Crystal structure of the dimerization domain of human filamin A

Citations Over TimeTop 19% of 2008 papers

Abstract

Reorganization of actin cytoskeleton plays a central role in many cell functions such as the maintenance of cell shape, cell division, cell locomotion, and phagocytosis.1 Filamins are actin-crosslinking proteins that give mechanical force to cells by binding to actin filaments and making the filamentous actin form bundles or gel networks.1 Filamin is also involved in various cellular processes by anchoring transmembrane receptors such as β-integrins and scaffolding signaling molecules of diverse functions.1, 2 In human, there are three genes, FLNA, FLNB, and FLNC for their encoded protein, filamin A, B, and C (FLNa, FLNb, and FLNc), respectively. FLNA and FLNB are expressed ubiquitously and FLNC is expressed in skeletal and cardiac muscle specifically.3, 4 Mutations in human filamin genes are associated with developmental regression and defective neuronal migration.5, 6 Periventricular disease in which cerebral cortical neurons fail to migrate and oto-palato-digital (OPD) spectrum disorders that show symptoms such as cleft palate, facial malformation, and bone dysplasia are caused by mutation of FLNA.7-10 Vertebrate filamin is a dimeric protein of V-shape and each 280 kDa subunit has one actin binding domain, 24 repeating homologous rod domains, and two hinge regions.1 Each rod domain has an immunoglobulin-like (Ig-like) folding that contains β-strands forming sandwiches of β-sheets.11 Actin binding domain has two calponin homology (CH1 and CH2) domains and allows filamins to cross-link filamentous actin into gels.12 First hinge region connects domains 15 with 16, and the second hinge that is a stretch of 35 amino acids in all vertebrates connects domains 23 with 24.1 Structures of various domains of each isoform have been reported. The complex structure of FLNa domain 21 and peptide from integrin β cytoplasmic tail shows that the peptide binds to the β-strands C and D of filamin domain.13 Complex structure of FLNa domain 17 (FLNa17) and glycoprotein Ibα (GP Ibα) tail also shows that the binding partner interacts with filamin similarly.14 Structure of FLNc domain 24 (FLNc24) revealed how dimerization is formed in the carboxy terminal of filamin.15 From these structures, it is suggested that the face consisting of strands C and D of filamin domains might be the general interaction site with other ligands.13 Here, we report the structure of FLNa domain 24 (FLNa24), compare the structure with FLNc24, and discuss how dimerization is formed in FLNa24. The open reading frame (ORF) of FLNa24 (residues 2559–2647) was cloned into the expression vector pGEX-4T-1 and was expressed in the Escherichia coli BL21 (DE3) strain. The cloned FLNa24 was expressed as fusion proteins with 26 kDa glutathione S-transferase (GST). Cells were grown at 20°C until an OD600 of 0.6 and then induced with 0.5 mM IPTG for 24 h. The protein was purified using GST column, GSTPrep™ FF 16/10 (GE Healthcare Bio-Science), and thrombin protease was added (20 units of thrombin per mg of protein) and incubated for 20 h at 20°C to cleave the GST tag off. After cleavage, the GST tag was separated from target protein using GST column, Glutathione Sepharose™ 4 Fast Flow (GE Healthcare Bio-Science). The purified fractions were pooled and dialyzed into 50 mM sodium phosphate, pH 8.0, containing 1 mM EDTA, 1 mM DDT, 1 mM PMSF, and 0.5 mM sodium azide. The dialyzed sample was applied to anion exchange column, HiPrep™ 16/10 Q FF (GE Healthcare Bio-Science) and eluted with a linear gradient from 0 to 500 mM NaCl in 50 mM sodium phosphate, pH 8.0. The purified fractions were analyzed by SDS-PAGE and concentrated by ultrafiltration using Amicon® Ultra. The concentrated sample was applied to gel filtration on Superdex™ 75 (GE Healthcare Bio-Science) with final buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA). Crystals of purified FLNa24 protein were grown by hanging drop vapor diffusion method at 20°C after mixing equal volumes (2 μL) of protein solution (30 mg/mL in final buffer) and the reservoir solution. Block shaped crystals appeared in 3–5 days in optimized reservoir solution consisting of 30% PEG3350, 0.1M Tris pH 8.2, and 0.2M lithium sulfate. For data collection, crystals were equilibrated in 20% glycerol cryo-protective solutions containing reservoir buffer and flash frozen in liquid