Structural and functional insights into CARDs of

0 downloads 0 Views 5MB Size Report
Jun 16, 2015 - 1 Structural illustration of zCARD models showing the structural folds and suggested/predicted residues involved in NOD1/2–RIP2 CARD– ...
Molecular BioSystems View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

PAPER

Cite this: DOI: 10.1039/c5mb00212e

View Journal

Structural and functional insights into CARDs of zebrafish (Danio rerio) NOD1 and NOD2, and their interaction with adaptor protein RIP2† Jitendra Maharana,*a Budheswar Dehury,b Jyoti Ranjan Sahoo,a Itishree Jena,a Aritra Bej,a Debashis Panda,c Bikash Ranjan Sahoo,a Mahesh Chandra Patraa and Sukanta Kumar Pradhana Nucleotide-binding and oligomerization domain-containing protein 1 (NOD1) and NOD2 are cytosolic pattern-recognition receptors (PRRs) composed of an N-terminal caspase activation and recruitment domain (CARD), a central NACHT domain and C-terminal leucine-rich repeats (LRRs). They play a vital role in innate immune signaling by activating the NF-kB pathway via recognition of peptidoglycans by LRRs, and ATPdependent self-oligomerization of NACHT followed by downstream signaling. After oligomerization, CARD/s play a crucial role in activating downstream signaling via the adaptor molecule, RIP2. Due to the inadequacy of experimental 3D structures of CARD/s of NOD2 and RIP2, and results from differential experimental setups, the RIP2-mediated CARD–CARD interaction has remained as a contradictory statement. We employed a combinatorial approach involving protein modeling, docking, molecular dynamics simulation, and binding free energy calculation to illuminate the molecular mechanism that shows the possible involvement of either the acidic or basic patch of zebrafish NOD1/2–CARD/a and RIP2–CARD in CARD–CARD interaction. Herein, we have hypothesized ‘type-I’ mode of CARD–CARD interaction in NOD1 and NOD2, where NOD1/2–CARD/a involve their acidic surfaces to interact with RIP2. Asp37 and Glu51 (of NOD1) and Arg477, Arg521 and Arg529 (of RIP2) were identified to be crucial for NOD1–RIP2 interaction. However,

Received 26th March 2015, Accepted 27th May 2015

in NOD2–RIP2, Asp32 (of NOD2) and Arg477 and Arg521 (of RIP2) were anticipated to be significant for

DOI: 10.1039/c5mb00212e

for protein–protein interactions. Altogether, our study has provided novel insights into the RIP2-mediated

www.rsc.org/molecularbiosystems

CARD–CARD interaction in zebrafish NOD1 and NOD2, which will be helpful to understand the molecular basis of the NOD1/2 signaling mechanism.

downstream signaling. Furthermore, we found that strong electrostatic contacts and salt bridges are crucial

Introduction Innate immunity offers the first line of defense against notorious pathogens by selective recognition of different microbial components. The defense mechanism depends on a specific set of germ line-encoded pattern recognition receptors (PRRs) i.e., toll-like receptors (TLRs), C-type lectin receptors (CLRs), absent in melanoma (AIM)-2-like receptors (ALRs), retinoic acid inducible gene (RIG)-I like receptors (RLRs), nucleotide-binding and oligomerization domain (NOD)-like receptors (NLRs), etc. These are widely distributed in the extracellular space as well as in the cytoplasmic a

Department of Bioinformatics, Orissa University of Agriculture and Technology, Bhubaneswar-751003, Odisha, India. E-mail: [email protected] b Biomedical Informatics Centre, Regional Medical Research Centre (ICMR), Bhubaneswar-751023, Odisha, India c Distributed Information Centre, Department of Agricultural Biotechnology, Assam Agricultural University, Jorhat-785013, Assam, India † Electronic supplementary information (ESI) available. See DOI: 10.1039/c5mb00212e

This journal is © The Royal Society of Chemistry 2015

region. Among the extracellular PRRs, TLRs and CLRs are present under membrane bound conditions (in cell membrane), and some TLRs (TLR3, 7, 8 and 9) are distributed in lysosomes as well as in endosomes. Intracellular PRRs comprehend the RLR, NLR and ALR families.1,2 The NLRs display a tripartite domain architecture, with a C-terminal ligand-binding domain (LBD), a centrally positioned NACHT domain (found in NAIP, CIITA, HET-E and TP1) and an N-terminal effector binding domain (EBD). The C-terminal LBD comprises of a fluctuating number of leucine-rich repeats (LRRs), which are involved in bacterialsensing during pathogenesis, whereas, the NACHT domain aid in nucleotide dependent oligomerization and the N-terminal EBD plays a pivotal role in the interaction with downstream effectors to persuade the NOD-signaling mechanism.3–6 NOD1 and NOD2 are well characterized NLRs that recognize bacterial peptidoglycan (PGN) fragments g-D-Glu-mDAP (iE-DAP) and muramyl dipeptide (MDP), respectively.7,8 Before the invasion of pathogens, NOD proteins remain inactive by folding their

Mol. BioSyst.

View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Paper

LRRs over the NACHT domain. Upon PGN-fragment recognition by the LBD/LRRs, the conformational alternations activate the receptor by inducing nucleotide-dependent oligomerization of the NACHT domain (called ‘molecular switch’).9–11 The molecular switch in turn transfers the signal to the adaptor protein receptorinteracting serine/threonine-protein kinase 2 (RIP2) through the caspase activation and recruitment domain (CARD) via homotypic CARD–CARD interaction.3,12–16 This interaction and signal transduction activates NF-kB and induces the production of proinflammatory cytokines for removal of bacterial pathogens from the hosts.17,18 The increasing evidence in support of the dysregulation of human NOD1 (hNOD1) and hNOD2 signaling results a number of deadly diseases. Furthermore, single nucleotide polymorphisms (SNPs) in NOD1 have been associated with eczema, atopic asthma and Inflammatory Bowel’s Disease (IBD)19 and specifically, E266K SNP has been associated to an increased risk to Helicobacter pylori infected peptic ulcer patients.20 Further studies have shown that certain mutations/variants i.e., R702W, G908R and L1007fsincC in hNOD2 were associated with Chron’s disease (CD).21–23 Several amino acid substitutions in the NACHT domain of hNOD2 have resulted in Blau syndrome (BS)24 and early-onset sarcoidosis (EOS).25 The interactions of NLRs with downstream adaptor proteins are generally mediated by death fold superfamily members viz., death domains (DD), death effector domains (DED), pyrin domains (PYD) and CARDs. Always a homotypic protein–protein interaction exists in-between the proteins of this superfamily which are crucial for inflammatory signaling mechanism. The members of this family are involved in macromolecular complex formation for the enhancement of function regulation or dysregulation.9,26 Generally, the death fold superfamily members rely on three different types of interactions.27 Among them, the interaction of Apaf1-prosaspase9 (PDB ID: 3GIS28) and Pelle and Tube DD–DD complexes29 shows type-I and type-II interaction, respectively. Whereas, the type-III interaction has not been observed in dimeric complexes; but exists in the structures of the PIDDosome.30 In spite of growing evidence for understanding the importance of the NLR family, several essential aspects such as pathogen recognition, nucleotide binding and protein–protein interaction within the signaling pathways are starting to emerge. Members of the NLR family are evenly distributed in lower and higher eukaryotes including fishes. The NOD-like receptor proteins have been identified in several fishes including rohu,31,32 zebrafish,33 trout34 and catfish.35 Current studies have shown the vulnerability of NLRs during IBD predisposition in a model organism zebrafish.33 So keeping these evidences in support of zebrafish as a model organism, it has been recognized as an attractive model organism to study in vivo significance of NLR proteins. In recent years, high-throughput computational tools have paved the way to identify critical residues in iE-DAP and MDP recognition as well as a differential mode of ATP binding in zebrafish NOD1 (zNOD1) and zNOD236–38 which is well supplemented by experimental evidences from the human counterpart.7,10,11

Mol. BioSyst.

