A defensive role against insect attack has been traditionally attributed to plant protease inhibitors. Here, evidence is described of the potential of a plant protease inhibitor, the potato carboxypeptidase inhibitor (PCI), to provide resistance to fungal pathogens when expressed in rice as a heterologous protein. It is shown that rice plants constitutively expressing the pci gene exhibit resistance against the economically important pathogens Magnaporthe oryzae and Fusarium verticillioides. A M. oryzae carboxypeptidase was purified by affinity chromatography and further characterized by mass spectrometry. This fungal carboxypeptidase was found to be a novel carboxypeptidase B which was fully inhibited by PCI. Overall, the results indicate that PCI exerts its antifungal activity through the inhibition of this particular fungal carboxypeptidase B. Although pci confers protection against fungal pathogens in transgenic rice, a significant cost in insect resistance is observed. Thus, the weight gain of larvae of the specialist insect Chilo suppressalis (striped stem borer) and the polyphagous insect Spodoptera littoralis (Egyptian cotton worm) fed on pci rice is significantly larger than that of insects fed on wild-type plants. Homology-based modelling revealed structural similarities between the predicted structure of the M. oryzae carboxypeptidase B and the crystal structure of insect carboxypeptidases, indicating that PCI may function not only as an inhibitor of fungal carboxypeptidases, but also as an inhibitor of insect carboxypeptidases. The potential impact of the pci gene in terms of protection against fungal and insect diseases is discussed.
As in the case of other plant defence proteins, the expression of genes encoding inhibitors of endoproteases is induced in response to pathogen infection (Cordero et al., 1994; Jongsma et al., 1994). In some cases, the pathogen-induced expression of protease inhibitors is itself dependent on, and a response to, the damage produced by the penetration and colonization by the pathogen in the host tissues (Cordero et al., 1994). In this situation, signals produced at the infected, wounded site trigger the activation of protease inhibitor genes. In other studies, the in vitro antifungal activity of inhibitors of serine and cysteine endoproteases has been described (Valueva and Molosov, 2004; Kim et al., 2005). However, direct evidence of antifungal effects provided by a recombinant plant protease inhibitor in planta is lacking.
The potato carboxypeptidase inhibitor (PCI), a 39-residue protein, accumulates in potato tubers as well as in wounded leaves of potato plants (Graham and Ryan, 1981; Villanueva et al., 1998). The PCI protein has been thoroughly characterized at the molecular level, and its three-dimensional structure has been determined (Clore et al., 1987). PCI inhibits all mammalian members of the A/B subfamily of carboxypeptidases. Thus, the C-terminus of the PCI protein interacts with carboxypeptidases in a substrate-like manner (Molina et al., 1994).
It is well known that carboxypeptidases play an important role in protein digestion in mammals and insects, acting to liberate free amino acids from the peptides produced by endopeptidase action, thus completing the digestive process and generating molecules that can be absorbed by the gut. Digestive carboxypeptidases are metallocarboxypeptidases that require Zn2+ for their activity (Rawlings and Barrett, 1995). Based on their substrate specifity, digestive carboxypeptidases have been classified as carboxypeptidase A (CPA) and carboxypeptidase B (CPB) enzymes. CPA has a preferred specificity for aliphatic (CPA1-type) and aromatic (CPA2-type) carboxy terminal residues, whereas CPB cleaves only the C-terminal basic residues lysine (Lys) and arginine (Arg). The three-dimensional structures of several mammalian and insect procarboxypeptidases, and their corresponding proenzymes, are available (Rees et al., 1983; García-Sáez et al., 1997; Estébanez-Perpiña et al., 2001; Bayés et al., 2003, 2005).
In this study, the in vitro antifungal properties of the PCI inhibitor against phytopathogens are reported. When introduced into rice, the pci gene confers protection against the agronomically important pathogens Magnaporthe oryzae and Fusarium verticillioides. A carboxypeptidase protein from M. oryzae was purified with a PCI-derived affinity column and further characterized by mass spectrometry. The PCI-affinity purified fungal carboxypeptidase was identified as a CPB (M. oryzae CPB, MoCPB), which was fully inhibited by PCI. Thus, the inhibitory properties of PCI against a fungal CPB, the first fungal CPB characterized, are reported in this work. The performance of insects feeding on pci rice was determined using two different species: a specialist pest of rice, the striped stem borer (Chilo suppressalis), and a generalist insect pest, the Egyptian cotton worm (Spodoptera littoralis). Moreover, amino acid sequence comparison and homology-based modelling of MoCPB revealed a high structural similarity of MoCPB with carboxypeptidases from the lepidopteran insects Helicoverpa zea and H. armigera. The potential impact of the pci gene, in terms of protection against fungal diseases and insect herbivory, is discussed.
Antifungal activity of PCI
Experiments were conducted in order to determine the extent to which PCI was capable of inhibiting the growth of fungal pathogens. For this, the PCI protein was obtained as described previously and its purity was confirmed by high-performance liquid chromatography (HPLC) analysis (Molina et al., 1992). The PCI protein was biologically active as judged by its ability to effectively inhibit bovine CPA (results not shown). To assay the in vitro antifungal activity of PCI, a microtitre plate assay was carried out (Cavallarin et al., 1998). In the presence of PCI, the growth of M. oryzae was impaired significantly (Figure 1a). After 48 h of incubation, a concentration of PCI of 20 µm was sufficient to produce 50% inhibition of M. oryzae growth. Increasing the concentration of PCI up to 45 µm resulted in 70% fungal growth inhibition. PCI also inhibited the growth of F. verticillioides, but to a lesser extent. Using a concentration of PCI of 40 µm, 40% inhibition of F. verticillioides growth was observed (results not shown).
Microscopic observations of M. oryzae cultures grown in the presence of PCI revealed severe changes in morphology, such as hyphal shortening and abnormal bed-like structures of M. oryzae hyphae (Figure 1b). Fungal cultures treated with bovine serum albumin (BSA) displayed normal growth behaviour, whereas fungal growth was fully inhibited by the presence of nystatin (0.1 µg/µL) in the medium (results not shown). From these results, it was concluded that low micromolar concentrations of PCI inhibited the in vitro growth of M. oryzae.
