The Polygalacturonase-Inhibiting Protein PGIP2 of Phaseolus vulgaris Has Evolved a Mixed Mode of Inhibition of Endopolygalacturonase PG1 of Botrytis cinerea

Abstract

Botrytis cinerea is a phytopathogenic fungus that causes gray mold in >1,000 plant species. During infection, it secretes several endopolygalacturonases (PGs) to degrade cell wall pectin, and among them, BcPG1 is constitutively expressed and is an important virulence factor. To counteract the action of PGs, plants express polygalacturonase-inhibiting proteins (PGIPs) that have been shown to inhibit a variety of PGs with different inhibition kinetics, both competitive and noncompetitive. The PG-PGIP interaction promotes the accumulation of oligogalacturonides, fragments of the plant cell wall that are general elicitors of plant defense responses. Here, we characterize the enzymatic activity of BcPG1 and investigate its interaction with PGIP isoform 2 from Phaseolus vulgaris (PvPGIP2) by means of inhibition assays, homology modeling, and molecular docking simulations. Our results indicate a mixed mode of inhibition. This is compatible with a model for the interaction where PvPGIP2 binds the N-terminal portion of BcPG1, partially covering its active site and decreasing the enzyme affinity for the substrate. The structural framework provided by the docking model is confirmed by site-directed mutagenesis of the residues that distinguish PvPGIP2 from the isoform PvPGIP1. The finding that PvPGIP2 inhibits BcPG1 with a mixed-type kinetics further indicates the versatility of PGIPs to evolve different recognition specificities.

The necrotrophic plant pathogen Botrytis cinerea is the causal agent of gray mold and attacks a broad range of host species. The fungus causes fruit and leaf rot as well as flower blight in the field and in greenhouses and is responsible for important losses in the post-harvest chain (Barrie, 1994). During the initial stages of infection, B. cinerea secretes many hydrolytic enzymes to penetrate the plant cell wall (Jarvis, 1977). Among them, the endopolygalacturonases (PGs) cleave the α-1,4 linkages between d-galacturonic acid (d-GalUA) residues in the homogalacturonan, the main component of pectin, and cause the separation of cells from each other and the maceration of host tissue. PGs of B. cinerea are encoded by a gene family of at least six members that are differentially expressed (Wubben et al., 1999). While Bcpg1 and Bcpg2 are expressed at a basal constitutive level, GalUA induces the expression of Bcpg4 and Bcpg6, and xylogalacturonic acid induces the Bcpg5 gene. The expression of Bcpg3 is induced by lowering the pH of the culture medium, irrespective of the carbon source. Differential expression of the gene family allows the degradation of the homogalacturonan under various environmental conditions and according to the catabolic demands of the fungus (Wubben et al., 2000). Bcpg1 is required for full virulence, and its inactivation by gene replacement produces mutants with a significant decrease in the secondary infection on tomato (Lycopersicon esculentum), apple (Malus domestica), and bean (Phaseolus vulgaris; ten Have et al., 1998). Similarly, PGs are important pathogenicity factors for other fungi, such as Alternaria citri (Isshiki et al., 2001) and Claviceps purpurea (Oeser et al., 2002), and bacteria, such as Agrobacterium tumefaciens (Rodriguez-Palenzuela et al., 1991) and Ralstonia solanacearum (Huang and Allen, 2000).

The complete hydrolysis of the homogalacturonan by fungal PGs can be hampered by polygalacturonase-inhibiting proteins (PGIPs) localized in the cell wall of many plants. PGIPs belong to the large family of the Leu-rich repeat (LRR) proteins (Mattei et al., 2001; Di Matteo et al., 2003). The LRR structure is shared by many plant proteins involved in the recognition of pathogens, such as the majority of the resistance gene products (Martin et al., 2003), and receptors for pathogen-associated molecular patterns, such as FLS2 (Gomez-Gomez et al., 2001). Moreover, several receptors involved in development and perception of hormones contain LRRs of the same type as PGIPs (Becraft, 2002; Szekeres, 2003).

