Molecular characterization, modeling, and docking analysis of late phytic acid biosynthesis pathway gene, inositol polyphosphate 6-/3-/5-kinase, a potential candidate for developing low phytate crops

Abstract

The coding sequence of inositol polyphosphate 6-/3-/5-kinase (GmIPK2) gene was identified and cloned from popular Indian soybean cultivar Pusa-16. The clone was predicted to encode 279 amino acids long, 30.97 kDa protein. Multiple sequence alignment revealed an inositol phosphate-binding motif, PxxxDxKxG throughout the IPK2 sequences along with other motifs unique to inositol phosphate kinase superfamily. Eight α-helices and eight β-strands in antiparallel β-sheets arrangement were predicted in the secondary structure of GmIPK2. The temporal analysis of GmIPK2 revealed maximum expression in the seed tissues during later stages of development while spatially the transcript levels were lowest in leaf and stem tissues. Endosperm-specific cis-regulatory motifs (GCN4 and Skn_1) which support high levels of expression, as observed in the developing seeds, were detected in its promoter region. The protein structure of GmIPK2 was modeled based on the crystal structure of inositol polyphosphate multikinase from Arabidopsis thaliana (PDB:4FRF) and subsequently docked with inositol phosphate ligands (PDB: 5GUG-I3P and PDB: 4A69-I0P). Molecular dynamics (MD) simulation established the structural stability of both, modeled enzyme and ligand-bound complexes. Docking in combination with trajectory analysis for 50 ns MD run confirmed the participation of Lys105, Lys126 and Arg153 residues in the formation of a network of hydrogen bonds to stabilize the ligand-receptor interaction. Results of the present study thus provide valuable information on structural and functional aspects of GmIPK2 which shall assist in strategizing our long-term goal of achieving phytic acid reduction in soybean by genetic modification of its biosynthetic pathway to develop a nutritionally enhanced crop in the future.

Electronic supplementary material

The online version of this article (10.1007/s13205-018-1343-7) contains supplementary material, which is available to authorized users.

Introduction

Soybean is a phenomenal crop with a unique nutrient profile. It is the only plant source which provides a complete protein with quality equivalent to animal protein yet its human consumption is found to be very limited, even in the countries where its production is very high. Besides, majority of its uses in the food industry owe primarily to its functional properties rather than nutritional gain. The above stated can be attributed to the presence of a variety of bioactive components including phytic acid (PA), saponins, protease inhibitors, isoflavones, lectins, etc. Amongst these, PA is probably the most well-known antinutrient that is found in abundance in cereal grains and legumes. It is a saturated cyclic acid whose metabolism has been substantiated to regulate phosphorous and myo-inositol homeostasis. Due to its highly negative chemical structure, it forms phytate–mineral–protein complexes which result in reduced bioavailability of phosphorus, other associated minerals and proteins and thus contribute to the debatable health effects associated with its consumption. Therefore, in a need to reduce the level of this compound several approaches were undertaken which blocked its constitutive synthesis, but often resulted in crop with deplorable agronomic performance (Bilyeu et al. 2008; Cichy and Raboy 2009; Feng and Yoshida 2004; Kuwano et al. 2009; Nunes et al. 2006). Recently trending reverse genetic approaches that target tissue-specific knockdown provide a promising alternative to eliminate this negative impact (Ali et al. 2013a, b). Research is thus underway to characterize PA biosynthetic pathway enzymes (Josefsen et al. 2007; Krishnan et al. 2015; Stiles et al. 2008; Sun et al. 2007; Sweetman et al. 2006, 2007) and study their expression patterns (Bhati et al. 2014; Fileppi et al. 2010; Suzuki et al. 2007) to generate adequate information for developing lpa soybean through metabolic engineering.

PA biosynthesis follows a succession of phosphorylation steps (Brearley and Hanke 1996a, b). One of its late pathway enzyme inositol polyphosphate 6-/3-/5-kinase (IPK2, EC 2.7.1.151) is a multiple specificity enzyme which phosphorylates the D-6, D-3, and/or D-5 positions on a variety of inositol polyphosphate substrates (InsP3/InsP4/InsP5) (Frederick et. al. 2005) that makes it a key enzyme in regulating PA turnover and is thus a crucial target for perturbing PA dynamics (Stevenson-Paulik et al. 2005). The enzyme has previously been studied in Arabidopsis (Stevenson-Paulik et al. 2002), yeast (Saiardi et al. 2000; York et al. 1999), mouse (Frederick et al. 2005) and human (Majerus 1992, 1996). A study has also been reported in soybean (Stiles 2007), however, it discusses only the gene’s expression and enzyme kinetics. Therefore, with the intention to extend the current knowledge on its transcript expression (both spatial and temporal) as well as to understand its structure, biochemical features and evolutionary history, we investigated this gene in soybean and compared it with those of another 19 homologs in plants through various bio-computational tools. Furthermore, to elucidate complete molecular and biochemical mechanisms regulating PA synthesis as well as inositol metabolism and signalling in plants, an insight into IPK2 protein’s structural features is required. Very less information regarding the three-dimensional structure of plant IPK2 is available. To facilitate the same, we determined the three-dimensional (3D) model of soybean GmIPK2 protein through homology modeling and performed its docking simulation with the substrate [1D-myo-inositol 1,4,5-trisphosphate (PDB: 5GUG-I3P) and 1D-myo-inositol-1,4,5,6-tetrakisphosphate (PBD: 4A69-I0P)] molecules.

Through this work, we made an effort to study IPK2 gene to lay a groundwork for achieving the main goal of our work which is, to develop low phytic acid soybean, a nutritionally and agriculturally important crop.

Materials and methods

Plant material

Field grown soybeans (Glycine max [L.] Merr. cv. Pusa-16) were procured from Division of Genetics, IARI, New Delhi for use in this study. For spatial and temporal expression profiling, root, stem, leaf, and flower tissues were collected from 30-day-old plants while developing seeds were collected regularly after flowering until maturation and sorted based on their sizes. After collection, the tissue samples were immediately frozen in liquid nitrogen and stored at − 80 °C until used.

PCR amplification, cloning and sequencing of partial GmIPK2 sequence

Total RNA was isolated from soybean seeds 8 mm in size using TRIzol reagent method (Invitrogen, USA). Approximately, 1 µg of isolated total RNA was used for single stranded cDNA synthesis by reverse transcription with RevertAid™ H Minus First Strand cDNA synthesis kit (ThermoFisher Scientific, USA). Prior to cDNA synthesis, the RNA was treated with RNase-free DNase I (Thermo Scientific, USA) for removal of any residual genomic DNA. Polymerase chain reaction (PCR) amplification was performed using 0.2 µg of the initial RT reaction optimized at 94 °C for 4 min; 35 cycles of 94 °C for 30 s, 62.5 °C for 30 s, 72 °C for 30 s; 72 °C for 10 min, using the GmIPK2 specific primers: IPK2F 5′-ATGCTCAAGATCCCGGAG-3′ and IPK2R 5′-CAGTTAGTCTGCGACACTAATTCAAGC-3′. On completion, 2 µl of amplified product was analyzed by electrophoresis on 1% agarose gel and purified using gel/PCR DNA Fragments Extraction Kit (DF100) by Geneaid. The purified amplicon was subsequently ligated in pGEM®-T Easy vector (Promega, USA) by TA cloning following protocol described in the manual and used directly for transformation of competent Escherichia coli DH5α cells. The recombinant plasmids were identified by blue-white screening and confirmed by restriction digestion with EcoRI to release the amplicon fragment. The nucleotide sequences of GmIPK2 thus isolated was determined by DNA sequence analysis of overlapping plasmid clones using universal primers (SP6 and T7) on an automated sequencer (ABI 3730xl DNA Analyzer, USA). The nucleotide sequence data was submitted to INSDC database GenBank.

