so deployed. The agreement of the PSD-95 BUNCH model with the the single particle analysis concentrated on the abundant tetrad experimental SAXS data is inferior compared to that of the Kir2.1NC4 complexes. Fig. A.5 shows the evolution of the tetrad envelope from a BUNCH model, Fig. 1(a,d). The fit of the BUNCH structures within the low resolution tomographic reconstruction. The final envelope is DAMMIN envelopes demonstrate this in real space, Fig. 1(b,e). This shown in Fig. 6(a,b,c,e). A copy of the Kir2.1NC4:PSD-95 model can be largely reflects the different modes of disorder displayed by the two docked into the tetrad envelope with the remainder of the model molecules. BUNCH refinements are based upon the adjustment of the being generated by the application of C4 molecular symmetry. The relative position and orientation of rigid body sub-domains. The SAXS docked model indicates that the tetrad is formed through an profile is dominated by domain structure. The greater variation in overhanging PDZ domain recruiting further Kir2.1NC4:PSD-95 complex. inter-domain structure in PSD-95 means that a single rigid body The clustered Kir2.1NC4 domains are brought into close physical contact domain (BUNCH) structure is less representative of the array of and the pore (C4) axis of each individual Kir2.1NC4 domain is inclined to structures in solution. The ab-initio DAMMIN dummy-atom model is the tetrad 4-fold rotation axis by an angle of 10.7°. representative of the overall average structure in solution (the EOM procedure also models this variation with an ensemble of domain structures). The DAMMIN model represents the data better but cannot 4. Discussion readily be decomposed into domains. Only in cases where there is little disorder and no inter-domain variation would BUNCH and 4.1. Validity of the structural models DAMMIN models represent the data equally well. The CTF correction of the nsTEM particle data sets was undertaken with reference to The protein expression constructs used in these experiments are scattering curves obtained by docking SAXS models into the nsTEM both derived from muridae, this was a consequence of the availability envelopes. The use of scattering curves calculated from high- of cDNA. The high resolution domain structures used in the resolution structural models to correct for the CTF in cryo-electron interpretation of the data here are also from a mixture of rat and microscopy has not proved useful; rather, measured scattering curves mouse: the cross species amino acid identity of the Kir2.1 and PSD-95 are generally required. nsTEM is limited to a lower resolution range proteins is 99.3% and 99.5% therefore the behavior and structure of the (circa ∞—20 Å) over which the calculation of SAXS curves from model proteins will be equivalent despite their mixed sources. The Kir2.1NC data is robust — as demonstrated in Fig. 1. Therefore the use of construct omits the trans-membrane domain. This protein does form calculated scattering curves in the CTF correction of nsTEM particles is a homotetrameric (Kir2.1NC4) domain and therefore is representative justified. The use of both DAMMIN and BUNCH based models to of the cytoplasmic channel domain in vivo [26,27,45,67]. Of the generate the CTF correction was explored and no difference was specific proteins studied here PSD-95 is most strongly associated with encountered in the correction calculated. With the exception of PSD- the synaptic membrane and Kir2.1 is expressed widely in various 95 alone each nsTEM envelope was refined from a reference free tissues. MAGUK proteins and Kir channels are found in numerous starting envelope. Docking of SAXS models was done in an automated S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 2385

Fig. 6. The ﬁnal nsTEM envelope of the (Kir2.1NC4:PSD-95)4 tetrad complex: (a–c) orthogonal views of the 25.8 Å resolution nsTEM envelope with the C4 symmetric model for

(Kir2.1NC4:PSD-95)4 ﬁtted, (e) the side view of the complex rotated by 45°. The outer contour encloses a volume corresponding to 872 kDa, the inner contour encloses approximately that volume occupied by the ordered domains. Generated using CHIMERA. (d) A segment of a micrograph showing tetrad particles and a longer extended aggregate; 2 boxes are 307.2×307.2 Å . (f) FSC and structure factor comparison: representation as for Fig. 3(e) comparing transformed DAMMIN models for the PSD-95:Kir2.1NC4 complex, C4 symmetry and with a Debye–Waller B-factor of 1451 applied. 2386 S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 fashion, as described in the Materials and methods section, therefore much stronger interaction than that afforded by an interaction with a minimizing the introduction of any interpretive bias. single PDZ domain.

