selection by phage display of a variant mustard trypsin inhibitor toxic against aphids

Selection by phage display of a variant mustard trypsininhibitor toxic against aphids

Luigi R. Ceci1,�, Mariateresa Volpicella2, Yvan Rahbe3, Raffaele Gallerani2, Jules Beekwilder4 and Maarten A. Jongsma4

1Istituto di Biomembrane e Bioenergetica, Sezione di Trani, Via Corato 17, 70059 Trani, Italy,2Dipartimento di Biochimica e Biologia Molecolare, Universita di Bari, Via Amendola 165/A, 70126 Bari, Italy,3INRA-INSA de Lyon, UMR BF2I (Biologie Fonctionelle Insectes & Interactions), 20 ave A. Einstein 69621 Villeurbanne

Cedex France, and4Plant Research International, PO Box 16, 6700 AA Wageningen, the Netherlands

Received 30 July 2002; revised 8 November 2002; accepted 13 November 2002.�For correspondence (fax þ39 080 5443317; e-mail [email protected]).

Summary

The mustard trypsin inhibitor, MTI-2, is a potent inhibitor of trypsin with no activity towards chymotrypsin.

MTI-2 is toxic for lepidopteran insects, but has low activity against aphids. In an attempt to improve the

activity of the inhibitor towards aphids, a library of inhibitor variants was constructed and cloned into the

pRlac3 phagemid vector. The library of 9.3� 107 independent colonies was created by randomisation of a

stretch of five consecutive codons in the reactive site. Repeated selection rounds against bovine trypsin and

chymotrypsin allowed the identification of novel, MTI-2 derived, antitrypsin and antichymotrypsin inhibi-

tors. Chy8, the selected variant with highest affinity for bovine chymotrypsin (Ki¼ 32 nM versus >1000 nM

for the wild-type) represents the strongest known recombinant chymotrypsin inhibitor of the MTI-2 family.

It is highly toxic to nymphs of the aphid Acyrthosiphon pisum, and moderately toxic to nymphs of Aphis

gossypii and Myzus persicae. The LC50 of 73 lg ml�1 towards A. pisum is the lowest value known among

chymotrypsin inhibitors. The aphicidal activity of Chy8 was improved eightfold compared to the wild-type

inhibitor. This demonstrates, for the first time, that bovine chymotrypsin provides a useful template to

select engineered proteins highly toxic against these aphids. The selected gene will allow the development

of transgenic crops that are protected against sucking insect pests.

Keywords: proteinase inhibitor, phage display, pest control, Acyrthosiphon pisum, Sinapis alba, molecular

evolution.

Introduction

Among the major plant defence strategies against herbi-

vores are inhibitors of proteases (Ryan, 1990). The domi-

nant concept explaining the mode of action against insects

is that protease inhibitors (PIs) expressed in plants act by

inhibiting protein digestion in the guts of insect larvae. This

results in amino acid deficiencies, and thereby leads to

serious developmental delays or mortality of larvae. Pro-

tease inhibitors are small proteins, currently described in

more than one hundred plants, which can be catalogued

according to their structure in several families (De Leo et al.,

2002; http://bighost.area.ba.cnr.it/PLANT-PIs). Several atte-

mpts to increase plant resistance to insects by overexpres-

sion of plant PIs have been carried out, but with mixed

success. In some cases, plants were indeed shown to be

protected from insect herbivores (Bell et al., 2001; De Leo

et al., 2001; Delledonne et al., 2001; Heath et al., 1997; Lee

et al., 1999). In other cases, however, plants became more

susceptible (De Leo et al., 1998). This can be explained

at least in part by the observation that after feeding on

PI-expressing plants or PI-containing diets, some insects

appear to adapt to the ingestion of PIs by producing new

and ‘PI-insensitive’ proteases (Brito et al., 2001; Jongsma

and Bolter, 1997; Jongsma et al., 1995a; Mazumdar-

Leighton and Broadway, 2001; Paulillo et al., 2000).

Mustard Trypsin Inhibitor 2 (MTI-2) is a trypsin inhibitor

isolated from mustard seeds (Menegatti et al., 1992) and

belongs to a family of PIs so far restricted to the Cruciferae

(De Leo et al., 2002). MTI-2 has been shown to inhibit in vitro

more than 75% of the protease activity of gut extracts from

Spodoptera exigua larvae (Volpicella et al., 2000). The

The Plant Journal (2003) 33, 557–566

� 2003 Blackwell Publishing Ltd 557

activity of the inhibitor expressed in transgenic plants has

been evaluated for different plant–insect combinations. Lar-

vae of Spodoptera littoralis, Mamestra brassicae and Plu-

tella xylostella reared on tobacco, Arabidopsis thaliana and

rapeseed displayed a variable level of susceptibility. The ef-

fects ranged from complete mortality of P. xylostella larvae

on A. thaliana, to almost complete insensitivity of M. bras-

sicae larvae on rapeseed plants (De Leo et al., 1998, 2001).