Molecular BioSystems

During the last decade, significant efforts were made to understand the mechanism of CARD–CARD interaction in NOD1/2 and RIP2 which has remained as a contradictory statement.39 Different workers have suggested the different modes of interaction that include surface charge interaction, acidic–basic surface interaction, multiple interface interactions, etc. which form the basis of CARD–CARD interaction.13–16 The unavailability of experimental structures hinders elucidating the mode of CARD–CARD interactions between NOD1/2 and RIP2. Hence, obtaining a three dimensional (3D) structure using high-throughput modeling tools establishes immense importance to understand and interpret the CARD–CARD interaction at the atomistic scale. To delineate the CARD–CARD interaction mechanism, we constructed 3D models for zNOD1–CARD, zNOD2–CARDa, and zRIP2–CARD through a combination of protein threading and a comparative modeling approach. We have established protein–protein docking refereeing the previous reports13–15 and MD simulation of these docked complexes to reveal key residues implicated in downstream signaling of innate immune response in zebrafish. Taken together, our study demonstrates structural and dynamic properties of zNOD1, zNOD2 and zRIP2 CARDs, and identifies the molecular features responsible for transmission of immune-signals through zRIP2 via CARD–CARD interaction.

Materials and methods Domain identification and model building The primary protein sequences of zNOD1, zNOD2 and zRIP2 were retrieved from the NCBI protein database (GenBank ID: XP_002665106; XP_697924 and NP_919392, respectively). The domain architectures of zNOD1, zNOD2 and zRIP2 were taken from previous studies.36–38 The secondary structures of all CARDs were predicted using PSIPRED (Fig. S1, ESI†).40 The template selections were performed using BLASTp41 search against the PDB database (http://www.rcsb.org/pdb/) to find experimental structures with maximum identity and lower E-value. The tertiary structure of zNOD1–CARD (13–103) was generated using MODELLER 9.14.42 As the BLASTp search for zNOD2–CARDa (3–96) and zRIP2–CARD (465–557) could not aid in finding an appropriate template for building homology models, the protein threading approach was followed for template identification. The primary sequences of zNOD2–CARDa and zRIP2–CARD were submitted to GeneSilico Metaserver2,43 ModLink+44 and SPARKS-X45 web servers. The resultant template which showed a good sequence to structure agreement in the results of all threading approaches was considered for model building. The resultant target-template sequence alignments were optimized by replacing gaps from structurally conserved regions to loop regions. The target-template sequence alignment and the template structure were subjected to MODELLER for generation of 50 numbers of rough models and the best model was evaluated based on the lowest discreet optimized protein energy (DOPE) score. The evaluated models were energy minimized in GROMACS 4.5.546 followed by side chain refinement in WHATIF.47

This journal is © The Royal Society of Chemistry 2015

View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Molecular BioSystems

Paper

Model quality assessment

Molecular docking of CARD–CARD complexes

Qualities of the final energy-optimized models were evaluated with respect to their stereo-chemical geometry and energy. PROCHECK48 was used to evaluate the geometry of the predicted models; ERRAT49 was used for valuing accuracy of the nonbonded atoms. ProSA50 was used to calculate the energy potential of each model. ProQ51 checked the quality of the predicted models, and the evaluation of bond lengths and bond angles were performed using MolProbity.52

The electrostatics is the major component of CARD–CARD interaction (suggested by Chou and co-authors).59 So to explore the possible mechanism behind CARD–CARD interaction, the CARDs of zNOD1 and zNOD2 were manually docked with zRIP2–CARD using Discovery Studio Visualizer (DSV) v4.0. The binding sites information was taken from Apaf1-procaspase9 interaction,28 CARD–CARD interaction of hNOD1–RIP2,13 and hNOD2–RIP2.14,15 In the first batch of interaction, the CARDs of zNODs were placed in Apaf1 and zRIP2–CARD in procaspase9, where the acidic patch of zNOD1/2–CARD/a faces the basic patch of zRIP2–CARD.13,14 Whereas in the second batch, the zRIP2–CARD was position in place Apaf1, and CARDs of zNOD1/2 were kept in the caspase9 position.15 A total of four docking complexes of zNOD1/2–CARD/a and RIP2–CARD were produced. The unrealistic molecular contacts (intermolecular bumps) between the residues of zNOD1/2 and zRIP2 CARDs formed during manual docking were resolved using DSV.

Molecular dynamics Molecular dynamics (MD) simulations of zNOD1–CARD, zNOD2– CARDa and zRIP2–CARD were carried out in GROMACS46 using the AMBER99SB-ILDN force-field53 to understand the structural stability, compactness of structure, residual fluctuation and dynamics of secondary structural elements within the protein. The simulation systems were solvated using TIP3P water models in separate cubic boxes with minimum distances of 10 Å between the protein surfaces and box edge. The simulation systems were neutralized by adding physiological ionic strengths (0.15 M) of counter ions. The atomic composition of different simulation systems have been listed in Table S1 (ESI†). The simulation systems were energy minimized using the steepest descent integrator with restraints of 1000 kJ mol 1 nm 1 force constant for a maximum of 5000 steps to remove steric conflicts and bad van der Waals contacts between side chain and main chain atoms. All the bond constructions were carried out using LINCS algorithm.54 The SETTLE algorithm was used to construct the geometry of water molecules.55 The calculation of long-range electrostatic interactions was performed using the particle mesh Ewald (PME) method. The energy minimized models were subjected for dual step position restrained equilibrium. The first round of equilibration involved an NVT simulation of 500 ps, by applying the velocity-rescaling thermostat, with the protein heavy atoms position restrained to allow for solvent molecules to reorient on the surface of the protein. NVT simulation was then followed by an NPT of 500 ps using the Parinello–Rahman barostat. The equilibrated systems were subjected for 50 ns unbiased MD simulation. The non-bonded interactions were calculated using a distance cut-off of 12 Å. The time-dependent secondary structure analysis of the models were performed using the STRIDE program integrated in the visual molecular dynamics (VMD 1.9.1).56 Two dimensional (2D) plots were generated using the Grace-5.1.23 program (http://plasma-gate. weizmann.ac.il/Grace/). Molecular visualizations were performed using PyMOL (academic license, http://www.pymol.org) and LigPlot + 1.4.4 (academic license).57 Calculation of the electrostatic potential The electrostatic potentials of zNOD1–CARD, zNOD2–CARDa and zRIP2–CARD models were calculated using the Adaptive Poisson–Boltzmann Solver (APBS).58 The calculation was performed with a grid spacing of 0.4 Å, at a temperature of 296 K, and a salt concentration of 0.15 M. The dielectric constants were set to e = 2 and e = 78 for protein and solvent, respectively.