Gene expression analysis by semi-quantitative reverse transcription PCR and quantitative real-time PCR

To analyze GmIPK2 gene expression in different soybean tissues, we first performed semi-quantitative reverse transcription PCR (RT-PCR) to obtain an expression pattern and then further estimated the transcript levels by quantitative real-time PCR (qRT-PCR). For RT-PCR, first strand cDNA was synthesized from total RNA using RevertAid™ H Minus First strand cDNA Synthesis Kit (Thermo Scientific, USA) and PCR was subsequently performed using the same pair of GmIPK2 specific primers and thermal cycling conditions which were described for its cloning. We further monitored the quantitative amplification of GmIPK2 by performing qRT-PCR analysis on a PikoReal 96 Real-Time PCR platform (ThermoFisher Scientific, USA). The reactions were set up using DyNAmo Flash SYBR Green qPCR Kit (Thermo Scientific, USA) with cDNA first strands as the template DNA. The expression of GmIPK2 was normalized to an endogenous control, the housekeeping gene phosphoenolpyruvate carboxylase (PEPCo) (Sugimoto et al. 1992; Tuteja et al. 2004). The primers for the gene (qIPK2F 5′-CGCGGATCCGCGTTGCAGAAGCTCAAG-3′ and qIPK2R 5′-TCCCCGCGGGGAGCGACACTAATTCAAG-3′) and the internal control (qPEPCoF 5′-CATGCACCAAAGGGTGTTTT-3′ and qPEPCoR 5′-TTTTGCGGCAGCTATCTCTC-3′) were designed using PrimerQuest tool by IDT, USA. The reactions were setup following the standard protocol provided in DyNAmo ColorFlash SYBR Green qPCR Kit (Thermo Scientific). The thermal profile used for PCR amplification was: 95 °C for 4 min; 40 cycles of 95 °C for 15 s, 60 °C for 30 s and fluorescence data collection. To minimize variation in the output, three technical replicates were carried out for each of the three biological replicates. The baseline data was collected for first 15 cycles to generate a baseline-subtracted plot of the logarithmic increase in fluorescence signal (∆Rn) versus cycle number. A standard fluorescence threshold (Rn) was set to 0.5 on the log fluorescence scale to determine the fractional cycle number (Ct value). The relative abundance of GmIPK2 was calculated using the 2-∆∆CT method (Livak and Schmittgen 2001). Dissociation curve analysis from 60 to 95 °C was also performed at the end of the assay to check for any non-specific amplification and/or contamination.

Sequence analysis and phylogenetic tree construction

Homologous IPK2 sequences from other plants were identified with National Center for Biotechnology Information (NCBI) Basic Local Alignment Search Tool (Protein BLAST, https://blast.ncbi.nlm.nih.gov/Blast.cgi) using the above deduced GmIPK2 sequence as a query. The primary sequence composition of IPK2 sequences were computed using PEPSTATS (Rice et al. 2000). The physio-chemical parameters of proteins were predicted using the ProtParam tool of ExPASy web server (Gasteiger et al. 2005). TargetP 1.1 Server using the cutoffs of 95% specificity was implemented for subcellular location prediction of GmIPK2 protein (Emanuelsson et al. 2000) and the outcome was compared to predictions obtained from MemType-2L (Chou and Shen 2007), WoLF PSORT (Horton et al. 2007), SubLoc v1.0 (Hua and Sun 2001) and CELLO v2.5 (Yu et al. 2006). Its transmembrane topology was predicted with PSIpred (Buchan et al. 2013), TMpred (Hofmann and Stoffel 1993) and NPS@ web programs (Rost et al. 1996). A final consensus was drawn manually and the topology was generated and visualized using Protter version 1.0 (Omasits et al. 2014). Presence of potential secretory signal peptides or mitochondrial targeting peptides was analyzed with SignalP 4.1 web server (Petersen et al. 2011). M-Coffee multiple sequence alignment (MSA) of the selected amino acid sequences was carried out to produce quality alignments which served as the basis for phylogenetic analysis (Wallace et al. 2006) to detect its evolutionary placement and phylogenetic similarity with other similar genes. The evolutionary tree was constructed using neighbor joining (NJ) clustering method to compute distances and poisson model for amino acid substitution in MEGA Version 6.0 (Molecular Evolutionary Genetic Analysis, Tamura et al. 2013). Bootstrap replications were set at 1000 to assess the degree of confidence for each clade of the observed tree. The final image was rendered with the Interactive Tree of Life server (iTOL) (Letunic and Bork 2016).

Promoter isolation and prediction of regulatory motifs

A motif search was carried out to define putative cis-elements in the promoter sequences involved in the regulation of IPK2 expression using PlantCARE (Lescot et al. 2002). To identify these cis-regulatory elements, around 2 kb upstream sequence of the IPK2 homologs were retrieved by NCBI’s nucleotide BLAST program and fed to PlantCARE web tool. The tool’s database identified regulatory elements in the isolated upstream sequences and each of the elements observed were analyzed with previously reported properties of the particular element. Unlike CDS, the regulatory sequences indirectly influence their immediate phenotype.

Secondary structure analysis and domain prediction

M-Coffee multiple sequence alignment (MSA) of the selected amino acid sequences was carried out and the alignment file was imported to Jalview (Waterhouse et al. 2009) to identify and shade the conserved amino acid sequences. Ungapped motifs were also detected using MEME web tool available on MEME suite 4.11.1 (Bailey et al. 2009). The motifs present were further verified using My Hits motif scan tool (Pagni et al. 2007). The domain composition was analyzed using CDD tool (http://www.ncbi.nlm.nih.gov/cdd) on NCBI server. A secondary structure consensus from amino acid sequences was built based on the joint prediction with SOPMA (nearest-neighbor method) and PHD (neural networks method) correctly predicting 82.2% of the residues for 74% of co-predicted amino acids (https://npsa-prabi.ibcp.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_seccons.html) (Geourjon and Deléage 1995; Rost and Sander 1993). Cysteine species and disulfide connectivity of protein sequences were determined using web tool DiANNA (Ferrè and Clote 2005). A secondary structure topology map of the 3D model was generated using ProMotif (http://www.ebi.ac.uk/thornton-srv/databases/pdbsum/Generate.html) and rendered using TopDraw (Bond 2003).

Homology modeling and quality assessment of predicted model

The 3D model of GmIPK2 was constructed by template-based homology modeling using automated comparative protein modeling servers, SWISS-MODEL (Biasini et al. 2014) and PHYRE2 (Kelley et al. 2015) as well as a standalone comparative modeling program MODELLER 9.16 (Webb and Sali 2016). Comparative modeling consists of five main steps: search for related protein structures, selection of one or more appropriate templates, target-template alignment, model building and model evaluation. The final step of comparative modeling was additionally performed with RAMPAGE (Lovell et al. 2003), VERIFY 3D (Eisenberg et al. 1997) and ProSA servers (Wiederstein and Sippl 2007) to evaluate the stereochemical and energetic properties of the obtained models. In addition to this, the quality of models can also be assessed by structural comparison to the template using the MatchMaker tool in UCSF Chimera (Pettersen et al. 2004), a molecular visualization software package and calculating the Cα root mean square deviation scores (RMSDs) for each of the comparative models. The secondary structures of the final and template proteins were also compared by pairwise 3D alignment using MATRAS 1.2 (Kawabata 2003). Agreement on the best model was made on the basis of the majority in best scores from these different quality analyses.

