Experimental and bioinformatic characterization of a recombinant polygalacturonase-inhibitor protein from pearl millet and its interaction with fungal polygalacturonases.

Polygalacturonases (PGs) are hydrolytic enzymes employed by several phytopathogens to weaken the plant cell wall by degrading homopolygalacturonan, a major constituent of pectin. Plants fight back by employing polygalacturonase-inhibitor proteins (PGIPs). The present study compared the inhibition potential of pearl millet PGIP (Pennisetum glaucum; PglPGIP1) with the known inhibition of Phaseolus vulgaris PGIP (PvPGIP2) against two PGs, the PG-II isoform from Aspergillus niger (AnPGII) and the PG-III isoform from Fusarium moniliforme (FmPGIII). The key rationale was to elucidate the relationship between the extent of sequence similarity of the PGIPs and the corresponding PG inhibition potential. First, a pearl millet pgip gene (Pglpgip1) was isolated and phylogenetically placed among monocot PGIPs alongside foxtail millet (Setaria italica). Upstream sequence analysis of Pglpgip1 identified important cis-elements responsive to light, plant stress hormones, and anoxic stress. PglPGIP1, heterologously produced in Escherichia coli, partially inhibited AnPGII non-competitively with a pH optimum between 4.0 and 4.5, and showed no inhibition against FmPGIII. Docking analysis showed that the concave surface of PglPGIP1 interacted strongly with the N-terminal region of AnPGII away from the active site, whereas it weakly interacted with the C-terminus of FmPGIII. Interestingly, PglPGIP1 and PvPGIP2 employed similar motif regions with few identical amino acids for interaction with AnPGII at non-substrate-binding sites; however, they engaged different regions of AnPGII. Computational mutagenesis predicted D126 (PglPGIP1)-K39 (AnPGII) to be the most significant binding contact in the PglPGIP1-AnPGII complex. Such protein-protein interaction studies are crucial in the future generation of designer host proteins for improved resistance against ever-evolving pathogen virulence factors.

Introduction

Pectin, a galacturonic acid-rich complex polysaccharide found in all plant cell walls (CWs), is composed of homogalacturonan (HG), xylogalacturonan, apiogalacturonan, rhamnogalacturonan I, and rhamnogalacturonan II (Vorwerk ). HG, a linear homopolymer of α-1,4-linked d-galactopyranosyluronic acid with varying extents of methylation and acetylation, is the most abundant component (Mohnen, 2008). Phytopathogens are known to produce endo- and exo-polygalacturonases (PGs) that can breakdown HG (Van den Brink and De Vries, 2011). As a counter stratagem, plants employ polygalacturonase-inhibitor proteins (PGIPs) to modulate PG activity leading to an accumulation of elicitor-active oligogalacturonides (De Lorenzo ). PGIPs are CW-bound glycoproteins belonging to the extracyoplasmic leucine-rich repeat (LRR) family of proteins (Shanmugam, 2005). PGIPs are very diverse in their PG-inhibition specificity and potential with varying degrees of inhibition (Gomathi and Gnanamanickam, 2004).

PG–PGIP complexes are considered a model protein–protein interaction system in the backdrop of plant–pathogen interactions (Misas-Villamil and Van der Hoorn, 2008). Although three-dimensional structures of many PGs have been elucidated to date (Pickersgill ; Federici ; Bonivento ), the only PGIP whose crystal structure has been solved is that of PvPGIP2 from Phaseolus vulgaris (Di Matteo ). Most of the data available on the PG–PGIP interactions has been a result of studies involving PvPGIP2. Previous studies employed targeted mutation of pg and pgip genes, and investigated the in vitro inhibition behaviour of the protein variants synthesized to identify the amino acid residues involved in the protein–protein interactions (Leckie ; Mattei ; Raiola ). Amide-exchange mass spectrometry in combination with protease protection and fluorescence spectrometric analysis was employed to deduce the amino acids of AnPGII, a PG isoform II from Aspergillus niger, required for interaction with PvPGIP2 (King ). The availability of advanced bioinformatic tools for protein homology modelling and docking have been exploited in in-depth analysis of PG–PGIP complexes in silico and found to be in conformity with the experimental results (Lim ; Maulik ).

In contrast to the magnitude of literature available on dicot PGIPs, information available in case of monocots is meagre. Although PGIPs from wheat and rice have been tested for inhibition against various PGs (Jang ; Kemp ; Janni , 2013), no efforts have gone into understanding their mode of inhibition and the underlying structural basis of their interaction with PGs. In addition, no attempt has been made towards characterization of PGIPs from millets, small-grained gramineous monocots. Pearl millet [Pennisetum glaucum (L.) R. Br.; synonym: Cenchrus americanus (L.) Morrone], is among the most important cereal crops grown in the semi-arid tropical regions of Africa and the Indian subcontinent (Sehgal ).

In the present study, the gene encoding pearl millet PGIP was isolated and expressed heterologously as a maltose-binding protein (MBP) fusion in Escherichia coli. The purified recombinant fusion protein was employed in in vitro inhibition studies against two fungal PGs, AnPGII and FmPGIII (PG isoform III from Fusarium moniliforme isolate PD). PvPGIP2, the most wide-spectrum and potent inhibitor of fungal PGs (Farina ), has been shown to inhibit AnPGII (Stotz ) and is reported to be ineffective against FmPGIII (Sella ). A study of the inhibition profile of pearl millet PGIP against the same two PGs was carried out to elucidate the relationship between the extent of sequence similarity and the corresponding ability to inhibit PG. Furthermore, in the present study, in silico protein modelling, docking, and mutation analyses were carried out to explain the in vitro results, gain an understanding of the underlying structural basis of interaction, and predict the putative amino acids involved. To the best of our knowledge, this is the first report on the production of recombinant PGIP from millets and exploration of its inhibitory potential.

Materials and methods

Plant material

Seeds of pearl millet (P. glaucum) cultivar IP18296, obtained from the International Crops Research Institute for the Semi-Arid Tropics, Hyderabad, India, were used in this study. The seeds were germinated on moist germination sheets; 2-d-old seedlings were harvested and stored at –80 °C till further use.

Isolation of nucleic acids

Total RNA and genomic DNA were extracted from 2-d-old pearl millet coleoptiles using a Total Plant RNA Isolation kit (Sigma) and a GeneJET™ Plant Genomic DNA Purification Mini kit (Thermo Scientific), respectively, as per the manufacturer’s instructions.

