Deciphering the Binding of Salicylic Acid to Arabidopsis thaliana Chloroplastic GAPDH-A1

Abstract

Salicylic acid (SA) has an essential role in the responses of plants to pathogens. SA initiates defence signalling via binding to proteins. NPR1 is a transcriptional co-activator and a key target of SA binding. Many other proteins have recently been shown to bind SA. Amongst these proteins are important enzymes of primary metabolism. This fact could stand behind SA's ability to control energy fluxes in stressed plants. Nevertheless, only sparse information exists on the role and mechanisms of such binding. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was previously demonstrated to bind SA both in human and plants. Here, we detail that the A1 isomer of chloroplastic glyceraldehyde 3-phosphate dehydrogenase (GAPA1) from Arabidopsis thaliana binds SA with a K_D of 16.7 nM, as shown in surface plasmon resonance experiments. Besides, we show that SA inhibits its GAPDH activity in vitro. To gain some insight into the underlying molecular interactions and binding mechanism, we combined in silico molecular docking experiments and molecular dynamics simulations on the free protein and protein-ligand complex. The molecular docking analysis yielded to the identification of two putative binding pockets for SA. A simulation in water of the complex between SA and the protein allowed us to determine that only one pocket-a surface cavity around Asn35-would efficiently bind SA in the presence of solvent. In silico mutagenesis and simulations of the ligand/protein complexes pointed to the importance of Asn35 and Arg81 in the binding of SA to GAPA1. The importance of this is further supported through experimental biochemical assays. Indeed, mutating GAPA1 Asn35 into Gly or Arg81 into Leu strongly diminished the ability of the enzyme to bind SA. The very same cavity is responsible for the NADP⁺ binding to GAPA1. More precisely, modelling suggests that SA binds to the very site where the pyrimidine group of the cofactor fits. NADH inhibited in a dose-response manner the binding of SA to GAPA1, validating our data.

Keywords: biacore; glyceraldehyde 3-phosphate dehydrogenase; molecular docking; molecular dynamics; protein ligand interaction; salicylic acid; surface plasmon resonance.

Figure 1

Production and purification of recombinant glyceraldehyde 3-phosphate dehydrogenase (GAPA1). In (A), a detection of GAPA1 with anti-6xHis antibodies in a soluble fraction of E. coli lysate via the peroxidase activity of secondary antibodies. In (B), a SDS-PAGE of GAPA1 purified from a soluble fraction of E. coli lysate using Ni-NTA resin. The protein band was visualised by stain-free imaging (Bio-Rad). Shading at the bottom of the track is due to the loading buffer pigments.

Figure 2

Assessment of GAPA1 binding to immobilised 3-aminoethyl salicylic acid (3-AESA) in surface plasmon resonance (SPR) assay. In (A), a range of GAPA1 concentrations were tested for their ability to bind immobilised 3-AESA. K_D of this interaction was calculated by fitting the binding profiles to a 1:1 Langmuir interaction model (Supplementary Figure S1A). Binding experiments were performed 3 times and led to a mean K_D value of 16.7 ± 5 nM. The Langmuir modelling allowed extracting the response at equilibrium (Req) values for each tested protein concentration. Req values were plotted against GAPA1 concentrations. In (B), GAPA1 ability to bind 3-AESA was assessed in the presence of an excess of SA or 4-HBA. For that, the concentration of GAPA1 was set at 200 nM. Samples were pre-incubated with an indicated concentration of SA on ice for 1 h prior to the binding test. Trend lines are generated for visualisation purposes.

Figure 3

Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) activity (NAD-dependent) of GAPA1 pre-incubated with SA or 4-hydroxybenzoic acid (4-HBA) on ice for 1 h. Graphs display averages of 3 replicates. Error bars show SD. Statistical differences were confirmed by ANOVA (panel A) and by a two-tailed Student’s t test (panel B): * (α = 0.05), *** (α = 0.001).

Figure 4

Relaxation of the protein structure in water. Root mean square deviations (RMSD) are shown for GAPA1 (GROMACS package v5.0.7). In RMSD, each structure from a trajectory is compared to a reference structure; here, the reference used is the structure of the protein after the pressure (NPT) equilibration step. The time frames corresponding to the 5 clusters are shown in colour. Clustering was performed for the simulation time 40 nsec to 600 nsec.