PGIPs inhibit and modulate the activity of fungal PGs and promote the accumulation of elicitor-active oligogalacturonides at least in vitro (Cervone et al., 1989; De Lorenzo et al., 2001; De Lorenzo and Ferrari, 2002). A direct role for PGIPs in plant defense was demonstrated recently. The overexpression of PGIPs in tobacco (Nicotiana tabacum), tomato, or Arabidopsis (Arabidopsis thaliana) confers enhanced resistance to B. cinerea infection (Powell et al., 2000; Ferrari et al., 2003; Manfredini et al., 2005).

The presence of small gene families coding for PGIPs accounts for the different inhibiting activities present in plants (Frediani et al., 1993; Stotz et al., 1993; D'Ovidio et al., 2004). Not only do PGIPs from different plants differ in their inhibitory activity, but PGIPs from a single plant often exhibit specificities against PGs from different fungi or different PGs from the same fungus (Desiderio et al., 1997). In P. vulgaris for instance, four genes code for PGIPs (PvPGIP1–4), and among them PvPGIP2 inhibits PGs from both Fusarium moniliforme (FmPG) and Aspergillus niger (AnPGII), whereas PvPGIP1 is effective only against AnPGII (Leckie et al., 1999). PvPGIP3 and PvPGIP4 are less efficient inhibitors of fungal PGs but inhibit PGs from the insects Lygus rugulipennis and Adelphocoris lineolatus (D'Ovidio et al., 2004). Recently, it was shown that PvPGIP2 is also the most effective inhibitor of BcPG1 (Manfredini et al., 2005).

Variability of recognition and function of PGIPs is not only reflected by their specificity but also by the different inhibition kinetics against fungal PG: tomato PGIP inhibits AnPGII in a noncompetitive manner (Stotz et al., 2000); PGIPs from bean (Lafitte et al., 1984) and raspberry (Rubus idaeus; Johnston et al., 1993) are noncompetitive inhibitors of PGs from Colletotrichum lindemuthianum and B. cinerea, respectively; PGIPs from pear (Pirus communis; Abu-Goukh et al., 1983) and orange (Citrus sinensis) fruits (Barmore and Nguyen, 1985) are competitive inhibitors of PG from Diplodia natalensis. Importantly, PvPGIP2 was previously found to act competitively against FmPG (Federici et al., 2001) and with a noncompetitive mechanism against AnPGII (King et al., 2002). This suggests that PvPGIP2 is a versatile protein capable of recognizing different structural motifs in its various protein partners. Here, we show that PvPGIP2 inhibits BcPG1 activity with a mixed-type kinetics. This is compatible with a model where the inhibitor and the substrate bind the enzyme at the same time. Molecular docking simulations and site-directed mutagenesis allowed us to build a model of the complex, which accounts for the mixed-type mechanism of inhibition.

RESULTS

PvPGIP2 Inhibits BcPG1 with a Mixed-Type Mechanism

BcPG1, expressed in Pichia pastoris and purified as already described (Manfredini et al., 2005), has a molecular mass of 36 kD and a pI of 8.1. Thermal stability of the enzyme was determined by following the loss of enzyme activity at 30°C, 37°C, 45°C, and 50°C as a function of time. An increase of temperature from 30°C to 37°C decreased the rate of hydrolysis to 40% after 180 min of incubation. At 50°C, the rate of hydrolysis was reduced to 60% after only 5 min of incubation and was almost completely abolished after 30 min (Fig. 1A). The pH optimum was determined by measuring the enzyme activity in the range of values from 3 to 8. BcPG1 activity showed a pH-dependent curve with a broad increase at acidic pH and a maximum at pH 5.5 (Fig. 1B). At pH higher than 5.5, the activity sharply decreased. BcPG1-specific activity, using polygalacturonic acid as a substrate, was 998 reducing group units (RGU)/mg, a value twice higher than that of FmPG (500 RGU/mg; Caprari et al., 1996) and twice lower than that of AnPGII (2,000 RGU/mg; Armand et al., 2000).

Figure 1.