In contrast to the preceding leaf-eating insect pests,

aphids are exclusively feeding on plant phloem-sap, which

usually displays a high free/bound amino acid ratio. Con-

sequently, they are thought not to rely on extensive protein

digestion for their nitrogen supply. However, a limited

number of serine PIs were shown to promote deleterious

effects on different aphid species (Casaretto and Corcuera,

1998; Quillien et al., 1998; Rahbe and Febvay, 1993; Rahbe

et al., 1995; Tran et al., 1997). Also, a modified oryzacystatin-

I (Oc-IDD86) and a chicken egg white cystatin were shown to

reduce the growth and survival of nymphs of Myzus persi-

cae in artificial diet assays (Cowgill et al., 2002). Within the

Bowman-Birk family of bifunctional serine proteinase inhi-

bitors, aphid toxicity was further demonstrated to be sup-

ported by the antichymotrypsin head (Rahbe et al., 2002b).

This led us to test the possibility that a chymotrypsin

specificity could also drive aphid toxicity in an independent

family of protease inhibitors.

Severalauthorshave reportedon the planned use ofphage

display to select improved inhibitors against pest proteases

(Beekwilder et al., 1999; Jongsma et al., 1995b; Koiwa et al.,

1998; Koiwa et al., 2001), but none so far have achieved the

aim of improved activities against target pests. Only rece-

ntly a direct comparison between two Helicoverpa prote-

inases with different sensitivities to four different plant PIs

was possible (Volpicella et al., 2002). It allowed mapping of

residues differing in the structures of two closely related

proteinases. It revealed a concentration of mutations

around the proteolytic groove which may be involved in

preventing access of the inhibitors. Earlier, we demonstrat-

ed the efficient selection on trypsin of phage particles dis-

playing wild-type MTI-2 from a large population of phages

exhibiting an inactive mutant (Volpicella et al., 2001). These

results prompted us to construct a large library of MTI-2

active site variants to be displayed on phage particles and

used for the selection of inhibitor variants with desirable

specificity and improved affinity. In this paper, the isolation

and characterisation of a novel inhibitor of plant origin with

a new protease specificity and with the strongest aphicidal

activity described in the literature to date is reported.

Results

Design and construction of a phage display library

In small proteinaceous trypsin inhibitors, enzyme recogni-

tion and binding affinity is primarily mediated by the P1 and

surrounding residues (Laskowski and Kato, 1980). In MTI-2,

the P1 residue is located at Arg-20 of the mature protein

(Ceciliani et al., 1994). Therefore, we chose to randomise

the DNA coding for amino acids Ala-18 to Phe-22. The 15

nucleotides encoding this stretch were replaced by NNK

codons (N¼A, C, G, T; K¼G, T), introduced by PCR into the

region of mti-2 coding for the mature protein (Figure 1). The

randomised genes were cloned into vector pRlac3, which

has been demonstrated to stably harbour trypsin-inhibitor

libraries (Beekwilder et al., 1999). The NNK codon rando-

misation results in the representation of all 20 amino acids

at each position, and leads potentially to about 34 million

different DNA sequences, encoding 205¼ 3.2 million differ-

ent polypeptides. A total of 93 million independent colonies

were estimated to be present in the entire library, predicting

a close to full representation of all peptide combinations

(Koscielska et al., 1998). Twenty clones were taken for sequ-

encing and no biases to particular nucleotides or codons

were observed (not shown).

Selection of MTI-2 variants against trypsin

and chymotrypsin

The MTI-2 phage display library was used to select MTI-2

variants active against bovine trypsin and chymotrypsin, as

Figure 1. Sequence of the mti-2 gene regioncoding for the mature MTI-2.The mti-2 sequence is reported in codons withthe corresponding amino acids above. The fivecodons to be randomised are in bold. Mutatednucleotides generating the MluI restriction siteare in capital letters (see Experimental proce-dures for details). The location of the MTI2-RANDOM primer is shown. The NcoI and NotIrestriction sites of the pRlac3 vector, and theMluI site are in italics. nnm is the complemen-tary code for nnk codons in the 50 directedmutagenic primer.

558 Luigi R. Ceci et al.

� Blackwell Publishing Ltd, The Plant Journal, (2003), 33, 557–566

described in the Experimental procedures. The amounts of

phage particles used and recovered for each cycle are

reported in Table 1. In terms of phage recovery (output/

input), the population of phages selected on trypsin gra-

dually enriched 450-fold in four selection rounds for MTI-2

variants that bind to trypsin. For chymotrypsin, a 300-fold

enrichment was observed between rounds 1 and 4. In round

3, the 100-fold enrichment of round 2 was partially lost, but

regained in round 4. Phage particles recovered from the

fourth cycle of both selections were analysed in more detail.

