Mutants of common bean alpha-amylase inhibitor-2 as an approach to investigate binding specificity to alpha-amylases

Despite the presence of a family of defense proteins, Phaseolus vulgaris can be attacked by bruchid insects resulting in serious damage to stored grains. The two distinct active forms of a-amylase inhibitors, a-AI1 and a-AI2, in P. vulgaris show different specificity toward a-amylases. Zabrotes subfasciatus a-amylase is inhibited by a-AI2 but not by a-AI1. In contrast, porcine a-amylase is inhibited by a-AI1 but not by a-AI2. The objective of this work was to understand the molecular basis of the specificity of two inhibitors in P. vulgaris (a-AI1 and a-AI2) in relation to a-amylases. Mutants of a-AI2 were made and expressed in tobacco plants. The results showed that all the a-AI2 mutant inhibitors lost their activity against the insect a-amylases but none exhibited activity toward the mammalian a-amylase. The replacement of His33 of a-AI2 with the a-AI1-like sequence Ser-Tyr-Asn abolished inhibition of Z. subfasciatus a-amylase. From structural modeling, the conclusion is that the size and complexity of the amylase-inhibitor interface explain why mutation of the N-terminal loop and resultant abolition of Z. subfasciatus a-amylase inhibition are not accompanied by gain of inhibitory activity against porcine a-amylase.


Introduction
Alfa-amylase inhibitors (α-AI) occur naturally in many food plants and are particularly abundant in cereals and legumes (Franco et al., 2002).Insects gain access to food sources when they evolve amylases that are not affected by inhibitors present in the food source, and plants become resistant when they evolve inhibitors effective against these insect enzymes.Genes encoding these inhibitors have been used to make transgenic crops by gene transfer technology because their expression in plants is harmful to target insects and pests, interfering with their digestion, absorption or reproduction, whereas no antinutritional or toxic effects were observed in rats fed on transgenic pea expressing the insecticidal bean α-amylase inhibitor (Pusztai et al., 1999).
The growth of larvae of two seed feeding beetles, C. maculates and C. chinensis, is inhibited when the diet of the larvae contains relatively low levels of common bean α-AI (Ishimoto & Kitamura, 1989).Pea transgenic seeds containing up to 0.1%-1.2% of α-AI1 were resistant to cowpea weevils and azuki bean weevils (Shade et al., 1994).Furthermore, in field trials, peas expressing α-AI1 were protected against damage by the pea weevil (Pusztai et al., 1999), suggesting that gene transfer mechanism can be used to make legumes resistant to bruchids.Common beans resistant to Z. subfasciatus have already been produced by traditional breeding techniques but presently there are no inhibitors in beans that are effective against Acanthoscelides obtectus, another important storage bean insect pest.A better understanding of the molecular basis of specificity of α-AI and their target amylases therefore has potential practical benefits.
In the last ten years, several studies of the interaction between α-amylase inhibitors and their target enzymes have been carried out (Pueyo et al., 1993;Mirkov et al., 1995;Grossi-de-Sá et al., 1997;Takahashi et al., 1999).Based on the pea lectin structure (Einspahar et al., 1986) and their own sequence alignments, Mirkov et al. (1995) constructed an α-AI1 mutant by replacement of the 188 WSY 190 triplet with GNV producing an inactive inhibitor.The same triplet was considered later by Grossi- de-Sá et al. (1997) in substitutions using the corresponding α-AI2 sequence.The α-AI2triplet was also changed to the α-AI1-like sequence.In both cases the specificity of the inhibitor remained unchanged.
The PPA/α-AI1 structure confirmed that interactions occur primarily by contacts with two inhibitor loops and revealed an extensive network of direct hydrogen bonds.Based on these structures, ZSA and the four pairwise α-AI1 or α-AI2 complexes with PPA or ZSA were modeled (Silva et al., 2000).Although none of the interface parameters analyzed were, alone, able to explain the specificity of inhibitors for their cognate enzymes, lists of interface forming residues were defined as targets for experimental efforts toward specificity alteration via site-directed mutagenesis.
The main goal of this work was to understand the molecular basis of the specificity of α-AI1 and α-AI2 inhibitors in P. vulgaris in relation to α-amylases.