A, Thermal stability of BcPG1. The enzyme activity is shown at different temperatures as a function of time. B, Enzymatic activity of BcPG1 as a function of pH. The following buffers were used: 50 mm sodium citrate (from pH 3.0 to 4.0), 50 mm sodium acetate (from pH 4.5 to 6.0), and 50 mm phosphate buffer (from pH 6.5 to 8.0).

The enzyme activity was measured at different concentrations of polygalacturonic acid (from 0.01 up to 15.0 mg/mL). The K_m value, calculated by nonlinear regression analysis of the experimental values, was 0.24 mg/mL (Table I). This is an intermediate value between the K_m values determined for AnPGII (<0.15 mg/mL; Armand et al., 2000) and FmPG (0.56 mg/mL; Federici et al., 2001). The experiments were then performed using the same amount of enzyme (0.02 ng/μL) in the presence of increasing concentrations of PvPGIP2 (0.03 and 0.07 ng/μL, respectively). As shown in Table I, the values of V_max decreased in the presence of increasing concentrations of PvPGIP2. The values of K_m instead increased by adding increasing amounts of PvPGIP2. The observed variation of both K_m and V_max is characteristic of a mixed mode of inhibition. This type of inhibition differs from the noncompetitive one because the dissociation constant of the substrate from the enzyme-inhibitor-substrate complex is different from the dissociation constant of the substrate from the enzyme-substrate complex (Fersht, 1999).

Table I.

Kinetic parameters of BcPG1 enzyme activity in the presence of PvPGIP2

	Km	Vmax
	mg/mL	RGU/mL
PG (0.02 ng/μL)	0.24	20.7
PG + PvPGIP2 (0.03 ng/μL)	0.89	6.47
PG + PvPGIP2 (0.07 ng/μL)	2.51	4.49

A Homology Model for the Structure of BcPG1

A homology model shows that BcPG1, like AnPGII and FmPG, folds as a parallel right-ended β-helix. A deep cleft on one side of the β-helix contains the residues involved in catalysis and substrate binding (Fig. 2). Three Asp residues (Asp-161, Asp-182, and Asp-183) and His-204 are responsible for the hydrolysis of the α-1,4 glycosidic bonds within the homogalacturonan. The residues Arg-237 and Lys-239 are in proximity of the catalytic Asp residues and bind the substrate at the reducing and nonreducing ends of the target glycosidic bond, respectively (defined as subsites +1 and −1, respectively; van Santen et al., 1999; Armand et al., 2000; Fig. 2). A number of other residues are predicted by molecular dynamics simulations to bind the substrate at subsites −4, −3, −2, −1, +1, +2, and +3 (Cho et al., 2001). Residues corresponding to each subsite in different PGs are listed in Table II. The catalytic residues and residues at subsites +1 and −1 are completely conserved, while a degree of variability among the different PGs is tolerated at the other subsites. The model structure of BcPG1 was superimposed to the structures of AnPGII and FmPG (Fig. 3A; van Santen et al., 1999; Federici et al., 2001). These enzymes are both inhibited by PvPGIP2 with different inhibition kinetics. The root mean square deviations (RMSDs) among equivalent Cα are below 2.2 Å, and the identity in the structure-based alignment is 36.9% for 328 overall aligned residues, demonstrating the conservation of the PG fold. The β-helix secondary structure elements are completely conserved, while differences are found in the length of the β-strands and in the length and conformation of the interstrand loops. The regions that form the boundaries of the active site cleft are of particular interest. A small 3₁₀ helix (residues 149 to 152) is followed by a loop whose length and conformation varies among the three PGs (Fig. 3A). This loop is followed by the conserved HNTD motif (residues 158 to 161), where the starting His residue has been demonstrated to play a role in PvPGIP2 recognition of FmPG (Federici et al., 2001). This residue is also predicted to be part of subsite +2 for substrate binding (Table II). The electrostatic potential surfaces of BcPG1, AnPGII, and FmPG are shown in Figure 3B. The active site cleft of FmPG appears narrower and deeper than that of AnPGII and BcPG1. This is mainly caused by the loop mentioned above, which is located at the crevice of the active site cleft and is considerably longer in FmPG with a seven-residue insertion (Fig. 3A). Also, the charge distribution appears nonequivalent in different areas surrounding the cleft, reflecting the lack of conservation of surface-localized charged residues. Some of the residues predicted to form subsites for the interaction with the substrate are charged, nonconserved, and solvent exposed (Table II). These variations likely play a role in defining the specific activities and pH optima of the different enzymes.