The resolution of the initial nsTEM envelopes obtained were The Kir2.1NC4:PSD-95 model indicates that PDZ1, PDZ2 are close determined by an FSC=1/2 criterion to be in the resolution range of to the folded C-terminal domains of Kir2.1NC4 (corresponding to 17–26 Å (Figs. A.2(d), A.3(d), A.4(e), A.5(f)); and the agreement of the residues 190–365 of the Kir2.1 sequence) and that and the GK domain models with the SAXS curve is good to a resolution of approximately is in close contact with these domains: in all cases a “turret helix” 15 Å (Fig. 1(a,d)) suggesting a similar resolution for the SAXS models. A (residues 357–365 of the sequence) is close to each respective preferred orientation was observed for all of the particles studied. In the domain. In contrast PDZ3 appears to be held further away from these

final cycle of the independent (Kir2.1NC4:PSD-95)4 refinement the folded channel domains, although within interaction distance of the maximum and minimum number of particles assigned to individual flexible C-terminus of a Kir2.1NC monomer (corresponding to classes were 159 and 23 respectively; and for Kir2.1NC4 713/161, in both residues 366–428 of the channel sequence). Fig. A.6 includes a cases views close to the direction of the 4-fold axis were more numerous schematic representation of the 1:1 complex illustrating the relative over side views. The particles contributing to the PSD-95 nsTEM positions of these domains. Fig. A.7 highlights the position of the independent envelope also showed a biased orientation (103/18) with turret helix and illustrates the range of structural variations of the end-on views in the minority, while for Kir2.1NC4:PSD-95 (792/156) flexible C-terminal regions that are consistent with the SAXS data. The any preferred orientation was slightly less pronounced. Preferred binding of all four termini may have implications for the assembly and orientations are common for nsTEM on carbon grids. Representative trafficking of the Kir2.1 channel. An association of the four C-termini back projections and aligned class averages for each particle set are of Kir2.1 with PSD-95 is likely to screen an RR motif on the C-terminus shown in Figs. A.2–5. The minimum number of particles in each set is of the channel domain. Similar motifs have been implicated as an sufficient to give a robust class average for the angular sampling endoplasmic reticulum retention signal for NMDA receptors [70]. intervals used. The subsequent multi-particle averaging procedure The formation of extended clusters apparently proceeds firstly produced a convergent refinement and ensured the consistency of each through the association of 1:1 Kir2.1NC4:PSD-95 complexes which of the 4 envelopes determined. The resolutions of each of the envelopes then further interact. This is illustrated schematically in Fig. A.6; was found to be improved or equivalent to those from the independent overhanging PDZ1/2 domains in the 1:1 complex can make exchange refinement as judged by the FSC curve. interactions between complexes, and the form of the 1:1 complex is

Each of the domains located in the SAXS model analysis is from a preserved in the tetrad. The (Kir2.1NC4:PSD-95)4 model indicates that high-resolution structural study (better than 2.5 Å for the X-ray models). the Kir2.1NC4 tetramers are brought into physical contact within the Although isolated specific domains could not be assigned without tetrad. Furthermore this interaction appears to be under some strain ambiguity, rigid body multidomain SAXS models for both Kir2.1NC4 and as demonstrated by the 10.7° inclination of the individual Kir2.1NC4 PSD-95 can readily be fitted within the nsTEM envelopes for Kir2.1NC4, pore axis to the tetrad C4 axis. PSD-95 and Kir2.1NC4:PSD-95. In a similar way the model for the Kir2.2 and Kir2.1 channels are homologues and the Kir2.1NC Kir2.1NC4:PSD-95 complex as a whole can readily be positioned within (cytoplasmic domain) crystal structure was used in the structure the (Kir2.1NC4:PSD-95)4 envelope (Fig. 6(a,b,c,e)). The stepwise model solution of Kir2.2 by molecular replacement [36]. A model for clustering fitting procedure used here lends increased robustness to this procedure: of full-length Kir2.x channels can be made using the Kir2.2 structure the (Kir2.1NC4:PSD-95)4 envelope was fitted with 4 copies of the (PDB ID: 3JYC)(Fig. 7(a)): Kir2.24 was positioned in the equivalent site unmodified Kir2.1NC4:PSD-95 model. The Kir2.1NC4:PSD-95, Kir2.1NC4 