Figure 5

Docking positions of SA in GAPA1. The 10 best scores docking poses are shown in the average structure of cluster 3. SA is in green and the protein is in magenta. Docking was calculated by AUTODOCK VINA [33,34] implemented in PyMOL. Structures were visualised using CHIMERA [35].

Figure 6

Molecular dynamic simulations of the complex between GAPA1 and SA. In (A), the distance between SA and Asn80 was calculated throughout a course of 20 nsec simulation; the starting docking corresponds to the best scored docking of SA in the average structure of cluster 1 (SA in pocket A). In (B), the distance between SA and Trp319 was calculated throughout a course of 20 nsec simulation; the starting docking corresponds to the best scored docking of SA in the average structure of cluster 3 (SA in pocket F). In (C), a position of SA is shown immediately after the temperature and pressure equilibrations using GROMACS package v5.0.7. In (D), a position of SA in the pocket is shown after a 20 nsec simulation in water. The starting docking used to generate panels (C) and (D) corresponds to the best scored docking of SA to the pocket A in the average structure of cluster 1. Structures were visualised using CHIMERA [35] and LigPlot⁺ [36]. H-bonds and distances (in nm) are shown as dashed green lines. Red spoked arcs represent protein residues making nonbonded contacts with the ligand. In (E), the distances between SA and each residue of GAPA1 were calculated during the 10 last nanoseconds of the simulation of SA docked to pocket A of the average structure of cluster 1.

Figure 7

Twenty nsec simulations of the complexes between SA and mutated GAPA1. SA was docked in pocket A before simulation started. This docking of SA corresponds to the docking scored 1 in Table 1 for the average structure of cluster 1. The in silico mutations are performed on the average structure of cluster 1. Simulations were run 5 times from 5 independent energy minimisation steps. Criteria for assuming ligand leaving the pocket: distance to Asn80 greater than 1 nm.

Figure 8

Alignment of different GAPDHs. Regions around Asn35 and Arg81 (marked with red triangles) of AtGAPA1 are shown. Plant NADP-dependent chloroplastic GAPDHs that are involved in photosynthesis are highlighted in green. AtGAPCP2 and AtGAPCP1 are NAD-dependent enzymes that function in chloroplasts. Animal and other plant GAPDHs are the cytosolic ones, while NAD-dependent ones are involved in glycolysis. Alignment was performed by ClustalX2. Conservation score is depicted using standard colour scheme of ESPript [37]. Residues are shown on red background if strictly conserved. At, Arabidopsis thaliana; Nt, Nicotiana tabacum; Zm, Zea mays; So, Spinacia oleracea; Os, Oryza sativa; Pp, Physcomitrella patens; Ph, Petunia hybrid; Hs, Homo sapiens; Mm, Mus musculus; Oc, Oryctolagus cuniculus; Dm, Drosophila melanogaster.

Figure 9

The SA binding pocket is the pocket where the adenine part of the NAD(P) cofactor fits. Structure in magenta: the protein structure in complex with SA after a 20 nsec simulation (such as in Figure 6A); structure in blue: the crystallographic structure corresponding to PDB 3K2B; orange: NAD; green: SA. (A) Overlapping between 3K2B and our simulation. (B) Close-up of A. Structures were visualised using CHIMERA [35].

Figure 10

Comparison of the abilities of wild-type (WT) GAPA1 (red line) and point-mutated GAPA1 Asn35Gly (pink line) and GAPA1 Arg81Lys (green line) to bind immobilised 3-AESA in SPR assays. All proteins were injected in 50 nM concentration. In these sensorgrams, the signal from the mock-coupled surface was subtracted. The response values of report points (*) after the beginning of the dissociation phase were extracted. The corresponding relative responses are indicated in Resonance Units (RU). The report points are indicated by *.

Figure 11

The ability of WT GAPA1 to bind immobilised 3-AESA is inhibited in the presence of NADH. GAPA1 concentration was set at 200 nM concentration. The response values of report points (*) after the beginning of the dissociation phase were extracted. The corresponding relative responses are indicated in Resonance Units (RU). The report points are indicated by *. In these sensorgrams, the signal from the mock-coupled surface was subtracted.