This journal is © The Royal Society of Chemistry 2015

Molecular dynamics of CARD–CARD complexes The four different CARD–CARD complexes (two each of zNOD1–RIP2 and zNOD2–RIP2) were subjected to MD simulation in GROMACS to understand the stability of complexes, complex compactness, and molecular interactions under dynamic conditions. All the MD simulation procedures and parameters were kept the same as described for apo systems. The atomic compositions of different complex systems are provided in Table S1 (ESI†). The trajectory analysis was performed using the tools incorporated in the GROMACS suite. The g_hbond program was used for hydrogen bond (H-bond) analysis and g_dist was used for mapping the distance between the monomers as well as intermolecular atomic distances between interacting residue pairs. Contact map analysis was performed using g_mdmat to observe the average intermolecular residual contacts formed during the course of MD simulation. Principal component analysis (PCA) was performed using g_covar and g_anaeig programs to infer the global structural motions of each complex. Binding energy calculation A total of 250 snapshots of each complex were collected from the last 5 ns MD trajectory for the computation of the absolute binding free energy. The snapshots were extracted using the ‘trjconv’ tool implemented in GROMACS46 and the electrostatic properties were computed by using APBS.58 The calculation of binding energy was performed to find out energetically stable zNOD1/2–RIP2 CARD– CARD complexes by employing the MM/PBSA method60 according to the following equation. DGbinding = Gcomplex

(Gprotein + Gligand)

The details of the procedures for binding energy calculation were followed from our previous studies.36,37

Results Domain identification and modeling Domain analysis revealed that zNOD1 possesses three discrete domains viz., CARD, NACHT and LRR, whereas zNOD2 includes

Mol. BioSyst.

View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Paper

Molecular BioSystems

an additional CARD domain in the N-terminal region (i.e., CARDa and CARDb). CARDa and CARDb were connected by a 9 residues linker, which forms a tandem. The CARD domain of zRIP2 was found to be located in the C-terminal region (Fig. S2, ESI†). PSI-BLAST search for zNOD1–CARD against the PDB database displayed few hits, out of which the NMR structure of hNOD1–CARD domain (PDB ID: 2DBD) showed 54% sequence identity, and was considered for homology modeling. Due to low sequence identity with templates from PDB for zNOD2–CARDa and zRIP2–CARD, we employed a protein threading protocol for model building (adopted from our earlier study).37 The resultant templates suggested by different threading programs are listed in Table S2 (ESI†). The 3D models of zNOD2–CARDa and zRIP2– CARD were built based on the NMR structure of human ICEBERG (PDB ID: 1DGN61). The target-template alignment was manually restrained to minimize the mismatches and gaps. Validation of CARD models The energy optimized models of zNOD1–CARD, zNOD2–CARDa and zRIP2–CARD were verified stereo-chemically using various structure validation tools (see Materials and methods). The accuracy of dihedral angles (F/C) was validated using Ramachandran plot, where zNOD1–CARD, zNOD2–CARDa and zRIP2–CARD showed 95.40, 90.60 and 89.50% of residues in the most favored region with no residue in the disallowed region. The overall G-factors of all the models were found to be within the acceptable range (0.5). The specified scores of stereo-chemical properties of the torsion angles were within the acceptable range.48 The ERRAT score provided accuracy of the non-bonded atoms, where our predicted models had scored better than the acceptable value (50%).49 ProSA Z-scores of the predicted models were in agreement with the experimental PDB structures of similar sizes.50 ProQ analysis indicated that the qualities of models were ‘extremely good’ based on their LGscores.51 Bond length and bond angle analysis of the proposed models in MolProbity showed that none of the residues were having bad side chains or main chain conflicts. The details of the model validation report are provided in Table 1 and Fig. S3 (ESI†). Structural overview and electrostatic surface potentials of CARDs All the modeled CARD domains retain a similar structural fold and consist of six numbers of a-helices. The helices were closely packed,

Table 1

where the intermediate helices (a2–a5) formed an antiparallel bundle, and helices a1 and a6 positioned on top of helix a4 and a5, respectively. The helix a1 showed a sharp kink near the conserved ‘arginine’ residue (Arg22 of zNOD1; Arg11 of zNOD2; Arg477 of zRIP2) and forms two smaller helices (Fig. 1A–C). Although the CARDs retain a similar type of structural fold, they differ in surface charge distribution. Electrostatic surface potential calculations using APBS showed differential surface charge distribution in CARDs. The zNOD1–CARD and zNOD2– CARDa showed a lesser positive patch (a1 and a4) and a wider negative patch (a2, a3, a5, and a6), whereas RIP2–CARD displayed a broader positive patch (a1, a4, a5 and a6) and a smaller negative patch (a2 and a3) (Fig. 1E–G). According to previous reports, either one of the two combinations NOD1/2 and RIP2–CARD was taking part in NOD1/2 mediated CARD– CARD interaction. But, from surface electrostatics of CARDs, it can be presumed that NODs may require a negative patch to interact with RIP2. Stability parameters of CARDs Long range MD simulations were performed to explain the structural stability, compactness of structure, residual fluctuation and stability of secondary structural elements of zNOD1– CARD, zNOD2–CARDa and zRIP2–CARD models. The stability of the models was determined by calculating the backbone RMSD, compactness of structure analyzed by radius of gyration (Rg) and the residual fluctuation was observed from Ca RMSF. The RMSD of zNOD1–CARD gained stability after 5 ns and NOD2–CARDa became stable after 35 ns; however an unstable graph was observed in zRIP2–CARD up to the 50 ns time scale (Fig. 2A). The radius of gyration of zNOD1–CARD and zNOD2– CARDa retained a stable plateau, whereas an unstable gyraduis was observed in zRIP2–CARD (Fig. 2B). The Ca RMSF of zNOD1– CARD showed less fluctuation in comparison to zNOD2–CARDa and zRIP2–CARD (Fig. 2C). The C-terminal region (a6) of zNOD2– CARDa showed maximum fluctuation (B4.5 Å) whereas, minimal fluctuations were observed in a1, loops between helix a1–a2 and a3–a4 (Fig. 2D). Further, it was noticed that a1 and a6 regions of zRIP2–CARD portrayed very high fluctuation (Fig. 2E). Analysis of the secondary structure from MD trajectories using STRIDE displayed that both zNOD1–CARD and zNOD2–CARDa retained a conserved secondary structure with diminutive disruption

Model validation scores of zNOD1–CARD, zNOD2–CARDa and zRIP2–CARD

Servers PROCHECK

ERRAT ProSA ProQ MolProbity

Mol. BioSyst.

Most favored regions (%) Additionally allowed regions (%) Generously allowed regions (%) Disallowed regions (%) Overall G-factor Overall quality Z-Score LG score MaxSub Cb deviations 40.25 Å (%) Residues with bad bonds (%) Residues with bad angles (%)

zNOD1–CARD

zNOD2–CARDa

zRIP2–CARD

95.40 4.60 0.00 0.00 0.21 98.80 5.77 4.40 0.59 0.00 0.00 0.00

90.60 8.20 1.20 0.00 0.02 91.86 4.54 3.40 0.51 0.00 0.00 0.00

89.50 8.10 2.30 0.00 0.12 100.00 4.44 1.94 0.34 0.00 0.00 0.00

This journal is © The Royal Society of Chemistry 2015

View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Molecular BioSystems

Paper

Fig. 1 Structural illustration of zCARD models showing the structural folds and suggested/predicted residues involved in NOD1/2–RIP2 CARD–CARD interaction. (A) zNOD1–CARD (B) zNOD2–CARDa and (C) zRIP2–CARDa; (D) multiple sequence alignment of different CARD domains: different residue colors in the structure as well as in alignment indicate finding of different groups (red: Manon et al., 2007, green: Wagner et al., 2009, blue: Fridh and Rittinger, 2012, brown: Mayle et al., 2014 and white: present study). Red and blue color boxes indicate acidic and basic patches of zNOD1/2 or, RIP2– CARD respectively. Electrostatic surface potential calculation of zNOD1–CARD (E), zNOD2–CARDa (F) and zRIP2–CARD (G) Negative surface potential designated in red color and positive in blue surface color, the yellow lines indicate acidic/basic patches involved in CARD–CARD interaction.