in the (Kir2.1NC4:PSD-95)4 tetrad model using a least-squares align- and PSD-95 envelopes each having been fitted with the unmodified ment (Cα coordinates of residues 190–366 of each the Kir2.24 domains Kir2.1NC4 and PSD-95 SAXS models: thus all four nsTEM envelopes were superimposed). The inclination of the Kir2.24 model to the tetrad reported in this study are fitted with a combination of two multidomain C4 axis was then set to 0° by a simple rotation of 10.7°. The (Kir2.24:PSD- SAXS models. The resolution of the docked models is challenging to 95)4 tetrad was then generated by the application of 4-fold rotational describe with a single number as each comprises high-resolution symmetry. In order to reduce stereochemical clashes (leaving one structures, SAXS refinement and docking, and require a Debye–Waller Kir2.24 inter-molecular domain Cα–Cα distance b3.0 Å) the (Kir2.24: B-factor to account for the affects of negative stain. The agreement of the PSD-95)4 tetrad radius was increased by 8.8 Å. docked models with the nsTEM envelope is summarized via the The PDZ domains within PSD-95 bind promiscuously and PSD-95 graphical data shown in Figs. 3(e), 4(e), 5(e) and 6(f). In each graph is also known to bind and cluster shaker type Kv channels [5] and to the agreement of the natural logarithm of the structure factor (Ln(I)) stabilize the channels in the membrane [71]. Kv channels are also derived from the nsTEM envelope (blue) and Ln(I) derived from the tetrameric and a subset has PDZ target sequences at their C-termini. docked structure (green) are compared versus resolution. In each case However Kv channels possess a radically different overall form when the resolution at which these lines diverge correlates with the tailing off compared with Kir channels [72] and also associate with Kvβ proteins. of the FSC curve (red), indicating that in each case that the docked model The clustering of Kv1.2 by PSD-95 has been observed to be influenced by is an appropriate representation of the nsTEM envelope. co-expression of Kvβ proteins [6]:Kvβ proteins are not predicted to be

interaction partners of PSD-95. It is notable that the structures of Kv1.24: 4.3. Implications of the models determined Kvβ4 [72,73] (PDB IDs: 2A79, 3LNM) could be accommodated within the (Kir2.14:PSD-95)4 tetrad model developed here (Fig. 7(d)) with The resultant (Kir2.1NC4:PSD-95)4 tetrad model, Kir2.1NC4:PSD-95 little overlap. Starting from the complete Kv1.24:Kvβ4 crystal structure model, along with the SAXS models for PSD-95 and Kir2.1NC4 form a Kvβ4 requires a relative rotation of about the pore axis of Kv1.24 by consistent set. A hypothesis arising from these models is that the C- approximately 45° to fit into the center of the (Kir2.14:PSD-95)4 Fig. 7 terminal −eSeI motif of Kir2.1NC4 may bind simultaneously to each of (d). Andersen's syndrome mutations (shown in yellow in Figure A.7) the PDZ domains in PSD-95. Solution binding and NMR studies using occur to residues away from the central pore in the Kir2.1 channel and a single and multiple PDZ domain constructs have shown that sequences number of these mutations could directly affect interactions between Kv corresponding to the C-termini of Kir2.1NC4 bind to PSD-95 PDZ1 and and Kir channels in the arrangement hypothesized in Fig. 7. PDZ2 [41] and to PDZ1,2 and 3 of the related protein SAP97 [64] with varying affinity. As the PSD-95 PDZ domains are connected by flexible 5. Conclusions linkers it seems unlikely that the PDZ domains act cooperatively in binding. However the independent interaction of multiple PDZ domains PSD-95 forms a loose preferred structure which is preserved on with a single channel tetramer will be enhanced by avidity resulting in a interaction with Kir2.14 channels. The domain models of PSD-95 are S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 2387

Fig. 7. Full-length tetrad models (Kir2.x4:PSD-95)4 and docking of the Kv1.24:Kβ4 structure: column (a) views of the model for the full-length (Kir2.x4:PSD-95)4 complex, Kir2.24 are shown as blue/cyan cartoons, PSD-95 in pink/magenta. Column (b) equivalent views of the (Kir2.x4:PSD-95)4 including a representation of the volume in the center of the complex

(gray surface). (c) The excluded volume alone (gray surface). (d) Docking of the crystal structure of Kv1.24:Kβ4 into the central volume: The Kir1.24 paddle chimera/Kvβ4 complex is shown (yellow/green cartoon) docked. Generated using CHIMERA.