Isolation and cloning of partial pgip genes

The cDNA was prepared by means of a two-step AMV RT-PCR kit (Qiagen) as per the manufacturer’s instructions, using total RNA from pearl millet coleoptiles as template. For PCR amplification of the partial pgip genes, primers Par1For (5ʹ-CTCGACCTCTCCTTCAACTC-3ʹ)/Par1Rev (5ʹ-ATGCCGCC GTAGATGGCGTT GTG-3ʹ) and Par2For (5ʹ-TGCGACTGGTA CGACGTCGACTG-3ʹ)/ Par2Rev (5ʹ-TCGCCACCTGCGCCGGG ATG-3ʹ) were designed based on the consensus region obtained by alignment of known monocot pgip gene sequences (GenBank accession nos AM180652–AM180657, NP_001147231, and XP_002439099) (Clone Manager Professional 9, Sci-Ed software). The amplification using Taq DNA polymerase (Merck Biosciences) resulted in ~400bp (Pglpgip1p) and ~500bp (Pglpgip2p) bands, respectively. They were gel purified using a QIAquick Gel Extraction kit (Qiagen), cloned in pTZ57R/T (InsTA clone PCR Cloning kit; Fermentas), and sequenced at Eurofins MWG Operon, Germany. Nucleotide BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi) searches were performed using default parameters to ascertain the gene identity.

Southern blot analysis

Pearl millet genomic DNA (10 µg) was digested with restriction endonucleases separately: ApoI (New England Biolabs, Germany), AhdI, BamHI, EcoRI, HindIII, KpnI, MscI, NaeI, SacI, XbaI, and XhoI (Thermo Scientific) and separated electrophoretically on a 0.7% agarose gel as described by Sambrook and Russell (2001). The DNA was transferred to a positively charged nylon membrane (Roche) as per the manufacturer’s instructions. As a hybridization probe, PCR-amplified Pglpgip1p was random-prime labelled using a DIG High Prime DNA Labeling and Detection Starter kit II (Roche). Pre-hybridization, hybridization, and chemiluminescent detection were performed as described previously (Wagenknecht and Meinhardt, 2011). Hybridization was carried out at 63 °C.

Inverse PCR

Pearl millet genomic DNA (3 µg) was digested using ApoI (New England Biolabs) and the digest mix was cleaned up using a NucleoSpin® Gel and PCR Clean-up kit (Macherey-Nagel). A 500ng aliquot of the cleaned-up digest was self-ligated using a Rapid DNA Ligation kit (Thermo Scientific).

The following two sets of inverse PCR primers, Inv1A (5ʹ-AC GCCTTCAGCTTCAACCTCTC-3ʹ)/Inv1B (5ʹ-TTGTGCGACA GCACTAGGGATG-3ʹ) and Inv2A (5ʹ-AGAGGTAGATCTGGT CGGCG-3ʹ)/Inv2B (5ʹ-CGCCAATTTCGCGCACC-3ʹ) were designed based on the Pglpgip1p and Pglpgip2p sequences, respectively. The PCR was carried out using Phusion® Hot Start High-Fidelity DNA polymerase (Finnzymes) with 30ng of self-ligated DNA as template. To the blunt-end PCR product of ~2kb obtained with Inv1A/Inv1B, an ‘A’ overhang was attached using Taq DNA polymerase (Thermo Scientific). It was further cloned in the pGEM-T Easy vector (Promega) and sequenced at Eurofins MWG Operon, Germany.

Sequence assembly

Sequence assembly (Clone Manager Professional 9, Sci-Ed software) was carried out using sequences obtained by inverse PCR and the Pglpgip1p sequence to determine the open reading frame (ORF) (Pglpgip1). Nucleotide BLAST searches were performed to ascertain the gene identity.

Bioinformatic analysis of Pglpgip1

The deduced amino acid sequence of Pglpgip1 (PglPGIP1) was further subjected to various in silico analyses using different bioinformatics tools: the signal peptide was identified using SignalP 4.1 (Petersen ) and the targeted protein localization was predicted using WoLF PSORT (Nakai and Horton, 1999). The assignment of domains in PglPGIP1 was performed based on the NCBI conserved domain search analysis (Marchler-Bauer ) and on a comparative amino acid sequence alignment with PvPGIP2 domain architecture (Di Matteo ) using the T-Coffee tool (Notredame ). The putative N-glycosylation sites were determined using NetNGly 1.0 (http://www.cbs.dtu.dk/services/NetNGlyc/).

The PglPGIP1 and other known monocot and dicot PGIP sequences (protein accession numbers are summarized in Supplementary Table S1 at JXB online) from the National Center for Biotechnology Information database were aligned with MUSCLE version 3.7 (Edgar ) and gaps and/or poorly aligned regions were removed with Gblocks version 0.91b (Talavera and Castresana, 2007). A phylogenetic tree was generated using the maximum-likelihood method employed in PhyML version 3.0 aLRT (Anisimova and Gascuel, 2006) on the http://www.phylogeny.fr platform (Dereeper ) using default settings. Internal branch consistency was appraised using the bootstrapping method with 100 bootstrap replicates. The tree was rendered using TreeDyn version 198.3 (Chevenet ).

The nucleotide sequence upstream of the ORF was submitted to PlantCARE (Lescot ) to determine the presence of plant-specific cis-elements.

Construction of the PglPGIP1 and FmPGIII expression plasmids

The multistep cloning strategy involved in construction of the PglPGIP1 expression plasmid (pET-22b::MBP-IEGR-PglPGIP1-6×His-Strep-tag® II, where IEGR is the factor Xa protease cleavage site and 6×His is a hexa-histidine tag) (construct E), vector control (pET-22b::MBP-IEGR-6×His-Strep-tag® II) (construct F) and the FmPGIII expression (pET-22b::FmPGIII-Strep-tag® II) (construct G) plasmids have been detailed in Supplementary Table S2 at JXB online (Supplementary Fig. S1 at JXB online provides a pictorial representation of the constructs). Briefly, the Pglpgpip1 coding sequence was initially cloned in pET-22b(+) in frame with the vector-encoded 6×His sequence. Then, malE (MBP-encoding gene), followed by a factor Xa protease recognition site was cloned upstream of pgip, and a Strep-tag® II-encoding sequence was cloned downstream of 6×His, allowing the synthesis of the fusion protein MBP–IEGR–PglPGIP1–6×His–Strep-tag® II. The vector control was generated by eliminating the pgip sequence from construct E. FmPGIII was PCR amplified from the pGEMT-FmPGIII construct and subcloned upstream of the Strep-tag® II in pET22b(+)-Strep-tag® II, allowing the synthesis of the FmPGIII–StrepII fusion protein. DNA manipulations such as agarose gel electrophoresis and bacterial transformation were carried out using standard protocols (Sambrook and Russell, 2001). Restriction digestion (New England Biolabs), plasmid isolation (InnuPREP Plasmid Mini/Midi kit; Analytik Jena Biosciences), ligation (Rapid DNA Ligation kit; Thermo Scientific), and gel extraction (NucleoSpin® Gel and PCR Clean-up kit; Macherey-Nagel) were carried out using kits according to the manufacturer’s instructions.