Refinement of the predicted homology model

Molecular dynamics (MD) simulation was performed to optimize the obtained GmIPK2 model. Simulation of the model was conducted in explicit solvent using the GROMACS (Groningen Machine for Chemical Simulations) 4.5.5 package (Pronk et al. 2013). The model was solvated with simple point charge (SPC216) water in a cubic box with edges that were 0.7 nm from the molecular boundary. Initially, energy minimization (maximum number of steps: 1000) was performed to remove steric conflicts between the protein and water molecules, using the steepest descent integrator. The system was then equilibrated by optimizing the solvent molecules surrounding the energy-minimized model with NVT (constant Number of particles, Volume, and Temperature) and NPT (constant Number of particles, Pressure, and Temperature) ensembles using Berendsen thermostat and Parrinello-Rahman barostat, respectively. Finally, the system was simulated for 50 ns thrice maintaining the same temperature (300 K) and pressure (1 bar) using the Particle Mesh Ewald (PME) electrostatics method. After completion of production run, we collected the data using a post-processing tool, trjconv (strips out coordinates and correct for periodicity) and used this corrected trajectory as the input for studying the conformational stability of our simulated protein via g_rms and g_rmsf tools. We subsequently generated and analyzed the output plots using a simple plotting program, xmgrace (Turner 2005).

Active site prediction, molecular docking and MD simulation of the docked complexes

The refined protein model thus generated was docked to characterize the 3D structure of the complex, to gain an insight into the crucial amino acid residues also referred to as active residues that are involved in the complex formation. Based on the available literature for IPK2, 1D-myo-inositol-1,4,5-trisphosphate (PDB: 5GUG-I3P) and 1D-myo-inositol-1,4,5,6-tetrakisphosphate (PDB: 4A69-I0P) ligands were selected for optimized GmIPK2 protein receptor. Docking of these ligands was performed using AutoDock Vina (version 1.1.2) (Trott and Olson 2010) following the semi-flexible approach of docking. AutoDock Vina reads all the molecules in a simplified PDB file representation, termed PDBQT and thus the coordinates of both GmIPK2 protein and its ligands (downloaded from the RCSB protein data bank) were prepared using MGL tools (version 1.5.4) AutoDock tools (Morris et al. 2009) prior to docking. All the water and solvent atoms of the protein were removed and the polar hydrogen atoms were added. The GmIPK2 molecule was kept rigid while the ligands were allowed to rotate and explore more flexible binding pockets. Vina used a customized rectangular 3D cartesian grid for specifying the binding site of the protein and for efficient geometric scoring. The dimensions of the grid were customized to make sure that the size of the search space is large enough for the ligand to rotate in. Once set, the docking run produced ligand poses each with a definite binding energy (kcal/mol) calculated based on the scoring function used in Vina. The conformations with the lowest binding affinity were chosen and the interaction diagrams were generated using Discovery Studio Visualizer 4.1 (Accelrys Software Inc., USA 2013). The amino acid residues present at a distance of 2 Å were considered as the binding partners of the ligands. The active amino acid residues were also predicted by combining results of three different interface prediction web servers, CASTp (Dundas et al. 2006), FTSite (Kozakov et al. 2015) and FunFOLD2 (Roche et al. 2013) into a consensus. The final complexes were equilibrated by MD simulation by following the procedure described for protein model in explicit water in the previous section but by applying restraints on the ligands to prevent them from moving away from the binding site. Once equilibrated, we simulated the complexes for 50 ns and analyzed the resulting trajectories for their stability using g_rms, g_rmsf, g_hbond and g_mmpbsa tools.

Results and discussion

Cloning and sequencing of GmIPK2

We first performed a BLASTN analysis in the plant comparative genomics portal Phytozome v9.1 (Goodstein et al. 2012) with the soybean IPK2 gene sequence available in NCBI (GenBank: NM_001250522) to retrieve a 1241 bp transcript sequence. Based on the sequence data thus derived, we designed primers specific to amplify complete GmIPK2 coding sequence (CDS) as well as a fragment of its 3′ untranslated region (UTR) with the aim to design silencing construct in the future specific to this highly conserved region because of its substantial role in gene regulation. To accomplish this, we converted total RNA isolated from 8 to 10 mm developing seed stage to cDNA and performed PCR amplification from the synthesized cDNA template following the protocol described under sect. “PCR amplification, cloning, and sequencing of partial GmIPK2 sequence”. The amplicon was then cloned into pGEM-T easy vector system and introduced into the bacterial host E. coli (DH5α). The putative cDNA clones were verified by restriction analysis with EcoRI and sequence characterized to ~ 910 bp residues in length (Fig. 1). It contains a single open reading frame ~ 840 bp long which potentially encode a single polypeptide of 279 amino acid residues and a ~ 70 bp 3′ UTR fragment. We submitted the obtained CDS data to NCBI (GenBank: KF297702) and used the information as query to conduct protein homology search using PSI-BLAST algorithm. Amongst the sequences producing significant alignment, we shortlisted 30 plant IPK2 sequences based on percentage sequence identity to carry out further in silico analysis.

Fig. 1.

Agarose gel showing PCR amplification product of GmIPK2 gene isolated from developing seeds (8–10 mm) of Glycine max cv. Pusa-16. Lane M: 500 bp DNA ladder, lanes 1–3: ~ 910 bp GmIPK2 fragment

Spatial and temporal expression profiling

As discussed above, tissue-specific modulation of IPK2 gene to generate a lpa mutant is essential to evade any possible pleiotropic effects. Before this can be achieved it is vital to investigate its spatial expression profile in different tissues as well as its temporal expression profile in developing seeds, to provide an initial point for strategic achievement of a desired level of silencing.

To study the expression of GmIPK2 gene in different tissues of a soybean plant and during seed development, semi-quantitative as well as real-time PCR expression analysis was performed with total RNA isolated from root, stem, leaf and flower tissues of 30-day-old G. max plants and developing seeds ranging from 0 to 16 mm in size distributed in eight different progressive stages. Semi-quantitative PCR analysis revealed a differential pattern of GmIPK2 transcript expression across the set of experimental tissues analyzed, with the highest level of transcripts observed in seeds (Fig. 2a). The same was confirmed by steady-state qRT-PCR analysis which also detected highest level of transcripts in seeds (Fig. 2b). This suggests that GmIPK2 play a key role in PA biosynthesis in this tissue for use as a primary source of energy during germination. In both the analyses, same level of amplification was observed for PEPCo housekeeping gene. Since PA is also required for several other vital functions throughout the plant system, a basal level of it is observed in all the tissues as well. Thus, amongst the other tissues analyzed, expression of GmIPK2 was also recorded in roots and flowers, but at a level lower than that observed in the cotyledons. The strong presence of GmIPK2 transcripts in these tissues may be attributed to its role in regulating cytosolic calcium gradient which correlates with pollen germination, pollen tube growth (Franklin-Tong et al. 1996; Malho 1998; Pierson et al. 1994; Xu et al. 2005), root growth, and root hair development (Bibikova et al. 1997; Felle and Hepler 1997; Wymer et al. 1997). A low level of expression was also observed in stems and leaves as it is studied to be involved in regulating a vital function of axillary shoot branching by participating in auxin signaling (Zhang et al. 2007).

