Structural basis for the interaction between dynein light chain 1 and the glutamate channel homolog GRINL1A Marı´a F. Garcı´a-Mayoral1,Mo´ nica Martı´nez-Moreno2, Juan P. Albar3, Ignacio Rodrı´guez-Crespo2 and Marta Bruix1

1 Departamento de Espectroscopı´a y Estructura Molecular, Instituto de Quı´mica-Fı´sica Rocasolano, Consejo Superior de Investigaciones Cientificas (CSIC), Madrid, Spain 2 Departamento de Bioquı´mica y Biologı´a Molecular I, Universidad Complutense, Madrid, Spain 3 Proteomics Facility, Centro Nacional de Biotecnologı´a, Consejo Superior de Investigaciones Cientı´ficas, Madrid, Spain

Keywords Human dynein light chain 1 (DYNLL1) is a dimeric 89-residue that dynein light chain; glutamate channel is known to be involved in cargo binding within the dynein multiprotein homolog; NMR; protein–protein interactions complex. Over 20 protein targets, of both cellular and viral origin, have been shown to interact with DYNLL1, and some of them are transported Correspondence I. Rodrı´guez-Crespo, Departamento de in a retrograde manner along microtubules. Using DYNLL1 as bait in a Bioquı´mica y Biologı´a Molecular I, yeast two-hybrid screen with a human heart library, we identified Universidad Complutense, 28040 Madrid, GRINL1A (ionotropic N-methyl-d-aspartate-like 1A), Spain a homolog of the ionotropic glutamate receptor N-methyl d-aspartate, as a Fax: +34 913944159 DYNLL1 binding partner. Binding of DYNLL1 to GRINL1A was also Tel: +34 913944137 demonstrated using GST fusion and pepscan membranes. Progres- E-mail: [email protected] sive deletions allowed us to narrow the DYNLL1 binding region of M. Bruix, Departamento de Espectroscopı´a y Estructura Molecular, Instituto de GRINL1A to the sequence REIGVGCDL. Combining these results with Quı´mica-Fı´sica Rocasolano, CSIC, Serrano NMR data, we have modelled the structure of the GRINL1A–DYNLL1 119, 28006 Madrid, Spain complex. By analogy with known structures of DYNLL1 bound to BCL-2- Fax: +34 915642431 interacting mediator (BIM) or neuronal nitric oxide synthase (nNOS), the Tel: +34 917459511 GRINL1A peptide also adopts an extended b-strand conformation that E-mail: [email protected] expands the central b-sheet within DYNLL1. Structural comparison with the nNOS–DYNLL1 complex reveals that a glycine residue of GRINL1A (Received 12 January 2010, revised 11 March 2010, accepted 15 March 2010) occupies the conserved glutamine site within the DYNLL1 binding groove. Hence, our data identify a novel membrane-associated DYNLL1 binding doi:10.1111/j.1742-4658.2010.07649.x partner and suggest that additional DYNLL1-binding partners are present near this glutamate channel homolog.

Structured digital abstract l MINT-7713396: DYNLL1 (uniprotkb:P63167)andGRINL1A (uniprotkb:P0CAP1) bind (MI:0407) by nuclear magnetic resonance (MI:0077) l MINT-7713280, MINT-7713382: DYNLL1, (uniprotkb:P63167) physically interacts (MI:0915) with GRINL1A (uniprotkb:P0CAP1)bytwo hybrid (MI:0018) l MINT-7713416, MINT-7713439: GRINL1A (uniprotkb:P0CAP1) binds (MI:0407)toDYNLL1 (uniprotkb:P63167)bypeptide array (MI:0081) l MINT-7713307: GRINL1A (uniprotkb:P0CAP1) binds (MI:0407)toDYNLL1 (uniprotkb:P63167) by pull down (MI:0096)

Abbreviations DTT, dithiothreitol; DYNLL, dynein light chain; GKAP, guanylate kinase-associated protein; GRINL1A, ionotropic glutamate receptor N-methyl-D-aspartate-like 1A; GSH, reduced glutathione; GST, glutathione S-transferase; NMDA, N-methyl D-aspartate; nNOS, neuronal nitric oxide synthase; PAK1, p21-activated kinase 1; PSD, post-synaptic density.