Production and purification of PglPGIP1 and FmPGIII fusion proteins

Competent E. coli SHuffle® T7 Express [pLysSRARE2] was transformed with the above-described constructs. Expression was carried out in 2 l batch cultures incubated for 24h at 26 °C using auto-induction solutions ‘M’ and ‘5052’ as described by Studier (2005). The total protein was extracted by a freeze–thaw cycle inducing the lysozyme-mediated autolysis of the cells. Additionally, sonication on ice at 40% amplitude, three times for 1min each (10 s on/10 s off cycles) on a Branson Digital Sonifier 250-D (G. Heinemann Ultraschall- und Labortechnik, Germany) was performed. The soluble protein was obtained as supernatant by centrifugation of the cell lysate at 40,000 g for 30min at 4 °C.

All purification steps were carried out at 10 °C on a FPLC system (ÄKTAExplorer; GE Healthcare, Freiburg, Germany). A flow rate of 1ml min–1 was maintained throughout. The protein peaks were pooled appropriately after each purification step and concentrated using centrifugal concentrators (Vivaspin™20; Sartorius). The intermediate buffer exchanges and desalting steps were carried out using 5ml HiTrap columns (GE Healthcare).

The fusion protein MBP–IEGR–PglPGIP1–6×His–Strep-tag® II (rPglPGIP1) was purified in two steps. In the first step, Strep-tag® II-based affinity purification was performed on a 1ml Strep-Tactin Superflow Plus Cartridge (Qiagen) using 50mM NaH2PO4, 300mM NaCl (pH 8.0) as loading/wash buffer and loading buffer containing 2.5mM d-desthiobiotin (pH 8.0; IBA Lifesciences) as elution buffer according to the manufacturer’s instructions. The desalted eluates were reconstituted in cation-exchange column loading/wash buffer (buffer A: 50mM NaH2PO4, pH 8.0) and subjected to purification on a 1ml RESOURCE Q Cartridge (GE Healthcare). The matrix-bound proteins were eluted (buffer B: 50mM NaH2PO4, 1M NaCl, pH 8.0) by applying a stepwise NaCl gradient [0–22% (buffer A to B) in 15min, held for 2min; 22–24% in 5min, held for 2min; 24–28% in 5min, held for 2min; 28–100% in 20 min] to resolve the desired protein from contaminating proteins. The rPglPGIP1 eluted at 22–24% NaCl held for 5min.

Recombinant FmPGIII-Strep-tag® II (rFmPGIII) and the vector control MBP–IEGR–6×His–Strep-tag® II (rVC) were purified by single-step affinity purification on a Strep-Tactin Superflow Plus Cartridge as described above.

Protein production and purification were monitored by immunoblot analysis of extracted proteins using Strep-Tactin®–horseradish peroxidase conjugate as probe (IBA Lifesciences). The chemiluminescence detection of blots was carried out as described previously (Wagenknecht and Meinhardt, 2011). The protein concentration of different samples was determined using a BCA protein assay kit (Pierce) with BSA as the standard.

PGIP activity assays

AnPGII (5ng; kind gift from Mr Madhusudhan, University of Mysore, Mysore, India) and rFmPGIII (36ng) were incubated separately in a reaction volume of 200 µl with 0.1mg ml–1 of polygalacturonic acid substrate (Sigma) at 30 °C in 50mM sodium acetate buffer (pH 4.2 and 4.6, respectively). PG activity was determined by reducing end-group analysis according to Anthon and Barrett (2002). PGIP activity was assayed by measuring the activity of the PGs pre-incubated with rPglPGIP1 for 20min at 30 °C. PGIP activity was expressed as the percentage reduction in the number of reducing ends (in µkat mg–1 of protein) liberated by PGs in the presence and absence of PGIP. rVC served as the control.

The effect of various parameters such as inhibitor concentration (0.316–12.64nM rPglPGIP1), substrate concentrations (0.025–0.25mg ml–1) and pH (3.5, 4.0, 4.5, and 5.0) on enzyme inhibition was determined. The kinetic parameters were computed by fitting the Michaelis–Menten equation on initial rate experimental data by non-linear fitting using OriginPro7 (Originlab). In separate experiments, the temperature and pH stability of rPglPGIP1 were studied by pre-incubating them separately for 1h at temperatures ranging from 20 to 100 °C, and for 16h at pH values of 2.0–11.0 at 4 °C, respectively, after which they were reconstituted in the appropriate assay buffer and their inhibition potential was assayed at 30 °C. All experiments were performed twice each in triplicate. The data of a representative experiment was subjected to Tukey’s honestly significant difference (HSD) test following analysis of variance at P<0.05.

Homology modelling of PglPGIP1 and FmPGIII

The initial homology models of PglPGIP1 and FmPGIII were generated using the MODELLER 9.12 package (Eswar ), with the X-ray crystallographic structures of PvPGIP2 [Protein Data Bank identity (PDB ID): 1OGQ] and endo-polygalacturonase I (FmPGI) from F. moniliforme isolate FC-10 (PDB ID: 1HG8) serving as templates, respectively. The quality of the generated models was assessed using the Verify3D server (Eisenberg ). Post-refinement of structural models was carried out using the KoBaMIN server (Rodrigues ), and further geometric accuracy of the constructed models was evaluated using MolProbity 4.02b (Davis ). The obtained models were further energy minimized using GROMACS 3.0 (Lindahl and Hess, 2001) with a GROMOS96-53a6 force field to remove geometric inaccuracies and steric clashes with 1000 steepest descent steps terminated at the convergence of maximum force <1000 kJ mol–1.

Electrostatic charge distribution analysis

The adaptive Poisson–Boltzmann Solver program (Unni ) was employed to calculate the electrostatic distribution of the following individual proteins and complexes: PglPGIP1, AnPGII, FmPGIII, PglPGIP1–AnPGII and PglPGIP1–FmPGIII. PyMOL version 1.2r3pre (PyMOL Molecular Graphics System; Schrödinger, LLC) was used for the visualization of the surface representation.

Docking and energy minimization

To obtain accurate conformations of PglPGIP1–AnPGII and PglPGIP1–FmPGIII protein complexes, the docking simulation was carried out in two steps; in the first step, the protein pairs were subjected to GRAMM-X (Tovchigrechko and Vakser, 2006) for the initial global search of conformation of the two proteins in complex form and subsequently, such output was resubmitted to the Rosetta 3.4 server (Lyskov and Gray, 2008) for further optimization of docking results based on rigid body orientation and side-chain conformation. The three-dimensional structure of AnPGII (PDB ID: 1CZF) used in the present docking study was retrieved from the PDB, Research Collaboratory for Structural Bioinformatics.

For energy minimization, molecular dynamics simulations at room temperature were carried out using GROMACS 3.0 (Lindahl and Hess, 2001) with a GROMOS96-53a6 force field for bound PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes. Protein in complexed form was immersed in a cubic waterbox of space water model with box edges 1 Å from the complexed protein periphery. The overall electric charge of the system was compensated. The system was pre-equilibrated, which involved minimization using the steepest descent method. A molecular dynamics simulation was performed at constant volume and temperature for 1 ns. Snapshots were collected every 2 ps.

Protein–protein interaction map

Hydrogen bond, ionic, and hydrophobic interactions across protein–protein interfaces were determined by the Protein Interaction Calculator (PIC) server (Tina ). Interactions were visualized and analysed using molecular modelling programs PyMOL 1.6 and CHIMERA 1.8 (Pettersen ).