Figure 2.

Ribbon representation of the homology-modeled BcPG1 oriented as to highlight the active site cleft. The reaction is catalyzed by residues D161, D182, D183, and H204. Residues important for substrate binding, based on experimental data and predictions obtained from various PGs, are also shown. Their correspondence to predicted substrate binding subsites is shown in Table II.

Table II.

Residues at predicted substrate binding subsites in different PGs

Subsite	AnPGII	FmPG	BcPG1
−4	His-132	Gln-133	Ser-111
−3	Arg-223	Lys-244	Arg-214
	Lys-127	His-128	Phe-106
−2	Lys-127	His-128	Phe-106
	Asn-207	Thr-218	Asn-188
−1	Lys-258	Lys-269	Lys-239
	Ser-229	Ser-240	Ser-210
	Tyr-291	Tyr-302	Tyr-272
	Asn-178	Asn-189	Asn-159
+1	Arg-256	Arg-267	Arg-237
	Lys-258	Lys-269	Lys-239
	His-223	His-234	His-204
+2	His-177	His-188	His-158
	Asp-282	Thr-293	Ser-263
+3	Asp-282	Thr-293	Ser-263

Figure 3.

A, Conservation of the PG fold; structural superposition among AnPGII (red), FmPG (blue), and BcPG1 (green). The arrow indicates a variable loop at the crevice of the active site where a seven-residue insertion is found in FmPG. B, Electrostatic potential surfaces of BcPG1, AnPGII, and FmPG. Positive charges are shown in blue, and negative charges are shown in red. Differences can be seen, around the active site cleft, in both molecular shapes and charge distributions.

Molecular Docking Simulation of the BcPG1-PvPGIP2 Interaction

The interaction between BcPG1 and PvPGIP2 was analyzed by molecular docking simulations. From the distribution of docking orientations obtained in the rigid-body docking step, normalized interface propensity (NIP) values per residue were obtained (Fernandez-Recio et al., 2004). In Figure 4, A and B, the surfaces of BcPG1 and PvPGIP2, respectively, are colored according to the NIP values of their residues. The NIP value represents the propensity of a given residue to be at the protein-protein interface in the docking simulations. A wide area with high NIP values (maximum NIP 0.84) is found in PvPGIP2 in the concave side of the LRR domain predicted to interact with protein partners (Fig. 4B). Two clusters of residues have been identified: the first one is formed by residues Phe-80, Thr-109, Tyr-82, Ser-133, and Tyr-107, while the second one is formed by residues Phe-201, Lys-225, Gln-224, and Asn-247. These clusters largely lie in the β-sheet B1 (Di Matteo et al., 2003). Importantly, the region predicted to be critical for protein-protein interaction includes residue Gln-224 (space-filled representation and colored green in Fig. 4B) demonstrated to be necessary for the interaction of PvPGIP2 with FmPG (Leckie et al., 1999). In addition to the docking simulations, an analysis based on the optimal docking area (ODA) method (Fernandez-Recio et al., 2005) was performed on the surface of PvPGIP2 in search of residues potentially involved in protein-protein interactions. The ODA value represents the hypothetical gain in energy if the corresponding surface patch were buried upon protein binding. Thus, ODA negative values of a particular surface residue (residues with ODA below −10.0 kcal/mol) correlate well with its propensity to interact with other residues rather than with the solvent. This method identified two clusters of residues with ODAs <−10 kcal/mol: one around Leu-130, Thr-129, and Tyr-107 and the other around Lys-225. Both areas overlap with those identified by the NIP values.