Fig. 2.

a RT-PCR expression analysis of GmIPK2 gene in different plant tissues of Pusa-16 cultivar using soybean housekeeping gene PEPCo as an internal control. b Relative quantification of GmIPK2 transcript levels in the samples analyzed above by qRT-PCR, normalized to soybean housekeeping gene PEPCo. The leaf tissue was taken as calibrator. The data are mean of technical triplicates of each of the three biological replicates with error bars indicating standard deviation (SD)

We then analyzed its temporal expression pattern during eight progressive seed development stages. In both semi-quantitative as well as real-time PCR expression analysis, we observed that its expression increased as the development progressed and reached a peak value at the later stages of seed development (Fig. 3). This pattern of expression coincides evenly with the pattern of accumulation of PA which is linear throughout most of seed development (Raboy and Dickinson 1987). Similar results were obtained in a microarray transcriptome study conducted in the past in our laboratory (GEO: GSE69821). This observation can be explained by the increase in production of phosphorous compounds required to support growth and development during initial stages of seed development when the synthesis of phosphorous reserve, PA is minimal (Raboy and Dickinson 1987).

Fig. 3.

a RT-PCR expression analysis of GmIPK2 gene in developing seeds of Pusa-16 cultivar using soybean housekeeping gene PEPCo as an internal control. b Relative quantification of GmIPK2 transcript levels in the samples analyzed above by qRT-PCR, normalized to soybean housekeeping gene PEPCo. The 0–2 mm seed stage was taken as calibrator. The data are mean of technical triplicates of each of the three biological replicates with error bars indicating SD

The spatiotemporal analysis in summary identifies seed as the major tissue for its expression with maximum relative expression occurring during the later stages of its development. Thus, from the present study we can hypothesize that targeting the GmIPK2 gene expression during late seed development stages may provide a potential strategy for generating lpa soybean with enhanced nutritional value.

Characterization of regulatory motifs in IPK2 gene promoter

To understand the mechanism regulating the spatiotemporal expression pattern of GmIPK2 gene thus observed, we analyzed the promoter region of all its homologs and identified different cis-regulatory elements located therein. A 2 kb sequence upstream to the open reading frame was identified and subjected to PlantCARE analysis. Database search revealed the presence of many motifs related to seed-specific promoters, hormone-responsive cis-elements (HRE) and cis-elements responsive to stresses (DSRE) that together contribute to the differential regulation of our gene (Table 1). Light responsive elements (Arguello-Astorga and Herrera-Estrella 1996; Feldbrugge et al. 1996; Lois et al. 1989; Lopez-Ochoa et al. 2007; Sessa et al. 1995) were observed most frequently which suggest a probable diurnal regulation of IPK2 expression. The GCN4 and Skn_1 motifs highly conserved in the promoters of cereal seed storage protein genes were located within the IPK2 promoter. These cis-acting elements play a central role in controlling endosperm-specific gene expression (Onodera et al. 2001; Takaiwa et al. 1996; Washida et al. 1999; Wu et al. 1998). Multiple HREs particularly those known to be involved in abscisic acid (ABA) and gibberellic acid (GA3) sensing were also identified. A ABA-responsive element (ABRE) with the core sequence PyACGTG/TC and three GA-responsive elements (GARE) AAACAGA, TCTGTTG and TATCCAC/T (Niu et al. 2016; Mongkolsiriwatana et al. 2009; Yamaguchi-Shinozaki et al. 1989) were identified which previously were reported to correlate with PA accumulation during grain filling (Abe et al. 2003; Matsuno and Fujimura 2014). Aggarwal et al. 2015 reported that IPK2 is an ABA-induced gene which is antagonistically suppressed by GA3 underlining the crucial role played by these hormones in regulating PA pathway genes. Putative elements responsive to methyl-jasmonate, drought inducibility and anaerobic induction were also observed which help combat assorted types of abiotic factors that plants are exposed to under natural environment (Abe et al. 2003; Kim et al. 1993; Nguyen et al. 2003; Rouster et al. 1997; Shinozaki and Yamaguchi-Shinozaki 2000).

Table 1.

Potential cis-acting elements identified in the 5′ regulatory sequences of plant IPK2s

Classification	Name	Sequence	Source organism	References
Endosperm	GCN_4 motif	GAAGCCA, TGAGTCA	Oryza sativa	Takaiwa et al. (1996), Onodera et al. (2001)
	Skn-1 motif	CAAGCCA, TGTGTCA, GTCAT	Oryza sativa	Washida et al. (1999)
ABA	ABRE	CACGTG, TACGTG	Arabidopsis thaliana	Yamaguchi-Shinozaki et al. (1989)
Gibberellin	GARE	AAACAGA, TCTGTTG, TATCCAC/T	Brassica oleracea Oryza sativa	Mongkolsiriwatana et al. (2009), Jun Niu et al. (2016)
Light	Box 4	ATTAAT	Petroselinum crispum	Lois et al. (1989)
	CATT-motif	GCATTC	Zea mays	Arguello-Astorga and Herrera-Estrella (1996)
	G-Box	CACGTG	Pisum sativum, Arabidopsis thaliana, Zea mays	Sessa et al. (1995), López-Ochoa et al. (2007)
	ATCC/T-motif	AATCTAATCC/T	Pisum sativum, Arabidopsis thaliana	Arguello-Astorga and Herrera-Estrella (1996)
	TCT-motif	TCTTAC	Arabidopsis thaliana	Arguello-Astorga and Herrera-Estrella (1996)
	Box-I	TTTCAAA	Pisum sativum	Arguello-Astorga and Herrera-Estrella (1996)
	ACE	AAAACGTTTA	Petroselinum crispum	Feldbrugge et al. (1996)
	GA-motif	AAGGAAGA	Glycine max	Arguello-Astorga and Herrera-Estrella 1996
MeJA	CGTCA-motif TGACG-motif	CGTCA TGACG	Hordeum vulgare Hordeum vulgare	Kim et al. (1993), Rouster et al. (1997)
Drought	MBS	T/CAACTG	Arabidopsis thaliana	Shinozaki and Yamaguchi-Shinozaki (2000), Abe et al. (2003)
Anaerobic	ARE	TGGTTT	Zea mays	Nguyen et al. (2003)

Computation of physiochemical parameters and subcellular localization prediction

GmIPK2 sequence similarity search using PSI-BLAST revealed homology to IPK2 protein sequences from different plant sources showing maximum similarity with Glycine soja (98%) and Vigna radiata (74%) (Table 2). Primary sequence analysis of these homologs indicate that leucine is the most abundant amino acid which makes upto approximately 9–11 mol percent of its backbone residues whilst the percentage of tryptophan and methionine were found to be the least at approx. 1% across species. The isoelectric points (pI) of all IPK2 proteins were computed to be under 7 suggesting that they are most likely to precipitate in acidic buffers except Glycine soja, Medicago truncatula, Zea mays and Sorghum bicolor which has a pI of 7.1, 8.57, 10.48 and 8.26, respectively indicating their solubility in basic buffers. The calculated pI will be useful for empirical protein purification by isoelectric focusing and ion exchange chromatography. The extinction coefficient (EC) (Gill and Von Hippel 1989) of IPK2 proteins measured at 280 nm in water was found ranging from 25,440 to 34,380 M–1 cm–1 with respect to their concentration of aromatic amino acids (11–15%) and cystine (disulfide bonds). These EC values can be used to calculate protein concentration in a solution which in turn help in the quantitative study of biochemical interactions (protein–protein and protein–ligand). The instability indices (Ii) computed for selected IPK2 proteins are used to determine their in vivo half-lives (Guruprasad et al. 1990). Rogers et al., 1986 reported that proteins having Ii values greater than 40 have an in vivo half-life of less than 5 h while those proteins having Ii values less than 40 have a longer in vivo half-life of 16 h. Our study showed that Ii values of all the homologs are less than 40 and hence are thermally stable with a long half-life except for Lotus japonicas and Zea mays kinases that have a Ii above 40 which indicate their possible thermal instability. Thermostability of proteins result from a combination of several factors acting synergistically. Here we assess thermal stability of our proteins in direct proportionality with their aliphatic index (Ai) which is a measure of the relative volume occupied by aliphatic side chains (Ikai 1980). The Ai values determined for IPK2 kinases ranged from 68.79 to 96.70 with those from brassicaceae family showing lowest thermal stability. This in turn is indicative of their greater flexibility at a wide range of temperatures when compared to proteins of other families. Grand average of hydropathicity (GRAVY) number reflect the average hydropathy of a protein, the positively rated being hydrophobic and negatively rated being hydrophilic in nature Kyte and Doolittle (1982). GRAVY index for IPK2 kinases was found ranging from − 0.517 to − 0.013 indicating that these proteins will interact favourably with water except for Vigna radiata with an index of 0.045 and hence is a potential hydrophobic protein.