nitrogen. Diffraction data were acquired at 100K on the beamline 4A at the Pohang Light Source (PLS) in Korea. Diffraction images were taken at 1° oscillation for 360° and 180 image frames were used for data processing. The data set was processed using HKL200016 and the CCP4 suite of programs.17 Structure determination was performed by molecular replacement using the previously known human FLNc24 structure as a search model (PDB ID code: 1V05)15 with the program Phaser.18 CHAINSAW program in the CCP4 was used to mutate the search model into FLNa24 PDB file.19 Structure refinement and rebuilding were performed by the programs CNS,20 Refmac521 in CCP4, and Coot.22 Refinement was performed using all reflections in 50–1.65Å resolution shell leaving 5% of reflections for Rfree calculation. Rigid body refinement yielded Rwork and Rfree 45.3 and 44.7%, respectively. Refinement of structure was done using simulated annealing, individual B-factor refinement, and minimization implementing bulk solvent correction on all stages. Noncrystallographic symmetry (NCS) was implemented over two monomers with the restraint weight of 300 kcal/mole/Å2. Model adjustment and refinement were done iteratively, and Rwork and Rfree dropped to 30.3 and 31.9%, respectively. At this stage water molecules were picked and model was refined, and Rwork and Rfree dropped to 24.5 and 27.3%, respectively. Two sulfate ions (from the crystallization condition) were placed in extra electron density, and model was adjusted and refined. After this stage NCS restraint was released and refinement was done, yielding Rwork and Rfree values of 21.4 and 23.8%, respectively. Model was adjusted and refined further in CCP4 using Refmac5. Finally, Refmac refinement with TLS refinement23 resulted in Rwork and Rfree values of 19.2 and 21.9%, respectively. The final model had Rwork and Rfree of 19.2 and 21.9%, respectively. The data collection and refinement statistics are summarized in Table I. The final structure was verified using PROCHECK24 and all the residues were in the most favorable region, additionally allowed region, or generously allowed region of the Ramachandran plot. The atomic structure and structure factor were deposited with the PDB ID code 3CNK. Overall structure of the FLNa24 forms immunoglobulin-like fold (see Fig. 1). Strands A, B, E, and D form a β-sheet and strands C, F, and G form another β-sheet. A dimer is in the asymmetric unit and the RMSD of the two monomers is 0.3Å. The noticeable differences in main chain between the two monomers lie in six regions. These are in residues 2565–2566 (in helix), 2578–2580 (in β-strand B), 2595–2597 (in loop CD), 2610–2613 (in β-strand D and loop DE), 2622–2624 (in loop EF), and 2645–2647 (in the C-terminus). As it can be predicted from high sequence identity (69.7 and 67.4% for FLNb and FLNc, respectively), the structure is very similar to the structure of the same domain of FLNb (PDB code 2EED) and FLNc (see Fig. 1).15 The root mean square deviation (RMSD) between FLNa24 and FLNc24 is only 0.6Å (Table II). The RMSD between the dimers of FLNa24 and FLNc24 is 0.7Å. The difference in secondary structure between FLNa24 and FLNc24 lies only in the length of strand D. Strand D of FLNa24 has eight amino acids (Glu2603-Gly2610) and its amino acid length is two amino acids longer than that of FLNc24. RMSD value between FLNb24 and FLNc24 is 1.2Å with 70.5% sequence identity. Considering the small RMSD between FLNa24 and FLNc24, it is interesting to see the result of Himmel et al.30 that FLNa does not form dimer with other filamin isoforms while FLNb and FLNc may form heterodimer. Sjekloca et al.29 compared the structure of the domain 23 of FLNc with those of other similar immunoglobulin domains and showed that the structure has high structural similarity to domain 17 of FLNc and domain 21 of FLNa. When the structure of FLNa24 is compared with those of other similar immunoglobulin domains, it shows high structural similarity to the domain 23 of FLNc and domains 17, 19, and 21 of FLNa among other nondimerizing domains (Table II). Protein sequences and cartoon representation of domain 24 of filamins. A: Multiple sequence alignment of domain 24 of human filamin A, B, and C (hFLNa24, hFLNb24, and hFLNc24). Sequence alignment was generated by CLUSTALW25 and was colored by default according to the ClustalX coloring scheme.26 Secondary structure elements of FLNa24 are