The effect of PCI on hyphal growth was explored further using the selective stain Congo red, a dye which exhibits a strong affinity for β-glucans. Thus, Congo red binds to the chitin in fungal cell walls and allows growing hyphae to be distinguished from non-growing hyphae (Matsuoka et al., 1995). Areas with active hyphal growth have little chitin deposition at their tips and, accordingly, show reduced Congo red staining. In contrast, when cell growth is inhibited, high Congo red staining is observed at the non-growing hyphal tips. The results obtained by Congo red staining of PCI-treated M. oryzae cultures are shown in Figure 1c. When compared with control cultures, PCI-treated fungal cultures showed strong Congo red staining at the tips of hyphae, indicating that hyphal growth is arrested in M. oryzae cultures. Together, these results indicate that PCI inhibits the in vitro growth of fungal pathogens.
Transgenic rice plants expressing the pci gene display enhanced resistance to fungal pathogens
Much of the interest in antifungal proteins lies in the potential use of their corresponding genes to engineer crop plants that are more resistant to disease. The above results demonstrate that PCI exhibits in vitro antifungal properties against important fungal rice pathogens. Therefore, we examined whether expression of the pci gene in transgenic rice confers disease resistance. To this end, the pci gene (Villanueva et al., 1998) was cloned under the control of the constitutive maize ubiquitin 1 (ubi 1) promoter (Christensen and Quail, 1996). Transgenic rice plants were produced by Agrobacterium-mediated transformation using the hygromycin resistance gene as the selectable marker. Northern blot analysis of T0 hygromycin-resistant lines revealed the accumulation of pci transcripts in the leaves of T0 plants (Figure 2a). The level of pci mRNA varied in the different transgenic lines. No effect on plant morphology or fertility was observed in pci-expressing rice lines. Independent events accumulating pci mRNA at different levels were selected as the parental lines to obtain T2 homozygous progeny plants.
PCI is an inhibitor of bovine CPA activity (Vendrell et al., 2000). In this work, the production of functional PCI protein in rice tissues was monitored by assaying the inhibitory activity of tissue protein extracts against bovine CPA. In the enzymatic assay, bovine CPA was pre-incubated with increasing amounts of leaf protein extracts from transgenic or control, untransformed rice plants before the addition of the CPA substrate. Extracts from the leaves of untransformed rice plants contained certain intrinsic inhibitory activity against bovine CPA. In agreement with the inhibitory properties of PCI, extracts from pci lines, when compared with extracts from untransformed plants, strongly inhibited bovine CPA activity (Figure 2b). These studies confirm the production of functional PCI protein in rice tissues.
To examine the functional relevance of pci expression in rice, pci rice lines were tested for resistance to infection with M. oryzae. Resistance was first determined using the detached leaf assay (Coca et al., 2004). Four independent T2 homozygous pci lines (lines 7-34, 9-1, 10-27 and 2-13) were assayed. Leaves from control and transgenic plants were inoculated with different doses of M. oryzae spores (106, 105 or 104 spores/mL). Differences were clearly observed in the degree of disease symptoms caused by M. oryzae between pci and untransformed plants at the various concentrations of inoculum. At a concentration of inoculum of 105 spores/mL, lesions on control leaves appeared 2–3 days after inoculation, whereas no detectable disease symptoms were observed in leaves of pci lines maintained under the same experimental conditions (Figure 3a). At 6 days after inoculation, leaves from control plants were visibly damaged, whereas those from pci lines appeared to be much healthier (results not shown). Control rice plants transformed with the empty vector (pCAMBIA 1300) showed results essentially similar to those of wild-type plants (results not shown; similar results have been reported previously by Coca et al., 2004).
To substantiate the results of symptom development in planta, trypan blue staining and microscopic observations of M. oryzae-infected rice leaves were performed. In control leaves infected with M. oryzae spores, extensive fungal colonization was seen (Figure 3b, left panel). The pathogen growing on the leaves of the untransformed plants was able to produce conidia. In contrast, under the same experimental conditions, a few hyphae were observed growing on the leaf tissue of the pci lines (Figure 3b, right panel).
The ability of pci plants to block the in planta growth of M. oryzae was tested further by counting the number of spores produced in infected leaves. At 6 days after inoculation, the average number of spores collected from infected control leaves was 11 250 ± 1050 spores/mL. The numbers of spores collected from the pci transgenic lines were as follows: 3750 ± 200 spores/mL for line 2-13, 2750 ± 250 spores/mL for line 7-34 and 4500 ± 350 spores/mL for line 9-1 (Figure 3c).
In other studies, transgenic plants constitutively expressing certain transgenes have been shown to possess activated plant defence mechanisms, which are normally activated only during pathogenesis (Mittler et al., 1995). To explore the possible activation of the plant defence system in transgenic rice lines, the expression of the endogenous PR1b gene was analysed in pci lines, without pathogen challenge. The expression of PR1 genes has been widely used as an indicator of the induction of the defence response. No expression of the rice PR1b gene was detected in leaves of transgenic lines in the absence of the pathogen (Figure 3d). As expected, PR1b gene expression was observed in M. oryzae-infected leaves of wild-type plants.
To verify the results obtained using the detached leaf assay, transgenic and control rice plants were sprayed with M. oryzae spore suspensions. At 3–4 weeks after spraying plants with fungal spores, control plants became infected, whereas transgenic plants showed no symptoms of infection (results not shown). The differences in performance between control and transgenic plants were striking at 45 days after inoculation with M. oryzae spores (Figure 3e). By this time, almost all of the control plants had died, whereas the pci lines remained healthy. Results from these experiments confirmed that the transgenic lines showed enhanced resistance to M. oryzae infection. Of the four independent pci lines assayed, lines 7-34, 9-1 and 2-13 exhibited better protection against M. oryzae than any other pci lines assayed in this work. These lines also showed the highest levels of accumulation of pci transcripts.
We also examined whether pci could confer resistance to another important pathogen of rice, the fungus F. verticillioides. Seeds from pci plants were able to germinate in the presence of F. verticillioides (Figure 3f, top). In contrast, F. verticillioides was highly virulent in the seeds from wild-type plants, and none of them germinated (Figure 3f, bottom). The percentages of surviving seeds were 80%, 65% and 60% for pci lines 7-34, 2-13 and 9-1, respectively.
Together, the disease resistance assays demonstrated that expression of the pci gene in rice enhanced resistance against the important fungal pathogens M. oryzae and F. verticillioides, the causal agents of the blast and bakanae diseases of rice, respectively.