Fig. 2 Stability parameters of the CARD models as a function of simulation time (50 ns). Backbone RMSD of CARDs (A); Rg of zNOD1/2 and RIP2–CARD/s (B); Ca RMSF of zNOD1–CARD (C); zNOD2–CARDa (D); and zRIP2–CARD (E). The color legend denotes different CARDs and helical positions marked in RMSF graphs.

This journal is © The Royal Society of Chemistry 2015

Mol. BioSyst.

View Article Online

Paper

Molecular BioSystems

in helix a6 and a1, respectively, whereas in zRIP2–CARD, helix a3 and a6 utterly lost the helical component (Fig. S4, ESI†).

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Residue identification and molecular docking of CARDs As evidenced from current literature and sequence-structure analysis, we expected two potential binding sites (either acidic or basic) of NOD1/2 as well as RIP2 CARDs supposed to involve in CARD–CARD interaction for activation of NF-kB signaling. The studies by Manon et al.13 and Wagner et al.14 suggested that the acidic patch of NOD1/NOD2–CARD/a interacts with the basic patch of RIP2–CARD, which contradicts with the recent observation by Fridh and Rittinger.15 In the first batch, we identified residues spanning in helix a2 and a3 of zNOD1/2– CARD/a (Asp37, Glu48, Asp49 and Asp37 in zNOD1 and Asp32, Glu43, Asp44 and Leu46 in zNOD2) that may interact with the positively charged residues (Arg477, Arg521 and Arg521) covering the helix a1 and a4 of zRIP2–CARD (Fig. 1A–C). However, in the second batch, (as reported by Fridh and Rettinger, 2012) the positively charged residues; Arg24 and Arg62 in zNOD1–CARD and Arg11 and Arg60 (helix 1 and 4) in zNOD2–CARD (basic patch) and the negatively charged residues of zRIP2–CARD; Asp494, Glu505, Asp506, Glu508 and Asp525 were assumed to create an acidic patch which may pave the way for CARD–CARD interaction (Fig. 1A–C). A sequence alignment of all CARDs (NOD1, NOD2 and RIP2) from human, mouse and zebrafish showed the probable conserved critical residues supposed to involve in CARD–CARD interaction (Fig. 1D). Molecular docking was performed manually (see Materials and methods) using DSV. Initially four numbers of complexes were created i.e., two each for zNOD1 and zNOD2, where in zNOD1/2–RIP2 Complex-I, helix a1 and a4 of zNOD1/2–CARD/a faced helix a2 and a3 of zRIP2–CARD, and in Complex-II, helix a2 and a3 of zNOD1/2–CARD/a positioned adjacent to the helix a1 and a4 of zRIP2–CARD. The residues interacting in these complexes are shown in Fig. S5 and summarized in Table S3 (ESI†).

Fig. 3 Stability parameters of zNOD1/2–RIP2 CARD–CARD complexes. RMSD of zNOD1–RIP2 (A), zNOD2–RIP2 (B), Rg of zNOD1–RIP2 (C), zNOD1–RIP2 (D), distance between zNOD1 and RIP2 during the course of simulation time (E) and distance of zNOD2 and RIP2 (F). Complex-I labeled in red while Complex-II is in blue.

Fig. 4 Structural illustration displaying the backbone superimposition of principal component 1 of zNOD1/2–RIP2 CARD–CARD complexes. (A) zNOD1–RIP2 Complex-I, (B) zNOD1–RIP2 Complex-II, (C) zNOD2– RIP2 Complex-I and (D) zNOD2–RIP2 Complex-II. The initial conformations were colored in blue and the final in red while the intermediate ones were colored grey for distinction.

Stability parameter and PCA of CARD–CARD complexes The RMSD analysis was performed to elucidate the stability of the CARD–CARD complexes during the course of simulation time. The RMSD of zNOD1–RIP2 Complex-I appeared stable just after 5 ns of MD simulation, whereas Complex-II seemed unstable (Fig. 3A). Likewise in NOD2, a similar type of graph was observed; Complex-I retained a stable graph, however Complex-II seemed unstable (Fig. 3B). The gyradius graph also suggested a similar type of results (Fig. 3C and D). The calculations of the distance between the centre of mass of two proteins suggested that Complex-I (both in NOD1 and NOD2 complexes) showed a constant distance, whereas Complex-II presented varied distance (Fig. 3E and F). To extrapolate the global motion of the CARD–CARD complexes, we performed PCA of each complex from 50 ns MD trajectories. Comparative analysis showed the global motion of Complex-I was quite less than that of Complex-II (Fig. 4) and signified that Complex-I appeared more stable than Complex-II.

Mol. BioSyst.

Contact map analysis The smallest distance between the residue-pairs of CARD–CARD simulation systems was analyzed using g_mdmat. In Complex-I of zNOD1 and zNOD2–RIP2 complexes, more numbers of residual contacts were observed in between helix a2 and a3 of zNOD–CARD and helix a1 and a4 of zRIP2–CARD (Fig. 5A and C). However, in zNOD1–RIP2 Complex-II, some weak interactions were observed in between zNOD1–CARD and zRIP2–CARD (Fig. 5B), but in zNOD2– RIP2 Complex-II, the close residual contacts had been detected in between helix a1 and a4 of zNOD2–CARDa and helix a2 and a3 of zRIP2–CARD (Fig. 5D). The intermolecular contact map analysis suggested that the interaction might be stable in the negative interface of zNOD1/2–CARD/a (spanning negative charged residues of helix a2 and a3) and the positive surface of zRIP2–CARD (positive charged residues of helix a1 and a4). The output from contact map analysis correlates with the previous reports of Manon et al.13 and Wagner et al.14

This journal is © The Royal Society of Chemistry 2015

View Article Online

Paper

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Molecular BioSystems

Fig. 5 Contact maps illustrating the intra-molecular interactions in between zNOD1/2–RIP2 CARD–CARD complexes. zNOD1–RIP2 Complex-I (A), Complex-II (B), zNOD2–RIP2 Complex-I (C) and Complex-II (D). Colored distance scale shown at the bottom indicates inter-/intra-molecular residual contact distances.

H-bond analysis and interatomic distance calculation H-bonds are the major intermolecular forces for complex stability between the protein and its binding counterpart. To infer the nature of CARD–CARD interactions, we analyzed the intermolecular H-bonds formed in between zNOD1/2–CARD/a and zRIP2–CARD as a function of simulation time. H-bond analysis indicated that Complex-I had more number of H-bonds as compared to Complex-II. In zNOD1–RIP2, Complex-I retained an average of 4.94 numbers of H-bonds, however Complex-II showed almost half (2.52) of H-bonds compared to Complex-I (Fig. 6A and B). Similarly, in zNOD2–RIP2, Complex-II shows lesser number of average H-bonds (3.70) than that of Complex-I (4.24) (Fig. 6C and D). H-bond analysis depicted the charge–charge interaction between the positive patch of zNOD1/2–CARD/a and the negative patch of zRIP2–CARD which might be indispensable for CARD–CARD interaction.

This journal is © The Royal Society of Chemistry 2015

Fig. 6 Analysis of the average number of intermolecular H-bonds in zNOD1/2–RIP2 CARD–CARD complexes during 50ns MD simulation. (A) zNOD1–RIP2 Complex-I, (B) zNOD1–RIP2 Complex-II, (C) zNOD2–RIP2 Complex-I and (D) zNOD2–RIP2 Complex-II. In both cases Complex II shows the highest stability in terms of average number of H-bonds. The red color indicates Complex-I and blue Complex-II.