Figure 4.

Docking analysis of the interaction between BcPG1 and PvPGIP2. A, The surface residues of BcPG1 are colored according to their NIP values: red for NIP > 0.4 and blue for residues with NIP < 0.0; intermediate values are scaled from blue to red. B, The surface residues of PvPGIP2 are colored with the same scheme as in A. Residue Gln-224 is colored in green. C, Ribbon representation of the interaction between PvPGIP2 and BcPG1 as obtained by molecular docking simulations. BcPG1 is shown in cyan and PvPGIP2 in magenta. D, Surface representation of the BcPG1-PvPGIP2 complex highlighting the partial coverage of the active site cleft.

The same analysis performed on BcPG1 predicts a wide interacting surface in the N-terminal portion of the parallel β-helix, with smaller NIP values (maximum NIP 0.36). This area includes the N-terminal α-helix that caps the interior of the β-helix and extends to the boundaries of the active site cleft (Fig. 4A). The ODA values of BcPG1 are also less significant with respect to PvPGIP2, which indicates that the BcPG1 surface has smaller affinity for protein binding.

The desolvation energy used in the ODA and NIP analyses shown above was applied to rank the most promising docking solutions that were further refined with Biased-Probability Monte-Carlo optimization of the ligand interface side chains. The 20 lowest-energy solutions were visually inspected, and we selected the one shown in Figure 4, C and D. Here, the entire sheet B1 of PvPGIP2 is engaged in the interaction with the N-terminal portion of BcPG1. As a result of the interaction, the active site cleft of BcPG1 is partially buried by the C-terminal end of PvPGIP2, though access of substrate appears not to be completely hindered.

A Mutation at the Predicted Protein-Protein Interface Decreases the Affinity of PvPGIP2 for BcPG1

BcPG1 is also inhibited by the PGIP isoform 1 of P. vulgaris (PvPGIP1), but with a 100-fold reduced efficiency (Manfredini et al., 2005). PvPGIP2 and PvPGIP1 differ by only eight residues in their mature forms (Leckie et al., 1999). Our docking model predicts that four of these residues, namely, Val-152, Ser-178, Gln-224, and His-271, are located at the interface. Among them, Ser-178 is positioned at the bottom of the concavity and does not interact with BcPG1 counterparts. The remaining four variant residues (Leu-60, Gln-291, Ala-297, and Ala-311) are predicted to be outside the interface.

The contribution of the single amino acids to the interaction was studied by mutating variant residues of PvPGIP2 into the corresponding ones of PvPGIP1. A series of six mutated pvpgip2 genes were expressed in Nicotiana benthamiana. The encoded proteins were purified and used in inhibition assays of BcPG1 activity. As expected from the docking model, residues predicted to be outside of the protein-protein interface had no effect on inhibition (Fig. 5). Mutations corresponding to residues predicted to interact with BcPG1, namely, His271Gln, Gln224Lys, and Val152Gly, were also tested. The Gln224Lys mutant showed no differences with the wild type, while the mutation His271Gln determined only a minor decrease in the inhibition efficiency (Fig. 5). The mutation Val152Gly had instead a considerable effect on PvPGIP2 activity. The inhibition efficiency was found to be lower than the wild type at all amounts tested; the concentration of inhibitor necessary to obtain 50% inhibition of one BcPG1 activity unit was double with respect to the wild type (Fig. 5). While 100% inhibition was obtained using 16 ng of wild-type PvPGIP2, only 87% inhibition was reached using 80 ng of the Val152Gly variant protein.

Figure 5.

Inhibition efficiencies of wild-type PvPGIP2 and mutants expressed as percentage of inhibition of 1 agarose diffusion PG unit (APU) of BcPG1. Each data point is the average of three independent experiments. The experimental variation was below 5%. Mutants were prepared changing PvPGIP2 residues with the corresponding residues of PvPGIP1. The mutation Val152Gly located at the predicted docking interface has a considerable effect on PvPGIP2 efficiency. The mutations His271Gln and Gln224Lys, also at the predicted interface, have little or no effect. Mutations outside the predicted docking interface (Gln291Lys, Ala297Ser, and Ala311Ser) also have no effect.