Characterization of a PCI-susceptible carboxypeptidase from M. oryzae
To obtain more insight into the mechanisms by which PCI exerts its antifungal activity, mycelial extracts from M. oryzae were subjected to affinity chromatography on immobilized PCI. Eluted fractions were analysed by sodium dodecylsulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and mass spectrometry (Figure 4). A single protein of 29 kDa was eluted with buffer at pH 9.0 (Figure 4a,b). Mass spectrometry analysis showed that the 29-kDa protein was a short-chain dehydrogenase identified in the M. oryzae genome (accession number MG10909.4). The protein fractions eluted with buffer at pH 11.0 from the PCI-affinity column fractions contained two proteins with an estimated molecular weight of approximately 53 kDa (Figure 4a,b). These protein fractions were pooled and further analysed by matrix-assisted laser desorption ionization-time of flight-mass spectrometry (MALDI-TOF-MS). A search against the protein sequence databases revealed that the masses of 11 tryptic peptides matched perfectly with the internal fragments of the deduced amino acid sequence for a metallocarboxypeptidase gene identified in the M. oryzae genome (http://www.ncbi.nlm.nih.gov, accession number XP_359930) (Dean et al., 2005) (Table 1). The deduced amino acid sequence of the M. oryzae carboxypeptidase gene is shown in Figure 5. Taking into account that all tryptic peptides identified in this work showed experimental mass values that were in agreement with the theoretical values expected for this protein, the identification of the affinity-purified protein as a metallocarboxypeptidase of M. oryzae can be reported with a high degree of confidence.
Table 1. Peptides identified by mass spectrometric analysis of the potato carboxypeptidase inhibitor (PCI) affinity-purified Magnaporthe oryzae carboxypeptidase (fractions eluted at pH 11.0 in Figure 4). All tryptic peptides matched with the deduced protein encoded by a putative metallocarboxypeptidase identified in the M. oryzae gene (accession number XP_359930)
The experimental masses of the tryptic fragments are indicated.
The theoretical value correlates with the experimental mass values.
The structure of digestive carboxypeptidases in mammals and insects has been investigated extensively (Christianson and Lipscomb, 1989; Vendrell et al., 2000; Bayés et al., 2003, 2005; Bown and Gatehouse, 2004). Carboxypeptidases are known to be synthesized as precursor proteins, or procarboxypeptidases, comprising an N-terminal pro-domain (the pro-segment or activation segment) and the active enzyme moiety (or CPA moiety). A comparison of the deduced amino acid sequence of M. oryzae carboxypeptidase with that of bovine procarboxypeptidase A is shown in Figure 5[arrow on the alanine (Ala) residue in Figure 5 indicates the N-terminus of the mature bovine CPA1]. An N-terminal extension is also predicted for the M. oryzae carboxypeptidase. The N-terminal regions corresponding to the pro-region or activation segments of M. oryzae and bovine CPA differ markedly in length and amino acid sequence. Overall, the amino acid sequence identity of the M. oryzae carboxypeptidase relative to bovine CPA is only 37% (mature CPA enzyme). Sequence alignment of the M. oryzae carboxypeptidase with bovine CPA also reveals conserved amino acid residues characteristic for digestive carboxypeptidases (Figure 5). Thus, the M. oryzae carboxypeptidase sequence contains the residues histamine-69 (His69), glutamic acid-72 (Glu72) and His196 (residues are numbered according to the mature bovine CPA), which are highly conserved in the catalytic zinc site in carboxypeptidases. The active site residues (Arg127 and Glu270, numbering of bovine CPA) and substrate-binding amino acids [Arg71, asparagine-144 (Asn144), Arg145, tyrosine-198 (Tyr198) and aspartic acid (Asp256), numbering of bovine CPA] of CPAs are present in the deduced amino acid sequence of M. oryzae CPB.
Digestive carboxypeptidase enzymes are classified on the basis of their substrate specificity. Enzyme specificity is primarily determined by interaction of the C-terminal amino acid of the substrate with a binding pocket on the enzyme, the residue located at the bottom of this pocket (residue 255, numbering of bovine CPA1) defining the specificity of the carboxypeptidase (Titani et al., 1975). CPA has a preferred specificity for hydrophobic, aliphatic and aromatic amino acid residues, whereas CPB is highly specific for basic C-terminal residues of the substrate. In line with this, the residue located at position 255 of CPB is negatively charged (Asp), whereas, in CPA, residue 255 is an isoleucine. Because the M. oryzae carboxypeptidase identified in this work has Asp at position 255, it was predicted that this fungal carboxypeptidase was a member of the CPB family. Enzyme activity assays using typical substrates for CPB confirmed that the affinity-purified protein was a CPB. As shown in Figure 4a, the protein fractions eluted at pH 11.0 from the PCI-affinity columm hydrolysed the typical substrates for CPB, both Hippuryl-Arg (Figure 4a) and N-(3-[2-furyl]acryloyl)-Ala-Lys (FAAK) (results not shown). Most importantly, this carboxypeptidase activity was fully inhibited by PCI (Figure 4a). In other experiments, it was found that incubation of the mycelial protein extracts with PCI effectively inhibited MoCPB activity. Thus, a concentration of 2.5 µm of PCI reduced the total CPB activity of M. oryzae by 98.1% (residual activity of mycelial extracts pre-incubated with PCI, 1.9%). These results proved definitively that PCI binds, and inhibits, a CPB from M. oryzae (named MoCPB).
When the amino acid sequence of MoCPB was used to query the protein sequence database, a high degree of sequence identity of this M. oryzae carboxypeptidase with carboxypeptidases identified in the genome of several other fungal plant pathogens was observed. MoCPB had the highest sequence identity with a putative carboxypeptidase of Phaeosphaeria nodorum (anamorph Stagonospora nodorum), an important pathogen of wheat (accession number EAT91098), with 50% of sequence identity (63% similarity). The identity of MoCPB with a carboxypeptidase from Gibberella zeae (anamorph Fusarium graminearum), a pathogen that infects many crop plants, including many cereal species (accession number XP_388042), was found to be 40% (56% similarity). In all cases, the N-terminal pro-peptide was not considered in this comparison. Of interest, sequence comparisons against the National Center for Biotechnology Information (NCBI; Bethesda, MD, USA) database also revealed that MoCPB showed a high sequence identity with carboxypeptidases previously identified in the lepidopteran insects H. zea and H. armigera (Figure 5). The degrees of identity of MoCPB with these insect carboxypeptidases were 48% with a CPB from H. zea (HzCPB, accession number CAJ30028) and 42% with a CPA from H. armigera (HaCPA, accession number CAA06418).