consistent with the closer association of SH3GK and PDZ3, and the tetrameric cluster. The longer chain-linked particles observed could close association of PDZ1 and PDZ2 with a greater degree of freedom result from a different mode or stoichiometry of interaction of of motion between these two elements. This implies a level of Kir2.1NC4 and PSD-95. disorder in the relative positions of the PSD-95 domains, in contrast to Kir2.14 channels form ordered tetrameric clusters under the direction Kir2.1NC4 where the structure is portioned into ordered core of PSD-95 and it is likely that other MAGUKs will induce similar domains, and ﬂexible termini. A 1:1 interaction of PSD-95 with the structures in combination with Kir channels. Within these clusters

Kir2.14 channel occurs, this is likely to proceed through the the cytoplasmic domains of the channels are in very close physical interaction of all three PDZ domains plus the SH3GK domain within contact. A compelling argument for their validity in vivo is that PSD-95 with the four C-termini of a single channel tetramer. In the voltage gated Kv/Kvβ channel assemblies could be accommodated low resolution structures determined here the interactions of the within these complexes albeit with structural adjustment. Kir channel C-termini with PSD-95 are not resolved. A number of other channels are passive conductors of K+ ions, therefore in order to studies have clearly shown that the presence of a PDZ binding motif is selectively conduct K+ ions into the cell in vivo Kir channels must be necessary for Kir:PSD-95 association, notably that Kir2.3:PSD-95 blocked or closed until the membrane potential overcomes the K+ binding is disrupted by phosphorylation of the Ser residue in the diffusion gradient. This indirect mechanism of voltage sensing in Kir −eSaI motif [20] and that the last 3 residues of the −eSeI motif in channels is thought to take place through small intracellular co- Kir2.2 are required to pull down PSD-95 from brain extract [43]. While factors (polyamines and Mg2+ ions) blocking the channel pore we cannot say that C-terminal motifs are the sole determinants of the [74,75]. Polyamines have binding sites in the cytoplasmic domains of structures observed here, the aforementioned results do indicate that Kir2.1 channels [26,76] (shown in green in Fig. A.7) and the conforma- C-terminal motifs are likely to play a signiﬁcant role in complex tionally dependent release of polyamines from these has been proposed as formation. a mechanism of gating [27]. The co-clustering of Kv and Kir channels The model ﬁtted to the larger tetrad envelope is consistent with proposed here could therefore give a more direct mechanism for voltage

PDZ1/2 of the Kir2.14:PSD-95 complex recruiting further complexes sensing by Kir channels as voltage induced changes of conformation in Kv via exchanging an interaction with the C-terminus of a second Kir2.1 could induce conformational changes in Kir through the close physical channel. This arrangement allows the formation of a self-enclosed contact of the Kv and Kir channel domains. 2388 S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389

Acknowledgments [13] O. Prange, T.P. Wong, K. Gerrow, Y.T. Wang, A. El-Husseini, A balance between excitatory and inhibitory synapses is controlled by PSD-95 and neuroligin, Proc. Natl. Acad. Sci. U. S. A. 101 (2004) 13915–13920. This work was supported by the Biotechnology and Biological [14] T. Ohnuma, H. Kato, H. Arai, R.L. Faull, P.J. McKenna, P.C. Emson, Gene expression Sciences Research Council [first investigator grant, number BBS/ of PSD95 in prefrontal cortex and hippocampus in schizophrenia, Neuroreport 11 (2000) 3133–3137. B/05567]; and the Royal Society [research grant, number 23055]. OKS [15] M. Feyder, R.M. Karlsson, P. Mathur, M. Lyman, R. Bock, R. Momenan, J. was supported by a Biotechnology and Biological Sciences Research Munasinghe, M.L. Scattoni, J. Ihne, M. Camp, C. Graybeal, D. Strathdee, A. Begg, Council studentship [Quota Studentship, number A1/B/08277]; V.A. Alvarez, P. Kirsch, M. Rietschel, S. Cichon, H. Walter, A. Meyer-Lindenberg, S. LO'R