Mol. BioSyst.

View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Paper

Molecular BioSystems

Fig. 7 2D representation of molecular interaction of zNOD1–RIP2 (A) and zNOD2–RIP2 (B) stable CARD–CARD complexes and distance calculation of H-bond forming residues. The 2D representation was visualized by LigPlot+ and distances were calculated by gdist program implemented in GROMACS. The numbers in the yellow box represent the corresponding H-bonds in the graph as well as in interaction figure.

The representative structures of both CARD–CARD complexes from MD trajectory showed similar numbers of H-bonds (Fig. 7). The consistency of H-bonds was evaluated by calculating the interatomic distances between the H-bond forming atoms of the respective amino acids in the CARD counterparts. In zNOD1–RIP2 complex, a total of six H-bonds were formed in between Asp37 and Glu51 (negative patch) of zNOD1–CARD and Arg477, Arg521 and Arg529 (positive patch) of zRIP2–CARD (Table 2) with an average molecular distance of 2.2 Å. As visualized (in Fig. 7A), Asp37 forms three strong H-bonds with Arg477 and Arg521. From distance calculation, it can be suggested that Asp37 may interact with Arg477 with one additional side-chain: main-chain H-bond. Glu51 formed two H-bonds with Arg529, but the interaction seemed stable after 40 ns of simulation time. Apart from H-bonds, hydrophobic interactions were also observed between

Gln33, Leu40 and Met41 of zNOD1–CARD and Ala470, Arg471, Ile473 and Ala474 of zRIP2–CARD. A total number of six H-bonds were observed between Glu28, Ser29, Asp32 and Glu38 (negative patch) of zNOD2–CARDa and Arg471, Arg477 and Arg529 (positive patch) of zRIP2–CARD (Fig. 7B). Further, we also observed two stable H-bonds between Asp32 of zNOD2–CARD, and Arg477 and Arg521 of zRIP2–CARD. In addition, one stable main-chain: sidechain H-bond was also observed in between Ser29 and Arg477. But, the interaction in between Glu28 and Arg521 being an electrostatic interaction, Glu38 may form weak salt bridges with Arg471. Moreover, some hydrophobic contacts were noticed in between Ser23, Leu35, Ala36, Trp42, Tyr45 and Arg49 of NOD2 and Gln466, Gly467, Ala470, Ile473 and Thr515 of RIP2 (Fig. 7B). Altogether, H-bonds along with hydrophobic contacts and electrostatic interactions played a vital role in CARD–CARD interaction.

Table 2 Representation of interaction analysis of zNOD1–RIP2 and zNOD2–RIP2 stable CARD–CARD complexes. The residues found crucial in this study are represented in bold font

Interaction types

zNOD1–CARD

zRIP2–CARD

Length (Å)

zNOD2–CARDa

zRIP2–CARD

Length (Å)

Hydrogen bond

Asp37:OD1 Asp37:OD2 Asp37:OD2 Glu51:OE1 Glu51:OE2 Asp37:OD2

Arg521:NH2 Arg521:NH1 Arg477:NH1 Arg529:NH2 Arg529:NH1 Arg477:CD

2.87 2.91 2.92 2.92 3.05 3.43

Glu28:OE2 Asp32:OD2 Asp32:OD2 Glu38:OE1 Ser29:O Glu38:OE2

Arg521:NH2 Arg477:NH1 Arg521:NH1 Arg471:NH2 Arg477:NH1 Arg471:NE

3.17 2.99 2.92 3.04 3.32 3.38

Hydrophobic

Leu40 Leu40 Met41

Ala470 Ile473 Ala474

4.26 4.96 4.50

Leu35 Leu35 Ala36

Ala470 Ile473 Ala474

4.23 4.95 4.14

Mol. BioSyst.

This journal is © The Royal Society of Chemistry 2015

View Article Online

Molecular BioSystems

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

RMSF analysis To explore the residual fluctuations of different CARDs in apo and holo conformations, we calculated the Ca RMSF of the whole protein and presumed interacting residues participate in CARD– CARD interaction. In zNOD1–CARD, very low fluctuation was observed in apo and Complex-I with a little difference, whereas high fluctuation was observed in a6 region in Complex-II (Fig. 8A). The interacting amino acids Asp37 and Glu51 (in Complex-I) showed a lower value than that of apo conformation which suggested the complex stability throughout the simulation. In comparison to zNOD1–CARD, zNOD2–CARDa possessed a high degree of fluctuation in a6 and connecting loop regions (Fig. 8B). Among the H-bond forming residues (Glu28, Asp32 and Glu38), Asp32 only showed lower RMSF in the holo state. The higher fluctuation of Glu28 and Glu38 may be due to the unstable

Paper

interaction with Arg471 and Arg521. Unlike zNOD1–CARD and zNOD2–CARD, zRIP2–CARD possesses a higher RMSF value. Mostly, in apo condition the RMSF seemed much higher than the holo form and in Complex-I of zNOD1/2–RIP2 complexes, zRIP2–CARD showed a comparatively lower fluctuation than Complex-II. All H-bond forming residues of zRIP2–CARD (Arg477, Arg521 and Arg529) showed lower fluctuation in the zNOD1–RIP2 complex, whereas in the zNOD2–RIP2 complex, Arg521 showed a higher RMSF value than that of apo conformation which might be due to the electrostatic attractions (or, weak H-bond formation) with Glu28. Binding energy calculation The binding energy estimation is a significant biophysical characteristic to understand the activity of bio-(macro) molecules in response to counterparts (macro as well as micro-molecules of the cellular environment that enhance or hinder certain biochemical pathways) in a complex-cellular environment. The association and disassociation of bio-(macro) molecules mediated by bonded and non-bonded contribution (in terms of energy) of both counterparts provide the information regarding enhancement or, inhibition. Here, we computed the binding energy of four CARD–CARD complexes by employing the MM/PBSA method and performed a comparative study to isolate the energetically stable complexes. In zNOD1–RIP2 complexes, Complex-I rendered a stronger binding affinity as compared to Complex-II (Table 3). Similar orders of binding parameters were also computed for zNOD2–RIP2 complexes and indicated the dependency of binding energy values on polar and non-polar parameters. Between zNOD1 and zNOD2, the difference of binding energy is very high, where it can be suggested that the zNOD1–RIP2 complex is more stable than zNOD2–RIP2. By excavating each component of binding energy (polar and non-polar contributions), we assumed that the zNOD1–RIP2 CARD–CARD interaction might be weaker than that of zNOD2–RIP2, and showed agreement with a recent study of Mayle and coauthors that single-interface binding of zNOD1–CARD might be insufficient for RIP2 mediated signal transduction.16

Discussion

Fig. 8 Graphical illustration shows Ca RMSF of different CARDs in apo and holo systems. (A) zNOD1–CARD, (B) zNOD2–CARDa and (C) zRIP2–CARD. The residual fluctuations were depicted in the bar diagram and inserted in specific graphs. The color legend represents different simulation systems.

This journal is © The Royal Society of Chemistry 2015

With the activation of NOD1 and NOD2-LRR by the bacterial cell wall components (i.e., iE-DAP and MDP, respectively) follows the nucleotide dependent oligomerizations of NACHT domains, which exposes CARDs to transmit the signals through adaptor protein RIP2 for NF-kB signaling.6 In this context, CARD–CARD interactions play a crucial role in innate-immune signaling. CARDs belong to the death fold superfamily. Generally, the interactions mediated by death fold superfamily members are of three types i.e., type-I, II and III.27 The crystal structure of Apaf1-prosaspase9 (PDB ID: 3GYS) represents the type-1 interaction,28 where the interfaces involve charge–charge interaction between helices a2 and a3 of Apaf1 and helices a1 and a4 of caspase9. Recent studies have also shown similar modes of interactions in NOD1/2–RIP2 complexes.13–16 The type-II interaction has

Mol. BioSyst.