Finally, the deduced amino acid sequence of MoCPB was submitted to different secondary structure prediction servers (FUGUE, GenTHREADER and J-PRED2), and a consensus ranking for secondary structure was generated. Significant secondary structure compatibility scores and high identity sequence scores were observed for the carboxypeptidase moiety. The best templates corresponded to the human procarboxypeptidase A2 (HPCPA2), H. armigera procarboxypeptidase A and H. zea procarboxypeptidase B [Protein Data Bank (PDB) codes 1AYE, 1JQGA and 2C1C, respectively]. Thus, the secondary structure prediction for MoCPB demonstrated that, although the amino acid residues showing α-helices and β-sheets of MoCPB were different from those of the bovine and insect carboxypeptidases, the M. oryzae protein had a high similarity in the secondary structure elements of the carboxypeptidase moiety, with eight α-helices (α4–α11) mixed with eight β-sheets (β5–β12) (Figure 5).
Localization of PCI in M. oryzae cells
To determine the localization of the PCI-susceptible CPB in M. oryzae cells, the PCI protein was fluorescently labelled and subsequently used in antifungal assays. The antifungal properties of the Alexa-488-labelled PCI protein were similar to those of the unlabelled PCI, discounting the possibility that labelling of PCI with Alexa-488 had a negative effect on the activity of PCI (results not shown). Confocal laser scanning microscopy (CLSM) was then used to monitor the localization of fluorescence in M. oryzae cultures grown in the presence of the Alexa-488-labelled PCI. As shown in Figure 6a (right panels), Alexa-488-labelled PCI (green) distributed uniformly in the periphery of M. oryzae cells. No fluorescence was detected inside the fungal cells in Alexa-488–PCI-treated cultures or in M. oryzae spores. Fungal cultures incubated with only the Alexa-488 dye showed no fluorescence (results not shown).
Membrane permeabilization as a consequence of membrane interaction and pore-forming activities has been described for several antifungal proteins (Theis and Stahl, 2004). The observation that Alexa-488-labelled PCI protein accumulated in the periphery of the fungal cells prompted us to investigate whether PCI has an effect on membrane integrity. The SYTOX Green uptake assay was used for these studies. SYTOX Green, a high-affinity nucleic acid stain that fluoresces on nucleic acid binding, penetrates cells with compromised plasma membranes, but does not cross the membranes of non-compromised living cells. In M. oryzae cultures grown in the presence of PCI and then stained with SYTOX Green, no permeabilization could be detected (Figure 6b, right panel). As expected, the dye entered the ethanol-treated fungal cells (compromised membranes) and stained the nuclear structure (Figure 6c, left panel). In fungal cultures treated with the Aspergillus giganteus antifungal protein (AFP), a protein that induces membrane permeabilization in M. oryzae cells, SYTOX Green penetrated the fungal cells and stained the nuclear structure (results not shown; however, the same result has been reported previously by Moreno et al., 2006). Finally, experiments were performed on the localization of Alexa-568-labelled PCI (red) in fungal cells with compromised membranes (ethanol-treated). As found in M. oryzae cultures grown in the presence of Alexa-488-labelled PCI, Alexa-568-labelled PCI (red) accumulated at the periphery of the fungal cells. On incubation of fungal cultures grown in the presence of Alexa-568-labelled PCI with 70% ethanol, fluorescence of DNA-complexed SYTOX Green (green) at the nucleus was observed, whereas Alexa-568-labelled PCI (red) remained at the periphery of the fungal cells (Figure 6c, right panel). In control experiments, the Alexa-568-labelled antifungal AFP protein (Moreno et al., 2006) was able to enter ethanol-treated fungal cells (results not shown). No intracellular localization of Alexa-568-labelled PCI was observed in PCI- and ethanol-treated fungal cultures, indicating that PCI does not penetrate the fungal cell, even in conditions in which fungal cells have compromised membranes.
Together, the results obtained by CLSM indicate that PCI most probably interacts with a CPB located at the periphery of the fungal cells, either at the wall or at the plasma membrane. Inhibition of fungal growth by PCI is, however, not related to the membrane permeabilization of fungal cells.
Predicted three-dimensional structure of the PCI-sensitive MoCPB
Crystal structures are now available for several carboxypeptidases (Rees et al., 1983; García-Sáez et al., 1997; Estébanez-Perpiña et al., 2001; Bayés et al., 2005). In this work, the amino acid sequence of MoCPB was analysed using two protein structure prediction servers (3D-PSSM and PDB-BLAST), with default parameters. The best structure compatibility score was observed for HPCPA2. Accordingly, the HPCPA2 protein was used for structure modelling of the MoCPB protein. The deduced model of the MoCPB protein was evaluated using the modeller program (Pieper et al., 2004). Moreover, the predicted three-dimensional structure was evaluated using the Very3D program with good quality scores. A model structure of MoCPB was then generated (Figure 7a). The carboxypeptidase moiety of MoCPB shows the classical fold of members of the carboxypeptidase family, with a central twisted eight-stranded β-sheet flanked by eight α-helices, which together form a globular α/β protein (Vendrell et al., 2000). A representation of MoCPB superimposed on HPCPA2 is shown in Figure 7b. Except for the N-terminal pro-segment, the overall structures of the CPB moiety of the M. oryzae precursor protein and HPCPA2 are highly conserved. The three-dimensional structure of MoCPB obtained by homology-based modelling also showed a high similarity to the structure of the insect HzCPB previously described by Bayés et al. (2005) (PDB code 1JQGA). Thus, homology-based modelling studies reveal that the structure of the enzyme moiety in MoCPB is almost identical to the structure found in the active enzyme of the H. zea CPB protein.
Effect of transgenic expression of pci on insect growth
Many studies have focused on the effect of inhibitors of endoproteases of plant origin on insect digestive proteinases. The benefits of transgenic plants engineered to express inhibitors of endoproteases, such as inhibitors of trypsins and chymotrypsins, in terms of resistance to insect pests, are also well documented. Very little is known about the in vivo effects of plant carboxypeptidase inhibitors on insect digestive proteases.