Computational mutagenesis

Computational alanine-scanning mutagenesis was carried out using the Robetta alanine-scanning server (Kim ) in order to identify the energetically favourable amino acids at the interface that are important in the stability of the PglPGIP1–AnPGII complex. This program scans the protein–protein interface for the ‘hotspots’ and evaluates the changes in the binding free energy of the complex by replacing each of the interface residue with alanine at the binding region.

The relative free energy of binding (ΔΔGbinding) was calculated as follows:

Positive values of ΔΔGbinding higher than 1 kcal mol–1 indicate that replacement by alanine is predicted to destabilize the complex.

Results and discussion

Isolation of pearl millet pgip genes

Initially, two partial pgip sequences: Pglpgip1p (407bp, GenBank accession no. GU474543.1) and Pglpgip2p (497bp, GenBank accession no. JQ425039), which shared a nucleotide sequence identity of 84%, were isolated from pearl millet using primers based on the consensus sequence of known monocot pgips. PGIPs from the same plant, irrespective of whether they are of monocot or dicot origin, share significant identity (Favaron ; Ferrari ; Janni ).

An inverse PCR approach was employed for the isolation of the complete pgip coding sequence. For this, a Southern blot analysis was performed to identify appropriate restriction endonucleases for template preparation. About 10 enzymes were tested; however, they did not or only partially cut the genomic pearl millet DNA. Only ApoI yielded a continuous digestion pattern and distinct hybridizing fragments of appropriate size (Supplementary Fig. S2 at JXB online) and thus was selected to prepare the template. Of the two primer pairs used in the inverse PCR, only InvPGIP1A/InvPGIP1B amplified an intense fragment of ~2.0kb, which was further cloned and sequenced. The sequence assembly resulted in a 2295bp contig (GenBank accession no. JF421287) with an ORF of 1014bp (Pglpgip1) (Supplementary Fig. S3 at JXB online). The pearl millet coding sequence was found to be uninterrupted, i.e. without any introns, upon comparison of the DNA and cDNA sequences. Most known PGIP-encoding genes have been shown to lack introns; however, exceptions do exist but only in case of dicots such as Arabidopsis, peach, raspberry and Antirrhinum majus (De Lorenzo ).

The Southern blot analysis of pearl millet genomic DNA and the isolation of two PGIP-encoding partial sequences indicated the possible occurrence of pgip in pearl millet as a small multigene family. However, further genome organization studies are necessary to confirm this. It has been reported that most plant pgip genes exist as small multigene families clustered within a specific chromosomal region (De Lorenzo ). Brassica napus, in contrast, is the only plant reported so far to contain a large gene family with at least 16 pgips (Hegedus ).

The bioinformatic characterization of Pglpgip1 was carried out to understand the domain architecture of the deduced protein, its phylogenetic position among known PGIPs, and nature of cis-regulatory motifs upstream of the ORF. Only 130 (37%) amino acids were identical between PglPGIP1and PvPGIP2; however, they were found to have a global alignment score of 95 (Supplementary Fig. S4 at JXB online). A putative signal peptide of 27 aa was identified (Fig. 1) for possible apoplastic localization (82% affirmative), which is crucial in plant defence against the invading microbial PGs encountered at the CW (Di Matteo ). A central domain was detected in PglPGIP1, which contained 10 imperfect LRRs of an ~24 aa residue consensus sequence characteristic for PGIPs, xxLxLxx.NxLx..GxIPxxLxxL.xxL (Di Matteo ). The pearl millet PGIP also contained two putative β-sheets, B1 and B2, and also the 310-helix similar to PvPGIP2. The central LRR domain has been shown in PvPGIP2 to fold into a right-handed superhelix and is connected to the B1-sheet through loops or β-turns. The B2-sheet is a unique feature not found in other LRR proteins (Di Matteo ). In PglPGIP1, the central domain was flanked by short N- and C- terminal regions with conserved C residues as reported for PvPGIP2, known to be involved in disulfide linkages necessary for the structural integrity of the protein (Di Matteo ). Seven putative N-glycosylation sites were predicted by NetNGly1.0 with a consensus sequence of N-x-S/T (where x is any amino acid except P), for N-glycosylation. Of the seven sites, positions 85 and 297 were conserved in all four rice PGIPs (OsPGIP1–4) and both wheat PGIPs reported previously (Janni ). N297 was the lone residue to be conserved even in PvPGIP2, whose glycosylation positions have been experimentally mapped (Mattei ). N288 was conserved in all rice and wheat PGIPs except in OsPGIP1. N258 was conserved in OsPGIP2 and OsPGIP3, whereas N177 was conserved only in the latter.

Fig. 1.

Sequence organization of derived amino acid sequences of PglPGIP1. The displayed sequence organization is a result of alignment of PglPGIP1 with PvPGIP2 whose secondary structure has been determined (Di Matteo ). The LRR consensus sequence ‘xxLxLxx.NxLx..GxIPxxLxxL.xxL’ is shown. Putative residues contributing to form the secondary structure elements are based on PvPGIP2 and indicated in light grey (sheets B1 and B2) and dark grey (310-helix). Putative glycosylation sites are indicated by an asterisk. The conserved C residues are highlighted in black. Numbering of amino acid residues is shown on the left.

Phylogenetic analysis of the deduced protein, PglPGIP1, and other available monocot and dicot PGIP sequences using the maximum-likelihood approach clearly showed monocot and dicot PGIPs forming separate clusters (Fig. 2). PglPGIP1, was placed among the monocot PGIPs but the millet PGIPs formed a separate branch. PglPGIP1 was found to share from 50% to just over 60% identity with most monocot PGIPs. The identity with dicots was, however, between 35 and 47%.

Fig. 2.

Phylogenetic tree showing the affiliation of PglPGIP1 among other known monocot and dicot PGIPs. The deduced amino acid sequences of PglPGIP1 and other known PGIP sequences obtained from GenBank were aligned using MUSCLE version 3.7, curated using the Gblocks version 0.91b, and then submitted to PhyML version 3.0 aLRT for phylogenetic analysis, and the tree was rendered using TreeDyn version 198.3. The position of PglPGIP1 is highlighted by a box. The branch support values are represented at branch points and the branch length scale is shown below the tree. The protein accession numbers are summarized in Supplementary Table S1.