Table 2.

Physiochemical parameters of shortlisted plant IPK2 sequences computed using the ProtParam tool

Organism	Accession no.	Sequence length	MW	pI	EC	Ii	Ai	GRAVY	− R	+ R
Glycine max	AGW99177.1	279	30,979.2	6.29	29,910	29.72	89.07	− 0.18	35	30
Glycine soja	KHN19419.1	178	19,898.87	7.10	23,950	29.73	95.73	− 0.059	21	21
Vigna radiata	XP_022634636.1	286	31,482.10	6.59	28,420	38.17	93.67	0.045	29	27
Cajanus cajan	XP_020232596.1	267	29,463.71	6.86	28,420	33.70	91.31	− 0.132	31	30
Phaseolus vulgaris	XP_007133857.1	264	28,951.0	6.59	26,930	38.90	90.38	− 0.017	27	25
Lotus japonicus	AFK39224.1	283	31,348.76	5.66	32,430	42.36	87.74	− 0.139	39	30
Cicer arietinum	XP_004510840.1	296	33,061.70	6.12	26,930	31.59	91.11	− 0.250	38	33
Medicago truncatula	XP_003627882.1	426	47,961.8	8.57	29,910	31.89	85.02	− 0.344	49	53
Arachis hypogaea	ALT56981.1	297	32,684.01	6.45	28,420	36.41	84.01	− 0.227	34	32
Arachis duranensis	XP_015937330.1	297	32,883.25	6.45	28,420	37.35	84.01	− 0.248	35	33
Corchorus capsularis	OMO94774.1	292	32,498.97	5.84	28,420	40.75	92.43	− 0.159	38	33
Theobroma cacao	EOY22693.1	304	34,213.77	6.32	31,400	37.61	84.61	− 0.292	39	36
Herrania umbratica	XP_021285810.1	305	34,392.01	6.56	36,900	40.38	83.38	− 0.307	38	36
Durio zibethinus	XP_022738659.1	295	32,954.54	6.71	28,420	38.72	87.83	− 0.237	36	35
Solanum lycopersicum	XP_004235863.1	294	32,476.9	5.93	28,420	34.32	83.23	− 0.261	35	30
Solanum tuberosum	NP_001335929.1	408	44,295.0	5.96	28,420	20.21	93.19	− 0.254	50	43
Brassica rapa	XP_009112000.1	274	30,640.7	6.25	34,380	30.25	80.66	− 0.283	36	33
Capsicum baccatum	PHT32205.1	374	40,995.36	5.97	32,890	33.47	90.19	− 0.297	50	44
Nicotiana attenuata	XP_019263239.1	368	40,865.55	5.51	31,400	30.30	93.51	− 0.179	48	40
Capsicum annuum	PHT66064.1	374	41,015.41	5.91	32,890	32.27	92.01	− 0.277	50	43
Capsicum chinense	PHU00941.1	374	40,986.37	5.81	32,890	32.27	92.27	− 0.271	51	43
Brassica napus	XP_013749343.1	285	31,961.0	5.87	35,870	30.23	75.86	− 0.387	40	34
Arabidopsis lyrata	XP_020870073.1	300	33,671.08	5.82	31,400	29.76	83.47	− 0.324	42	35
Prunus avium	XP_021815221.1	281	31,100.28	6.20	31,400	26.98	86.41	− 0.209	34	30
Trifolium pratense	PNY08557.1	273	30,275.83	5.50	23,950	29.99	96.70	− 0.054	37	26
Arabidopsis thaliana	NP_200984.1	300	33,486.7	5.72	31,400	25.96	80.23	− 0.329	40	31
Prunus persica	XP_007209445.1	282	31,303.57	6.50	31,400	28.39	85.39	− 0.252	35	33
Lepidium latifolium	ACK86969.2	297	33,212.5	6.59	34,380	29.25	81.95	− 0.285	31	29
Zea mays	XP_008649440.2	240	26,665.45	10.48	29,450	65.74	68.79	− 0.517	22	37
Aegilops tauschii	XP_020147020.1	287	30,607.86	6.13	26,930	38.85	88.43	− 0.052	33	29
Sorghum bicolor	XP_002452184.1	322	34,550.64	8.26	28,420	39.34	90.93	− 0.013	32	34

MW molecular weight (g/mol), pI isoelectric point, EC extinction coefficient (M−1cm−1), Ii instability index, Ai aliphatic index, GRAVY grand average hydropathy, (−R) number of negative residues, (+ R) number of positive residues

Protein localization and target peptide predictions are significant studies as they aid in in silico protein function characterization as well as genome annotation. The acidic amino acid composition of IPK2 homologs determined by the physiochemical analysis conducted above suggest that they are cytoplasmic in nature as opposed to the basic amino acid composition of membrane proteins for their stability (Schwartz et al. 2001). Further sequence analysis based on TargetP scores (cTP: 0.163, mTP: 0.066, SP: 0.087, other: 0.906) also suggest that IPK2 kinases may be located anywhere in the cell besides chloroplast and mitochondria. The sequences were not predicted to have any signaling pre-sequence which was indicated by their low signal peptide (SP) score reinforced that they are soluble in nature. Besides, a consensus of predictions obtained from WoLF PSORT, CELLO v2.5, SubLoc v1.0 and MemType-2L servers (Table S1) also established the cytoplasmic character of IPK2 protein. However, TMpred, PSIPRED, NPS@ and DAS servers also identified a single consensus C-terminal transmembrane region positioned at 259–274. This could have been mistaken due to the presence of large regions of hydrophobic residues in the C-terminal soluble region, as hydrophobicity alone is used as the criterion to predict membrane-spanning regions. The absence of any transmembrane helix is also well documented in the hydrophobicity plot generated using waveTM server (Fig S1) (Pashou et al. 2004) as majority of amino acids show negative for hydrophobicity.