View Article Online

Paper Table 3

Molecular BioSystems MM/PBSA binding free energy (kJ mol 1) of zNOD1–RIP2 (A) and zNOD2–RIP2 (B) CARD–CARD complexesa

Polar contribution

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

DGbind

DGcoul

Nonpolar contribution DGps

DGpolar

DGvdW

DGnps

DGnonpolar

(A) zNOD1–RIP2 CARD–CARD complexes Complex I 52.49 (0.43) 925.94 (0.40) Complex II 19.90 (0.34) 277.86 (0.30)

966.29 (0.48) 353.64 (0.37)

40.35 75.77

75.01 (0.35) 49.56 (0.30)

17.83 (0.01) 6.31 (0.01)

92.84 55.87

(B) zNOD1–RIP2 CARD–CARD complexes Complex I 4.64 (0.37) 1689.73 (0.59) Complex II 90.29 (0.39) 320.49 (0.53)

1938.95 (0.64) 601.10 (0.58)

249.22 280.61

223.36 (0.31) 170.69 (0.32)

30.50 (0.01) 19.63 (0.01)

253.85 190.32

a DGbind = binding free energy; DGcoul = electrostatic energy (coulombic term); DGps = polar solvation energy; DGpolar = polar term (DGcoul + DGps); DGvdW = van der Waals energy; DGnps = nonpolar solvation energy; DGnonpolar = nonpolar term (DGvdW + DGnps). Numbers in parentheses indicate standard errors.

been observed in between Pelle and Tube DD–DD complexes, whereas, the interaction involved helix a4 and loop between a4 and a5 of one DD and the loop between a5 and a6 and helix a6.29 The type-III interaction has not been observed in dimeric complexes but exists in the structures of the PIDDosome30 and RIG1–CARDab.62 In a recent study, Mayle and co-authors have suggested that the interaction of NOD1–RIP2 is possibly like type-III.16 The crystal structure of Apaf1-procaspase9 has revealed that the interaction is a charge–charge type and the acidic surface (helices a2 and a3) of Apaf1 interacted with the basic surface (helices a1 and a4) of caspase9.28 Initially, Inohara and co-workers suggested that NOD1 is not only involved in interaction with RIP2 but also with procaspases 1, 2, 4 and 9 from their co-immuno precipitation experiment.63 Later, the same group confirmed the only requirement of RIP2 in NOD1-mediated NF-kB signaling and denied the involvement of caspases.64 Furthermore, Tanabe and coworkers observed in their extensive mutagenesis experiment that RIP2 is also essential for NOD2based NF-kB signaling and also suggested that both the CARDs of NOD2 are crucial for interaction with RIP2.3 In 2007, Manon et al. solved the first NMR structure of NOD1–CARD and suggested that the involvement of the acidic patch of NOD1 and the basic patch of RIP2 CARDs were essential for NF-kB signaling on the basis of co-immunoprecipitation experiment and in vivo NF-kB activation assays. They used NOD1–CARD NMR structure and homology models of RIP2–CARD for their mutagenesis experiment.13 Wagner and co-authors suggested that NOD2 possesses a similar type of surface to that of NOD1, as evidenced by the yeast two-hybrid experiment.14 The findings of Fridh and Rittinger,15 in 2012, contradict the findings of Manon et al.13 and Wagner et al.14 The group stated the involvement of the acidic patch (a2 and a3) of RIP2 in CARD–CARD interaction and also suggested that mostly NOD2–CARDa governs the CARD–CARD interaction in the case of NOD2 for activation of the NF-kB signaling pathway. Again another controversial statement by Mayle et al.16 suggested that a single interface is not sufficient for NOD1–RIP2 CARD–CARD interaction and proposed two different interfaces. In spite of some substantial efforts by efficient research groups in recent years, the understanding of CARD–CARD interfaces involved in NOD1/2–RIP2 interaction is still contradictory.38

Mol. BioSyst.

Henceforth, to understand the molecular mechanism behind the biophysical basis of CARD–CARD interaction, we have employed various structural bioinformatics approaches to disseminate the possible interfaces (acidic/basic) of zNOD1/2 and zRIP2 involved. From our study, we inferred the energetic stability in acidic–basic interfaces rather than the basic–acidic interface among zNOD1/ 2–RIP2 complexes (Fig. 9A and B). The acidic surface of zNOD1/ 2–CARD and the basic surface of zRIP2–CARD may be the possible CARD–CARD interaction interface, which is in agreement with the earlier findings of Manon et al.13 and Wagner et al.14 and contradicts with the findings of Fridh and Rittinger15 and Mayle et al.16 In NOD1, Manon et al.13 have shown that acidic residues, Glu53, Asp54 and Glu56 (helix a3) of NOD1–CARD and basic residues, Arg444, Arg483 and Arg448 (helices a1 and a4) of RIP2–CARD were crucial for the interaction and the residues Leu40, Val41, Asp42 and Leu44 cannot be ruled out. Our observation in zebrafish suggests that the residues, Asp37 and Glu51 (corresponding to Asp42 and Glu56 of NOD1), form a strong saltbridge with Arg477, Arg521 and Arg529 of zRIP2 (Arg444, Arg483 and Arg488 of RIP2) (Fig. 9A). In NOD2, Wagner et al.14 proposed that the CARD–CARD interaction in NOD2 requires similar interfaces alike NOD1. Furthermore, they have also suggested that the acidic residues, Glu69, Asp70 and Glu71, in tandem CARD of NOD2 are crucial for interaction with RIP2 in an electrostatic fashion similar to the interaction of Apaf1 and caspase-9.28 Our study predicts that the strong salt-bridge between Asp32 (helix a4) of zNOD2–CARDa/ab with Arg477 and Arg521 (helices a1 and a4 of zRIP2) are crucial for interaction (Fig. 9B). As suggested (by Proell et al.9) the first and last residues of CARDs, caspases or adaptor proteins (either from acidic or basic patch) might be crucial for interaction like Apaf1-procaspase9. We found the involvement of Asp37 (of zNOD1–CARD) and Asp32 (of zNOD2–CARDa) for interaction with Arg477 (first basic residue of zRIP2–CARD). Furthermore, in zNOD1, two strong salt bridges were observed between Glu51 (of zNOD1–CARD) and Arg529 (of zRIP2–CARD), however in zNOD2, the interaction Arg529 could not be established due to the presence of ‘Leu’ in the last acidic residue position of NOD2–CARDa. Altogether, our computational analysis (in terms of RMSD, gyradius, PCA, contact-map analysis, H-bond analysis, residual fluctuation and absolute binding free energy) predicted the

This journal is © The Royal Society of Chemistry 2015

View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Molecular BioSystems

Paper

Fig. 9 Illustrations of final interactions of zNOD1–RIP2 (A) and zNOD2–RIP2 (B) CARD–CARD stable complexes. The interacting residues represented in a colored stick model (red: NOD1/2–CARD/a and blue: RIP2–CARD) and green dotted lines indicate the H-bonds.

involvement of the acidic patch of zNOD1–CARD and zNOD2– CARDa and the basic patch of zRIP2–CARD are essential for zNOD1/2-mediated CARD–CARD interaction. In the current scenario, the in silico approaches viz., protein modeling, docking and MD simulation have revolutionized the understanding of molecular interactions between key biomolecules and their downstream ligands involved in important biochemical pathways. Although these computational methodologies are quite useful and have enormously contributed in providing atomistic insights to the biochemical interactions, the accuracy of such prediction depends on suitable computer algorithms, scoring functions and scalable hardware.65 In addition, the accuracy of docking predictions primarily depends on the initial conformations. Therefore, to understand the mode of CARD–CARD interactions, we docked both the CARDs manually in reference to previous studies13–16 and experimental structures of Apaf1-procaspase928 and the docked complexes were optimized using long term MD simulations which have resulted in achieving a desired output to a greater extent. Although, our study is purely a computational one and requires experimental validation in the near future, the findings will be helpful to understand the CARD– CARD interaction in NOD signaling pathway.