For each selection, DNA from 10 different colonies was

sequenced (Table 2a,b).

Analysis of selected clones

For 10 MTI-2 variants selected on trypsin and 10 on chymo-

trypsin, clonal phage stocks were generated, and the capa-

city to bind to immobilised trypsin or chymotrypsin was

assayed by phage ELISA. Phages selected on trypsin

showed similar binding activities as phages carrying

wild-type MTI-2 (Table 2a), indicating that trypsin-binding

phages were well represented. On the other hand, only two

phages selected on chymotrypsin showed significantly

enhanced binding to chymotrypsin as compared to the

wild-type MTI-2 phage (Table 2b). Phages displaying potato

Table 1 Panning the MTI-2 library against trypsin andchymotrypsin

Input Output Recovery

Trypsin selection1st round 1.0� 1013 1.0� 105 1.0� 10�8

2nd round 1.0� 1013 6.0� 106 6.0� 10�7

3rd round 8.4� 1013 3.6� 107 4.2� 10�7

4th round 4.0� 1012 1.8� 107 4.5� 10�6

Chymotrypsin selection1st round 1.0� 1013 2.0� 104 2.0� 10�9

2nd round 0.8� 1013 1.7� 106 2.1� 10�7

3rd round 2.8� 1013 2.2� 105 7.8� 10�9

4th round 1.0� 1013 6.0� 106 6.0� 10�7

Number of phages used (input) and selected (output) for eachround of selection has been reported. Recovery is given as output/input ratio.

Table 2 Sequences and activity of selected MTI-2 phage variants selected against trypsin and chymotrypsin after the 4th round ofselection(a)

Phage clone Protein sequence DNA sequence ELISA signal

Empty – – 45PI-2 – – 131WT APRIF GCA CCT CGG ATA TTT 222Try1–3 APRIF GCA CCT CGG ATA TTT 222Try 4–5 APRIF GCA CCT CGG ATT TTT 246Try 6–8 APRTL GCG CCG CGG ACT CTG 281Try 9 ST#PL AGT ACG TAG CCT TTG NDTry 10 Frameshift TTT TGG CAG AC- TTT NDLibrary XXXXX NNK NNK NNK NNK NNK 40

(b)

Phage clone Protein sequence DNA sequence ELISA signal

Empty – – 35PI-2 – – 81WT APRIF GCA CCT CGG ATA TTT 43Chy1 LPRLA CTG CCG CGT TTG GCG 46Chy2 PHGTT CCG CAT GGT ACG ACG 39Chy3 RHTKE AGG CAT ACG AAG GAG 43Chy4 NSRTQ AAT TCG AGG ACT CAG 115Chy5 PPTMQ CCG CCT ACG ATG CAG 34Chy6 PRDTP CCG CGG GAT ACT CCG 52Chy7 PRSTP CCT CGG TCT ACG CCG 35Chy8 APWRH GCT CCT TGG AGG CAT 112Chy9 LLLGI CTC CTA CTT GGC ATC 53Chy10 HHAHQ CAT CAT GCG CAT CAG 62Library XXXXX NNK NNK NNK NNK NNK 29

Empty: phage particles without recombinant proteins; PI2: phage particles carrying recombinant potato protease inhibitor PI2; WT: phageparticles carrying recombinant mustard protease inhibitor MTI-2; Library: phage particles carrying a fraction of the original library of MTI-2variants; #: TAG stop codon, coding for Gln in supE E. coli cells; ND not determined.


Phage-selected aphicidal protein 559

proteinase inhibitor 2 (PI-2) were used as positive control.

PI-2 is a double-headed inhibitor with subnanomolar affi-

nity towards trypsin and chymotrypsin (Beekwilder et al.,

2000). Phages Chy4 and Chy8 gave similar or higher ELISA

reactions as the PI-2 phage, indicating comparable affi-

nities for chymotrypsin. Mixed library phages (unselected)

showed no enhanced binding to any of the enzymes, com-

pared to negative control phages without any PI insert.

From analysis of the sequences of the 10 trypsin-selected

phages, it can be seen that two variants prevail: APRIF (five

times) and APRTL (three times) (Table 2a). The sel-

ected APRIF variants, which are identical to the wild-type

MTI-2, are encoded by two different nucleotide sequen-

ces, indicating that the wild-type amino acid sequence is re-

selected from a very large sequence space and must

represent an optimal conformation for bovine trypsin with

very few degrees of freedom. Both variants have the

arginine residue at the putative P1-position which is con-

sidered to be important for trypsin binding. The selected

frameshift variant does not produce MTI-2 and its presence

suggests that this clone still has a significant growth

advantage relative to trypsin inhibiting MTI-2 variants

despite the use of the pRLac3 vector (Beekwilder et al.,

1999).