Motif analysis and secondary structure characterization

Multiple sequence alignment (MSA) of the selected plant IPK2 homologs was performed using M-Coffee web server which is a meta-method for assembling MSA by combining the output of several individual methods into one to generate the best possible alignment. The consensus alignment thus generated revealed several significantly conserved motifs and sites unique to inositol phosphate kinases (Fig. 4). A signature inositol phosphate-binding motif, PxxxDxKxG was identified in all the aligned sequences (Odom et al. 2000; Saiardi et al. 1999) which confirms that they belong to inositol phosphate kinase (IPK) superfamily of IP kinases. Holmes and Jogl, 2006 state that the members of this superfamily share several strictly conserved signature motifs with each other and are predicted to assume the same overall fold, despite the low sequence conservation. The core catalytic tyrosine kinase motif, RxxxExxxY was also discovered in all the sequences which suggest that they are tyrosine-specific protein kinases (Cooper et al. 1984). IPK2 sequences from Solanaceae and Rosaceae families were found to contain a Glycine-rich consensus ATP-binding GxGxxG motif characteristic of protein kinase C (PKC) catalytic domain (Steinberg 2008). The classical PKC and plant CDPKs recognized phosphorylation S/TxK/R motif (Nishikawa et al. 1997; Neumann et al. 1996; Roberts and Harmon 1992) was also identified in some of the sequences speculating their role in lipid-dependent PA biosynthetic pathway. Such promiscuous kinase activity suggests that both lipid-dependent and independent pathways regulate PA biosynthesis as well as basic nuclear and cellular processes in plants (Josefsen et al. 2007; Stevenson-Paulik et al. 2002). A protein recognition LxxLL motif common to all of the aligned sequences indicate their participation in protein–protein interactions, regulating cell signalling, cell adhesion, and transcription (Plevin et al. 2005). Further analysis by MEME suite web server identified a total of 11 conserved ungapped motifs, with motif PxxxDxKxG being the most conserved amongst all IPK2 homologs as indicated by its lowest E value of 2.9e-487 (Fig S2). The obtained motifs were subjected to BLASTP analysis for conformation of their annotations which established that they all belong to IPK superfamily domain (CDD Acc: cl12283) and thus substantiate our previous results. We further explored GmIPK2 protein sequence by Motif Scan which recognized diverse protein kinase phosphorylation sites (Table S2). Since phosphorylation acts as a molecular switch in modulating protein function, structural rearrangement and cellular localization it can be suggested that GmIPK2 play a critical role in many biological regulatory events of signalling, proliferation, differentiation, and apoptosis. Its role in biological processes as diverse as mRNA export (York et al. 1999), DNA repair (Hanakahi et al. 2000), regulation of chromatin structure (Shen et al. 2003; Steger et al. 2003), maintenance of basal resistance to plant pathogens (Murphy et al. 2008) and apoptosis (Agarwal et al. 2009) have been studied in the past.

Fig. 4.

M-Coffee multiple sequence alignment diagram of selected plant IPK2 protein sequences rendered with Jalview. The sequence motifs shared amongst all the representatives are coloured according to percentage identity. The consensus row at the bottom shows the most frequent residue at each column or a ‘+’ if two or more residues are equally abundant

The secondary structure of a protein is more conserved than its nucleotide sequence and is, therefore, a prized source of information in understanding its classification, function, molecular evolution, and interaction with macromolecules (Reehana et al. 2013). In addition, secondary structure provides the first framework for homology based prediction of a protein 3D-model. Thus, in the current study, we inferred the secondary structure composition of IPK2 kinases from a three-state prediction done using NPS@ web server (Table 3). A high coil content was observed in most of the sequences including GmIPK2 while some showed them in nearly equal proportion with α-helix. This structural state can be justified based on the rich content of highly flexible glycine and kink inducing proline amino acid residues. The percentage of extended strands (% Ee) in all the kinases were found ranging from 13 to 29% except for the Poaceae family kinases which showed a low % Ee conformation (below 10%). PDBSum tool PROMOTIF analysis of GmIPK2 polypeptide identified total eight α-helices and eight β-strands arranged to form three antiparallel β-sheets, interspersed throughout by regions of coil or turn conformations (Fig. 5). We also recognized a varying number of bonded half-cystine pairs in all the IPK2 protein sequences using DIANNA server. It revealed the presence of 8 Cys residues in GmIPK2 and the most probable half-cystine pairs predicted by CYS-REC were 94–121, 123–276 and 158–186. These potential long-term interactions participate in stabilizing the native conformations of our proteins and may as well contribute to differences in their tertiary structures.

Table 3.

Three-state description of secondary structure content and disulfide pattern prediction of IPK2 sequences

Organism	α-Helix (%Hh)	Extended strands (%Ee)	Random coil (%Cc)	Disulphide bridge prediction
Glycine max	26.88	13.26	31.54	94–121, 123–276, 158–186
Glycine soja	34.27	29.78	35.96	20–117, 22–85, 161–175
Vigna radiata	29.72	17.13	24.48	127–190, 263–281
Cajanus cajan	25.84	13.11	29.21	119–177
Phaseolus vulgaris	24.62	21.21	28.79	116–176, 208–249
Lotus japonicus	25.44	14.49	26.86	94–215, 149–187, 261–283
Cicer arietinum	21.96	14.53	28.72	122–185, 217–279, 260–274
Medicago truncatula	18.54	15.02	31.69	98–262, 325–357
Arachis hypogaea	24.58	14.48	33.67	120–188, 223–288, 269–292
Arachis duranensis	24.58	14.48	33.33	223–288, 269–292
Corchorus capsularis	19.52	20.21	25.34	120–187, 222–267
Theobroma cacao	19.74	16.78	39.14	50–120, 187–222, 267–298, 288–296
Herrania umbratica	18.69	14.75	42.95	120–296, 222–298, 267–288
Durio zibethinus	18.98	22.37	26.10	93–120, 149–222, 267–288
Solanum lycopersicum	18.03	18.37	30.27	120–187, 222–267
Solanum tuberosum	17.40	17.65	26.23	120–187, 222–283
Brassica rapa	13.50	21.17	22.63	144–186, 221–264
Capsicum baccatum	17.65	17.38	30.75	120–187, 222–283
Nicotiana attenuata	21.47	15.76	25.82	156–223, 258–303
Capsicum annuum	19.52	16.58	30.48	120–187, 222–267
Capsicum chinense	20.32	16.31	30.21	120–187, 222–267
Brassica napus	18.25	18.25	24.56	120–220, 144–185, 264–281
Arabidopsis lyrata	22.00	17.00	23.33	120–187, 222–272
Prunus avium	17.44	17.44	27.05	120–222, 156–187
Trifolium pratense	21.61	15.75	31.87	173–205, 203–250
Arabidopsis thaliana	22.00	16.67	24.67	120–187, 197–272
Prunus persica	17.02	17.02	29.79	120–222, 156–187
Lepidium latifolium	17.85	17.17	28.62	78–269, 120–187, 222–285
Zea mays	16.25	10.42	44.17	101–203, 192–200
Aegilops tauschii	28.57	7.67	31.36	110–188, 124–267
Sorghum bicolor	30.12	12.42	26.40	13–300, 137–160

The secondary structure data were generated by joint prediction with SOPMA and PHD while disulfide bonding pattern was determined using DiANNA (DiAminoacid Neural Network Application) 1.1 server

Fig. 5.

Topology map of GmIPK2 generated using ProMotif. There are a total of eight α-helices (1–8) and three β-sheets (β-sheet 1: strands A1 and A2; β-sheet 2: strands B1, B2, B3 and B4; β-sheet 3: strands C1 and C2)

Evolutionary analysis

Conserved motif analysis of IPK2 protein sequences point at a distinct evolutionary association between these kinases. Phylogenetic analysis would provide a further basis to determine their relatedness as well as to understand their collective evolution from a common ancestor. Previous research has reported that IPK superfamily of kinases to which IPK2 belongs evolved from a common ancestor (Irvine and Schell 2001; Shears 2004). In our study, based on the alignment obtained in “Motif analysis and secondary structure characterization” of GmIPK2 protein and its homologs, we constructed a neighbor-joining phylogenetic tree using MEGA 6.0 software (Fig. 6). The tree topology derived was supported by high bootstrap values. The IPK2s were clustered into six well delineated groups. The clusters consist of members of the Poaceae, Brassicaceae, Malvaceae, Rosaceae, Solanaceae and Fabaceae families. The Poaceae family of monocots (Zea may, Sorghum bicolor and Aegilopus tauschii) was found to be most distantly related to G. max in comparison to the fabaceae family of eudicots (Glycine soja, Phaseolus vulgaris, Vigna radiata, Cajanus cajan, Trifolium pratense, Cicer arietinum, Medicago truncatula, Lotus japonicas, Arachis duranensis, and Arachis hypogaea) which are most closely related.