was supported by an Engineering and Physical Sciences Research G. Grant, A. Holmes, Association of mouse Dlg4 (PSD-95) gene deletion and human DLG4 gene variation with phenotypes relevant to autism spectrum Council Doctoral Training Grant studentship. The Science and Technol- disorders and Williams' syndrome, Am. J. Psychiatry 167 (2010) 1508–1517. ogies Facilities Council provided small angle X-ray scattering measure- [16] Z. Songyang, A.S. Fanning, C. Fu, J. Xu, S.M. Marfatia, A.H. Chishti, A. Crompton, A.C. ment time at Station 2.1 of the Daresbury Synchrotron Radiation Source. Chan, J.M. Anderson, L.C. Cantley, Recognition of unique carboxyl-terminal motifs by distinct PDZ domains, Science 275 (1997) 73–77. Electron micrographs were recorded at the Manchester Interdisciplinary [17] A.W. McGee, S.R. Dakoji, O. Olsen, D.S. Bredt, W.A. Lim, K.E. Prehoda, Structure of Biocentre Electron Microscopy facility, Dr. C.A. Shintre assisted with data the SH3-guanylate kinase module from PSD-95 suggests a mechanism for collection on the Jeol 1220 microscope. Light scattering analysis of PSD- regulated assembly of MAGUK scaffolding proteins, Mol. Cell 8 (2001) 1291–1301. 95 was performed in the Faculty of Life Sciences Biomolecular Analysis [18] H. Shin, Y.P. Hsueh, F.C. Yang, E. Kim, M. Sheng, An intramolecular interaction Facility at the University of Manchester with the assistance of Dr. T.A. between Src homology 3 domain and guanylate kinase-like domain required for Jowitt. The authors would like to acknowledge the generous advice of channel clustering by postsynaptic density-95/SAP90, J. Neurosci. 20 (2000) 3580–3587. Robert Ford, Clair Baldock (both Faculty of Life Sciences, University of [19] H. Yu, J.K. Chen, S. Feng, D.C. Dalgarno, A.W. Brauer, S.L. Schreiber, Structural basis Manchester) and Steven Ludtke (Baylor College of Medicine) on for the binding of proline-rich peptides to SH3 domains, Cell 76 (1994) 933–945. electron microscopy, small angle scattering and the use of EMAN [20] N.A. Cohen, J.E. Brenman, S.H. Snyder, D.S. Bredt, Binding of the inward rectifier K+ channel Kir 2.3 to PSD-95 is regulated by protein kinase A phosphorylation, software respectively. Neuron 17 (1996) 759–767. The PSD-95 pGEX-6P construct was made by MLL, the PSD-95 [21] Y. Kubo, J.P. Adelman, D.E. Clapham, L.Y. Jan, A. Karschin, Y. Kurachi, M. Lazdunski, purification was defined by TDH, SAXS samples for PSD-95 were C.G. Nichols, S. Seino, C.A. Vandenberg, International Union of Pharmacology. LIV. prepared by TDH. The Kir2.1NC construct was made by OKS with Nomenclature and molecular relationships of inwardly rectifying potassium channels, Pharmacol. Rev. 57 (2005) 509–526. fi fi advice from MLL, the Kir2.1NC4 puri cation was de ned by OKS, TDH, [22] J.J. Zaritsky, D.M. Eckman, G.C. Wellman, M.T. Nelson, T.L. Schwarz, Targeted SF and MG. SAXS samples for Kir2.1NC4 were prepared by SF and MG. disruption of Kir2.1 and Kir2.2 genes reveals the essential role of the inwardly nsTEM samples were prepared by SF and MG. Electron micrographs rectifying K(+) current in K(+)-mediated vasodilation, Circ. Res. 87 (2000) 160–166. were recorded by RFC. Initial SAXS analysis was performed by JGG and [23] K. Morishige, N. Takahashi, I. Findlay, H. Koyama, J.S. Zanelli, C. Peterson, N.A. LO'R. SMP directed the research and performed the remaining SAXS Jenkins, N.G. Copeland, N. Mori, Y. Kurachi, Molecular cloning, functional and EM analysis. expression and localization of an inward rectifier potassium channel in the mouse brain, FEBS Lett. 336 (1993) 375–380. [24] M.D. Ashen, B. O'Rourke, K.A. Kluge, D.C. Johns, G.F. Tomaselli, Inward rectifier K+ Appendix A. Supplementary data channel from human heart and brain: cloning and stable expression in a human cell line, Am. J. Physiol. 268 (1995) H506–H511. [25] B.K. Panama, M. McLerie, A.N. Lopatin, Functional consequences of Kir2.1/Kir2.2 Supplementary data to this article can be found online at doi:10. subunit heteromerization, Pflugers Arch. 460 (2010) 839–849. 