shown below the sequences. B: Cartoon representation of hFLNa24 and hFLNc24. Left: dimeric structure of hFLNa24, right: monomeric structure of hFLNc24. Figures were generated using PyMOL.27 The interface between two monomers in dimer structure of FLNa24 is formed by mainly strands C and D as shown in FLNc24 structure [Fig. 1(B)]. It is very similar to the dimerization interface of FLNc24 in its size, the characteristics of composing amino acid residues, and dimerization pattern.15 The interface buries 18% (977Å2) of the solvent accessible surface of the monomer. The interface (Asn2857-Gly2610) contains 42% of polar amino acids and strand D region contains more polar residues than stand C region does. Putative hydrogen bond formation between the two C strands and that between the two D strands is distinctive [Fig. 2(A,B)]. There are six hydrogen bondings between the main chains of the two D strands, and five water molecules connect the two D strands through hydrogen bonding in addition. The number of hydrogen bonding between the main chains of the two C strands in FLNa24 is same as that in FLNc24. Strand Cs form no direct hydrogen bonding between two monomers but four water molecules are involved in hydrogen bonding network. Instead, three residues (Leu2591, Val2592, and Gly2593) of the strand C are contributing to the hydrophobic interaction between two monomers. It is interesting to see that dimerization of domain 24 is formed by antiparallel association of each domain. It is speculated that the hinge region, which flanks the domain 24, gives room for the domain 24 region to be flexible so that the dimerization domains form antiparallel structure while other rod domains of whole filamin dimer run parallely along the filamin rod axis. Dimerization interface of domain 24 of filamin A. A: Hydrophobic interaction and putative hydrogen bonding between the two C strands. Hydrophobic stacking is formed by residues Leu2591, Val2592, and Gly2593. Interaction between the two C strands is featured by this hydrophobic interaction as it is revealed in domain 24 of filamin C by Pudas et al.15 Four water molecules are involved in the hydrogen bonding network. B: Putative hydrogen bonding network between the two D strands. There are six hydrogen bondings between protein residues, and additionally five water molecules are involved in the hydrogen bonding network. Figure 2(A,B) was generated using PyMOL. For clarity, protein residues are shown as sticks and water molecules are represented as black spheres. The interaction of filamin with other proteins is regulated by different mechanism such as receptor occupancy, phosphorylation, and proteolysis.2 The potential phosphorylation sites for the 11 domains of FLNc were reported previously as the conserved three residues, (Ser, Phe) Pro (Phe, Tyr, Thr).29 The putative phosphorylation site, Ser2640-Pro2641-Tyr2642 is also found in carboxy terminus of FLNa24. Different binding partners interact with domain 24 or tandem domains including domain 24 of filamins. Domain 21–24 of filamin can interact with various proteins including presenilins.31 Domain 24 of FLNa interacts with the small GTPase RalA32 and Granzyme B.33 For the human FLNc, domain 24 alone is enough and sufficient for dimerization.29 In FLNa24 structure, strands C and D are forming dimerization interface as previously shown in FLNc24 structure.15 Interestingly, in the complex structure of FLNa17 and GPIbα, it was the strands C and D of FLNa17 that were interacting with the GPIbα peptide.14 These show that domain 24 may have a role in interaction with other binding molecules in addition to the dimerization role. Elucidating how domain 24 interacts with other binding partners and which part of domain 24 is involved would give further information on the features of the residues of FLNa that are involved in the interaction. We thank the Korea Basic Science Institute (KBSI) and the National Center for Inter-University Research Facilities (NCIRF) at Seoul National University for the use of NMR spectrometers.

Related Papers

• → The filamins(2011)452 cited
• → Filamin A Regulates Cardiovascular Remodeling(2021)54 cited
• → New insights into the versatile roles of platelet FlnA(2012)41 cited
• → Filamin-A Increases the Stability and Plasma Membrane Expression of Polycystin-2(2015)17 cited
• Heterogeneity of Platelet Functional Alterations in Patients With Filamin A Mutations(2013)