DISCUSSION

In this work, we have compared BcPG1 from B. cinerea with two characterized fungal PGs, AnPGII and FmPG. The specific activity of BcPG1 is intermediate between the more efficient AnPGII (Armand et al., 2000) and the less efficient FmPG (Federici et al., 2001). The shapes of the active site clefts of the three PGs considerably differ: the cleft of FmPG is narrower and deeper than that of AnPGII and BcPG1. This might influence the access of substrate or the dissociation of the product from the active site, explaining the low specific activity of FmPG. A narrowed active site cleft with respect to that of AnPGII is observed also in the crystal structure of the isoform AnPGI and is correlated to the processive behavior of this enzyme (van Pouderoyen et al., 2003).

The pH optimum of BcPG1 (5.5) is close to that of FmPG (5.2; Caprari et al., 1996) but differs by more than one pH unit from that of AnPGII (4.2; Armand et al., 2000). The structure of PG from Aspergillus aculeatus, determined both at pH 4.5 (active) and at pH 8.0 (inactive) shows that the RMSD between equivalent Cα in the two structures is below 0.5 Å, suggesting that the backbone is rigid and does not allow significant pH-dependent conformational changes (Cho et al., 2001). However, changes in pH may determine different side chain conformations or ionization states of residues involved in substrate binding. The deep cleft of PGs can accommodate approximately seven to eight GalUA units. An oligogalacturonide consisting of eight residues was docked into the active site cleft of A. aculeatus PG, allowing prediction of residues involved in substrate binding at all subsites (Cho et al., 2001). Some of the residues contacting the substrate, especially those at the +1 and −1 positions, are completely conserved in the PG family (see Table II). Other residues are variable: for instance, Asp-282 of AnPGII, at subsite +2, is replaced by Thr-293 in FmPG and by Ser-263 in BcPG1; His-132 of AnPGII, at subsite −4, is replaced by Gln-133 in FmPG and by Ser-111 in BcPG1; Lys-127 of AnPGII, at subsite −3, is replaced by His-128 in FmPG and Phe-106 in BcPG1 (Table II). FmPG and BcPG1 have similar substrate binding subsites, and this is consistent with their similar pH optima and specific activities. Ionizable residues of AnPGII involved in binding the substrate are likely responsible for both the lower pH optimum and the higher specific activity as compared to FmPG and BcPG1.

In this work, we have studied the interaction of BcPG1 with PvPGIP2. The strong inhibition of BcPG1 by PvPGIP2 is responsible of the reduced susceptibility to B. cinerea infections of transgenic tobacco and Arabidopsis plants overexpressing PvPGIP2 (Manfredini et al., 2005). We have demonstrated that PvPGIP2 inhibits BcPG1 with a mixed-type mode. This is compatible with a model where the substrate and the inhibitor bind the enzyme at the same time and the binding of the inhibitor affects the dissociation constant of the substrate-enzyme complex.

The analysis of the ODA and the NIP indicated a strong capacity of the inhibitor to form protein-protein interactions and predicted the presence of a large area in the concave side of the LRR solenoid capable of interacting with other proteins. This area is located in the sheet B1 found to be hypervariable in the PGIP family of proteins (D'Ovidio et al., 2004) and includes the residue Gln-224 that confers to PvPGIP2 the specificity toward FmPG (Leckie et al., 1999). The same analysis, performed on the homology model of BcPG1, indicates a reduced propensity of the enzyme to interact with other proteins, consistent with the notion that PGs do not evolve to maximize their protein-protein interacting properties; on the contrary, they tend to escape PGIP recognition.