1016/j.bbamem.2011.06.021. [26] S. Pegan, C. Arrabit, W. Zhou, W. Kwiatkowski, A. Collins, P.A. Slesinger, S. Choe, Cytoplasmic domain structures of Kir2.1 and Kir3.1 show sites for modulating gating and rectification, Nat. Neurosci. 8 (2005) 279–287. [27] M. Osawa, M. Yokogawa, T. Muramatsu, T. Kimura, Y. Mase, I. Shimada, Evidence References for the direct interaction of spermine with the inwardly rectifying potassium channel, J. Biol. Chem. 284 (2009) 26117–26126. [1] S.N. Gomperts, Clustering membrane proteins: it’s all coming together with the [28] L.S. Wood, T.D. Tsai, K.S. Lee, G. Vogeli, Cloning and functional expression of a PSD-95/SAP90 protein family, Cell 84 (1996) 659–662. human gene, hIRK1, encoding the heart inward rectifier K+-channel, Gene 163 [2] S.E. Craven, D.S. Bredt, PDZ proteins organize synaptic signaling pathways, Cell 93 (1995) 313–317. (1998) 495–498. [29] G. Andelfinger, A.R. Tapper, R.C. Welch, C.G. Vanoye, A.L. George Jr., D.W. Benson, [3] M.L. Leyland, C. Dart, An alternatively spliced isoform of PSD-93/chapsyn 110 KCNJ2 mutation results in Andersen syndrome with sex-specific cardiac and binds to the inwardly rectifying potassium channel, Kir2.1, J. Biol. Chem. 279 skeletal muscle phenotypes, Am. J. Hum. Genet. 71 (2002) 663–668. (2004) 43427–43436. [30] S. Bendahhou, E. Fournier, S. Gallet, D. Menard, M.M. Larroque, J. Barhanin, [4] Y. Horio, H. Hibino, A. Inanobe, M. Yamada, M. Ishii, Y. Tada, E. Satoh, Y. Hata, Y. Corticosteroid-exacerbated symptoms in an Andersen's syndrome kindred, Hum. Takai, Y. Kurachi, Clustering and enhanced activity of an inwardly rectifying Mol. Genet. 16 (2007) 900–906. potassium channel, Kir4.1, by an anchoring protein, PSD-95/SAP90, J. Biol. Chem. [31] C.W. Lu, J.H. Lin, Y.S. Rajawat, H. Jerng, T.G. Rami, X. Sanchez, G. DeFreitas, B. 272 (1997) 12885–12888. Carabello, F. DeMayo, D.L. Kearney, G. Miller, H. Li, P.J. Pfaffinger, N.E. Bowles, D.S. [5] E. Kim, M. Niethammer, A. Rothschild, Y.N. Jan, M. Sheng, Clustering of Shaker-type Khoury, J.A. Towbin, Functional and clinical characterization of a mutation in K+ channels by interaction with a family of membrane-associated guanylate KCNJ2 associated with Andersen–Tawil syndrome, J. Med. Genet. 43 (2006) kinases, Nature 378 (1995) 85–88. 653–659. [6] A.M. Tiffany, L.N. Manganas, E. Kim, Y.P. Hsueh, M. Sheng, J.S. Trimmer, PSD-95 [32] M. Tristani-Firouzi, J.L. Jensen, M.R. Donaldson, V. Sansone, G. Meola, A. Hahn, S. and SAP97 exhibit distinct mechanisms for regulating K(+) channel surface Bendahhou, H. Kwiecinski, A. Fidzianska, N. Plaster, Y.H. Fu, L.J. Ptacek, R. Tawil, expression and clustering, J. Cell Biol. 148 (2000) 147–158. Functional and clinical characterization of KCNJ2 mutations associated with LQT7 [7] W. Wong, L.C. Schlichter, Differential recruitment of Kv1.4 and Kv4.2 to lipid rafts (Andersen syndrome), J. Clin. Invest. 110 (2002) 381–388. by PSD-95, J. Biol. Chem. 279 (2004) 444–452. [33] S. Pegan, C. Arrabit, P.A. Slesinger, S. Choe, Andersen's syndrome mutation effects [8] D.R. Marks, D.A. Fadool, Post-synaptic density perturbs insulin-induced Kv1.3 on the structure and assembly of the cytoplasmic domains of Kir2.1, Biochemistry channel modulation via a clustering mechanism involving the SH3 domain, J. 45 (2006) 8599–8606. Neurochem. 103 (2007) 1608–1627. [34] N.M. Plaster, R. Tawil, M. Tristani-Firouzi, S. Canun, S. Bendahhou, A. Tsunoda, M.R. [9] C. Aoki, I. Miko, H. Oviedo, T. Mikeladze-Dvali, L. Alexandre, N. Sweeney, D.S. Donaldson, S.T. Iannaccone, E. Brunt, R. Barohn, J. Clark, F. Deymeer, A.L. George Jr., Bredt, Electron microscopic immunocytochemical detection of PSD-95, PSD-93, F.A. Fish, A. Hahn, A. Nitu, C. Ozdemir, P. Serdaroglu, S.H. Subramony, G. Wolfe, SAP-102, and SAP-97 at postsynaptic, presynaptic, and nonsynaptic sites of adult Y.H. Fu, L.J. Ptacek, Mutations in Kir2.1 cause the developmental and episodic and neonatal rat visual cortex, Synapse 40 (2001) 239–257. electrical phenotypes of Andersen's syndrome, Cell 105 (2001) 511–519. [10] X. Chen, C. Winters, R. Azzam, X. Li, J.A. Galbraith, R.D. Leapman, T.S. Reese, [35] A. Kuo, J.M. Gulbis, J.F. Antcliff, T. Rahman, E.D. Lowe, J. Zimmer, J. Cuthbertson, Organization of the core structure of the postsynaptic density, Proc. Natl. Acad. Sci. F.M. Ashcroft, T. Ezaki, D.A. Doyle, Crystal structure of the potassium channel U. S. A. 105 (2008) 4453–4458. KirBac1.1 in the closed state, Science 300 (2003) 1922–1926. [11] A. Fujita, Y. Kurachi, SAP family proteins, Biochem. Biophys. Res. Commun. 