Construction of the mutants by PCR mutagenesis
The PCR (polymerization chain reaction) techniques, as described in Mole et al. (1989) andLandt et al. (1990), were used to construct the mutants (M1, M2, M3 e M4) of the α-AI2 gene.In each case, the substitution assumed the same correspondent amino acid sequence like in α-AI1 (Table 1).For construction of the first mutant (M1), the cDNA (complementary deoxyribonucleic acid) of α-AI2 on the pUC8 plasmid (Grossi- de-Sá et al., 1997) was used as template; the amino acids serine (Ser) and tyrosine (Tyr) were inserted between positions 32 and 33, and His33 was replaced by asparagine (Asn) (SYH33N).These residues are involved in the N-ter-minal loop and this mutation was repeated in all other mutants.The M1 mutant was used as template to construct the M2 mutant where Gln71 was replaced by histidine (Q71H); this residue is involved in binding to α-amylase but is outside of the two main interacting loops.The M1 mutant was also used to construct the M3 mutant, but in this case the additional mutation involved the C-terminal loop, where the His175, Glu176, Tyr177 and Phe179 were replaced by tyrosine, glutamine, tryptophan and tyrosine, respectively.The M4 triple mutant was constructed to contain all the mutations introduced in the other mutants.The primers used to perform the mutations were α-AI2PR1: 5'CATAGAGT CCACTGTATTATAGGATAGTTGTAAGTAGCC3'; α-AI2PR2: 5'TGCTTCGCGGTGAGTGGTGATATT CA3' and α-AI2PR3: 5'GCCATCTCAGGGGTTTATC AATGGAGCTATGAAACGAGAGACGTG3', with the mutated nucleotides in bold letters.The primers of the 3' and 5' α-AI2 sites, in a sense (5'CCGAATTCAT GGCTTCCTCCAACTTA3') or in antisense orientation, were combined in each case to obtain the megaprimer (intermediary step) or the whole gene.
The PCR products were digested by EcoRI, gel purified and ligated into pGEMT-easy (Promega Corporation, Madison, Wis., USA).In each cloning step, several clones of each construction were sequenced to confirm that no errors had been produced.To guarantee maximum fidelity, the Pfu enzyme (Promega) was used in the reactions.The BsmI and PstI sites were introduced (by PCR) into the ends of both the wild-type and mutants versions of the α-AI2 gene to facilitate the fusion of this gel purified fragment (750 bp) into plasmid pTA-2 (Grossi- de-Sá et al., 1997), containing the PHA-L (phytohemaglutinin) (536 bp) seed-specific promoter.Finally, the HindIII-EcoRI chimeric (PHA/wild-type or PHA/α-AI2 mutants) cDNA fragments were cloned in respective sites into the plant transformation vector pCAMBIA 1390 (Cambia GPO, Camberra, Australia) containing the nopaline synthase transcriptional (NOS) terminator.

Tobacco transformation
The above mentioned vectors were introduced into tobacco (Nicotiana tabacum cv.Xanthi) to express the protein in the seeds.The binary vectors were transferred to Agrobacterium tumefaciens GV3101PMP90 (Koncz & Schell, 1986) using heat-shock method (Brasileiro & Carneiro, 1998).Sterilized leaf explants were transformed by a simplified version of the leaf-disc method (Horsh et al., 1985).Five different cultures of Agrobacterium harboring each construction were grown in 5 mL YEB medium (Vervliet et al., 1975) supplemented with rifampicin (100 µg/mL), gentamicin (50 µg/mL) and kanamycin (100 µg/mL), for 16 hours, 28±2 o C. For co-transformation experiments, 1 mL of the Agrobacterium culture (A600nm = 0.1) was added to 20 mL MS liquid medium (Murashige & Skoog, 1962).Twenty leaf explants were incubated with the Agrobacterium culture in a Petri dish for 5 min, at room temperature.Explants were then immediately placed on MS solid medium (0.7% agar) for two days in darkness at 28±2 o C. For regeneration and selection, the explants were transferred to MS solid medium (0.65% agar) containing 1 mg/mL benzylaminopurine (BAP), 500 µg/L cefotaxime and 30 µg/mL hygromycin and maintained under a 16 hours photoperiod at 25±2 o C. Untransformed explants were placed onto the same medium with or without hygromicin as the negative and positive controls, respectively.Shoots regenerated on selection medium were excised at the base and placed in Magenta GA7 boxes containing rooting medium (the same medium as the regeneration medium but without BAP).The hygromycin-resistant plants tested by PCR using both 3' and 5' α-AI2 and hygromycin primers were transferred to soil and grown in a greenhouse at 25±10 o C and 50% humidity.The mature seeds were collected after 4 months.