2340 FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS M. F. Garcı´a-Mayoral et al. DYNLL1 and GRINL1A interaction

Introduction Dynein light chain, which has two isoforms termed receptors for either glycine or c-amino butyric acid DYNLL1 and DYNLL2 in mammals, is a small co-localize with the post-synaptic scaffolding protein dimeric protein that was initially described as a mem- gephyrin, another DYNLL1-associated protein [19,20]. ber of the myosin V [1,2] and dynein molecular motors Our proteomic studies have shown that, in rat brain, [3–5]. Originally, DYNLL1 was thought to bind to cer- DYNLL1 can associate with several NMDA receptors, tain protein cargos and transport them in a retrograde such as the NR3A-2 isoform [19]. manner along microtubules, bound to the dynein In this paper, we describe the screening of a human machinery. However, DYNLL1 can also be found in heart library using human DYNLL1 as bait. Several its soluble form and not associated with the large independent clones of the potential ionotropic gluta- dynein motor or microtubules [3,6]. In addition, cer- mate receptor-like N-methyl-d-aspartate-like 1A tain viruses are known to hijack the dynein machinery (GRINL1A) were retrieved corresponding to potential during their infective cycles, and several viral proteins DYNLL1-associated proteins. Then, using the pepscan are known to bind to DYNLL1 directly, such as rabies technique, we narrowed down the DYNLL1 binding virus P protein [7], African swine fever p54 protein [8] site, and showed that it is localized close to the cytosolic or Ebolavirus protein VP35 [9]. C-terminus of GRINL1A. Further complementary Numerous crystal and NMR atomic structures of approaches, including GST fusion, yeast two-hybrid dynein, both in the presence and absence of peptide assays and NMR titrations, confirmed the interaction ligands, are now available [10–13]. Sequence inspection and allowed fine mapping of the DYNLL1 binding site of DYNLL1-associated proteins followed by yeast within GRINL1A. Finally, all the experimental evi- two-hybrid assay, pepscan, site-directed mutagenesis dence was combined to build a structural model of and pull-down assays have revealed that DYNLL1 DYNLL1 bound to a GRINL1A peptide. associates with its target proteins essentially through the sequence motifs (R K)STQT and (K R)(D E)- ⁄ ⁄ ⁄ Results TGIQVDR [11,14,15]. In both cases, the polypeptide stretches that associate with DYNLL1 adopt an Identification of GRINL1A as a new extended conformation and form an additional DYNLL1-interacting protein b-strand that extends a pre-formed b-sheet, with the glutamine residue occupying an invariant position in Human DYNLL1 was fused in-frame to the binding the DYNLL1 binding groove. domain of GAL4 in the prey vector pGBT9, which DYNLL1 and DYNLL2 are also known to associ- was then used to transform yeast. After mating with ate on the cytosolic face of the plasma membrane, a yeast pre-transformed human heart library, mostly at the post-synaptic density, where they seem positive clones were selected and analysed by auto- to be involved in protein clustering in the proximity of mated DNA sequencing. Several overlapping clones the N-methyl d-aspartate (NMDA) receptor–post- were identified that included residues 144–463 of the synaptic density 95 (PSD-95) complex. For instance, human GRINL1A gene (Fig. 1A, accession number DYNLL2, and to a lesser extent DYNLL1, bind to a GI:238064959), which encodes a protein of 463 amino PSD-95-associated protein guanylate kinase domain- acids homologous to ionotropic glutamate receptors associated protein (GKAP) [16]. DYNLL1 also binds [21]. In order to identify potential DYNLL1 binding to neuronal nitric oxide synthase (nNOS), another sites within the GRINL1A sequence, the pepscan assay PSD-95-associated protein also present in the post-syn- was used. Overlapping dodecapeptides covering resi- aptic density [17,18]. Likewise, at inhibitory synapses, dues 156–466 of the prey protein were synthesized on

A

Fig. 1. Identification of GRINL1A as a DYNLL1-interacting protein and fine map- ping of the binding site. (A) Using a yeast two-hybrid screen, residues 144–463 of GRINL1A (C-terminal end) were shown to B bind to DYNLL1. (B) Pepscan analysis of the putative DYNLL1 binding site within the GRINL1A sequence.

FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS 2341 DYNLL1 and GRINL1A interaction M. F. Garcı´a-Mayoral et al. a cellulose membrane, which was subsequently incu- A bated with purified recombinant DYNLL1. The bind- ing assay of DYNLL1 to GRINL1A revealed three potential binding regions, which are all close to the C-terminus of the protein (Fig. 1B). The three stretches comprised spot C4 [residues ASASLRERIRHL(342– 353)], spot C30 [residues TREIGVGCDLLP(420–431)] and spot D7 [residues VMPSRNYTPYTR(441–452)]. None of the peptides has a consensus DYNLL1-bind- ing site such as KSTQT or KDTGIQVDR [14,15], and no glutamine residues were found in these three positive spots. We considered spot C4 to be a false-positive, and B discarded it on the basis that peptides enriched in His, Lys and Arg amino acids typically bind to the antibody against hexahistidine used in the develop- ment of the pepscan assay. With that in mind, we fused the 100 C-terminal amino acids of GRINL1A (residues 363–463) to glutathione S-transferase (GST), and performed an in vitro binding assay to DYNLL1 (Fig. 2A). Purified GST–GRINL1A(363- 463) was bound to a glutathione–agarose resin, and a solution of purified recombinant DYNLL1 was C passed through the column. Elution of GST– GRINL1A(363–463) from the column using reduced glutathione (GSH) allowed us to detect DYNLL1 associated with GRINL1A by Coomassie staining (Fig. 2A, lanes B and C) or by using DYNLL1- specific antibodies (data not shown). This observa- tion demonstrates that the binding site of DYNLL1 within GRINL1A is located between residues 363 and 463. DYNLL1 did not interact with GST alone bound to the glutathione–agarose resin (data not Fig. 2. Binding of DYNLL1 to a GST–GRINL1A construct and fine shown). Next, we analysed this interaction in detail mapping of the binding site. GST–GRINL1A(363-463) was using a yeast two-hybrid approach, with wild-type expressed and purified in Escherichia coli, and extensively dialy- DYNLL1 fused to the bait vector pGBT9 and vari- sed. (A) Purified protein was loaded onto a GSH–agarose resin. ous GRINL1A constructs fused to the prey vector Lane A, flow-through (unbound protein). Purified recombinant pGAD. Positive interactions were confirmed by DYNLL1 was allowed to bind to the immobilized GST– means of the X-Gal assay using double transfor- GRINL1A(363–463), and, after extensive washing, the proteins were eluted with GSH (lanes B and C). A control of DYNLL1 mants that were able to grow in the absence of Trp, only was loaded in lane D. (B) Yeast two-hybrid screen of several Leu and His. Sequential shortening narrowed the GRINL1A constructs performed with wild-type DYNLL1. Positive binding site of DYNLL1 within the GRINL1A poly- binding is represented by a ‘+’ symbol. (C) Comparison between peptide to residues 365–463 (Fig. 2B). We produced GRINL1A and a set of known peptide sequences from diverse several point mutations, including the glutamine DYNLL1-interacting targets. embedded in the consensus DYNLL1-binding motif PSQT (Q433G), and residues contained within spots D7 and D8 (R445G, Y447G and the double mutant and 423, or the positive spot C30 in the pepscan assay Q433G ⁄ R445G), and determined that these mutants (residues 420–431) is indeed the binding site and resi- still bind to DYNLL1. Finally, we shortened dues 421 and 422 (absent in the yeast two-hybrid con- GRINL1A and tested the segment 423–463 for struct) are necessary for binding. DYNLL1 binding. As shown in Fig. 2B, this construct To test this hypothesis, we decided to narrow down did not result in a positive interaction. This indicates the binding site using three partially overlapping syn- that the binding site is located between residues 365 thetic peptides in solution that were allowed to bind