Earlier evaluation of cis-regulatory elements of various plant pgip genes in silico detected elements responsive to light, wounding, salicylic acid, abscisic acid, fungal elicitors, ethylene, and various other motifs with unknown functions (Kumar ; Lu ). The 1078bp sequence upstream of the Pglpgip1 ORF was analysed in silico for the identification of cis-elements and showed the presence of numerous important regulatory motifs. The consensus motif sequences and their respective functions retrieved from PlantCARE have been summarized in Supplementary Table S3 at JXB online (Supplementary Fig. S5 at JXB online shows the position of the cis-elements in the upstream sequence). The presence of elements responsive to plant stress hormones (e.g. ACGT-containing abscisic acid response element, TC-rich repeats, TCA element, CGTCA motif, ethylene response elements) and anoxic stress (anaerobic response element) indicates a role for PGIP in pearl millet’s response to biotic and abiotic stress. In bean cultivar ‘Pinto’, PvPpgip2 was found to be upregulated in response to wounding, oligogalacturonides, glucan derived from Phytophthora megasperma f. sp. glycinea and salicylic acid (D’Ovidio ). A slightly higher accumulation of poplar Pdpgip4 transcripts was observed over that of Pdpgip2 transcripts at 1 and 4h after treatment with salicylic acid and H2O2, respectively (Cheng ). However, similar expression levels were reported for both pgip genes in case of methyl jasmonate treatment. In Chinese cabbage, pgip expression levels were increased in response to abiotic stresses such as water-logging, salt, cold treatment, and mechanical wounding (Ahsan ). Significant numbers of elements were found to be responsive to light (e.g. Box 1, G-Box, Specificity protein 1). The effect of light on transcription as well as post-transcriptional processes in plant growth and development has been reported previously (Gilmartin ). Plant defence genes have also been proposed to be circadian regulated, and biotic stress responses have been shown to be modulated by a light-sensing network in Arabidopsis (Mullineaux ; Genoud ; Wang ). A plant’s ability to acclimatize to changes in quality and intensity of light has been found to be critical in mounting an effective response against invading pathogens (Kulheim ). Given the presence of numerous light-responsive elements in Pglpgip1 and other known pgips, and the fact that pgips are known to be induced by both biotic and abiotic stresses, the relationship between light-induced pgip expression and plant defence makes perfect sense.

Production of recombinant PglPGIP1 and FmPGIII

rPglPGIP1 was produced in order to assess its potential inhibition against fungal PGs (rFmPGIII and AnPGII). Our initial attempts to synthesize these proteins in eukaryotic expression systems such as Nicotiana benthamiana and Pichia pastoris were not successful (data not shown). Furthermore, PglPGIP1 production in E. coli faced protein solubility and purification problems. Hence, several solubility and affinity purification fusion tags were tried out for the soluble production and purification of PglPGIP1 (data not shown). The N-terminal MBP and C-terminal Strep-tag® II combination of solubility and affinity purification tags, respectively, were found to be optimal for its expression in E. coli SHuffle® T7 Express (pLysSRARE2) and efficient purification. FmPGIII expression faced no solubility issues, and hence only a C-terminal Strep-tag® II was incorporated for purification purpose. Expression with pET-22b::MBP-IEGR-PglPGIP1-6×His-Strep-tag® II in E. coli SHuffle® T7 Express (pLysSRARE2) and purification of the resulting recombinant fusion protein by coupling an anion-exchange column downstream of Strep-Tactin Superflow Plus affinity matrix resulted in a homogenous pure protein, rPglPGIP1, of 79.1kDa (Supplementary Fig. S6A at JXB online). The two-step purification resulted in 505- and 1323-fold purification with protein recoveries of 37 and 33%, respectively.

The single-step affinity purification of rFmPGIII and rVC on a Strep-Tactin Superflow Plus Cartridge resulted in relatively homogenous pure proteins of 37.9kDa (Supplementary Fig. S6A) and 46.3kDa (Supplementary Fig. S6B), respectively. Single-step purification of rFmPGIII resulted in 320-fold enrichment with a protein recovery of 72%.

Immunoblot analysis confirmed the synthesis and successful purification of the recombinant proteins (Supplementary Fig. S6). In addition, tandem mass spectrometric analysis of trypsin-digested peptides of purified recombinant proteins was carried out. Subsequent correlation of the resulting spectral data using SEQUEST algorithm (Eng ) against a database containing information on theoretical digest of the respective fusion proteins confirmed their identity (data not shown).

In vitro inhibition profile of rPglPGIP1

The in vitro inhibition of rPglPGIP1 against AnPGII and rFmPGIII was carried out in order to compare it with the known inhibition profile of PvPGIP2 against the same two PGs. Intact MBP–PglPGIP1 fusion protein was employed in the inhibition studies as protease factor Xa digestion for cleaving off MBP resulted in random degradation of protein (data not shown). rPglPGIP1 displayed a differential activity profile against the fungal PGs used in the present study, with only a partial inhibition observed against AnPGII (Fig. 3A) and no inhibition against rFmPGIII. Hence, the effect of various parameters on AnPGII inhibition by rPglPGIP1 has been presented below. The rVC, as expected, showed no inhibition of either enzyme. Possibly, the lack of glycosylation of rPglPGIP1 produced in E. coli could be a factor responsible for the limited inhibition observed.

Fig. 3.

AnPGII inhibition assay. (A) Effect of inhibitor concentration. AnPGII (5ng) was assayed with and without inhibitor (rPglPGIP1/rVC) over a concentration range of 0.316–12.64nM and a graph with the enzyme activity over inhibitor concentration was plotted. (B) pH optimum. AnPGII (5ng) was assayed with and without inhibitor at a concentration (rPglPGIP1/rVC) of 3.16nM and a graph with the enzyme activity over pH units was plotted to determine the pH optima of inhibition. (C) pH stability. AnPGII (5ng) was assayed with and without inhibitor pre-incubated for 16h at pH values of 2.0–10.0 at 4 °C upon reconstitution in the assay buffer [at a concentration (rPglPGIP1/rVC) of 3.16 nM] and a graph with the enzyme activity over pH units was plotted to determine the pH stability of inhibitor. (D) Temperature stability. AnPGII (5ng) was assayed with and without inhibitor pre-incubated for 1h at temperatures ranging from 20 to 100 °C [at a concentration (rPglPGIP1/rVC) of 3.16 nM] and a graph with the enzyme activity over temperature was plotted to determine the temperature stability of inhibitor. The data points are means of a single experiment carried out in triplicates. Results are shown as means±standard error. Means designated with the same letter are not significantly different according to Tukey’s HSD test at P<0.05.

PvPGIP2 has been reported to competitively inhibit FmPGI by binding at its substrate-binding site (Leckie ), whereas FmPGIII, which shares 91.7% amino acid identity with FmPGI, evaded inhibition by the same PGIP (Sella ). Although a structural explanation is lacking, an attempt was made to explain the observed effect using amino acid sequence alignment. The chief difference between the FmPGIII and FmPGI consist of one short stretch of substitutions in the N-terminal region and five substitutions, two of which are close to the active site. These polymorphic sites may have led to subtle changes in the protein structure, thus preventing the accessibility of the substrate-binding site of FmPGIII for interaction with PvPGIP2 (Sella ). As mentioned earlier, although sharing 37% amino acid identity, the global alignment score of PglPGIP1 and PvPGIP2 is very high due to the degeneracy of residues in the LRR domain. Hence, lack of potency of PglPGIP1 against FmPGIII is understandable. Assessment of the PglPGIP1–FmPGI interaction could further be useful for a deeper comparative profiling of inhibition with PvPGIP2.