269 [36] X. Tao, J.L. Avalos, J. Chen, R. MacKinnon, Crystal structure of the eukaryotic strong (2000) 1–6. inward-rectifier K+ channel Kir2.2 at 3.1 A resolution, Science 326 (2009) [12] A.E. El-Husseini, S.E. Craven, D.M. Chetkovich, B.L. Firestein, E. Schnell, C. Aoki, D.S. 1668–1674. Bredt, Dual palmitoylation of PSD-95 mediates its vesiculotubular sorting, [37] D.A. Doyle, A. Lee, J. Lewis, E. Kim, M. Sheng, R. MacKinnon, Crystal structures of a postsynaptic targeting, and ion channel clustering, J. Cell Biol. 148 (2000) complexed and peptide-free membrane protein-binding domain: molecular basis 159–172. of peptide recognition by PDZ, Cell 85 (1996) 1067–1076. S. Fomina et al. / Biochimica et Biophysica Acta 1808 (2011) 2374–2389 2389

[38] J.F. Long, H. Tochio, P. Wang, J.S. Fan, C. Sala, M. Niethammer, M. Sheng, M. Zhang, [58] M.V. Petoukhov, D.I. Svergun, Global rigid body modeling of macromolecular Supramodular structure and synergistic target binding of the N-terminal tandem complexes against small-angle scattering data, Biophys. J. 89 (2005) 1237–1250. PDZ domains of PSD-95, J. Mol. Biol. 327 (2003) 203–214. [59] P. Bernado, E. Mylonas, M.V. Petoukhov, M. Blackledge, D.I. Svergun, Structural [39] H. Tochio, F. Hung, M. Li, D.S. Bredt, M. Zhang, Solution structure and backbone characterization of flexible proteins using small-angle X-ray scattering, J. Am. dynamics of the second PDZ domain of postsynaptic density-95, J. Mol. Biol. 295 Chem. Soc. 129 (2007) 5656–5664. (2000) 225–237. [60] S.J. Ludtke, P.R. Baldwin, W. Chiu, EMAN: semiautomated software for high- [40] G.A. Tavares, E.H. Panepucci, A.T. Brunger, Structural characterization of the resolution single-particle reconstructions, J. Struct. Biol. 128 (1999) 82–97. intramolecular interaction between the SH3 and guanylate kinase domains of [61] E.F. Pettersen, T.D. Goddard, C.C. Huang, G.S. Couch, D.M. Greenblatt, E.C. Meng, PSD-95, Mol. Cell 8 (2001) 1313–1325. T.E. Ferrin, UCSF Chimera—a visualization system for exploratory research and [41] S. Pegan, J. Tan, A. Huang, P.A. Slesinger, R. Riek, S. Choe, NMR studies of analysis, J. Comput. Chem. 25 (2004) 1605–1612. interactions between C-terminal tail of Kir2.1 channel and PDZ1,2 domains of [62] D. Svergun, C. Barberato, M.H.J. Koch, CRYSOL — a program to evaluate X-ray PSD95, Biochemistry 46 (2007) 5315–5322. solution scattering of biological macromolecules from atomic coordinates, J. Appl. [42] M.N. Rasband, E.W. Park, D. Zhen, M.I. Arbuckle, S. Poliak, E. Peles, S.G. Grant, J.S. Crystallogr. 28 (1995) 768–773. Trimmer, Clustering of neuronal potassium channels is independent of their [63] G. Harauz, M. van Heel, Exact filters for general geometry three dimensional interaction with PSD-95, J. Cell Biol. 159 (2002) 663–672. reconstruction, Optik 73 (1986) 146–157. [43] D. Leonoudakis, L.R. Conti, S. Anderson, C.M. Radeke, L.M. McGuire, M.E. Adams, [64] B.T. Goult, J.D. Rapley, C. Dart, A. Kitmitto, J.G. Grossmann, M.L. Leyland, L.Y. Lian, S.C. Froehner, J.R. Yates III, C.A. Vandenberg, Protein trafficking and anchoring Small-angle X-ray scattering and NMR studies of the conformation of the PDZ complexes revealed by proteomic analysis of inward rectifier potassium channel region of SAP97 and its interactions with Kir2.1, Biochemistry 46 (2007) (Kir2.x)-associated proteins, J. Biol. Chem. 279 (2004) 22331–22346. 