2342 FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS M. F. Garcı´a-Mayoral et al. DYNLL1 and GRINL1A interaction to purified 15N-labelled recombinant DYNLL1. We protein ratios using the 15N-labelled protein and unla- tested peptides with the sequences VETREIGVGCD belled peptides. The 15N-HSQC spectrum of free DY- LLPS(418–432), LLPSQTGRTREIVMP(429–443) and NLL1 was used for spectral assignment and to SRNYTPYTRVLELTM(444–458). Strong binding monitor changes upon addition of the peptides. The was observed for the first peptide only (see below). spectral assignment confirmed that, under the experi- Interestingly, a sequence comparison with the nNOS mental conditions used, DYNLL1 is a symmetric stretch known to bind to DYNLL1 revealed a clear dimer. When DYNLL1 was titrated with peptides sequence homology, with acidic, basic and b-branched LLPSQTGRTREIVMP and SRNYTPYTRVLELTM, amino acids occupying similar positions (Fig. 2C). no changes in resonance chemical shifts or in the Remarkably, in the GRINL1A sequence, a glycine signal intensity were observed, indicating that no inter- residue appears at the position of the glutamine action takes place. However, titration with peptide residue that is present in many DYNLL1-binding VETREIGVGCDLLPS results in large chemical shift proteins. changes for numerous resonances, providing evidence for a specific interaction (Fig. 3). For simplicity, we refer to the monomers as A and B, respectively. The NMR titration of DYNLL1 with GRINL1A peptides residues participating in the interaction are: (a) I8– In order to identify specific DYNLL1-interacting E15, (b) W54–H68 and (c) Y75–F86 of monomer A, sequences in GRINL1A, we performed NMR titra- and (d) N33–I38 and (e) K43–K48 of monomer B. tions with the three overlapping synthetic peptides These regions are found mainly in the N-terminal loop spanning the residues 418–458 of GRINL1A. We (a) and the central b-sheet (b, c) of monomer A, and recorded 15N-HSQC spectra at increasing peptide: helix a2 (d, e) of monomer B.

Fig. 3. NMR titration of DYNLL1 with the GRINL1A VETREIGVGCDLLPS peptide. Superposition of 15N-HSQC spectra of DYNLL1 recorded at various DYNLL1: GRINL1A peptide VETREIGVGCDLLPS ratios during titration. The spectra correspond to free DYNLL1 (red), a 1 : 1 ratio of protein:peptide (cyan) and excess peptide (protein:peptide ratio of approximately 1 : 8) (blue). The inset indicates the chosen spec- tral region, with selected labelled residues in the slow exchange regime upon complex formation. One resonance per residue is observed for free DYNLL1 (red) and in the presence of excess peptide (blue), while two resonances appear for the under-saturated complex at the 1 : 1 ratio (cyan).

FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS 2343 DYNLL1 and GRINL1A interaction M. F. Garcı´a-Mayoral et al.

The exchange regime for the residues mentioned A above is slow in the NMR time scale. Two reso- nances per residue are observed at peptide:protein ratios of 0.5 and 1, corresponding to the free and bound forms, respectively. Their intensities vary according to the population fractions of the free and bound forms of DYNLL1 at the various peptide: protein ratios. A unique set of resonances is obser- ved at a molar ratio of 2 : 1, when all the peptide is bound to the protein dimer, confirming the 2 : 1 stoichiometry of the functional complex under these sample conditions.

Model of the DYNLL1–GRINL1A VETREIGVGCDLLPS peptide complex B Haddock docking calculations produced structures for the complex that were classified into four clusters. Analysis of the various clusters allowed us to select the one with the highest Haddock score as the most representative model of the DYNLL1–GRINL1A VETREIGVGCDLLPS peptide complex. In this clus- ter, the GRINL1A peptide is aligned anti-parallel to the interacting b3 strand of one DYNLL1 monomer, as observed in all other DYNLL1 complexes reported so far [10,11,15,22]. In two of the other clusters, an alternative peptide orientation was obtained, and the anti-parallel orientation was observed in the remaining C cluster, but the interacting surface was shifted, which disrupted the hydrogen bond network and reduced the buried interface area. Figure 4A shows the 20 lowest-energy conformers from the selected cluster modelled using the Haddock docking calculations. The rmsd values for the 20 lowest-energy conformers of the best cluster are 0.71 ± 0.11 A˚ for the backbone and 1.11 ± 0.09 A˚ for the heavy atoms. The mean total energy of the ensemble is )8146 kcalÆmol)1, being )314 kcalÆmol)1 the intermolecular contribution. The corresponding values for the van der Waals’ and electrostatic terms Fig. 4. Structural model of the DYNLL1–GRINL1A VETREIG- )1 are )887 and )8523 kcalÆmol for the complex, and VGCDLLPS peptide complex. (A) Superposition of the 20 lowest- )1 )56 and )259 kcalÆmol for the intermolecular contri- energy conformers from the selected cluster modelled using Had- butions, respectively. The mean protein–peptide buried dock docking calculations. Each DYNLL1 monomer is displayed in a surface area is 1512 A˚2. different colour (blue and pink). The VETREIGVGCDLLPS peptide of Biochemical and mutational analysis of several GRINL1A is shown in green. (B) Lowest-energy conformer of the family, showing the electrostatic surface of the protein. Positively DYNLL1 partners allowed the identification of two con- charged residues are shown in blue, negatively charged residues in sensus sequences (K ⁄ R)XTQT and G(I ⁄ V)QV(D ⁄ E), red, and uncharged residues in white. A stick representation is which contain a conserved glutamine surrounded used for the peptide. (C) Lowest-energy conformer of the family, by hydrophobic residues, as DYNLL1-interacting using ribbon and stick representations for the protein and the pep- sequences [14,15]. This glutamine forms strong hydro- tide, respectively. Selected side chains with important roles in the gen bonds with two highly conserved DYNLL1 resi- interactions are shown. Residues participating in salt-bridge interac- dues, Q35 and K36, at the beginning of helix a2. tions are labelled. In all cases, only the peptide that occupies the binding cavity in the front view of the complex is shown. Structural analysis of our best cluster shows