Effect of inhibitor concentration on AnPGII activity

No inhibition was observed at the lowest rPglPGIP1 concentration tested, but a gradual increase in inhibition was observed with increasing inhibitor concentration (Fig. 3A). A significant increase in inhibition was observed at rPglPGIP1 concentrations of 0.632, 1.26, and 3.16nM with inhibition being 7, 16, and 26%, respectively. However, a further increase of the rPglPGIP1 concentration (up to 6.32 and 12.64nM) resulted in marginally increased inhibition reaching a value of 28% at most. PGIP concentrations of 3.16nM and/or 1.26nM were used in further studies.

Although the pattern of inhibition of both rPglPGIP1 and PvPGIP2 against the same two PGs was similar, bean PGIP showed complete inhibition against AnPGII at concentrations lower than that used in the present study (Leckie ; Sella ). As observed in the present study, such partial inhibitions of various PGs have been reported in several PG/PGIP systems (Gomathi and Gnanamanickam, 2004). PGIPs from cotton showed 50 and 25% inhibition against PGs from Aspergillus niger and Verticillium dahlia, respectively (James and Dubery, 2006). Wheat PGIP, on the other hand, showed just over 5% inhibition of PGs from A. niger and F. moniliforme (Kemp ). PGIPs are very diverse in their inhibition specificity and potential, i.e. PGIPs isolated from a single plant can inhibit different PGs, the inhibitors isolated from different plants have been found to inhibit the same PG, some PGIPs have the ability to inhibit a multitude of PGs, and the degree of inhibition has also been found to vary greatly (Gomathi and Gnanamanickam, 2004). PGIPs have also been noted to have the potential to inhibit in vitro the PGs of those microbes that are non-pathogenic or to which they are generally not exposed, as in our case (Gomathi and Gnanamanickam, 2004). The presence of PGIPs with different inhibition potentials could be a ploy employed by the host ensuring the accumulation of elicitor-active oligogalacturonides.

Mode of AnPGII inhibition

The control protein, rVC (1.26 and 3.16nM) had no effect on the kinetic parameters of AnPGII (Table 1). rPglPGIP1 (1.26 and 3.16nM), however, was found to decrease the V max without affecting the K m of AnPGII. Hence, the mode of inhibition was established to be non-competitive in nature. PGIPs depending on the PG are known to employ competitive, non-competitive, and mixed modes of inhibition, thus indicating that PGIPs are capable of recognizing different structural motifs in its various protein partners (Sicilia ). PGIPs from different sources have been shown to display similar as well as different inhibition kinetics against the same enzyme. For example, BcPG from Botrytis cineria was competitively inhibited by pear PGIP but in a mixed mode by the inhibitor from bean (Abu-Goukh and Labavitch, 1983; Sicilia ). However, AnPGII was non-competitively inhibited by PGIPs of tomato and bean (Stotz ), as in our case.

Table 1.

The kinetic parameters of AnPGII with and without rVC and rPglPGIP1

AnPGII (5ng) was assayed using a substrate concentration range of 0.025–0.25mg ml–1 with and without inhibitors (rPglPGIP1/rVC) at concentrations of 1.26 and 3.16nM. The kinetic parameters were calculated by fitting the Michaelis–Menten equation on initial rate experimental data by non-linear fitting using OriginPro 7 (Originlab).

	K m (mg ml–1)	V max (μkat mg–1 protein)
AnPGII	0.091	28.5
AnPGII+rVC (1.26nM)	0.091	28.5
AnPGII+rVC (3.16nM)	0.092	28.6
AnPGII+rPglPGIP1 (1.26nM)	0.092	24.3
AnPGII+rPglPGIP1 (3.16nM)	0.091	21.3

pH optima of AnPGII inhibition

The AnPGII inhibition by rPglPGIP1 (3.16nM) over a pH range of 3.5–5.0 showed the pH optima for inhibition to be between 4.0 and 4.5 with 25 and 22% inhibition, respectively (Fig. 3B). However, no inhibition was observed at pH 3.5 and 5.0. The pH optima for inhibition of A. niger PG by PGIP-I and PGIP-II of guava fruit was determined to be 4.2, whereas that for PGIP-III was 4.4 when assayed over a pH range of 4.0–5.5 (Deo and Shastri, 2003). A study involving interaction of PvPGIP2 with five isoforms of A. niger PGs (AnPGI, AnPGII, AnPGA, AnPGB, and AnPGC) showed PvPGIP2 to greatly reduce the activity of three of the PG isoforms—AnPGI, AnPGII and AnPGB—at all tested pH values. For AnPGA and AnPGC, PvPGIP2 displayed inhibition only at pH 4.75 and below pH 4.2, respectively. At pH 5.0, however, PvPGIP2 was found to activate the two PG isoforms (Kemp ). This diversity of PGIP activity in different pH environments is crucial in countering the multitude of PGs encountered from various pathogens with different pH optima. In addition, such differential PGIP activity has been proposed to be involved in mounting a better host defence response through the generation of a steady-state concentration of biologically active oligogalacturonides (Kemp ).

pH and thermal stability of rPglPGIP1

rPglPGIP1 (at 3.16nM) was found to be stable over a wide range of pH values from 4.0 to 8.0 (Fig. 3C). The inhibition potential decreased by approximately 50% at pH 3.0 and 9.0, with no observed inhibition at other tested pH values. Guava PGIPs were reported to be stable only over a narrow pH range of 2.0–4.0 (Deo and Shastri, 2003), but the PGIP from chilli retained >50% activity at pH 3.0 and 8.0 (Shivashankar ).

rPglPGIP1 (at 3.16nM) was found to be equally active from 20 to 50 °C (Fig. 3D). The activity dropped very slightly to 22% at 60 °C and significantly at 70 and 80 °C to 11 and 3%, respectively. No inhibition was observed beyond this temperature. Earlier studies on thermal tolerance of the protein have shown it to be relatively thermostable. PGIPs from ‘Bartlett’ pear (Abu-Goukh ) and orange (Barmore and Nguyen, 1985) retained significant activity at 60 °C. The latest study on PGIP isolated from tomato was found to retain partial inhibitory activity, even at 100 °C (Schacht ).

In silico analysis of PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes

In silico protein modelling, docking, and computational mutagenesis studies were carried out to account for the differential behaviour of pearl millet PGIP against AnPGII and FmPGIII, and for the limited similarity in the inhibition of PglPGIP1 and PvPGIP2 against the same two PGs. It was necessary to predict the underlying structural basis of interaction as the two PGIPs share an amino acid identity of just 37%.