Protein extraction
The soluble wild type and recombinant proteins expressed in tobacco seeds, were extracted as previously described (Grossi- de-Sá et al., 1997) by grinding 150 mg of dry seeds, in an ice-cold mortar with 1.0 mL 50 mM Tris-HCl, pH 8.0, containing 0.1% Triton X-100, 30 mM NaCl and 1% 2-mercaptoethanol.The homogenates were twice centrifuged for 10 minutes at the top speed of a bench centrifuge and the supernatants stored at -20 o C or immediately used to measure the inhibitory activity.Protein concentration was determined according to Bradford (1976) using bovine serum albumin as standard.

Alpha-amylase inhibition assay
The α-amylase activity was measured using the dinitrosalicylic acid (DNS) method adapted from Bernfeld (1955) and Ishimoto & Chrispeels (1996) using 1% soluble starch as substrate.To measure inhibitory activity of tobacco seed extracts against either PPA or ZSA expressed in Sf9 insect cells, the solutions containing 200 µg of the total extract (previously analyzed in immunoblots) and one unit α-amylase activity (defined as the amount of this enzyme that increased the absorbance at 546 nm by 0.1 absorbance units during 25 minutes) were pre-incubated at room temperature for 20 minutes prior to the addition of 250 µL of the substrate solution followed by incubation at 37 o C for 10 minutes.The reactions were stopped by the addition of 500 µL of DNS reagent followed by color development, placing the tubes in boiling water for 10 minutes.After addition of 5 mL distilled water, the absorbance was read at 546 nm.Alfa-amylase inhibitor-1 purified from P. vulgaris was used as positive and negative controls in assays with PPA and ZSA, respectively.Assays were carried out in duplicate.Duplicate inhibition values differed by no more than 10%.

Computational methods
In order to study the behavior of mutant M1 in structural terms, four model complexes, PPA-α-AI2, PPA-M1, ZSA-α-AI2 and ZSA-M1, using MODELLER (Sali & Blundell, 1993) were constructed.Experimental results show that the M1 complexes are merely hypothetical, i.e., do not actually form.Since the patterns of inhibition of M2-4 were identical to that of M1, these other mutants were not studied further.The structure of PPA in complex with α-AI1, PDB code 1DHK (Bompard-Gilles et al., 1996) was used to construct the PPA-α-AI2 and PPA-M1 model complexes while the structure of TMA in complex with α-AI1, PDB code 1VIW (Nahoum et al., 1999) was used for modeling ZSA-α-AI2 and ZSA-M1.Alignments were adjusted so that insertions and deletions in the model proteins relative to the templates could be smoothly accommodated.In each case, ten variant models were constructed and refined against MODELLER's objective function which measures the models agreement with template-derived restraints, in combination with the standard CHARM energy function.
The best performing model (that with the lowest objective function) for each of the complexes was taken as the respective final model.Each final model complex was analysed, and compared with the crystal structure of PPA in complex with wild-type α-AI1, using the protein-protein interaction server (Jones & Thornton, 2001).This server determines which residues in a protein complex contribute with the surface areas of the interface and measures geometrical characteristics of the interface which can then be compared with typical values (Jones & Thornton, 1996).Visualization of protein structures was carried out using O (Jones et al., 1991) and Pymol (Delano, 2001) programs.