In order to determine the effect of pci expression in rice on the growth of insect pests, a series of feeding bioassays was carried out. Thus, the growth rates of the larvae of two lepidopteran insects, C. suppressalis and S. littoralis, fed on pci plants, were determined. For feeding bioassays with S. littoralis, second-instar larvae (L2) were fed on a diet consisting of leaves from either control or transgenic rice plants only. Daily observations allowed the relative growth rates of larvae and foliar consumption to be determined. Data from three feeding experiments are presented in Figure 8a,b. Larvae fed on a diet of leaves from pci plants grew faster than those fed on control leaves. Differences between pci and control plants became significant 2 days after infestation. At 4 days after infestation, the weight gains for larvae fed on pci lines 7-34, 9-1 and 2-13 were 41.2%, 32.6% and 32.2%, respectively (Figure 8a). The increased growth rate of S. littoralis larvae fed on transgenic plants was accompanied by an increase in the consumption of leaf material (Figure 8b).
Transgenic plants expressing the pci gene were also evaluated in terms of their effects on the growth of C. suppressalis larvae. Data from typical feeding experiments with transgenic and control plants are presented in Figure 8c. Fourteen days after infestation, larvae fed on pci plants showed significantly higher weights than those fed on control plants; 35% and 24% higher weight gains were observed for larvae fed on pci lines 9-1 and 2-13, respectively (Figure 8c). The consumption of rice expressing PCI resulted in a decrease in the level of CPB activity present in the gut extracts of C. suppressalis (Figure 8d).
Thus, the results obtained in bioassay experiments revealed that the weight gain of larvae of both insect species (C. suppressalis and S. littoralis) fed on pci plants was significantly larger than that of larvae fed on wild-type plants. The negative effect of pci expression in transgenic rice was more conspicuous with a polyphagous insect (S. littoralis) than with a specialized insect (C. suppressalis).
Plants have evolved distinct molecular mechanisms to defend themselves against pathogen and/or insect attack. The induction of the expression of genes encoding protease inhibitors in response to insect feeding has been described in numerous plant families. The production of these protease inhibitors by plants is aimed to interfere with the insect's digestive system. Potent inhibition of insect digestive proteases has been demonstrated for many plant protease inhibitors when added to artificial diets. These in vitro effects of plant protease inhibitors have been confirmed in vivo on several plant–insect systems by the use of transgenic plants (Hilder et al., 1987; Johnson et al., 1989; Duan et al., 1996; Vila et al., 2005). This study provides, for the first time, direct evidence that a plant protease inhibitor, PCI, exhibits in vivo antifungal activities against plant pathogens. An important aspect of this study was the observation that, although the production of PCI in rice plants confers pathogen resistance, it may be counterproductive in terms of resistance to insect pests.
Initially, we demonstrated that a purified preparation of PCI exhibits in vitro antifungal activity against plant pathogens. M. oryzae was chosen for these studies because this fungus causes rice blast, the most important fungal disease of cultivated rice. Breeding for durable resistance to this fungus is difficult as a result of the highly dynamic manner by which the blast pathogen population responds to a resistant rice cultivar. By contrast, F. verticillioides causes the bakanae disease of rice, a seed-borne disease that seriously affects seedling growth. This fungus not only causes a decrease in quantity, but also in quality, of rice by producing toxins, namely fumonisins, that affect human and animal health (Nelson et al., 1993). These two fungi, M. oryzae and F. verticillioides, also cause diseases in a large number of cereals and grasses. The repeated use of hazardous agrochemicals for the control of these fungal diseases has several drawbacks, such as their lack of specificity, increased incidence of the development of resistance on prolonged application and adverse impacts on human health and the environment.
In this study, transgenic rice lines constitutively expressing the pci gene were generated. All showed normal morphology and were fertile. The functionality of the plant-produced PCI protein was confirmed by the ability of protein extracts from leaves of transgenic plants to inhibit bovine CPA. Most importantly, transgenic rice plants showed enhanced resistance to the fungal pathogens M. oryzae and F. verticillioides. That the transgene product itself does not influence the expression of endogenous defence mechanisms was supported by the observation that no expression of the endogenous PR1b gene was detected in pci rice lines in the absence of the pathogen. Collectively, our molecular and phenotypic data indicate that the enhanced resistance observed in PCI-expressing rice plants is the result of the antifungal activity of PCI.
A PCI-derived affinity column was used to purify a PCI-susceptible carboxypeptidase from M. oryzae. Mass spectrometry revealed that the PCI-bound protein corresponded to a carboxypeptidase encoded by a gene identified in the M. oryzae genome (accession number XP_359930) (Dean et al., 2005). Even though two proteins were eluted from the affinity chromatography column using buffer at pH 11.0, all the tryptic peptides identified by mass spectrometry corresponded to the protein encoded by this gene. Whether the low-molecular-weight protein corresponds to a truncated form of the high-molecular-weight protein remains to be determined. Differences in glycosylation might also explain the two protein bands (four consensus glycosylation sites are present in the M. oryzae amino acid sequence).
Amino acid sequence analysis of the M. oryzae carboxypeptidase revealed that the residues critical for substrate binding and cleavage activity of carboxypeptidases, including an Asp residue at position 255, critical for CPB specificity, were present in this fungal carboxypeptidase. Using substrates specific for CPB, it was demonstrated that the affinity-purified M. oryzae carboxypeptidase exhibited CPB activity. Enzyme assays also revealed that a low concentration of PCI efficiently inhibited the activity of this fungal CPB. On the basis of these results, it is suggested that the PCI-susceptible carboxypeptidase of M. oryzae described here is a CPB, the first fungal CPB to be characterized. The potent inhibitory activity of PCI towards MoCPB indicates that PCI exerts its antifungal activity through the inhibition of this specific MoCPB.
With regard to the function of MoCPB, several possibilities can be envisaged. Firstly, our antifungal assays using fluorescently labelled PCI revealed that this PCI-susceptible CPB was located either at the plasma membrane or cell wall of the fungal cell. In this respect, it is well known that the cell walls of fungi contain proteins and certain enzymes involved in cell wall biosynthesis and fungal growth, such as chitin synthase, which are synthesized as precursor proteins that are proteolytically processed to their corresponding mature active forms (Machida and Saito, 1993). Therefore, it is possible that PCI could inhibit a CPB involved in the processing of proteins or enzymes that participate in the process of biosynthesis of the fungal cell wall. The abnormal morphology of M. oryzae hyphae grown in the presence of PCI (and the lack of PCI fluorescence in fungal spores), together with the observation that hyphal growth is arrested in PCI-treated fungal cultures, support this possibility. Secondly, it is possible that MoCPB may be involved in the pathogenicity. The M. oryzae genome contains a large, diverse set of proteins that are predicted to be secreted, including enzymes involved in the degradation of the plant cell wall (Dean et al., 2005). Interestingly, the predicted secreted proteome of M. oryzae includes the CPB identified in this work (Dean et al., 2005). A third, not exclusive possibility is that MoCPB could play a role in extracellular protein degradation, facilitating the amino acid uptake necessary for the pathogen's growth and development. Further studies are required to determine whether MoCPB is involved in the pathogenicity during the infection of rice plants by M. oryzae.