Conclusion We have employed several in silico methodologies to illuminate the molecular interactions and possible binding interfaces of CARDs in zNOD1/2–RIP2 mediated CARD–CARD interaction involved in the NOD signaling pathway. The 3D structure prediction of CARDs of zNOD1, zNOD2 and zRIP2 was performed using a combined approach of protein threading and homology modeling. The dynamic behavior of CARDs in the apo and holo state was studied in aqueous solution using MD simulations. Herein, our study

This journal is © The Royal Society of Chemistry 2015

highlights that the CARD–CARD interaction is driven by an electrostatic component, and both zNOD1 and zNOD2 use an acidic interface to bind zRIP2. In addition, we have hypothesized that the conserved residue of helix a4 (Asp37, Glu51 in zNOD1 and Asp32 in zNOD2) might play a crucial role in interaction and conserved ‘Arg’ of RIP2 (Arg477, Arg521 and Arg529 for zNOD1 and Arg477 and Arg521 for zNOD2) are vital in mediating the CARD–CARD interaction in zebrafish. Although, our prediction is based on existing literature (which still remains contradictory) that needs further experimental validation for legitimacy in the near future. In conclusion, our study has demonstrated the possible biophysical and molecular basis of CARD–CARD interaction that will be useful for understanding NOD-mediated NF-kB signaling in zebrafish as well as in the development of novel therapeutics.

Competing interest The authors have declared that no competing interests exist.

Acknowledgements The authors are thankful to Dr Sachinandan De, Principal Scientist, ABTC, National Diary Research Institute, Karnal, Haryana, for his suggestions and helpful discussions during the manuscript preparation. The authors also thanks Dr Mahendra Kumar Modi, Coordinator, Distributed Information Centre, Department of Agricultural Biotechnology, Assam Agricultural University, Jorhat, Assam, for rendering computational facility.

References 1 S. Akira, S. Uematsu and O. Takeuchi, Cell, 2006, 124, 783–801. 2 O. Takeuchi and S. Akira, Cell, 2010, 140, 805–820.

Mol. BioSyst.

View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Paper

3 T. Tanabe, M. Chamaillard, Y. Ogura, L. Zhu, S. Qiu, J. Masumoto, P. Ghosh, A. Moran, M. M. Predergast, ˜ez, EMBO ´n G. Tromp, C. J. Williams, N. Inohara and G. Nu J., 2004, 23, 1587–1597. ˜ ez, ´n 4 N. Inohara, M. Chamaillard, C. Mc Donald and G. Nu Annu. Rev. Biochem., 2005, 74, 355–383. ˜ ez, Immunity, ´n 5 T. D. Kanneganti, M. Lamkanfi and G. Nu 2007, 27, 549–559. 6 J. M. Wilmanski, T. Petnicki-Ocwieja and K. S. Kobayashi, J. Leukocyte Biol., 2008, 83, 13–30. 7 H. Laroui, Y. Yan, Y. Narui, S. A. Ingersoll, S. Ayyadurai, M. A. Charania, F. Zhou, B. Wang, K. Salaita, S. V. Sitaraman and D. Merlin, J. Biol. Chem., 2011, 286, 31003–31013. 8 C. L. Grimes, Z. LdeAriyananda, J. E. Melnyk and E. K. O’Shea, J. Am. Chem. Soc., 2012, 134, 13535–13537. 9 M. Proell, S. J. Riedl, J. H. Fritz, A. M. Rojas and R. Schwarzenbacher, PLoS One, 2008, 3, e2119. 10 B. Zurek, M. Proell, R. N. Wagner, R. Schwarzenbacher and T. A. Kufer, Innate Immun., 2012, 18, 100–111. 11 J. Mo, J. P. Boyle, C. B. Howard, T. P. Monie, B. K. Davis and J. A. Duncan, J. Biol. Chem., 2012, 287, 23057–23067. 12 P. Rosenstiel, K. Huse, A. Till, J. Hampe, S. Hellmig, C. Sina, S. Billmann, O. von Kampen, G. H. Waetzig, M. Platzer, D. Seegert and S. Schreiber, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 3280–3285. ˜ ez, J. P. Simorre and S. Cusack, ´n 13 F. Manon, A. Favier, G. Nu J. Mol. Biol., 2007, 365, 160–174. 14 R. N. Wagner, M. Proell, T. A. Kufer and R. Schwarzenbacher, PLoS One, 2009, 4, e4931. 15 V. Fridh and K. Rittinger, PLoS One, 2012, 7, e34375. 16 S. Mayle, J. P. Boyle, E. Sekine, B. Zurek, T. A. Kufer and T. P. Monie, J. Biol. Chem., 2014, 2289, 22900–22914. 17 M. Hasegawa, Y. Fujimoto, P. C. Lucas, H. Nakano, K. Fukase, ˜ez and N. Inohara, EMBO J., 2008, 27, 373–383. ´n G. Nu 18 H. L. Rosenzweig, K. T. Galster, S. R. Planck and J. T. Rosenbaum, Invest. Ophthalmol. Visual Sci., 2009, 50, 1746–1753. 19 S. Weidinger, N. Klopp, L. Rummler, S. Wagenpfeil, N. Novak, H. J. Baurecht, W. Groer, U. Darsow, J. Heinrich, A. Gauger, T. Schafer, T. Jakob, H. Behrendt, H. E. Wichmann, J. Ring and T. Illig, J. Allergy Clin. Immunol., 2005, 116, 177–184. 20 P. Hofner, Z. Gyulai, Z. F. Kiss, A. Tiszai, L. Tiszlavicz, ´th, D. Szo ˜ke, B. Molna ´r, J. Lonovics, Z. Tulassay and G. To ´ndi, Helicobacter, 2007, 12, 124–131. Y. Ma 21 M. Chamaillard, S. E. Girardin, J. Viala and D. J. Philpott, Cell. Microbiol., 2003, 5, 581–592. 22 J. Hampe, A. Franke, P. Rosenstiel, A. Till, M. Teuber, K. Huse, M. Albrecht, G. Mayr, F. M. De La Vega, J. Briggs, ¨sler, B. Sipos, ¨nther, N. J. Prescott, C. M. Onnie, R. Ha S. Gu ¨lsch, T. Lengauer, M. Platzer, C. G. Mathew, M. Krawczak U. R. Fo and S. Schreiber, Nat. Genet., 2007, 39, 207–211. 23 R. K. Weersma, A. Zhernakova, I. M. Nolte, C. Lefebvre, J. D. Rioux, F. Mulder, H. M. van Dullemen, J. H. Kleibeuker, C. Wijmenga and G. Dijkstra, Am. J. Gastroenterol., 2008, 103, 621–627.

Mol. BioSyst.