The initial homology models of PglPGIP1 (Supplementary Fig. S7A at JXB online) and FmPGIII (Supplementary Fig. S7B) were generated using already known structures of PvPGIP2 (PDB ID: 1OGQ) and FmPGI (PDB ID: 1HG8) serving as templates, respectively. The quality of the models was assessed using Verify3D. Scores predicted for each residue of our constructed model were >0.1, indicating that all of the residues were located in favourable structural environments. After post-refinement with the KoBaMIN program, the final models were revalidated using MolProbity. In the present study, the MolProbity score for both the models was >85%, representing a good quality model. Summarized results from KoBaMIN and MolProbity are given in Supplementary Table S4 at JXB online. Models further energy minimized using GROMOS96-53a6 force field to remove any local strains were finally used for the docking studies.

Docking studies of PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes

To predict the conformation and the putative interactions between PglPGIP1 with AnPGII and FmPGIII, two different protein–protein docking programs, GRAMM-X and Rosetta 3.4, were used. Firstly, using GRAMM-X, fast-Fourier-transformation-based unrestrained rigid body docking was performed, which generated the top 10 solutions of the complexes. In the absence of the detailed data on the binding mode or mutational studies on PglPGIP1, we selected the docked complex based on the present in vitro inhibition studies, as well as available experimental evidence from PvPGIP2 interactions with the two enzymes (Leckie ; Federici ; King ; Di Matteo ; Sella ; Maulik ; Benedetti ). For further optimization, selected models from GRAMM-X were subjected to Rosetta 3.4. This docking algorithm searches a set of conformations from a given starting conformation for the optimal fit between the two partners. It employs a Monte Carlo search followed by simultaneous optimization of side-chain conformations. The resulting ‘decoys’ obtained from the docking simulations were ranked using an energy function dominated by van der Waals interactions, an implicit solvation model and an orientation-dependent hydrogen bonding potential. Selection of the best docked conformation obtained from Rosetta 3.4 was based on: (i) docking score being an overall measure of the energy of the complex; (ii) an interface score representing the score of the complex minus the total score of each partner in isolation; and (iii) involvement of any residues in the interaction evident from the experimental studies from PvPGIP2. The best docked conformations of the PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes are shown in Fig. 4, and the docking and interface scores are presented in Supplementary Table S5 at JXB online.

Fig. 4.

Protein docking analysis. Docked poses of PglPGIP1–AnPGII (A) and PglPGIP1–FmPGIII (B) complexes. PglPGIP1 interacts through its solvent-exposed concave cavity with AnPGII and FmPGIII at their N- and C-termini (circled in black), respectively. The substrate-binding site in FmPGIII appears to be more exposed compared with that of AnPGII. (This figure is available in colour at JXB online.)

From the docked protein complexes, it was suggested that AnPGII (Fig. 4A) and FmPGIII (Fig. 4B) both interact with PglPGIP1 at its concave surface, but their binding orientations differ. The residues at the β-strand/β-turn motif of the central LRR domain constitute the solvent-exposed concave surface of the PGIP and this region determines the binding specificity for PGs (Kobe and Deisenhofer, 1994). LRRs are known to be versatile protein recognition domains present in over 14,000 proteins (Matsushima and Miyashita, 2012). In case of the PglPGIP1–AnPGII complex, the concave site of PglPGIP1 interacts with the N-terminal site of AnPGII, whereas in the PglPGIP1–FmPGIII complex, the concave site of PglPGIP1 interacts with the C-terminal site of FmPGIII. The substrate-binding site in FmPGIII appeared to be more exposed compared with that of AnPGII. In terms of the docking score, binding of PglPGIP1 with AnPGII was predicted to be stronger in comparison with FmPGIII. These in silico results are consistent with the in vitro outcomes. Interaction of PvPGIP2 with AnPGII and FmPGI showed some residues important for interaction with one PG to be dispensable for interaction with the other, suggesting that different but overlapping subsets of residues are vital in binding different ligands (Leckie ). The molecular docking simulation of the BcPG1–PvPGIP2 complex showed the B1-sheet of PvPGIP2 to interact with the N-terminus of BcPG1, and the active site of BcPG1 was partially buried by the PvPGIP2 C-terminus (Sicilia ). Analysis of the PvPGIP2–FmPG1 (now referred to as FpPG) complex showed the residues present at both the convex and concave side of the PGIP N-terminus to be involved in interaction with the loops surrounding the active site of the PG (Benedetti ). These studies further demonstrated the structural flexibility and the versatility of PGIP binding interactions with the various PGs.

Evaluation of protein–protein interactions in PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes by PIC analysis

Energy-minimized PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes were analysed using the PIC server to predict all possible types of interactions at the protein–protein interfaces, and the putative amino acid residues involved were mapped (Table 2). It was apparent from the data that the protein–protein contacts in both complexes were mediated through all types of interactions, i.e. ionic, hydrophobic, and hydrogen bonds, whereas the ionic and hydrophobic interactions were predominating at the surface. However, in the docked as well as energy-minimized structures, the PglPGIP1–AnPGII complex was found to have a stronger binding interaction than that recorded in the PglPGIP1–FmPGIII complex.

Table 2.

Protein interaction analysis of PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes using the PIC

The residue pairs involved in the interacting complexes, sorted according to the type of interaction, are shown (the PGIP residue numbering followed excludes the putative signal peptide).

Hydrophobic interactions
In PglPGIP1–AnPGII complex		In PglPGIP1–FmPGIII complex
PglPGIP1	AnPGII	PglPGIP1	FmPGIII
F54	Y130	W105	A306
M100	W85	W243	I332
I102	W85	L268	A330
W105	A43
F124	P56
F129	A40
A172	P56
V175	A36
Side-chain H-bonding interactions
T28	S234	R153	N266
H79	T64	D222	T332
N145, N147	E83
S195	E54
Q219	E54
Ionic interactions
D31	R233	D56	K269
D42	K124	D290	K300
D50	K127
R74	E83, E84
L77	D62
H79	D62
D126	K39
R240	E54

Electrostatic surface charge distribution on PglPGIP1, AnPGII, and FmPGIII individually, and the PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes

To illustrate the charge distributions of molecules, a three-dimensional electrostatic surface potential was generated separately on PglPGIP1 (Fig. 5A), AnPGII (Fig. 5B), and FmPGIII (Fig. 5C), as well as on the PglPGIP1–AnPGII (Fig. 5D) and PglPGIP1–FmPGIII (Fig. 5E) complexes.

Fig. 5.

Electrostatic surface potential of individual proteins and protein complexes. Electrostatic potential maps of PglPGIP1 (A), AnPGII (B), FmPGIII (C), PglPGIP1–AnPGII (D), and PglPGIP1–FmPGIII (E) complexes on which surface colours are fixed at red (–5) or blue (+5). Marked regions display the difference in charge distributions in surface maps of the individual proteins.

The surface comparison of AnPGII and FmPGIII suggested that the structures not only differed with respect to charge distribution but also differed in shape. This difference in the surface potential of the individual enzymes was also reflected in their predicted differential binding interactions with PglPGIP1. Once again, the PglPGIP1–AnPGII complex was more compact than PglPGIP1–FmPGIII, with the active site cleft of FmPGIII exposed to a greater extent, which further explicated the in vitro results. This is consistent with an earlier report that electrostatic and van der Waals interactions play a significant role in the proper recognition and discrimination of PGs by PGIPs (Maulik ).