Results and Discussion
The major interaction between α-AI1 and PPA, as revealed by the crystal structure of the complex, is via two loops of α-AI1 that make numerous hydrogen bonds with the amino acids in the substrate cleft of PPA (Bompard-Gilles et al., 1996;Silva et al., 2000).To assess the role of the two loops for the specificity of α-AI-amylase interaction, chimeric amylase inhibitor was made containing specific changes based on modeling experiments (Silva et al., 2000).Analyzing the full set of previously identified interface-forming residues, four mutants were designed (Table 1).The mutated amino acids were chosen for their putative importance in electrostatic and hydrophobic aspects of the enzymeinhibitor interaction.One option would have been the simple mutation of putatively important residues to amino Table 1.Aminoacid sequence involved on the mutations (1) .
2 Outside of the loops acids of different physicochemical characteristics, or simply to alanine, and the monitoring of the consequent loss of ZSA inhibition.Instead, the chosen residues were mutated to their α-AI1 counterparts so that, in addition to loss of ZSA inhibition, possible acquisition of PPA inhibition could be monitored.Through this dual assay approach, it was expected that an enhanced understanding of the detailed importance of the mutated residues would be achieved.
The PCR products of wild-type and mutant genes were ligated to 5' regulatory sequence of PHA-L, because of known success of this seed-specific promoter in expressing the α-AI activity in plants (Altabella & Chrispeels, 1990).Therefore, all chimeric construction as well as wild-type form of α-AI2 were expressed in tobacco seeds.Since α-AI1 needs to be proteolytically processed to be active, plant expression, leading to proper glycosylation and proteolytic processing, was essential (Altabella & Chrispeels, 1990;Pueyo et al., 1993).Approximately 15-20 independent transgenic plants (tested by PCR) were obtained for each construction.
Four transgenic lines for each construction were used to perform immunoblots with serum raised against α-AI1.The level of transgene expression was approximately equal to all mutants.The similar polypeptide profiles observed in the 12-15 kDa range to tobacco seeds expressing the recombinant proteins and wild-type inhibitor demonstrate that mutations did not destabilize the proteins to a point where they were degraded by endogenous proteases (Figure 1).
To investigate the specificity of the recombinant proteins accumulated in tobacco seeds in relation to α-amylases, total extracts were tested for inhibitory activity against PPA and ZSA.To circumvent possible stability problems, the proteins expressed in tobacco seeds were tested immediately after the preceding extraction.In agreement with previous reports (Suzuki et al., 1993;Grossi-de-Sá et al., 1997, Grossi-de-Sá & Chrispeels, 1997) α-AI1 inhibited PPA but not ZSA, and α-AI2 inhibited ZSA, but not PPA.However, all the mutants lost inhibitory activity against ZSA and none of the mutants acquired specificity toward PPA (Figure 2).Because the level of the expression varied between mutant and wild-type proteins (Figure 1), different concentrations were used in the assays to assess the activity of the mutated proteins.
The high sequence identity between modeled proteins and templates (78% between α-AI1 and α-AI2; 61% for the comparison of ZSA and TMA) along with careful manual positioning of insertions and deletions, ensure reliable models.In the cases of the ZSA complexes, the present models should be significantly more accurate than those previously analysed (Silva et al., 2000) since they were based on the TMA-α-AI1 (Nahoum et al., 1999)   complex rather than on the structure of α-AI1 alone.Comparison of the structures of complexed and uncomplexed α-AI1 models reveals significant differences in the conformations of loops that interact with α-amylase.
Attempts to model a ZSA-M1 complex revealed unnaturally close atomic contacts between ZSA and M1 in the region of the latter's α-AI1-like two residues insertion.Comparison with the TMA-α-AI1 crystal structure shows that these clashes arise mainly from the substitution of Thr291 (TMA numbering) with the larger Asp and from different backbone structure of the loop around residue 293 which is one residue shorter in ZSA than in TMA (Figure 3).As well as the steric aspect, it is notable that Asp 291 in ZSA is largely buried in the hypothetical interface with M1, distant from any positive charge with which to interact.In contrast, in the modeled complex with α-AI2, charge compensation is provided in the form of His33 (Figure 3).These results are consistent with the inhibition of TMA (but not ZSA) by α-AI1 and ZSA (but not TMA) by α-AI2 (Suzuki et al., 1993;Nahoum et al., 1999).
The inability of α-AI2 or M1 to inhibit PPA is consistent with the geometric analyses of the hypothetical complexes obtained from the protein-protein interaction server.The server measures surface complementarity as a gap volume index (Jones & Thornton, 1996).Lower values indicate better complementarity and typical values for experimentally determined enzyme-inhibitor complexes are 2.2±0.5.The values obtained for the two crystal structures of α-amylase-inhibitor complexes are 2.2 and 2.3.In sharp contrast, the hypothetical PPA-M1 and PPA-α-AI2 complexes score 2.7 and 3.0, respectively, at the limit or outside the typical range.These poorer complementarity values presumably contribute to the lack of PPA inhibition by α-AI2 or M1, although other factors may also be involved.
The most striking finding reported is the absence of inhibition of ZSA by M1 which differs from wild-type α-AI2 only in the replacement of His33 with Ser-Tyr-Asn.The explanation obtained through modeling involved combined steric and electrostatic effects.Hence, the alteration in α-AI2 structure produced a drastic change in inhibitory behavior through two distinct mechanisms.This result follows other examples of apparently fine-tuned α-amylase inhibitor specificity (Franco et al., 2000).The multiple factors underlying amylase inhibitor specificity have been reviewed recently (Franco et al., 2002).
The lack of appearance of PPA inhibition even in M4 is explained by consideration of the complexity of the PPA-α-AI1 interface.Eight inhibitor segments containing one to ten residues each contribute to the interface with PPA.This is more than double the number of inhibitor segments typically involved in enzymeinhibitor interfaces (3±0.9 segments).The size of the interface area between PPA-α-AI1 (1,430 Å 2 ) is much larger than the average for enzyme-inhibitor complexes (785±75 Å 2 ).These analyses show that the interface between PPA and α-AI1 is larger and more complex than typical enzyme-inhibitor interfaces.This naturally makes more difficult the task of precisely determining all the factors involved in conferring specificity for α-AI1 or α-AI2 on a given α-amylase (Silva et al., 2000).Even in the most-changed mutant M4, only 41% of the inhibitor surface area in the PPA-α-AI1 complex has been grafted into the α-AI2 environment.