2344 FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS M. F. Garcı´a-Mayoral et al. DYNLL1 and GRINL1A interaction that, although this glutamine is not present in the 95, nNOS, shank and GKAP. The fact that GRINL1A GRINL1A target, the interaction still occurs in may associate with the NMDA receptor subunits and the canonical mode (Fig. 4B). The peptide lies along the that its C-terminus can bind DYNLL1 tightly shows grooves at each side of the DYNLL1 dimer interface, that this protein member of the dynein machinery can extending the central b-sheet through a hydrogen bond be targeted to the post-synaptic density through its network that connects residues H68–F62 in the swapped association with at least three independent proteins b3 strand of DYNLL1 monomer A with residues (GKAP, nNOS and GRINL1A), and agrees with the R421–C427 of the peptide in an anti-parallel manner, laminar organization revealed by negative staining and similar to that described for other complexes (Fig. 4C). immunogold labelling [26]. Due to symmetry constraints, both peptides bind to each half-site in a parallel manner, which contributes to A non-consensus sequence within GRINL1A is enhanced specificity, and interactions are similar if responsible for DYNLL1 binding in the canonical monomers A and B are exchanged; consequently inter- mode actions concerning only one peptide are discussed. The residue structurally equivalent to the glutamine In this study, we have identified GRINL1A as a novel present in the consensus sequences, G426, is close to interacting physiological partner of DYNLL1. Using the side chain of K36 in monomer B; however, the various biochemical and biophysical approaches, we different nature of glycine does not allow interactions have delimited the protein fragment responsible for this with the residues capping helix a2. Stabilizing salt- interaction, and found that it spans residues V418– bridge interactions are found between the carboxylate S432. This sequence does not contain the glutamine resi- group of E422 and the amine groups of K43 and K44 due conserved in canonical type DYNLL1-interacting of monomer B, and the side chains of R421 and D12 sequences. So far, structural information is only avail- of monomer A. These residues point towards the inte- able for two DYNLL1 complexes with peptides that rior of the peptide binding groove, facing opposite lack the conserved glutamine, the X-ray structure of sides of the central b-sheet (Fig. 4C). As mentioned p21-activated kinase 1 (PAK1) [22] and the docked previously, the construct starting at residue 423 did model of myosin Va [2], which is involved in actin- not bind to DYNLL1, and these results corroborate mediated intracellular transport. However, despite this the importance of residues 421 and 422 for binding. distinctive feature compared with many DYNLL1 tar- gets, sequence alignment of GRINL1A peptide VET- REIGVGCDLLPS with the interacting portion of the Discussion nNOS peptide, which belongs to the GIQVD type of consensus sequence, reveals a high degree of similarity Biological significance (Fig. 2C). Our NMR data provide clear evidence for Human GRINL1A was initially identified as an iono- direct interaction between DYNLL1 and GRINL1A tropic glutamate receptor-like gene that mapped to peptide VETREIGVGCDLLPS, and confirm a binding 15q22.1 [21]. Subsequent analysis of the stoichiometry of 2 : 1 (peptide:DYNLL1 dimer). This transcription unit revealed that the gene comprises at result, and the fact that only one set of resonances is least 28 exons, and its organization proved to be much observed in the spectra for free and fully bound more complex than previously anticipated [23]. DYNLL1, prove that, under the experimental condi- Sequence comparison has revealed that some tran- tions used in this study, DYNLL1 is a symmetric dimer. scripts of GRINL1A are significantly similar to those Moreover, we have mapped the interaction surface, of both the neuromuscular junction protein yotiao and which is similar to that reported for other DYNLL1 the N-termini of the NR2 and NR3 NMDA receptor complexes that do or do not include the conserved glu- subunits [23]. In addition, GRINL1A is severely down- tamine. The exchange regime of the resonances indicates regulated in patients with sporadic Alzheimer’s disease that the interaction is strong and in the sub-micromolar [24]. It has been recently reported that GRINL1A and range. Estimated dissociation constants in the low and subunits of the NMDA receptor associate at the sub-micromolar range calculated by NMR titration plasma membrane and are able to co-immunoprecipi- or isothermal titration calorimetry (ITC) have been tate [25]. Thus, GRINL1A might be enriched in the reported for a few DYNLL1–target peptide complexes, post-synaptic density, a specialized electron-dense ranging from 0.6 to 1 lm for Bim and swallow to 3 lm structure underneath the post-synaptic plasma mem- for dynein intermediate chain, 10 lm for nNOS and brane of excitatory synapses, co-localizing with the 100 lm for PAK1 [22,27,28]. In addition, a sub-micro- NMDA receptors and close to proteins such as PSD- molar dissociation constant of 0.9 lm has also been

FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS 2345 DYNLL1 and GRINL1A interaction M. F. Garcı´a-Mayoral et al. reported for the dynein–intermediate chain complex dimer surface, and extends the b-sheet dimer interface. based on changes in the Trp fluorescence [29]. Residues R421–C427 of the GRINL1A peptide adopt a b-strand conformation as assumed by residues K229–V235 of nNOS peptide, and the pattern of Comparison of the DYNLL1–GRINL1A peptide hydrogen bonds that link the peptide b-strand anti- complex with other DYNLL1–peptide complexes parallel to the swapped b3 strand of DYNLL1 is that The 3D model for the DYNLL1–GRINL1A VET- expected from sequence alignments with peptides con- REIGVGCDLLPS peptide complex obtained using the taining the GIQVD consensus motif. Haddock docking calculations is consistent with the Figure 5A shows the superposition of our model titration data and in agreement with the DYNLL1 and that of the DYNLL1–nNOS complex determined complexes of known structure. The peptide occupies by NMR spectroscopy. The backbone rmsd obtained the two identical concave channels on each side of the when the dimer is used for the superposition is 0.88 A˚, a value that indicates a high conformational similarity, and no important differences in peptide side chain con- A formations are detected. The PAK1 crystal complex (Fig. 5B) is also quite similar (rmsd value 1.41 A˚). In addition to the main backbone b-sheet hydrogen bonds, the DYNLL1–GRINL1A peptide docked structure shows several intermolecular salt-bridge inter- actions, particularly involving the N-terminus of the peptide. These interactions may increase the stability and binding affinity and play important roles in bind- ing specificity. Similar electrostatic interactions to those established by E422 of the GRINL1A peptide are maintained by the aspartate residues of nNOS and Bim peptides, and mutation of this residue in the Bim pep- tide significantly reduces the binding affinity [15,30]. Additionally, D230 of nNOS has been found to form a B hydrogen bond with the hydroxyl group of T67. This interaction is also present in other related complexes. Several studies have shown that the aspartate at position i)4 relative to the conserved glutamine is important for binding, and its substitution by other res- idues significantly decreased DYNLL1 binding [22,31]. Charge–charge interactions with D12 are similarly maintained by K229 of nNOS and K5 of Bim peptides. Hydrophobic interactions play a crucial role in high- affinity binding of DYNLL1 to most of its targets [11]. Some important contacts are: I423 with V66, L84 and F73, V425 with F62, F73, Y75 and L84, and C427 with G63, Y75, Y77 and A82. The latter are also pres- ent in the nNOS–DYNLL1 complex for the equivalent Fig. 5. Comparison of the DYNLL1–GRINL1A VETREIGVGCDLLPS I233 and V235 residues of nNOS. Other interactions, peptide modelled complex with other DYNLL1 complex structures. (A) Superposition of the lowest-energy conformers of the DYNLL1– although probably less important, are also expected to GRINL1A VETREIGVGCDLLPS peptide modelled complex and the contribute to the binding affinity, for example those solution structure of the DYNLL1–nNOS peptide complex. (B) between T420 and the backbone atoms of H68, L429 Superposition of the lowest-energy conformers of the DYNLL1– and Y77 in strand b4, and between P431 and the GRINL1A VETREIGVGCDLLPS peptide modelled complex and the aliphatic part of Q80 side chain in the loop connecting crystallographic structure of the DYNLL1–PAK1 peptide complex. In strands b4 and b5. both cases, the DYNLL1 dimer from the modelled complex is As mentioned above, conserved Q234 in nNOS and shown in pink and the DYNLL1 NMR ⁄ crystal structure is shown in blue. The GRINL1A VETREIGVGCDLLPS peptide is shown in green, other GIQVD- and KXTQT-containing peptides is and the nNOS ⁄ PAK1 peptides are shown in purple. Selected side replaced by a glycine in GRINL1A. Structural analyses chains with important roles in the interactions are shown. of known DYNLL1complexes have established the

2346 FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS M. F. Garcı´a-Mayoral et al. DYNLL1 and GRINL1A interaction importance of this glutamine, and a role in binding formed MATCHMAKER library, the yeast nitrogen base specificity has been proposed [11]. As shown here, the without amino acids (SD medium) and yeast transforma- Q ⁄ G substitution does not impair the interaction, indi- tion system 2 were purchased from BD Clontech (Moun- cating that the glutamine residue is not absolutely tain View, CA, USA). Pure synthetic peptides with the required for binding and is most likely involved in GRINL1A C-terminus sequences VETREIGVGCDLLPS increasing binding affinity. G426 of GRINL1A occu- (418–432), LLPSQTGRTREIVMP(429–443) and SRNYT- pies the equivalent position to this glutamine, which is PYTRVLELTM(444–458) were purchased from Thermo typically involved in hydrogen bonds with the nitro- Scientific. gen of K36 and the carboxylate of E35 that form an N-terminal cap for helix a2 in monomer B. Glycine, Yeast two-hybrid screen with its minimal side chain, cannot form such interac- tions in the complex described here. However, the short Saccharomyces cerevisiae strain Y190 (MATa, ura3-52, his3-200, ade2-101, lys2-801, trp1-901, leu2-3, 112, gal4D, distance between G426 and G63 ⁄ K36 (< 5 A˚) enables gal80D, cyhr2, LYS2::GAL1 -HIS3 -HIS3,MEL1; van der Waals’ contacts to be established similarly to UAS TATA URA3::GAL1 -GAL1 -lacz) was used for all yeast those maintained by G424 (G232 in nNOS) with K36 UAS TATA two-hybrid assays. The pre-transformed MATCHMAKER and the Y65 ring. The intrinsic flexibility of the glycine library is a high-complexity cDNA library cloned into a residue may facilitate subtle structural rearrangements yeast GAL4 activation domain (AD) vector and pre-trans- that enhance binding, compensating for interactions formed into Saccharomyces cerevisiae host strain Y187 lost through the Q fi G substitution. (MATa, ura3-52, his3-200, ade2-101, trp1-901, leu2-3, 112, In summary, this study has identified a peptide ) gal4D, met , gal80D, URA3::GAL1UAS-GAL1TATA-lacz,- sequence within the GRINL1A protein that adds to MEL1). All cloning was performed in DH5a Escherichia the growing list of DYNLL1 target sequences lacking coli cells, and protein expression was performed in the the conserved glutamine that is the usual hallmark of protease-deficient BL21 (DE3) E. coli strain. DYNLL1 pro- DYNLL1 binding sequences, yet binds to DYNLL1 at tein was used as bait in a yeast two-hybrid screen in order the same binding site and in similar fashion. to identify new DYNLL1-interacting proteins. We ampli- A hierarchy in the binding affinity of DLC8 targets fied full-length DYNLL1 cDNA by PCR, introducing has been proposed, with a decreasing order of affinity EcoRI and SalI sites at the 5¢ and 3¢ ends of the cDNA. depending on the presence of both the conserved gluta- This PCR product was ligated into pGBT9 plasmid (BD mine and the aspartate at position i-4, the presence of Clontech) in-frame with the DNA binding domain of the the conserved glutamine only, or the presence of the yeast transcription factor GAL4. This construction was aspartate only [22]. The GRINL1A peptide VET- used to transform the yeast strain Y190 (Mata) in order to REIGVGCDLLPS target lacks both of these residues, identify positive interactions with proteins of a human and still binds to DYNLL1 with high affinity, suggest- heart cDNA library (MATCHMAKER pre-transformed 6 ing that binding specificity is not as strong as previ- library). The library comprised > 2 · 10 independent ously thought. This is in agreement with the wide clones inserted in the pACT2 plasmid in yeast strain Y187 (Mata). Yeasts were allowed to mate, and positive colonies variety of DYNLL1-interacting partners and its bio- ) ) ) were selected by plating on Leu ⁄ Trp ⁄ His ⁄ SD plates in logical role as a multi-functional protein. Further the presence of 10 mm 3-aminotriazole (TDO plates). structural work is needed to shed light on the molecu- Approximately 150 positive colonies were picked, and the lar basis leading to DYNLL1 target recognition. interaction was confirmed by white ⁄ blue screening using X-b-Gal as the substrate. The DNA of yeasts that dis- Experimental procedures played a positive interaction was isolated and amplified by PCR using oligonucleotides that annealed in the pACT2 Materials plasmid. The PCR fragments were subsequently sequenced, and the data were analysed using public databases. 15 [ N]-labelled (NH4)Cl was purchased from Cambridge Iso- tope Laboratories Inc. (Andover, MA, USA) Glutathione Sepharose 4 Fast Flow resin was obtained from Amersham b-galactosidase assay Pharmacia Biotech (GE Healthcare Europe GmbH, Barce- In order to confirm the interaction between DYNLL1 and l l l l lona, Spain). -leucine, -tryptophan, -histidine, -lysine, various fragments of GRINL1A, we performed a colony d uracil, adenine, X-b-Gal (5-bromo-4-chloro-3-indolyl-b- - lift filter assay [32]. Colonies grown on plates without histi- galactopyranoside) and gluthatione were purchased from dine (TDO plates) were re-grown on fresh plates with histi- Sigma-Aldrich (Barcelona, Spain). 3-aminotriazole was dine at 30 C for 2 days, and then replicas of the plates obtained from FLUKA (Sigma-Aldrich). The pre-trans- were taken using with sterile Whatman filters, and yeasts

FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS 2347 DYNLL1 and GRINL1A interaction M. F. Garcı´a-Mayoral et al. were allowed to grow over the filter in the same conditions. tion with recombinant DYNLL1 (0.13 lm) was performed For the b-galactosidase assay, cells were subjected to a free- overnight at room temperature. Subsequently, the mem- ze ⁄ thaw cycle in liquid nitrogen to cause lysis. Then lysates brane was incubated for 2 h at room temperature with a were incubated with the substrate of the b-galactosidase, commercial antibody against the hexahistidine tag present X-b-Gal, at 30 C, and the appearance of blue spots was in the recombinant protein (1 : 100 000 dilution in TBS). observed after 1–6 h. Development of the membrane was performed by enhanced chemiluminiscence according to the manufacturer’s instruc- tions. The intensity of each spot was quantified using a Recombinant expression of DYNLL1 and NMR UVI-tec digital image analyser (UVItec, Cambridge, UK) sample preparation and the software UVIband V97. In all cases, spots corre- Cloning of DYNLL1 in pET-23, with a 6 · His tag at the sponding to the dodecapeptides synthesized onto the same C-terminus of the protein, and expression of the recombi- membrane were compared with each other. Controls with nant protein were performed as described previously [18]. antibody in the absence of recombinant DYNLL1 were Three samples of 15N-labelled DYNLL1 were prepared performed in order to be able to subtract non-specific to final concentrations of approximately 50 lm in 90% binding due to reactivity of the antibody against certain

H2O ⁄ 10% D2O aqueous solutions in 100 mm potassium synthetic peptides. phosphate buffer, 1 mm dithiothreitol (DTT), pH 7.0. DTT was added to prevent disulfide-mediated protein aggrega- NMR experiments tion. Concentrated solutions of the three GRINL1A pep- tides were prepared by dissolving the peptides in 100 mm Each 15N-labelled DYNLL1 protein sample was titrated potassium phosphate buffer solutions containing approxi- with an unlabelled GRINL1A peptide. Titrations were per- mately 16 mm DTT to prevent cysteines from forming formed by recording series of 15N-HSQC spectra at 25 Cin intermolecular disulfide bridges. The final concentrations of a Bruker Avance 800 MHz spectrometer (Bruker, Rheinstet- the peptide solutions were 4.6, 6.0 and 2.7 mm for ten, Germany) equipped with a z-gradient cryoprobe for the peptides VETREIGVGCDLLPS, LLPSQTGRTREIVMP free protein and for increasing peptide:protein ratios (0, 0.5, and SRNYTPYTRVLELTM, respectively. 1.0, 2.0, 4.0 and 8.0). The spectral assignment of DYNLL1 amide proton resonances was determined from published data obtained under similar sample conditions [11,12]. Cloning and expression of GST–GRINL1A Chemical shift perturbation analysis was performed using 15 1 We cloned various fragments of GRINL1A into the weighted average values for N and H chemical shifts pGEX-2T plasmid by digestion of pACT2-GRINL1A with according to the following equation: EcoRI ⁄ XhoI, and ligation into pGEX2T digested with the qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 2 15 2 same restriction endonucleases. These constructions were Ddav ¼ ðDd H) þ½ðDd N) =10: used to transform BL21 (DE3) E. coli competent cells. We purified the recombinant proteins using the GSH–Sepha- Haddock modelling rose resin, according to the manufacturer’s instructions. All recombinant proteins were dialysed against NaCl ⁄ Pi to Docking of DYNLL1 with interacting GRINL1A peptide remove GSH, and these proteins were used directly to test VETREIGVGCDLLPS was performed using the Haddock the interaction with DYNLL1 in vitro. software web portal (available at http://haddock.chem.uu. nl/services/HADDOCK/) using ambiguous interaction restraints derived from the NMR titration experiments. The Binding of recombinant DYNLL1 to peptide coordinates for DYNLL1 were taken from the PDB struc- libraries synthesized on cellulose membranes ture of DYNLL1 bound to nNOS peptide (PDB ID 1F96). Purification of recombinant DYNLL1 has been described The coordinates for the GRINL1A VETREIGVGCDLLPS previously [18] Mapping studies were performed using over- peptide were modelled using the alignment mode within lapping dodecapeptides corresponding to a fragment of Swiss Model (http://swissmodel.expasy.org/) with the PDB GRINL1A prepared by automated spot synthesis (Abimed, structure of the nNOS chain D peptide complexed with Langenfeld, Germany) on an amino-derivatized cellulose DYNLL1 (PDB ID 1F96). The two-first residues of the membrane, with their C-termini immobilized via a poly- target GRINL1A VETREIGVGCDLLPS peptide were ethylene glycol spacer and their N-termini acetylated. We excluded from this alignment and were not modelled. performed the DYNLL1 binding as previously reported Similar docking calculations were run with the full-length [14,33]. The cellulose membranes were coated with 1% GRINL1A VETREIGVGCDLLPS peptide structure built non-fat dried milk in TBS (50 mm Tris, pH 7.0, 137 mm from an anti-parallel b-sheet conformation using Pymol NaCl, 2.7 mm KCl) for 4 h at room temperature. Incuba- software tools (pymol.org). In order to simplify and speed