Analysis of PglPGIP1–AnPGII interaction by computational alanine-scanning mutagenesis

The residues at the β-strand/β-turn motif of PvPGIP2 have been shown to interact with AnPGII at the D110 α-helix, opposite the substrate-binding site, which fits perfectly with the non-competitive mode of inhibition (Stotz ). To identify the hotspot residues involved in the interaction of PglPGIP1 and AnPGII, which also follows the non-competitive mode of inhibition, a computative alanine mutagenesis of residues at the interface was carried out. Virtual scanning was performed over all interface residues and changes in the binding free energy were calculated upon alanine substitution of residues at protein–protein interfaces.

From the existing experimental data on PvPGIP2, the amino acids H104, Y105, Y107, D131, V152, F201, Q224, and K225 have been reported to be important in interaction with AnPGII (Leckie ; Casasoli ; Spinelli ; Benedetti ). Alignment of PvPGIP2 and PglPGIP1 sequences identified PglPGIP1 residues T99, M100, I102, D126, N147, Q196, Q219, and I220, respectively, at positions corresponding to the PvPGIP2 residues mentioned above. Interestingly, computational mutagenesis predicted six of them to be hotspot residues significantly involved in interaction, except for I102 and I220, with ΔΔGbinding values of 0.61 and 0.87 kcal mol–1, respectively (Fig. 6A). In addition, all the identified residues could be pinned down to the β-strand/β-turn solvent-exposed region except for D31, N145, and R240 found localized at the N-terminus, LRR-4 and LRR-8 in the central domain (Fig. 1), respectively.

Fig. 6.

Computational alanine mutagenesis of PglPGIP1–AnPGII interface residues. The plot displays the contribution of individual interacting residues from PglPGIP1 (the PGIP residue numbering followed excludes the putative signal peptide) (A) and AnPGII (B) in the stability of the PglPGIP1–AnPGII complex. Interface residues were defined as those residues with a side chain having at least one atom within a sphere with 4 Å radius of an atom belonging to the other partner in the complex and binding hotspots defined as those residues that show ∆∆Gbinding >1 kcal mol–1.

The information on AnPGII residues involved in interaction with PGIPs is very limited. Studies on conformational changes in AnPGII–HG–PvPGIP2 using amide-exchange mass spectrometry identified four residues (E95, G104, D110, and I139) to be involved in PvPGIP2 binding (King ). The residues were found to lie opposite the substrate-binding site around the underside of the barrel near the D110 α-helix, consistent with the reported non-competitive mode of inhibition. However, the docking pose and protein–protein interaction analysis as well as the computational alanine-scanning mutagenesis of the PglPGIP1–AnPGII complex (Fig. 6A, B) identified that the N-terminal region of AnPGII interacts with the concave surface of PglPGIP1. Many of the identified residues were found to be localized mainly at the small α-helix, and β-strands of β-sheet PB1 at the N-terminus of AnPGII away from the substrate-binding surface, which again is consistent with the observed non-competitive mode of inhibition.

Most residues of PglPGIP1 and AnPGII identified by protein–protein interaction analysis as involved in interaction were also identified as significantly important upon computational alanine-scanning mutagenesis (Table 2 and Fig. 6). The D126 (PglPGIP1)–K39 (AnPGII) interaction was predicted by alanine mutation studies to be the most significant binding contact in the PglPGIP1–AnPGII complex.

The above results predicted the preferential use of very similar motif regions in PG recognition by the two PGIPs with few identical amino acids. PvPGIP2 shares 99 and 88% identity with PvPGIP1 and GmPGIP3, respectively, at the amino acid level. However, assessment of their inhibition against AnPGII and FmPGI, unexpectedly showed PvPGIP2 and GmPGIP3 to share similar inhibition profiles, but PvPGIP2 and PvPGIP1 did not (Leckie ; D’Ovidio ). Docking studies of these complexes proposed that not just sequence similarity but also conservation of key structural features are crucial in preserving the function mediated by appropriate protein-–protein interactions (Maulik ). Even though PglPGIP1 and PvPGIP2 employed similar motifs for interaction with AnPGII at non-substrate-binding sites, they engaged different regions of AnPGII. Subtle differences in the overlapping residues as well as recruitment of additional, completely different amino acids could be responsible for the observed binding disparity.

In silico alanine mutation studies predicted D126 (PglPGIP1)–K39 (AnPGII) interaction as the single most important binding contact in the PglPGIP1–AnPGII complex. This could be important as the D126 position is an evolutionarily conserved residue in PGIPs. The importance of a single amino acid in protein–protein interaction has been observed earlier. PvPGIP1 gained the ability to inhibit FmPGI through a single amino acid substitution of K224 into the corresponding amino acid of PvPGIP2, a Q (Leckie ). Hence, in silico analysis of the PglPGIP1–AnPGII complex is a good starting point for further experimental mutational analysis to arrive at the actual residues involved in protein–protein interactions.

In conclusion, the present study, together with earlier literature, suggests that the PG–PGIP interactions are complex, and structural and mutational analyses of various PG–PGIP complexes would be needed before a comprehensive generalized conclusion could be drawn about structure–function correlation. Pearl millet is afflicted by many fungal and bacterial diseases. Downy mildew disease alone accounts for annual production losses in the range of 20–40% (Singh, 1995). Such protein–protein interaction studies will be crucial from the perspective of generation of designer host proteins with improved combat potential against the ever-evolving pathogen virulence factors.

Supplementary data

Supplementary data are available at JXB online.

Supplementary Table S1. List of plant PGIPs used in the protein phylogenetic analysis.

Supplementary Table S2. Construction of PglPGIP1, vector control and FmPGIII expression plasmids for expression in Escherichia coli SHuffle® T7 Express (pLysSRARE2).

Supplementary Table S3. Upstream sequence analysis of the Pglpgip1 gene.

Supplementary Table S4. Refinement of PglPGIP1 and FmPGIII models.

Supplementary Table S5. Docking and interface scores of PglPGIP1–AnPGII and PglPGIP1–FmPGIII complexes.

Supplementary Fig. S1. Pictorial representation of the PglPGIP1, vector control and FmPGIII expression plasmids for expression in Escherichia coli SHuffle® T7 Express (pLysSRARE2).

Supplementary Fig. S2. Southern blot analysis of pearl millet total DNA.

Supplementary Fig. S3. Nucleotide and derived amino acid sequences of the pearl millet Pglpgip1 gene.

Supplementary Fig. S4. Alignment of PglPGIP1 and PvPGIP2 sequences using the T-Coffee multi-alignment tool.

Supplementary Fig. S5. Upstream cis-regulatory elements in the Pglpgip1 gene.

Supplementary Fig. S6. Purification of recombinant PglPGIP1, vector control and FmPGIII fusion proteins synthesized in Escherichia coli SHuffle® T7 Express (pLysSRARE2).

Supplementary Fig. S7. Homology modelling of PglPGIP1 and FmPGIII.