14117–14128. [44] P. Klemmer, A.B. Smit, K.W. Li, Proteomics analysis of immuno-precipitated [65] T. Nakagawa, K. Futai, H.A. Lashuel, I. Lo, K. Okamoto, T. Walz, Y. Hayashi, M. synaptic protein complexes, J. Proteomics 72 (2009) 82–90. Sheng, Quaternary structure, protein dynamics, and synaptic function of SAP97 [45] M. Nishida, R. MacKinnon, Structural basis of inward rectification: cytoplasmic controlled by L27 domain interactions, Neuron 44 (2004) 453–467. pore of the G protein-gated inward rectifier GIRK1 at 1.8 A resolution, Cell 111 [66] D. Korkin, F.P. Davis, F. Alber, T. Luong, M.Y. Shen, V. Lucic, M.B. Kennedy, A. Sali, (2002) 957–965. Structural modeling of protein interactions by analogy: application to PSD-95, [46] J.D. Thompson, T.J. Gibson, D.G. Higgins, Multiple sequence alignment using PLoS Comput. Biol. 2 (2006) e153. ClustalW and ClustalX, Curr. Protoc. Bioinformatics Chapter 2 (2002) Unit 2 3. [67] M.H. Nanao, W. Zhou, P.J. Pfaffinger, S. Choe, Determining the basis of channel- [47] The CCP4 suite: programs for protein crystallography, Acta Crystallogr. D Biol. tetramerization specificity by X-ray crystallography and a sequence-comparison Crystallogr. 50 (1994) 760–763. algorithm: Family Values (FamVal), Proc. Natl. Acad. Sci. U. S. A. 100 (2003) [48] W. Kabsch, Solution for best rotation to relate 2 sets of vectors, Acta Crystallogr. A 8670–8675. 32 (1976) 922–923. [68] N. Blom, S. Gammeltoft, S. Brunak, Sequence and structure-based prediction of [49] G.N. Murshudov, A.A. Vagin, E.J. Dodson, Refinement of macromolecular eukaryotic protein phosphorylation sites, J. Mol. Biol. 294 (1999) 1351–1362. structures by the maximum-likelihood method, Acta Crystallogr. D Biol. Crystal- [69] K. Heusser, B. Schwappach, Trafficking of potassium channels, Curr. Opin. logr. 53 (1997) 240–255. Neurobiol. 15 (2005) 364–369. [50] F.W. Studier, Protein production by auto-induction in high density shaking [70] S. Standley, K.W. Roche, J. McCallum, N. Sans, R.J. Wenthold, PDZ domain cultures, Protein Expr. Purif. 41 (2005) 207–234. suppression of an ER retention signal in NMDA receptor NR1 splice variants, [51] A.P. Golovanov, G.M. Hautbergue, S.A. Wilson, L.Y. Lian, A simple method for Neuron 28 (2000) 887–898. improving protein solubility and long-term stability, J. Am. Chem. Soc. 126 (2004) [71] D.G. Jugloff, R. Khanna, L.C. Schlichter, O.T. Jones, Internalization of the Kv1.4 8933–8939. potassium channel is suppressed by clustering interactions with PSD-95, J. Biol. [52] C. Boulin, R. Kempf, M.H.J. Koch, S.M. Mclaughlin, Data appraisal, evaluation and Chem. 275 (2000) 1357–1364. display for synchrotron radiation experiments — hardware and software, Nucl. [72] S.B. Long, E.B. Campbell, R. Mackinnon, Crystal structure of a mammalian voltage- Instrum. Methods Phys. Res. A 249 (1986) 399–407. dependent Shaker family K+ channel, Science 309 (2005) 897–903. [53] A.V. Semenyuk, D.I. Svergun, Gnom — a program package for small-angle [73] S.B. Long, X. Tao, E.B. Campbell, R. MacKinnon, Atomic structure of a voltage-dependent scattering data-processing, J. Appl. Crystallogr. 24 (1991) 537–540. K+ channel in a lipid membrane-like environment, Nature 450 (2007) 376–382. [54] D.I. Svergun, M.V. Petoukhov, M.H. Koch, Determination of domain structure of [74] B. Fakler, U. Brandle, E. Glowatzki, S. Weidemann, H.P. Zenner, J.P. Ruppersberg, proteins from X-ray solution scattering, Biophys. J. 80 (2001) 2946–2953. Strong voltage-dependent inward rectification of inward rectifier K+ channels is [55] V.V. Volkov, D.I. Svergun, Uniqueness of ab initio shape determination in small- caused by intracellular spermine, Cell 80 (1995) 149–154. angle scattering, J. Appl. Crystallogr. 36 (2003) 860–864. [75] Z. Lu, Mechanism of rectification in inward-rectifier K+ channels, Annu. Rev. [56] M.B. Kozin, D.I. Svergun, Automated matching of high- and low-resolution Physiol. 66 (2004) 103–129. structural models, J. Appl. Crystallogr. 34 (2001) 33–41. [76] H.T. Kurata, W.W. Cheng, C. Arrabit, P.A. Slesinger, C.G. Nichols, The role of the [57] D.I. Svergun, Restoring low resolution structure of biological macromolecules from cytoplasmic pore in inward rectification of Kir2.1 channels, J. Gen. Physiol. 130 solution scattering using simulated annealing, Biophys. J. 76 (1999) 2879–2886. (2007) 145–155.