The chymotrypsin-selected variants are not dominated

by a particular sequence (Table 2b). Only in the Chy8

(APWRH) and Chy9 (LLLGI) mutants the aromatic or hydro-

phobic residues in P1-position (W and L, respectively)

correspond to the expected residues for chymotrypsin

inhibition (Laskowski and Kato, 1980). The phage ELISA,

however, showed that only Chy4 and Chy8 bound signifi-

cantly above background to chymotrypsin.

Recombinant expression of MTI-2 variants and

Ki determination

MTI-2 variants Try7, Chy4 and Chy8 were produced in the

yeast Pichia pastoris, purified and characterised in terms of

their apparent equilibrium constants (Ki) against trypsin

and chymotrypsin. Results are shown in Table 3 together

with published values for wtMTI-2 and the PLI mutant

(Volpicella et al., 2001). PLI is an MTI-2 mutant in which

Arg-20 was replaced by Leu-20 by site-directed mutagen-

esis, rendering it an inhibitor of chymotrypsin. It is the only

other known recombinant chymotrypsin inhibitor of the

MTI-2 family. The selected trypsin binders, represented

by wtMTI-2 and Try7, have similar low, picomolar inhibition

constants for trypsin (Ki¼ 0.010 and 0.052 nM, respec-

tively), and do not inhibit chymotrypsin at all. Variant

Chy8 has an affinity for chymotrypsin (Ki¼ 32 nM), which

is twice better than the PLI mutant (Ki¼ 60 nM). Trypsin

inhibition is almost lost in Chy8. Unexpectedly, the Chy4

variant does not inhibit chymotrypsin at all, though it

mediates strong binding of phages to chymotrypsin.

Activity of selected chymotrypsin inhibitors

against aphids

The effects of wtMTI-2, PLI and Chy8 on growth and survival

of the pea aphid Acyrthosiphon pisum is displayed in

Figure 2 and Table 4. At the lowest dose of 20 mg ml�1, the

chymotrypsin inhibitor Chy8 causes significantly stronger

growth inhibition compared to the wtMTI-2 and PLI mutant

(Fisher’s PLSD test post-ANOVA versus wild-type, P¼0.0062). The inhibitory concentration required to reach 50%

growth reduction during the experimental period (IC50) of

Chy8 (75 mg ml�1) and PLI (198mg ml�1) was five- and two-

fold lower compared to the wtMTI-2 (366 mg ml�1). In addi-

tion, Chy8 and PLI induced complete mortality after a few

days at a dose of 500mg ml�1 (Figure 2b). At this high dose,

the median lethal time, i.e. the time required to reach 50%

mortality in the exposed population (median LT50) of Chy8,

was 1.50� 0.097 days, while that of wtMTI-2 was 3.50�3.67 days (log-rank test w21 df ¼ 52:08, P< 0.0001). This dif-

ference in mortality rate was also very significant for the

Chy8 mutant at dose 100 mg ml�1 (LT50¼ 2.44� 0.54 days for

Chy8, and >8 days for wtMTI-2, w21 df ¼ 13.8, P¼ 0.0002).

These results demonstrate a two-step improvement in toxi-

city (Table 4), first by changing the specificity of wtMTI-2 to

chymotrypsin (PLI), and second by improving the affinity for

chymotrypsin twofold by phage display selection (Chy8).

The toxicity of the wild-type and Chy8 inhibitors was

also assayed against the aphids M. persicae and A. gossy-

pii. In Table 4, IC50, LC50 and mean LT50 values observed for

these two species and A. pisum are listed and a comparison

is made with PsTi-2 (Rahbe et al., 2002b). Chy8 strongly

inhibited growth of M. persicae (IC50¼ 145mg ml�1), but

A. gossypii was less sensitive (IC50¼ 500mg ml�1). Never-

theless, Chy8 performed better than wtMTI-2 and PsTi-2

against both aphids. The effect of Chy8 on growth of

M. persicae was more pronounced than on its survival.

Discussion

Chy8, the MTI-2 variant selected by phage display against

chymotrypsin, was revealed as the strongest chymotrypsin

Table 3 The apparent equilibrium dissociation constants (Ki) ofselected MTI-2 variants for bovine trypsin and chymotrypsin

Trypsin (nM)b Chymotrypsin (nM)b,c

awtMTI-2 0.010� 0.002 >1000Try7 0.052� 0.009 >1000aPLI 34� 9 60� 10Chy4 18� 5 >1000Chy8 161� 96 32� 3

aData of wtMTI-2 and PLI from Volpicella et al. (2001).bValues are from at least three separate determinations.cThe assay does not permit to accurately determine Ki greaterthan 1000 nM.



inhibitor and the most potent aphicidal agent within the

MTI-2 family of PIs. Feeding assays with nymphs of

A. pisum, A. gossypii and M. persicae also showed that

Chy8 is more toxic to aphids than PsTi-2, a strong chymo-

trypsin inhibitor of the Bowman-Birk family. These results

support previous findings indicating that some chymotryp-

sin inhibitors, in contrast to trypsin inhibitors, are active

against aphids (Quillien et al., 1998; Rahbe et al., 2002b).