Fig. 6.

Phylogenetic tree showing evolutionary relationship of 31 plant IPK2 sequences divided into different clades, colour coded to indicate the plant family to which they belong. The posterior probability values are indicated corresponding to every node

Three-dimensional model construction

Model building

To derive structural information about GmIPK2 protein we built its theoretical model by following homology modeling approach since no X-ray crystal or NMR structure of it is available. The homology modeling technique takes advantage of structural conservation found in similar proteins that have evolved from a common ancestor. Yeast IPK2 protein was the first member of inositol multikinase family whose crystal structure was determined (Holmes and Jogl 2006), but it was found to show a very low sequence similarity with GmIPK2 protein. As we know, aligning two sequences can be a difficult process if the sequence similarity is low, we therefore, used the above derived GmIPK2 protein sequence as a query in PSI-BLAST to find more similar sequences amongst the PDB database proteins with resolved 3D structures to use as a potential template. The only closest homologous sequence available in PDB was that of chain A of Arabidopsis thaliana inositol phosphate multikinase (PDB:4FRF) which showed 55% sequence identity with an e value of 9e-99. The initial comparative models, i.e., GmIPK2-S and GmIPK2-P were built using fully automated SWISS-MODEL and PHYRE2 servers, respectively, which also identified 4FRF_A as the most reliable template using sensitive hidden markov model searches and used the same as the structural input. A global quality estimation score (GMQE) of 0.62 was provided by SWISS-MODEL which indicates a reasonably reliable structure. Homology model was additionally built using MODELLER 9.16 program from the X-ray crystal structure coordinates of the previously identified template structure (4FRF_A). The software generated five different models by optimizing the objective function of spatial restrains in a cartesian space. Three different energy scores viz. molpdf, DOPE and GA341 were computed for each of these models and compared to one another to select the best 3D structure (Table S3). Model 4 (GmIPK2-M4), with the lowest molpdf and DOPE scores of 1734.48291 and − 27846.58203 respectively, was chosen as the principal conformational structure.

Structure validation

We then assessed the accuracy and reliability of all the three predicted models viz. GmIPK2-S, GmIPK2-P and GmIPK2-M4 using various online diagnostic tools (Table 4). RAMPAGE server which evaluates the 3D-structures based on Ramachandran plot calculations showed variable distribution of torsion angles in all the models. Fig S3a shows Ramachandran plot for GmIPK2-S model, showing 92.5% residues in favourable region and 4.9% in allowed region with just 2.7% residues in outlier region of the plot, which reflects its superior backbone geometry. We then utilized ProSA-web computational engine to analyze overall quality of the models based on their z scores and local quality based on their residue energies. GmIPK2-S model showed a better z score of − 6.56 which is displayed in the energy distribution plot derived from a group of experimentally determined protein structures of similar size. The Z score thus observed was very much within the range of scores typically found for native conformations of this group which indicates a good overall quality of the modelled protein (Fig S3b). Moreover, its residue energy was computed to be largely negative which further reflects that the local regions in the protein are modelled well (Fig S3c). The accuracy of 3D-models was further analyzed from energy profiles obtained by Verify3D program. Figure S3(d) shows 3D-1D profile for GmIPK2-S model with 97.82% of the model residues showing an average 3D-1D profile score > = 0.2 and hence validate that majority of its amino acid sequence reconcile to its environment in the 3D structure. Additionally, the pairwise 3D structural alignment of GmIPK2-S model with template protein, 4FRF_A using MATRAS 2.1 program revealed that both the structures shared 91.3% secondary structure identity (Fig S4a) and the average distance between the Cα backbone atoms of their 3D structures, i.e., root mean square deviation (RMSD) measured through superimposition was 0.44 Å (between 196 atom pairs) (Fig S4b). Based on the majority of winning scores, GmIPK2-S (Fig. 7) was chosen as the best comparative model for energy minimization and further analyses.

Table 4.

Validation parameters computed for the energy-minimized 3D models of GmIPK2 protein built using different programs

Model	RAMPAGE Percentage of residues in favoured region	ProSA Z score	Verify3D Percentage of residues with 3D-1D score ≥ 0.2
GmIPK2-S	92.5	− 6.56	97.82
GmIPK2-P	82.2	− 6.28	84.95
GmIPK2-M4	87.7	− 5.84	79.93

Fig. 7.

Homology model of GmIPK2_S protein rendered using PyMOL

Molecular dynamics simulation

We subsequently performed MD simulation on our predicted GmIPK2-S model using GROMOS96 53A6 force field to compute its stability and dynamics. Initial potential energy minimization of solvated model showed that the maximum force dropped below the defined value of 1000 kJ mol^− 1 nm^− 1 in nearly 500 steps. The protein structure was then subjected to 50 ns of equilibration run at a constant temperature and pressure to obtain its molecular trajectory. The trajectory thus obtained was used to determine RMSD of Cα backbone atoms of the model using its starting structure as the reference, to determine its convergence towards an equilibrium state. Figure 8a shows RMSD as a function of simulation time. We observed that the protein stabilized around 10 ns of production run and converged to ~ 0.55 nm at 50 ns. The initial increase of RMSD could be attributed to the restraints in the system applied in the equilibration phase and their release later at the beginning of production phase. Besides RMSD, we also calculated root mean square fluctuations (RMSFs) to study mobility of the protein structure to draw an idea of its flexibility regions. From the RMSF plot (Fig. 8b), we identified Leu (75, 135, 272), Asp (76, 79, 127), Ala (77), Ser (78, 130–132), Gly (80, 136), His (84), Lys (126, 154), Glu (129, 230), Arg (128, 153) residues to be more flexible. This implies that the mentioned residues show greater movement from their native position, i.e., are dynamic in nature and thus are functionally more relevant. Thus, overall, the simulation results highlight the stable and reliable nature of our protein model and find it fit to be used for further active site predictions.

Fig. 8.

a Root mean square deviations and b Root mean square fluctuations of the Cα backbone atoms in GmIPK2_S model over 50 ns MD simulation

Molecular docking with inositol phosphates

Active site predictions

Once the protein model was refined, a comparative study was performed to detect possible binding pocket residues using CASTp, FTSite and FunFOLD2 binding site prediction servers (Fig S5). Lys105, Thr110, Lys122, Lys126, Ser130, Lys138, Ile139, Pro140, Arg153, Lys154, Gln157, and Ser219 were determined as possible active site residues in GmIPK2-S model. Similar predictions were also made in template protein (4FRF_A) which identified Arg104, Thr105, Pro108, Phe137, Lys149, Arg152, His216, Asn218, Ser219, Gln242, and Val246 as the probable binding site residues. From these studies, we could decipher that residues Lys, Arg, Pro, Thr, Gln and Ser are highly conserved in active sites of functionally identical model and template proteins.