Molecular BioSystems

24 C. Miceli-Richard, S. Lesage, M. Rybojad, A. M. Prieur, ¨fner, M. Chamaillard, H. Zouali, S. Manouvrier-Hanu, R. Ha G. Thomas and J. P. Hugot, Nat. Genet., 2001, 29, 19–20. 25 N. Kanazawa, I. Okafuji, N. Kambe, R. Nishikomori, M. Nakata-Hizume, S. Nagai, A. Fuji, T. Yuasa, A. Manki, Y. Sakurai, M. Nakajima, H. Kobayashi, I. Fujiwara, H. Tsutsumi, A. Utani, C. Nishigori, T. Heike, T. Nakahata and Y. Miyachi, Blood, 2005, 105, 1195–1197. 26 F. Di Virgilio, Pharmacol. Rev., 2013, 65, 872–905. 27 K. Kersse, M. J. Bertrand, M. Lamkanfi and P. Vandenabeele, Cytokine Growth Factor Rev., 2011, 22, 257–276. 28 H. Qin, S. M. Srinivasula, G. Wu, T. Fernandes-Alnemri, E. S. Alnemri and Y. Shi, Nature, 1999, 399, 549–557. 29 T. Xiao, P. Towb, S. A. Wasserman and S. R. Sprang, Cell, 1999, 99, 545–555. 30 H. H. Park, Y. C. Lo, S. C. Lin, L. Wang, J. K. Yang and H. Wu, Annu. Rev. Immunol., 2007, 25, 561–586. 31 B. R. Sahoo, B. Swain, M. R. Dikhit, M. Basu, A. Bej, P. Jayasankar and M. Samanta, Appl. Biochem. Biotechnol., 2013, 170, 1282–1309. 32 J. Maharana, B. Swain, B. R. Sahoo, M. R. Dikhit, M. Basu, A. S. Mahapatra, P. Jayasankar and M. Samanta, Fish Physiol. Biochem., 2012, 39, 1007–1023. 33 S. H. Oehlers, M. V. Flores, C. J. Hall, S. Swift, K. E. Crosier and P. S. Crosier, Dis. Models & Mech., 2011, 4, 832–841. 34 M. Chang, T. Wang, P. Nie, J. Zou and C. J. Secombes, Fish Shellfish Immunol., 2011, 30, 118–127. 35 Z. Sha, J. W. Abernathy, S. Wang, P. Li, H. Kucuktas, H. Liu, E. Peatman and Z. Liu, Dev. Comp. Immunol., 2009, 33, 991–999. 36 J. Maharana, B. R. Sahoo, A. Bej, M. C. Patra, B. Dehury, G. K. Bhoi, S. K. Lenka, J. R. Sahoo, A. K. Rout and B. K. Behera, Mol. BioSyst., 2014, 10, 2942–2953. 37 J. Maharana, M. C. Patra, B. C. De, B. R. Sahoo, B. K. Behera, S. De and S. K. Pradhan, J. Mol. Recognit., 2014, 27, 260–275. 38 J. Maharana, B. R. Sahoo, A. Bej, I. Jena, A. Parida, J. R. Sahoo, B. Dehury, M. C. Patra, S. R. Martha, S. Balabantray, S. K. Pradhan and B. K. Behera, PLoS One, 2015, 10, e0121415. 39 J. P. Boyle, R. Parkhouse and T. P. Monie, Open Biol., 2014, 4, 140178. 40 D. W. Buchan, F. Minneci, T. C. Nugent, K. Bryson and D. T. Jones, Nucleic Acids Res., 2013, 41, W349–W357. 41 S. F. Altschul, W. Gish, W. Miller, E. W. Myers and D. J. Lipman, J. Mol. Biol., 1990, 215, 403–410. 42 N. Eswar, B. Webb, M. Marti-Renom, M. S. Madhusudhan, D. Eramian, M. Y. Shen, U. Pieper and A. Sali, Curr. Protoc. Protein Sci., 2007, 2, 1–30. 43 M. A. Kurowski and J. M. Bujnicki, Nucleic Acids Res., 2003, 31, 3305–3307. 44 O. Fornes, R. Aragues, J. Espadaler, M. A. Marti-Renom, A. Sali and B. Oliva, Bioinformatics, 2009, 25, 1506–1512. 45 Y. Yang, E. Faraggi, H. Zhao and Y. Zhou, Bioinformatics, 2011, 27, 2076–2082. ´ll, R. Schulz, P. Larsson, P. Bjelkmar, R. Apostolov, 46 S. Pronk, S. Pa M. R. Shirts, J. C. Smith, P. M. Kasson, D. van der Spoel, B. Hess and E. Lindahl, Bioinformatics, 2013, 29, 845–854.

This journal is © The Royal Society of Chemistry 2015

View Article Online

Published on 16 June 2015. Downloaded by University of Iowa on 18/06/2015 03:58:06.

Molecular BioSystems

47 G. Chinea, G. Padron, R. W. Hooft, C. Sander and G. Vriend, Proteins, 1995, 23, 415–421. 48 R. A. Laskowski, M. W. MacArthur, D. S. Moss and J. M. Thornton, J. Appl. Crystallogr., 1993, 26, 283–291. 49 C. Colovos and T. O. Yeates, Protein Sci., 1993, 2, 1511–1519. 50 M. Wiederstein and M. J. Sippl, Nucleic Acids Res., 2007, 35, W407–W410. 51 B. Wallner and A. Elofsson, Protein Sci., 2003, 12, 1073–1086. 52 V. B. Chen, W. B. Arendall 3rd, J. J. Headd, D. A. Keedy, R. M. Immormino, G. J. Kapral, L. W. Murray, J. S. Richardson and D. C. Richardson, Acta Crystallogr., Sect. D: Biol. Crystallogr., 2010, 66, 12–21. 53 K. Lindorff-Larsen, S. Piana, K. Palmo, P. Maragakis, J. L. Klepeis, R. O. Dror and D. E. Shaw, Proteins, 2010, 78, 1950–1958. 54 B. Hess, H. Bekker, H. J. C. Berendsen and J. Fraaije, J. Comput. Chem., 1997, 18, 1463–1472. 55 S. Miyamoto and P. A. Kollman, J. Comput. Chem., 1992, 13, 952–962. 56 W. Humphrey, A. Dalke and K. Schulten, J. Mol. Graphics, 1996, 14, 33–38.

This journal is © The Royal Society of Chemistry 2015

Paper

57 R. A. Laskowski and M. B. Swindells, J. Chem. Inf. Model., 2011, 51, 2778–2786. 58 N. A. Baker, D. Sept, S. Joseph, M. J. Holst and J. A. Mc Cammon, Proc. Natl. Acad. Sci. U. S. A., 2011, 98, 10037–10041. 59 J. J. Chou, H. Matsuo, H. Duan and G. Wagner, Cell, 1998, 94, 171–180. 60 D. Spiliotopoulos, A. Spitaleri and G. Musco, PLoS One, 2012, 7, e46902. 61 E. W. Humke, S. K. Shriver, M. A. Starovasnik, W. J. Fairbrother and V. M. Dixit, Cell, 2000, 103, 99–111. 62 A. Peisley, B. Wu, H. Xu, Z. J. Chen and S. Hur, Nature, 2014, 509, 110–114. 63 N. Inohara, T. Koseki, L. del Peso, Y. Hu, C. Yee, S. Chen, ˜ez, J. Biol. ´n R. Carrio, J. Merino, D. Liu, J. Ni and G. Nu Chem., 1999, 274, 14560–14567. 64 N. Inohara, T. Koseki, J. Lin, L. del Peso, P. C. Lucas, F. F. Chen, Y. Ogura and G. Nunez, J. Biol. Chem., 2000, 275, 27823–27831. 65 E. Yuriev, M. Agostino and P. A. Ramsland, J. Mol. Recognit., 2011, 24, 149–164.

Mol. BioSyst.