Thus, a new class of genes is available for increasing plant

defence against aphids, and phage display selection of

more active inhibitors can assist in improving these plant

defence strategies.

Aphid-targeted toxins and inhibitory peptides are known,

and have been reported to be active when expressed in

Figure 2. Toxicity of purified recombinant MTI-2 inhibitors against the pea aphid.Dose–response of larval growth inhibition, compared to aphids reared on control diet (a), and evolution of survival during whole larval life at dose 500 mg ml�1 (b);wild-type inhibitor (circles/dotted lines), and chymotrypsin-directed mutants: PLI (rhombuses/dashed lines) and Chy8 (squares/solid lines); bars representstandard error of mean.

Table 4 Toxicity of Chy8 and MTI-2 inhibitors against aphids

Acyrthosiphon pisum Myzus persicae Aphis gossypii

IC50, mean�SE (mg ml�1)Chy8 75 145 Approximately 500PLI 198 ND NDMTI2wt 366 ND Not inhibitedPsTi-2 280 >800 >800

LC50, mean�SE (mg ml�1)Chy8 73 (54–99) Approximately 500 Approximately 475PLI 151 (110–206) ND NDMTI2wt 567 (271–ND) ND >500PsTi-2 310 Not lethal Not lethal

LT50, days, Kaplan–Meier estimate (at 500 mg ml�1)A. pisum (median) A. pisum (mean) M. persicae (mean) A. gossypii (mean)

Chy8 1.50� 0.097 1.41� 0.71 5.70� 0.14 3.50� 0.36PLI 1.50� 0.45 2.14� 0.25 6.30� 0.09 NDMTI2wt 3.50� 3.67 3.69� 0.25 6.41� 0.06 3.76� 0.31PsTi-2 3.50� 2.00 2.77� 0.11 Not lethal� Not lethal�

�800 mg ml�1.IC50 and LC50 values observed for A. pisum, M. persicae and A. gossypii aphids fed on inhibitor-containing diets. Data for the Bowman-Birkinhibitor from pea (PsTi-2; Rahbe et al., 2002b) are provided for comparison.Median and mean values indicate times at which 50% of the population is alive and the mean survival times, respectively.



transgenic plants (Jouanin et al., 1998). The observed activ-

ities, however, are nowhere near the highly active (nano-

molar range) microbial Bacillus thuringiensis toxins

available for lepidopteran and some coleopteran pests.

The lack of available toxins against the large plant-sucking

order of hemipteran insects has negatively affected the

economic success of Bt toxin-engineered plants like potato,

because the need to spray insecticides was not significantly

reduced. High throughput screening of semi-purified toxins

is difficult in this insect group owing to its peculiar feeding

physiology. Optimisation of lead compounds therefore

seems an attractive alternative strategy.

The P1 residue of canonical PIs represents the main

determinant in establishing the specificity of the inhibited

protease (Laskowski and Kato, 1980). It lies in an exposed

loop, surrounded by amino acids which are important for

the stability of the complex between PI and protease. This

has been extensively investigated for the ‘double-headed’

Bowman-Birk inhibitors by screening synthetic combina-

torial peptide libraries of 11 amino acids (McBride and

Leatherbarrow, 2001). Unfortunately, the same approach

cannot be carried out for MTI-2. A synthesised circular

oligopeptide corresponding to residues 17–26 of MTI-2

(CAPRIFPTIC), and containing the P1 residue at Arg-20

reactive site, does not show any activity toward trypsin

(Ceci, unpublished results). This result may be explained by

the fact that due to a more complex fold, the loop contain-

ing the MTI-2 reactive site is not joined by these terminal

cysteine residues (Ruoppolo et al., 2000).

In the case of the MTI-2 inhibitor, we decided to address

a combinatorial analysis of the reactive site region by tak-

ing advantage from phage display technology (Scott and

Smith, 1990). Previous studies for the plant PIs PI-2 (a

potato serine proteinase inhibitor), MTI-2 and scN and

scL (two soybean cysteine-proteinase inhibitors) demon-

strated that the capacity of PI-displaying phage particles to

bind to immobilised proteinases is strictly related to the

binding affinity of the molecules for the inhibited enzymes

(Jongsma et al., 1995a; Koiwa et al., 1998; Volpicella et al.,

2001). For the three PI types, it was also demonstrated

that phage libraries displaying mixes of active and inactive

(or less active) inhibitors at different ratios allowed the

enrichment of phage particles displaying active inhibitors

(Jongsma et al., 1995b; Koiwa et al., 1998; Volpicella et al.,

2001). Phage display selections of large randomised

libraries also allowed the identification of improved inhi-

bitors (Beekwilder et al., 2000; Koiwa et al., 2001).