Docking and residue interaction analysis

Molecular recognitions are vital to many biological processes. However, experimental determination of structures of molecular interactions is cost intensive, demand time, and expertise. We therefore, chose computational molecular docking to model our protein–ligand binding and characterize the interactions between its binding pocket residues and known active ligands. Multiple substrate specificities have been described previously for IPK2 gene product, catalyzing primarily 5GUG-I3P and 4A69-I0P (Stevenson-Paulik et al. 2002). We thus docked these centroid ligands into the binding cavity of GmIPK2 protein using molecular docking program Autodock Vina based on a semi-flexible docking approach with the scaling factor defined within 0.1 nm to predict their bound geometry. VINA uses its iterated local search global optimizer algorithm to produce 9 different poses of which pose 1 corresponding to each ligand was identified as the best binding mode based on their lowest binding affinity score of − 6.2 kcal/mol for 5GUG-I3P and − 5.8 kcal/mol for 4A69-I0P computed by VINA’s default statistical scoring function. We further inspected the molecular interactions between these protein substrate poses using Discovery Studio to predict functionally important amino acid residues and found that both the ligands were stabilized in their active site area by strong hydrogen-bonding interactions (Fig. 9a, b). Lys105 and Lys126 were identified as H-donors to the phosphate group oxygen of both 5GUG-I3P and 4A69-I0P while Arg153 a H-acceptor for hydrogen bonds formation suggesting that these binding pocket residues may play a pivotal role in enzymes function and protein structure stability. Based on the previous work conducted by Holmes and Jogl, we hypothesize that side chains of these amino acid residues may assist in inducing conformational changes on inositol phosphate binding, enabling the enzyme to interact with differently phosphorylated inositol polyphosphates in different orientations, thus endorsing its substrate versatility. Besides Ser217 was also found to form stabilizing hydrogen bond with I0P ligand. Gln157 and Pro140 were observed to form an unconventional carbon–oxygen hydrogen bond with 5GUG-I3P and 4A69-I0P respectively indicating their possible contribution to ligand binding affinity and ligand recognition (Klaholz and Moras 2002). Lengths and angles of hydrogen bonds stabilizing the GmIPK2-I3P and GmIPK2-I0P complexes are enlisted in Table 5a, b respectively. Moreover, our analysis indicates that Thr110, Lys138 and Lys154 show non-bonding interactions with the ligands. 2D protein GmIPK2-I3P/I0P ligand interaction diagrams (Fig. 9c, d) were also generated using Discovery Studio.

Fig. 9.

Molecular docking of substrates to the GmIPK2_S homology model. Hydrogen-bonding interactions of a 5GUG-I3P and b 4A69-I0P substrates with residues in the active site of GmIPK2_S protein. The substrate is depicted in stick representation with carbon atoms coloured turquoise, oxygen atoms red, hydrogen atoms yellow and phosphorus atoms blue. The interacting protein side chains are represented as maroon sticks in a and green sticks in b. c, d 2D-schematic representation of the interactions shown in a, b respectively drawn using Discovery Studio Visualizer

Table 5.

Hydrogen bonds between the active site residues of GmIPK2 and its substrates (a) 5GUG-I3P and (b) 4A69-I0P along with their distances and angles measured using Accelrys Discovery Studio Visualizer 4.1

(a)
GmIPK2			5GUG-I3P		Distance (A°)	Angle (Degree°)
Residue	Atom	Chemistry	Atom	Chemistry
Lys105	HZ3	H-Donor	O7	H-Acceptor	2.074	49.9029
Lys126	HZ3	H-Donor	O14	H-Acceptor	1.852	44.2482
Arg153	O	H-Acceptor	H39	H-Donor	2.168	45.6080
Gln157	CA	H-Donor	O18	H-Acceptor	3.348	52.1869

(b)
GmIPK2			4A69-I0P		Distance (A°)	Angle (Degree°)
Residue	Atom	Chemistry	Atom	Chemistry
Lys105	HZ2	H-Donor	O18	H-Acceptor	2.693	47.8602
Lys105	HZ3	H-Donor	O24	H-Acceptor	2.078	28.0171
Lys126	HZ3	H-Donor	O28	H-Acceptor	2.179	35.7367
Lys126	O	H-Acceptor	H40	H-Donor	2.121	39.3959
Pro140	CA	H-Donor	O4	H-Acceptor	3.732	60.3277
Arg153	O	H-Acceptor	H36	H-Donor	2.187	16.0359
Ser217	HG	H-Donor	O27	H-Acceptor	2.899	22.5987

Molecular dynamics simulation of GmIPK2-I3P and GmIPK2-I0P complexes

We subsequently subjected our docked protein–ligand complexes to MD simulation using GROMOS96 43A1 force field to understand their stability and dynamics. The trajectories obtained were utilized to construct their respective RMSDs, RMSFs and H-bond interactions. The GmIPK2-I3P and GmIPK2-I0P complexes exhibited a deviation between ~ 0.45–0.65 nm (Fig S6) and ~ 0.45–0.58 nm (Fig S7a), respectively, that converged to ~ 0.6 and ~ 0.55 nm, respectively, at 50 ns. This suggests that the structures were stabilized following simulation. From their RMSF plots, we observed fluctuations up to ~ 0.38 nm in the GmIPK2-I3P complex (Fig S6b) and ~ 0.29 nm in the GmIPK2-I0P complex (Fig S7(b)) which reveal the characteristic regional flexibilities of functional significance in each complex. We also analyzed the hydrogen bonds which participate in the maintenance of these complexes. Three main hydrogen bonds help to stabilize the ligands, 5GUG-I3P and 4A69-I0P, within the enzyme’s active site, two of them acting as acceptors and only one as a donor. In particular, the hydrogen bonds formed with Lys105 and Lys126 and the oxygen atom of each ligand’s phosphate group, were constantly held. Figure 10 shows the time evolution of these main H-bonds. Other less significant hydrogen bonds were formed sporadically between the ligands and the protein molecule. Lastly, we calculated the binding free energies (ΔG_binding) of both the complexes using g_mmpbsa for their respective trajectories. ΔG_binding values of − 12.4 and − 11.9 kcal/mol were estimated for GmIPK2-I3P and GmIPK2-I0P complexes respectively. This indicates that the binding of ligand molecules is thermodynamically favourable and thus validates the reliability of our simulation.

Fig. 10.

Time evolution of main hydrogen bonds formed between the GmIPK2_S protein and the 5GUG-I3P and 4A69-I0P ligands over 50 ns MD simulation

Conclusion

The present work describes expression analysis and molecular characterization of IPK2 multifunctional kinase involved in the late phase of PA biosynthesis. To initiate the study, we identified and cloned the partial gene sequence of IPK2 from G. max cv. Pusa-16. IPK2 transcripts, when assessed during seed development showed a predominant expression in the later stages of its development which consequently suggests that perturbing IPK2 gene at this stage can be a viable strategy for manipulating PA levels in soybean seeds to achieve a lpa trait. Computational analysis of the gene further highlighted its molecular features including cis-acting promoter elements potentially regulating its observed expression; primary and secondary structural features shedding light on its physicochemical features, conserved functional motifs and major structural elements; and its evolutionary relationship, which can be used to design several experimental studies. The 3D model of IPK2 protein developed can assist in the experimental determination of its 3D model as well as can be considered as a working model for generating hypothesis to make more accurate predictions of protein function and catalytic mechanism in the future. The docking studies performed subsequently can be used for structure-based designing of potent IPK2 inhibitors, useful in studying IPK2 active site and substrate selectivity and for studying PA biosynthesis pathway in detail. In conclusion, the obtained results provide very important preliminary data needed to manipulate PA content in soybean seeds as well as other crops for improving their nutritional quality by biotechnological intervention in the future.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Compliance with ethical standards

Conflict of interest

The authors declare that there is no conflict of interest regarding the publication of this article.