2348 FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS M. F. Garcı´a-Mayoral et al. DYNLL1 and GRINL1A interaction up the calculations, and due to the twofold symmetry axis 4 Harrison A & King SM (2000) The molecular anatomy of the complex, only one peptide was docked in the of dynein. Essays Biochem 35, 75–87. DYNLL1 dimer. 5 King SM (2000) The dynein microtubule motor. During the first step of rigid-body energy minimization, Biochim Biophys Acta 1496, 60–75. 1000 structures were generated, of which 200 were kept for 6 King SM, Barbarese E, Dillman JF III, Patel-King RS, the second semi-flexible simulated annealing step and final Carson JH & Pfister KK (1996) Brain cytoplasmic and

flexible water refinement. Semi-flexible residues were flagellar outer arm dyneins share a highly conserved Mr automatically defined from an analysis of intermolecular 8000 light chain. J Biol Chem 271, 19358–19366. contacts. Active and passive residues for the protein inter- 7 Raux H, Flamand A & Blondel D (2000) Interaction of action interface were selected on the basis of the chemical the rabies virus P protein with the LC8 dynein light shift perturbation data and mean residue solvent accessibil- chain. J Virol 74, 10212–10216. ity. Residue solvent accessibilities were calculated using the 8 Hernaez B, Diaz-Gil G, Garcia-Gallo M, Ignacio molmol program [34], and a 30% cut-off value was chosen Quetglas J, Rodriguez-Crespo I, Dixon L, Escribano to define the solvent-accessible surface. Selected active resi- JM & Alonso C (2004) The African swine fever virus dues for the protein were R60, N61, G63, Y65 and T67 dynein-binding protein p54 induces infected cell apopto- from monomer A, and N33, K36, K44 and K48 from sis. FEBS Lett 569, 224–228. monomer B. Passive residues were I8, K9, N10, D12, E15 9 Kubota T, Matsuoka M, Chang TH, Bray M, Jones S, and Q80 from monomer A. All residues in the peptide Tashiro M, Kato A & Ozato K (2009) Ebolavirus VP35 T420-S432 were considered active. In the case of the full- interacts with the cytoplasmic dynein light chain 8. length peptide, residues V418-S432 were considered active. J Virol 83, 6952–6956. Half of the ambiguous interaction restraints were randomly 10 Liang J, Jaffrey SR, Guo W, Snyder SH & Clardy J deleted for each docking trial. Final cluster analysis was (1999) Structure of the PIN ⁄ LC8 dimer with a bound performed to evaluate the structure quality of the docked peptide. Nat Struct Biol 6, 735–740. complexes. The patterns of intermolecular interactions for 11 Fan J, Zhang Q, Tochio H, Li M & Zhang M (2001) the full-length peptide and the modelled peptide were very Structural basis of diverse sequence-dependent target similar, and no frame shift is induced by addition of two recognition by the 8 kDa dynein light chain. J Mol Biol residues at the N-terminus. 306, 97–108. 12 Wang W, Lo KW, Kan HM, Fan JS & Zhang M (2003) Structure of the monomeric 8-kDa dynein light Acknowledgements chain and mechanism of the domain-swapped dimer This work was supported by grants from the Minis- assembly. J Biol Chem 278, 41491–41499. terio de Ciencia e Innovacio´n BFU2009-10442, 13 Williams JC, Roulhac PL, Roy AG, Vallee RB, CTQ2008-00080 ⁄ BQU and CSD2006-00023. We would Fitzgerald MC & Hendrickson WA (2007) Structural like to thank Fernando Roncal (CBM, Madrid) for and thermodynamic characterization of a cytoplasmic synthesis of the pepscan cellulose membranes and dynein light chain–intermediate chain complex. Proc Peter Rapali (Budapest, Hungary) for careful revision Natl Acad Sci USA 104, 10028–10033. of the manuscript. 14 Rodriguez-Crespo I, Yelamos B, Roncal F, Albar JP, Ortiz de Montellano PR & Gavilanes F (2001) Identifi- cation of novel cellular proteins that bind to the LC8 References dynein light chain using a pepscan technique. FEBS Lett 503, 135–141. 1 Espindola FS, Suter DM, Partata LB, Cao T, Wolenski 15 Lo KW, Naisbitt S, Fan JS, Sheng M & Zhang M JS, Cheney RE, King SM & Mooseker MS (2000) The (2001) The 8-kDa dynein light chain binds to its targets light chain composition of chicken brain myosin-Va: via a conserved (K ⁄ R)XTQT motif. J Biol Chem 276, calmodulin, myosin-II essential light chains, and 8-kDa 14059–14066. dynein light chain ⁄ PIN. Cell Motil Cytoskeleton 47, 16 Naisbitt S, Valtschanoff J, Allison DW, Sala C, Kim 269–281. E, Craig AM, Weinberg RJ & Sheng M (2000) 2 Hodi Z, Nemeth AL, Radnai L, Hetenyi C, Schlett K, Interaction of the postsynaptic density-95 ⁄ guanylate Bodor A, Perczel A & Nyitray L (2006) Alternatively kinase domain-associated protein complex with a light spliced exon B of myosin Va is essential for binding the chain of myosin-V and dynein. J Neurosci 20, 4524– tail-associated light chain shared by dynein. 4534. Biochemistry 45, 12582–12595. 17 Jaffrey SR & Snyder SH (1996) PIN: an associated pro- 3 King SM (2008) Dynein-independent functions of tein inhibitor of neuronal nitric oxide synthase. Science DYNLL1 ⁄ LC8: redox state sensing and transcriptional 274, 774–777. control. Sci Signal 1, pe51.

FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS 2349 DYNLL1 and GRINL1A interaction M. F. Garcı´a-Mayoral et al.

18 Rodriguez-Crespo I, Straub W, Gavilanes F & Ortiz 26 Valtschanoff JG & Weinberg RJ (2001) Laminar orga- de Montellano PR (1998) Binding of dynein light nization of the NMDA receptor complex within the chain (PIN) to neuronal nitric oxide synthase in the postsynaptic density. J Neurosci 21, 1211–1217. absence of inhibition. Arch Biochem Biophys 359, 27 Hall J, Hall A, Pursifull N & Barbar E (2008) Differ- 297–304. ences in dynamic structure of LC8 monomer, dimer, 19 Navarro-Lerida I, Martinez Moreno M, Roncal F, and dimmer–peptide complexes. Biochemistry 47, Gavilanes F, Albar JP & Rodriguez-Crespo I (2004) 11940–11952. Proteomic identification of brain proteins that interact 28 Song C, Wen W, Rayala SK, Chen M, Ma J, Zhang M with dynein light chain LC8. Proteomics 4, 339–346. & Kumar R (2008) Serine 88 phosphorylation of the 20 Fuhrmann JC, Kins S, Rostaing P, El Far O, Kirsch J, 8-kDa dynein light chain 1 is a molecular switch for its Sheng M, Triller A, Betz H & Kneussel M (2002) dimerization status and functions. J Biol Chem 283, Gephyrin interacts with dynein light chains 1 and 2, 4004–4013. components of motor protein complexes. J Neurosci 22, 29 Nyarko A, Hare M, Hays TS & Barbar E (2004) The 5393–5402. intermediate chain of cytoplasmic dynein is partially 21 Roginski RS, Mohan Raj BK, Finkernagel SW & disordered and gains structure upon binding to light- Sciorra LJ (2001) Assignment of an ionotropic gluta- chain LC8. Biochemistry 43, 15595–15603. mate receptor-like gene (GRINL1A) to human chromo- 30 Puthalakath H, Villunger A, O’Reilly LA, Beaumont some 15q22.1 by in situ hybridization. Cytogenet Cell JG, Coultas L, Cheney RE, Huang DC & Strasser Genet 93, 143–144. A (2001) Bmf: a proapoptotic BH3-only protein 22 Lightcap CM, Sun S, Lear JD, Rodeck U, Polenova T regulated by interaction with the myosin V actin & Williams JC (2008) Biochemical and structural char- motor complex, activated by anoikis. Science 293, acterization of the Pak1–LC8 interaction. J Biol Chem 1829–1832. 283, 27314–27324. 31 Lajoix AD, Gross R, Aknin C, Dietz S, Granier C & 23 Roginski RS, Mohan Raj BK, Birditt B & Rowen L Laune D (2004) Cellulose membrane supported peptide (2004) The human GRINL1A gene defines a complex arrays for deciphering protein–protein interaction sites: transcription unit, an unusual form of gene organiza- the case of PIN, a protein with multiple natural tion in eukaryotes. Genomics 84, 265–276. partners. Mol Divers 8, 281–290. 24 Jacob CP, Koutsilieri E, Bartl J, Neuen-Jacob E, 32 Breeden L & Nasmyth K (1985) Regulation of the yeast Arzberger T, Zander N, Ravid R, Roggendorf W, HO gene. Cold Spring Harb Symp Quant Biol 50, Riederer P & Grunblatt E (2007) Alterations in 643–650. expression of glutamatergic transporters and receptors 33 Martinez-Moreno M, Navarro-Lerida I, Roncal F, in sporadic Alzheimer’s disease. J Alzheimers Dis 11, Albar JP, Alonso C, Gavilanes F & Rodriguez- 97–116. Crespo I (2003) Recognition of novel viral sequences 25 Roginski RS, Goubaeva F, Mikami M, Fried-Cassorla that associate with the dynein light chain LC8 identified E, Nair MR & Yang J (2008) GRINL1A colocalizes through a pepscan technique. FEBS Lett 544, 262–267. with N-methyl d-aspartate receptor NR1 subunit and 34 Koradi R, Billeter M & Wu¨ thrich K (1996) molmol:a reduces N-methyl d-aspartate toxicity. Neuroreport 19, program for display and analysis of macromolecular 1721–1726. structures. J Mol Graph 14, 29–32 & 51–55.

2350 FEBS Journal 277 (2010) 2340–2350 ª 2010 The Authors Journal compilation ª 2010 FEBS