Assuming that MTI-2 follows the canonical model of

serine protease inhibition, we decided to randomise the

five consecutive amino acids surrounding the P1 residue

Arg-20. To test whether the MTI-2 phage display library was

capable of generating improved inhibitors against aphids,

bovine chymotrypsin was used as selection target. The use

of a species-specific target enzyme was not possible in this

case, because no chymotrypsin activity was detected in

aphid midguts.

Trypsin is a natural target for MTI-2, and was chosen as a

positive control. We found very strong trypsin inhibitors

after four selection rounds with equilibrium dissociation

constants ranging from 10 to 50 pM. The wild-type MTI-2

protein sequence dominated the selected clones (Table 2a,

50% of the selected clones: APRIF) and was encoded by two

different DNA sequences. The only other sequence variant

with activity (Try6–8: APRTL) was found three times among

10 examined clones. It differed from the wild-type in only

two positions, but was five times less effective with a Ki of

52 pM, so that apparently a low degree of sequence diver-

gence already creates a high loss in affinity.

When chymotrypsin was used as a selection target, the

fraction of recovered phages improved two orders of mag-

nitude compared to the initial ratio, but was about seven-

fold lower compared to trypsin. The efficiency of phage

recovery is considered to be an indicator of the affinity of

the binding variants for the target molecule so that less

strong binders were expected (Markland et al., 1996). The

phage ELISA indicated the presence of only two variants

with strong chymotrypsin affinity. Chy8, as recombinant

protein, was shown to be an inhibitor of chymotrypsin with

an equilibrium dissociation constant of 32 nM. This Ki is

about twofold better than that obtained by simply modify-

ing the P1 residue from Arg to Leu (Volpicella et al., 2001).

The behaviour of recombinant Chy4 variant is different: it

mediates a strong ELISA signal, but shows no detectable

inhibition. Probably this variant binds to a different exosite

on the chymotrypsin surface without hampering substrate

digestion by chymotrypsin. Chy8 is the strongest inhibitor

selected against chymotrypsin, but still has a Ki which is

1000-fold higher compared to wtMTI-2 on trypsin (Table 3).

An inhibitor with a poor Ki will enrich at a slower rate

compared to inhibitors with a good Ki. This may explain

the presence of a rather large fraction of phages with poor

binding to chymotrypsin and the lack of a consensus

sequence among the selected clones.

When Chy8 was fed to the aphid A. pisum, it was found to

be at least fivefold more toxic than the chymotrypsin inhi-

bitors of the Bowman-Birk family (Quillien et al., 1998;

Rahbe and Febvay, 1993; Rahbe et al., 2002b). It also

affected the growth of the aphids M. persicae and A. gos-

sypii, but less than A. pisum, as was also observed for the

pea inhibitor (Table 4). Interestingly, when A. pisum was

reared on artificial diet containing the MTI-2-PLI mutant,

which has an antichymotrypsin activity between those of

MTI-2 wild-type and Chy8 (Volpicella et al., 2001), an inter-

mediate effect on both growth and mortality of insects was

observed. From these results it appears that aphicidal

activity of MTI-2 mutants is correlated to their inhibitory

activity against bovine chymotrypsin. This quantitative

relationship cannot be generalised to a Ki value only, since



the pea Bowman-Birk inhibitor PsTi-2, which has an inhi-

bitor constant against bovine chymotrypsin of 1.3 nM

(Pouvreau et al., 1998), is not more efficient against

A. pisum than Chy8 (Table 4). This may be explained by

the fact that the affinity for a molecular target is not the only

determinant of toxicity of a metabolite for a living organism

(Witt et al., 2001). Transport and stability processes are

among the important pharmacokinetic traits determining

the final toxicity outcome. Little is known about peptide

transport in insects, although delivery to internal tissues in

immunologically active forms has been demonstrated for

immunoglobulins (Allingham et al., 1992), a lectin (Fitches

et al., 2001), oryzacystatin-I (Rahbe et al., 2002a) and PsTi-2

(Deraison-Manuel, personal communication).