Among the highest ranking solutions of the docking simulations, we selected the orientation shown in Figure 4 because it involved the surfaces predicted to have propensity for protein-protein interactions and explained the mixed mode of inhibition observed. This orientation predicts that PvPGIP2 binds the enzyme with its concave face, which is completely buried upon formation of the complex. The N-terminal end of the enzyme is engaged in the complex, and the active site cleft is partially buried by the inhibitor. The cleft, however, is still capable of binding the substrate. Consistently, the predicted substrate binding residues at subsites −2, −1, +1, +2, and +3 are still solvent exposed after binding of the inhibitor. Two residues at subsites −4 and −3, Ser-111 and Phe-106 are instead completely buried by PvPGIP2. This may cause a decreased affinity of the enzyme for the substrate, reflected in the increased K_m. On the other hand, the decrease of V_max can be explained by the reduced number of interactions between the enzyme and the substrate that likely renders more difficult to reach the transition state free energy.

Site-directed mutagenesis on the residues that distinguish PvPGIP2 from PvPGIP1 confirmed the structural framework provided by the docking model. The mutation of residues predicted to be located outside of the PvPGIP2-BcPG1 interface (Gln291Lys, Ala297Ser, and Ala311Ser) had no effect on the inhibition of BcPG1 activity. The mutation of residues His-271 and Gln-224, predicted to interact with BcPG1 residues (Gln-129 and Thr-41, respectively) through H-bond interactions, had only a minor effect probably because the corresponding residues of PvPGIP1 (Gln-271 and Lys-224, respectively) are still capable of interacting with the BcPG1 counterparts. The mutation of Val-152 into the corresponding Gly of PvPGIP1 had instead a remarkable effect in reducing the inhibitory capability of the mutant. The residue Val-152 is positioned in the middle of a row of three hydrophobic residues of PvPGIP2 (Val-128, Val-152, and Leu-175) that are engaged in stacking interactions and located at the predicted PvPGIP2-BcPG1 interface. The mutation of Val-152 into Gly increases the local flexibility of the PvPGIP2 main chain and may cause a distortion of the surface with a reduced affinity for BcPG1. This hypothesis is consistent with the observation that the residue following Val-152 is also a Gly. Importantly, the Val152Gly mutation was already found to increase by fivefold the dissociation constant of the PvPGIP2-AnPGII complex, suggesting that the presence of a Gly in this position plays a role in determining the reduced affinity of PvPGIP1 toward different PGs (Leckie et al., 1999).

In conclusion, we have shown that the mechanism of inhibition played by PvPGIP2 on BcPG1 is of a mixed type; this was corroborated by docking simulations and site-directed mutagenesis. The finding that PGIP performs different types of inhibition suggests that the inhibitor adapts the recognition capabilities of its wide concave surface in many ways and against different epitopes of different PG ligands. We suggest that the docking approach employed here is applied to other PG-PGIP couples, thus providing an additional tool to study the versatility of PGIP in evolving new recognition capabilities.

MATERIALS AND METHODS

BcPG1 and PvPGIP2 Purification

The BcPG1 gene was cloned into Pichia pastoris and purified as described (Manfredini et al., 2005). PvPGIP2 was overexpressed in Nicotiana benthamiana plants infected with a modified potato virus X as described (Leckie et al., 1999) and purified to homogeneity as described (Di Matteo et al., 2003). Protein concentrations were determined in parallel by the method of Bradford (1976) and by SDS-PAGE analysis comparing the intensity of the protein band with different amounts of bovine serum albumin (Sigma-Aldrich) used as a standard. PvPGIP2 mutants Val152Gly, Gln224Lys, His271Gln, Gln291Lys, Ala297Ser, and Ala311Ser were prepared as described (Leckie et al., 1999). Residues of PvPGIP2 are numbered according to D'Ovidio et al. (2004) where numbering starts with the first residue of the mature protein, which corresponds to residue 30 (a Gln) described in a previous article by Leckie et al. (1999).

BcPG1 Characterization

PG activity was determined by standard reducing end-group analysis (PAHBAH assay). A standard curve was prepared using monogalacturonic acid at different concentrations. The reaction mixture was prepared using 100 μL of polygalacturonic acid 3% (w/v), 30 μL of sodium acetate 0.5 m, pH 5, purified BcPG1, and water to a final volume of 300 μL. One activity unit (RGU) was defined as the amount of PG producing one microequivalent of reducing groups per minute at 30°C with 1.0% (w/v) polygalacturonic acid as substrate (Caprari et al., 1996).