Conclusions
1.The first interacting loop of the inhibitors is an important contributor to specificity but other factors also exist.
2. Computational analyses are useful for highlighting important interface residues and analyzing the results of their mutation.3. The interaction between α-AI and α-amylases is atypically complex so that additional work will be required to fully understand all specificity determinants.

Figure 2 .
Figure 2. Activities of α-AI1and α-AI2(a), extracted from Phaseolus vulgaris, α-AI2(b) and mutants, extracted from transformed tobacco seeds, in relation to ZSA and PPA.NT refers to seed extract from untransformed tobacco.The enzyme activity values were calculated for four transformed lines for each mutant.In each of (a) and (b) the highest activity obtained was defined as 100% and the remaining results expressed as percentages.

Figure 1 .
Figure 1.Immunoblot of the α-AI2 and mutants expressed in tobacco seeds.87.5 µg of the total protein was loaded in each lane.α-AI2a refers to an inhibitor purified from Phaseolus vulgaris and α-AI2b to an inhibitor expressed in tobacco seeds while M1, M2, M3 and M4 refer to α-AI2 mutants.NT refer to untransformed plant (negative control).Molecular weight standard are shown on the left.

Figure 3 .
Figure 3. Structural comparison of the ZSA-α-AI2 (white) and ZSA-M1 (green, thinner bonds and smaller atoms) complexes near the side of mutation.The additional two residues present in M1 compared to α-AI2 cause steric clashes with Asp 291 of ZSA as marked by the broad green dotted line.With α-AI2, in contrast, a favorable ionic interaction with His 32 is present.The extra residues do not clash with the Thr (black) that replaces Asp 291 in TMA.