Plant PIs have been regarded as defence molecules active

primarily in the guts of protein-digesting insects. The mech-

anism of action against aphids, however, remains unresol-

ved considering that no significant serine protease activity

has been detected in the guts. The fact that the activity

differs between different aphid species further complicates

our understanding (Casaretto and Corcuera, 1998; this

paper). More chymotrypsin inhibitors belonging to differ-

ent families and more inhibitor variants must be investi-

gated to firmly establish the positive correlation between

antichymotrypsin activities and aphicidal properties. It may

turn out that only with knowledge of the bona fide mole-

cular target of such inhibitors we may expect progress in

understanding the toxic properties of inhibitors, and in the

identification of variants more specific than those selected

with the bovine enzyme. The in vitro toxicity (LC50¼ 73mg

ml�1) of Chy8 against A. pisum is sufficient to predict that

transgenic plants expressing this protein at approximately

1% of total soluble protein in the phloem should be pro-

tected against this aphid. This would mark an important

step in the development of transgenic plants with protec-

tion against sucking pests. It would also provide a direct

demonstration of the potential role of endogenous pro-

tease inhibitors in the phloem in plant defence against

sucking pests.

Experimental procedures

Strains and materials

Xl-1 blue (Stratagene) and ElectroMAX DH12S cells (Life Technol-ogies) were used as bacterial host. GS115 (his4) P. pastoris strainwas obtained from Invitrogen. Unless specifically indicated, allDNA manipulations were according to standard procedures or asspecifically indicated in the manuals for the Quickchange site-directed mutagenesis kit (Stratagene), expression in P. pastoris(Invitrogen), Pwo DNA polymerase (Boehringer) and restrictionenzymes. All recombinant plasmids were checked by sequencing.Bovine trypsin (TPCK treated, chymotrypsin free) and chymotryp-sin (TLCK treated, trypsin free) were purchased from Sigma.

Oligonucleotides

The following primers were used. Figures above the sequencescorrespond to position in the mti-2 sequence reported in Figure 1.Figures above pRLAC3-1347 give sequence location in the pRLac3vector (Beekwilder et al., 1999). Restriction sites are underlined. Atpositions indicated as N, all four nucleotides are incorporated withthe same frequency. At positions indicated as M, either A or Cnucleotides are incorporated.



Construction of pRLac3-MTI2-MluI

The phagemid pRlac3-MTI-2 consisting of the phage display vectorpRlac3 carrying the gene for mature MTI-2 was used to constructthe phage library (Volpicella et al., 2001). A silent MluI site at aconvenient distance from the reactive site was introduced into mti-2 (Figure 1, codons 28 and 29 from MTI-2). For this purpose, theQuickchange site-directed mutagenesis kit was used with oligo-nucleotides MTI2-WL and MTI-2 WLR.

Randomisation of the mti-2 gene

A 1090-bp fragment overlapping with the vector DNA was ampli-fied with Pwo polymerase using pRlac3-MTI2 as a template andoligonucleotides PRLAC3-1347 and MTI2-RANDOM as primers.MTI2-RANDOM contains five codons of the mti-2 gene that arerandomised. Subsequently, the amplified fragment was purifiedthrough QIAquick PCR purification system (QIAGEN). The purifiedfragment was digested with ApaLI and MluI, after which it wasseparated by electrophoresis in 1% Seakem GTG agarose (FMCBioProducts). An 870-bp fragment was eluted from the gel usingthe QIAEX II gel extraction kit (QIAGEN).

The pRlac3-mti2-MluI vector was also digested with ApaLI andMluI restriction enzymes. Restricted vector was electrophoresedand purified as described for the PRLAC3-1347/MTI2-RANDOMPCR amplified fragment.

Large scale ligation was carried out using 3.6 mg of vector and5.4 mg of insert in 2.7 ml reaction volume using T4 DNA ligase(NEB) at 168C overnight. After ligation, DNA was concentrated byethanol precipitation at �808C for 2 h in the presence of0.1 mg ml�1 glycogen (Life Technologies) and 0.3 M NaAcpH¼ 5.2. After centrifugation, DNA pellets were washed twice with70% ethanol and dissolved in 30 ml of 0.1�TE.

Ligated DNA was used to transform ElectroMAX DH12S cells byelectroporation at 1.8 V, 25mFD, 200O. Transformed cells werecollected and plated on LB plates containing 0.1 mg ml�1 ampicillinand 2% glucose, and incubated overnight at 378C. Before plating,an aliquot was taken to determine the concentration of ampicillin-resistant colony forming units (cfu). The next day colonies wererecovered from plates with 40 ml of LB-broth with 2% glucose and15% glycerol, and stored at �808C as 0.75 ml aliquots.

Selection of the phage display library

0.4 ml of the library of bacteria (corresponding to 3.4� 1010 cfu)were used for phage particle production. Phage particles wereobtained by using the VCSM13 helper phage (Stratagene) asdescribed before (Beekwilder et al., 1999).