An important aspect of this study was the observation that, although the production of PCI in rice plants confers pathogen resistance, it may be counterproductive in terms of resistance to the lepidopteran insects C. suppressalis and S. littoralis. C. suppressalis (striped stem borer), a stem-feeding specialist of rice plants, represents one of the most devastating pests of cultivated rice. By contrast, S. littoralis (Egyptian cotton worm) is a leaf-feeding polyphagous insect that attacks a variety of commercially important crops, including cotton, rice, maize, legumes and grasses. Thus, feeding experiments revealed that the weight gain of larvae of C. suppressalis and S. littoralis fed on pci rice was significantly larger than that of larvae fed on wild-type plants. The apparent effect of the weight gain of larvae fed on transgenic rice suggests a PCI-induced mechanism of adaptation to compensate for the effect of PCI on insect digestive proteases. In line with this, several reports have established that insect larvae are able to adapt to the presence of inhibitors of endoproteases by replacing the inhibited enzymes by other protease inhibitor-insensitive proteases, these larvae then developing faster than larvae fed on controls. Thus, the phenomenon of overcompensation to dietary proteinase inhibitors, resulting in increased weight gain of the target pest, has been repeatedly described (Jongsma and Bolter, 1997; De Leo et al., 1998; Girard et al., 1998; Cloutier et al., 1999, 2000). It is also true that very few studies have focused on the effect of plant-produced inhibitors of carboxypeptidases on insect carboxypeptidases.
In this work, we constructed a three-dimensional model of MoCPB. Of interest, the predicted three-dimensional structure of MoCPB obtained by homology-based modelling revealed a high similarity with the crystal structure of a CPB from the lepidopteran insect H. zea (Bayés et al., 2005; PDB code 2C1C). These observations indicate that the in planta-produced PCI might function not only as an inhibitor of fungal carboxypeptidases, but also as an inhibitor of insect digestive carboxypeptidases. Aspects of inhibitor specificity towards insect and fungal carboxypeptidases, and carboxypeptidase–inhibitor complexes, should now be assessed.
In summary, the results presented here broaden the spectrum of action of plant proteinase inhibitors and highlight the role of the PCI protein in protection against fungal pathogens. As demonstrated, the antifungal activity of PCI relies on its ability to inhibit a specific fungal CPB: MoCPB. Although pci confers protection against fungal pathogens in transgenic rice, a significant cost in insect resistance is observed. Considering the versatility of insect responses to protease inhibitor consumption, strategies for protection against fungal diseases and insect pests based on transgenic plants expressing plant protease inhibitors must be evaluated together, and on a case-by-case basis.
Plant and fungal material
Transformation was carried out using the Mediterranean elite japonica rice (Oryza sativa L.) cultivar Ariete. Rice plants were grown at 27 ± 2 °C with an 18-h/6-h light/dark photoperiod. M. oryzae (anamorph Pyricularia oryzae) (PR09 isolate, kindly provided by D. Tharreau, CIRAD, Montpellier, France) was maintained on rice flour medium (20 g/L rice flour, 15 g/L agar and 2.5 g/L yeast extract) until the mycelium covered the surface of the plate. Spores were collected by adding sterile water to the surface of the mycelium. After filtration, spores were adjusted to the appropriate concentration with sterile water using a Bürker counting chamber. The fungus F. verticillioides, formerly known as F. moniliforme Sheldon, was grown on potato dextrose agar (Difco Laboratories, Detroit, MI, USA) plates. Conidial suspensions were prepared as described above and counted.
In vitro antifungal assays
The in vitro activity of PCI was determined using a pure preparation of PCI and the microtitre plate assay, as described previously (Cavallarin et al., 1998). Briefly, spores (50 µL, 106 spores/mL) were pre-germinated for 6 h, and then PCI was added to the desired final concentration. Fungal cultures were then incubated at 28 °C for the required period of time and the absorbance was read [optical density at 595 nm (OD595 nm)]. Fungal growth was also analysed microscopically to confirm the spectrophotometric data. Controls with nystatin (0.1 µg/µL) or BSA (10 µm) were carried out.
Congo red staining
Congo red staining was used to allow visualization of chitin deposition at the hyphal tips of M. oryzae. Pre-germination and treatment of fungal cultures with PCI were performed in 96-well microtitre plates, as described above. Following incubation with PCI at a final concentration of 40 µm, Congo red was added to the fungal cultures to a final concentration of 1 mm. After 10 min, the fluorescence was viewed by CLSM with a Leica TCS SP microscope (Heidelberg, Germany), using an excitation wavelength of 543 nm and an emission wavelength of 560–635 nm. Hyphal growth was determined by a lack of Congo red staining at the hyphal tip, whereas hyphae with arrested growth showed Congo red staining across the series of optical sections at the hyphal tip (Matsuoka et al., 1995).
SYTOX Green uptake assay
Magnaporthe oryzae spores were pre-germinated for 6 h in potato dextrose broth, as described above. PCI was added to a final concentration of 40 µm and the plates were incubated for 16 h at 28 °C. Control cultures without PCI were also assayed. SYTOX Green (Molecular Probes, http://www.proves.com, Leiden, The Netherlands) was then added (final concentration, 0.2 µm). After incubation for 10 min, fungal cells were analysed by CLSM. For the detection of SYTOX Green uptake, an excitation wavelength of 488 nm and an emission wavelength of 500–554 nm were used. To obtain compromised membranes, the fungal structures were incubated in 70% ethanol for 10 min at room temperature, washed in potato dextrose broth and then stained with SYTOX Green (Springer and Yanofsky, 1989).
PCI localization studies
The PCI protein was labelled with Alexa-568 or Alexa-488 according to the manufacturer's instructions (Molecular Probes). In vitro antifungal assays with Alexa-labelled PCI protein were carried out as indicated above. The localization of Alexa-568-labelled PCI in fungal cultures was determined by CLSM using an excitation wavelength of 578 nm and an emission wavelength of 603 nm. For experiments with Alexa-488-labelled PCI, an excitation wavelength of 495 nm and an emission wavelength of 519 nm were used.