Thermal stability was assayed at four different temperatures (30°C, 37°C, 45°C, and 50°C), incubating BcPG1 with substrate in a period from 5 to 180 min.

The effect of pH on activity was determined using the following buffers: 50 mm sodium citrate (from pH 3.0 to 4.0), 50 mm sodium acetate (from pH 4.5 to 6.0), and 50 mm phosphate buffer (from pH 6.5 to 8.0). Kinetic curves were determined as follows: initial rate measurements were made in 50 mm sodium acetate buffer (pH 5.0) at substrate (polygalacturonic acid) concentrations varying from 0.01 to 15.1 mg/mL. The reaction was initiated by the addition of enzyme (0.02 ng/μL) and monitored as for the standard PG assay. The same measurements were done in the presence of increasing amounts of inhibitor (0.03 and 0.07 ng/μL). Values of K_m and V_max were obtained by nonlinear regression analysis using the program Kaleidagraph (Synergy Software).

Inhibitory activity of PvPGIP2 site-directed mutants was determined using a modified agarose diffusion assay (Ferrari et al., 2003). In the gel diffusion assay, PG activity was expressed as APUs, with 1 APU defined as the amount of enzyme that produces a halo of 0.5-cm radius (external to the inoculation well) after 16 h at 30°C.

Homology Modeling of BcPG1

The best template for homology modeling was searched using the Web-based programs FUGUE (Shi et al., 2001) and HOMSTRAD (Stebbings and Mizuguchi, 2004). Three-dimensional homology models of BcPG1 were built with the template crystal structure of AnPGII (PDB entry 1cfz). Initial alignments were performed with FUGUE and annotated with JOY (Mizuguchi et al., 1998). Secondary structure predictions were performed using the program PSIPRED (McGuffin et al., 2000). The alignments were examined and manually modified in view of secondary structure predictions and conservation of residues known to play roles in catalysis and/or substrate binding. The final alignment between BcPG1 and AnPGII was used to generate 15 homology models using MODELER, version 6.2 (Sali and Blundell, 1993). The 15 models were ranked by analysis of their overall energy and geometric violations, and the best one was chosen. The model was further validated for its stereochemical quality using PROCHECK (Laskowski et al., 1993) and for sequence-environment compatibility using VERIFY3D (Eisenberg et al., 1997). Structural superpositions and RMSD calculations were performed using the SSM algorithm as implemented in the program COOT (Emsley and Cowtan, 2004). Figure 3B was prepared using GRASP (Nicholls et al., 1991). Figures were otherwise prepared using PyMOL (DeLano Scientific) and ICM (www.molsoft.com).

Docking Simulations between Polygalacturonase and PGIP

Sampling of the mutual rigid-body orientations between the molecules BcPG1 and PvPGIP2 (PDB code 1ogq) was performed by pseudo-Brownian Monte-Carlo minimization as implemented in the ICM-DISCO docking program (Fernandez-Recio et al., 2003). The desolvation energy based on accessible surface area with atomic solvation parameters optimized for rigid-body docking (Fernandez-Recio et al., 2004) was included in the final energy function in order to score all the rigid-body orientations (a total of 7,813 docking poses). NIP values were obtained from these rigid-body docking landscapes in order to identify significant protein binding sites (Fernandez-Recio et al., 2004). The 400 rigid-body docking poses with lowest binding energy were reduced to a total of 188 conformations after removing similar orientations (i.e. those with ligand backbone RMSD < 4 Å). These selected docking solutions were further refined with Biased-Probability Monte-Carlo optimization of the ligand interface side chains, as implemented in ICM-DISCO. The same scoring function as in rigid-body docking was used for the final rescoring of the refined conformations. After a visual inspection of the 20 lowest-energy solutions, we found a satisfactory docking conformation (ranked 15) with the PG active site partially occluded by the PGIP.