Production and molecular characterization of transgenic rice plants
The coding sequence for the pci gene (Villanueva et al., 1998) was cloned into the BamHI site of the pAHC17 plasmid DNA under the control of the promoter region, first intron and first exon of the maize ubi 1 gene and the nopaline synthase (nos) terminator sequences (Christensen and Quail, 1996). The pci (309 bp) DNA sequence comprised the entire coding sequence of the PCI precursor protein, including the N-terminal signal peptide and pro-region as well as the C-terminal extension (Villanueva et al., 1998). Next, the DNA fragment covering the entire cassette for expression of the pci gene was generated by polymerase chain reaction (PCR) using the primers 5′-CGGGGTACCAAGCTTGGGCTGCAGTGCAGCGTGACC-3′ (forward) and 5′-CGGGGTACCAAGCTTGTTTGACAGCTTATCATCGG-3′ (reverse) (sequences in italic indicate the HindIII restriction site introduced into the PCR primers to facilitate the subsequent cloning step). Following HindIII digestion, the entire cassette (ubi::pci::nos) was inserted into the HindIII-digested pCAMBIA 1300.
Transgenic rice lines were produced by Agrobacterium-mediated transformation of embryonic callus derived from mature embryos, as described previously (Sallaud et al., 2003). The expression vector construct was transferred to Agrobacterium tumefaciens EAH105 strain. The parent pCAMBIA 1300 vector already contains the hygromycin phosphotransferase gene (hptII), affording hygromycin resistance in the T-DNA region. Selected T0 plants were grown under glasshouse conditions to obtain homozygous transgenic lines in the T2 generation.
Plants were screened for expression of the pci gene by Northern blot analysis. For this, RNAs from the leaves of transgenic and control plants were separated in a formaldehyde-containing agarose gel, transferred on to nylon membranes (Hybond-N, Amersham Biosciences, Piscataway, NJ, USA) and hybridized to 32P-labelled pci probes. Hybridization was carried out in 40% formamide, 5 × SSPE [1 × SSPE = 0.18 m NaCl, 10 mm NaH2PO4, 1 mm ethylenediaminetetraacetic acid (EDTA), pH 7.5], 5 × Denhardt's and 0.5% SDS containing 100 µg/mL of salmon sperm DNA at 42 °C. Membranes were washed in 0.5 × SSPE, 0.1% SDS at 65 °C.
For reverse transcriptase-polymerase chain reaction (RT-PCR) of the rice PR1b gene, total RNA was extracted from the leaves of pci and wild-type rice plants. Leaves were collected from eight individual plants. As a control, RNA samples from M. oryzae-infected leaves of wild-type plants (48 h after inoculation with fungal spores) were also obtained. The first cDNA was synthesized with DNase-treated total RNA (1 µg) with Moloney Murine Leukemia Virus (M-MLV) RT (Roche, Mannheim, Germany). Aliquots of the resulting RT reaction product were used as template for RT-PCR analysis. The primers used for amplification reactions were for OsPR1b (forward, 5′-CTTGGCGAGAACCTCTTCTG-3′; reverse, 5′-GCCGGCTTATAGTTGCATGT-3′). The control of the rice actin 1 gene was performed using the forward primer 5′-CGACGAGTCTGACCCATCCA-3′ and reverse primer 5′-GTACCCGCATCAGGCATCTG-3′.
Disease resistance assays
To test the resistance to M. oryzae infection of transgenic plants, infection of detached leaves with different doses of M. oryzae spores (20 µL; 106, 105 and 104 spores/mL) was performed, as described previously (Coca et al., 2004). Briefly, the second leaves of 2-week-old soil-grown rice plants were placed into plate dishes with 1% w/v water agar containing 2 mg/L kinetine. Whatmann filter paper discs saturated with an M. oryzae PR9 spore suspension at the appropriate concentration were placed on to the upper face of the leaves for 36 h and then removed. The inoculated leaves were maintained in a chamber under humid conditions at 28 °C with 16 h light and 8 h dark for the required period of time. The development of disease symptoms with time was followed. Four independent T2 homozygous pci lines and at least 10 plants per line were assayed.
Trypan blue staining was used to confirm fungal colonization. For this, the fungal-inoculated leaves were fixed with formaldehyde (ethanol–formaldehyde–acetic acid, 80 : 3.5 : 5, v/v) by vacuum infiltration for 1 h at room temperature. After one change, the tissue was kept in the fixation solution overnight. The plant material was then stained with lactophenol blue solution for 6 h at room temperature, washed with water and observed with a Zeiss Axiophot microscope (Carl Zeiss, Thornwood, NY, USA) under bright-field illumination. Furthermore, the ability of the fungus to grow and produce spores was estimated by counting the number of spores collected from control and transgenic rice lines. Infection with M. oryzae spores was accomplished using the detached leaf assay with a spore suspension at a concentration of 1 × 106 spores/mL and five inoculations per leaf. Spores were harvested from each leaf in 1 mL of sterile water, 6 days after inoculation. Spores were counted in 10-µL aliquots of each sample. Three independent counts for each sample, and from three individual leaves for each independent transgenic line (lines 7-34, 2-13 and 9-1), were averaged to calculate the number of spores per millilitre.
The evaluation of resistance to blast of pci-expressing rice plants was also assayed by spraying soil-grown control and transgenic rice plants with M. oryzae spores. Plants at the three-leaf stage were sprayed with a spore suspension (106 spores/mL, containing 0.02% v/v Tween 20) until the leaves were covered with fine droplets. Following inoculation, the plants were maintained in a dew chamber with 90% relative humidity at 27 ± 2 °C under an 18-h/6-h light/dark photoperiod. The development of symptoms was monitored visually. These experiments were carried out at least three times with four independent transgenic lines.
To evaluate the resistance to infection by F. verticillioides, seeds from control and pci-expressing rice plants were immersed in a conidial suspension of this fungus (107 spores/mL), dried on sterile filter paper and placed on Murashige and Skoog medium without sucrose at 27 ± 2 °C under an 18-h/6-h light/dark photoperiod for the required period of time.
Purification of CPB from M. oryzae
Magnaporthe oryzae was grown in 50 mL of potato dextrose broth for approximately 18 days, and the mycelium was harvested by filtration. Mycelial protein extracts were prepared using 0.05 m Tris-HCl, pH 7.5, 0.45 m NaCl as the extraction buffer, at 4 °C for 30 min with continuous slow stirring. Crude protein extracts were centrifuged at 12 000 g for 15 min at 4 °C and the supernatant was taken. Ammonium sulphate was added to the crude protein extracts to reach 10% ammonium sulphate saturation, and incubated on ice for 20 min. The precipitate was removed by centrifugation. The supernatant was brought to 45% ammonium sulphate saturation and allowed to stand at 4 °C overnight. The samples were centrifuged as above, and the pellet was retained. The 10%–45% pellet was suspended in extraction buffer, concentrated (Centricon-30 filter units, Amicon Inc, Beverly, MA, USA) and resuspended in extraction buffer (7.5 mL). Protein concentrations were determined by the dye-binding assay (Bio-Rad, München, Germany).
The mycelial extract was applied to a column of immobilized PCI (0.1 mL of Sepharose coupled to PCI). The column was washed successively with 2 mL of 10 mm Tris-HCl, pH 7.5, 0.15 m NaCl; 1 mL of 50 mm Tris-HCl, pH 9.0; and 1 mL of 50 mm Na2HPO4, pH 11.0.
Mass spectroscopic and sequence data analyses
Pooled fractions from the affinity column were separated by 12.5% SDS-PAGE and stained with colloidal Coomassie blue following standard protocols. Protein bands were manually excised, reduced with dithiothreitol (DTT) and alkylated with iodoacetamide. Enzymatic digestion was performed overnight with trypsin at 37 °C, and the digested peptides were extracted with acetonitrile–trifluoroacetic acid (TFA). Proteins were identified by peptide mass fingerprinting and subsequent peptide fragmentation spectral analysis by MALDI-TOF-MS using a 4700 Proteomics Analyser MALDI-TOF-MS (Applied Biosystems, Proteomics Platform, Parc Cientific de Barcelona, Universitat de Barcelona, Barcelona, Spain). Mass spectra were searched against the NCBI protein sequence database (http://www.ncbi.nih.gov/).
The fold compatibility between the target and PDB entries was analysed using two protein structure prediction servers: 3D-PSSM (Kelley et al., 2000) and PDB-BLAST (http://bioserv.cbs.cnrs.fr), with default parameters. The best template was found to correspond to HPCPA2 (PDB code 1AYE). The three-dimensional model was built directly using modbase (Pieper et al., 2004), which relies on modeller for fold assignment, sequence structure alignment, model building and assessment (http://salilab.org/modeller). The best model obtained had a score of 0.98, a model being predicted to be good when the score is higher than a pre-specified cut-off of 0.7 (the probability for correct folding of the model is greater than 95%). The three-dimensional structure prediction was then evaluated using the verify3d program (Eisenberg et al., 1997). Molecular graphics images were produced using the UCSF Chimera package (Pettersen et al., 2004).
Carboxypeptidase activity assays
The inhibitory activity of in planta-produced PCI was determined by titrating bovine CPA (Sigma, St Louis, MO, USA) with increasing amounts of rice leaf extracts. For this, protein extracts were prepared from leaves of pci-expressing and wild-type plants using 150 mm Tris-HCl, pH 7.5 and 0.45 m NaCl as extraction buffer. Protein concentrations were determined using the Bio-Rad protein assay reagent and BSA as a standard. Protein extracts (50, 100 and 150 µg) were pre-incubated with bovine CPA (0.2 µg) for 30 min on ice. The CPA activity in total protein extracts obtained from rice leaves was assayed using benzoyl-glycyl-phenylalanine (Hippuryl-l-Phe, HPA), at a final concentration of 1 mm, or N-(4-methoxyphenyl-azoformyl)-l-phenylalanine (AAFP, Bachem, Bubendorf, Switzerland), at a final concentration of 0.1 mm, as substrates (Mock et al., 1996).
To determine the CPB activities of mycelial protein extracts, extracts were prepared as described above and used in enzyme assays (50 µg of the mycelial protein extract per assay). CPB activity was then assayed using the typical substrates for CPB: benzoyl-glycyl-arginine (Hippuryl-l-Arg, HA; at a final concentration of 1 mm) and FAAK (at a final concentration of 0.2 mm), as described previously (Bayés et al., 2005).
Insect bioassays of transgenic lines and enzyme assays with gut extracts
For feeding assays with S. littoralis, leaves from 3-week-old plants were placed on mesh-covered boxes lined with moist filter paper. The larvae were starved for 2 h prior to use. Forty second-instar (L2) S. littoralis larvae were placed in two boxes (20 larvae/box) after determining their weight. The leaf material used for bioassays was fresh every day, its weight being recorded before and after the feeding period. Observations of individual larvae and leaves were made every 24 h over a 4-day period to determine the foliage consumption and relative growth rate. For each box, the leaf area eaten was measured by computer-aided image analysis using Quantity One (Bio-Rad) software. Each experiment was carried out three times with three independent T2 homozygous lines (lines 7-34, 9-1 and 2-13).
Feeding trials with C. suppressalis larvae were essentially performed as described by Vila et al. (2005). L2 larvae were allowed to feed for 2 weeks on soil-grown control and transgenic plants. Larvae found inside the stems were then collected and recorded.
Enzyme assays with midgut extracts from S. littoralis and M. oryzae larvae were carried out as described previously (Tamayo et al., 2000; Vila et al., 2005). Thus, the total proteolytic activity of midgut extracts was determined using casein–resorufin (Roche) as substrate. The assay used 0.1% (w/v) casein–resorufin and 5 µg of midgut extract in a final volume of 200 µL (Tamayo et al., 2000). Inhibition assays were carried out to determine the inhibitory effect of PCI on the total proteolytic activity present in the gut extracts of S. littoralis and C. suppressalis larvae. For this, samples of gut extracts were pre-incubated with PCI (10 µm) according to the procedure described by Vila et al. (2005).
JQ and LV were recipients of a fellowship from the Generalitat de Catalunya. We thank Dra. Gloria Rossell (Instituto de Investigaciones Químicas Ambientales, IIQAB, Barcelona, Spain) for kindly supplying S. littoralis larvae. We are grateful to J. Lorenzo for providing us with the pure PCI protein and the Sepharose-PCI column, and to Monica Rodriguez de la Vega for valuable help on structure modelling. This research was supported by grants from the Ministerio de Ciencia y Tecnología (BIO2003-04936-C02-01 and BIO2006-05583 to BSS, and BIO2004-05879 to FXA). We also acknowledge the support of the PICASSO exchange programme of the French Ministry of Foreign Affairs and the Spanish Ministerio de Ciencia y Tecnología. We thank the ‘Xarxa de Referència en Biotecnología’ from the Generalitat de Catalunya for substantial support.