Selection was carried out with proteases immobilised on immu-notubes essentially as described by Griffiths et al. (1994). Bovinetrypsin or chymotrypsin were coated on immunotubes (Maxisorp75 mm� 12 mm, Nunc) by using 4 ml of either trypsin solution(0.2 mg ml�1 in 100 mM Tris–HCl pH 7.8, 0.25% glutaraldehyde) orchymotrypsin solution (0.2 mg ml�1 in 100 mM Tris–HCl pH 7.8,0.25% glutaraldehyde). Binding of proteases was allowed for 45min at room temperature. After three washes with PBS buffer,tubes were filled with PBS buffer supplemented with 2% w/v non-fat dry milk, left at room temperature for 2 h, and washed againwith PBS. 1� 1013 phages in 4 ml of PBS buffer with 2% w/v non-fatdry milk and 0.1% v/v Tween 20 were then added to the coatedtubes and left for 30 min at room temperature with gentle rotation,followed by 90 min without agitation. After removing the phagesuspension, tubes were rinsed 10 times with PBS buffer with0.1% v/v Tween 20 and 10 times with PBS buffer. Bound phageswere removed by adding 1 ml of 0.1 M HCl–glycine pH 2.2 andincubating for 30 min at room temperature with gentle rotation.Eluted phage suspensions were recovered and neutralised with0.5 ml of 1 M Tris pH 7.4. Recovered phage particles were thenamplified in E. coli as described by (Beekwilder et al., 1999). Onemillilitre of amplified phages (between 1� 1012 and 1� 1013 cfu)were used in the subsequent rounds of selection, for a total of fourrounds.

From the second round onwards, tube washing after incubationwith phages was increased to 20 times with PBS buffer with 0.1% v/v Tween 20, and 20 times with PBS.

Analysis of protease binding of phages

Selected phages were used to infect Xl-1blue E. coli cells, andisolated colonies were picked. Clones were grown overnight inLB broth supplemented with ampicillin and glucose. Subse-quently, 15% glycerol was added, and clones were stored at�808C. Plasmid DNA was isolated for sequencing using Big-Dyeterminator sequencing (Perkin Elmer). Individual clones were usedto generate phages as described by Beekwilder et al. (1999). ELISAof phages were performed as already described (Beekwilder et al.,1999) using the Pharmacia Phage ELISA kit.

Expression and activity of MTI-2 mutants

Selected antitrypsin and antichymotrypsin MTI-2 variants wereexpressed in the yeast P. pastoris as already described for thenative MTI-2 (Volpicella et al., 2000). Coding regions were obtainedfrom the phagemide vectors by amplification of a 234-bp fragmentusing the plasmid from the selected phage as template and oli-gonucleotides MTI2-XhoI and MTI2-UGA as primers, and then



transferred to the pPIC9 vector for secreted expression in P. pas-toris. Expression was usually performed in 100 ml in 1 l flasks.

Purification of recombinant proteins was carried out by gelfiltration through a Sephadex G50 column (3 cm� 100 cm) equili-brated with 50 mM Na-phosphate buffer, pH 7.2.

Activities against trypsin and chymotrypsin were determined byusing the substrates benzoyl-DL-Arg-p-nitroanilide (BApNA) andN-succinyl-L-Ala-L-Ala-L-Pro-L-Leu-p-nitroanilide (SAAPLpNA), re-spectively, as described (Volpicella et al., 2000, 2001).

Aphid bioassays

For assessing their oral toxicity towards the pea aphid, recombi-nant proteins eluted in water from phenyl-sepharose chromato-graphy were assayed for protein content and lyophilised. Wild-type inhibitor and the PLI and Chy8 mutants were tested between20 and 500 mg ml�1. Acyrthosiphon pisum (clone LL01, collected onalfalfa) was used in artificial diet bioassays as described by Rahbeand Febvay (1993). Three batches of 18 individual new-born larvaewere put on control and treated diets, and aphids were scoredevery day for survival, and finally weighed after 8 days at thenearest microgram. Growth data were analysed by one-wayANOVA, and modelled according to an exponential decay fit togive IC50 (inhibition concentration 50, dose inhibiting larval growthby 50% as compared to controls). Survival data, with censoredobservations for surviving individuals at the end of experiment,were analysed by the non-parametric Kaplan–Meier statistics toreach significance of treatments (Log-rank test) and median ormean survival times (LT50) with their corresponding standarderrors. All data were analysed with STATVIEW v5 MacOS (SASInstitute Inc.).

Acknowledgements

The work described in this paper was part of EU project FAIR6-CT98-4239 and programme 338 of the Dutch ministry of Agricul-ture, Nature Management and Fisheries.

References

Allingham, P.G., Kerlin, R.L., Tellam, R.L., Briscoe, S.J. and Stand-fast, H.A. (1992) Passage of host immunoglobulin across themid-gut epithelium into the haemolymph of blood-fed buffaloflies Haematobia Irritans Exigua. J. Insect Physiol. 38, 9–17.

Beekwilder, J., Rakonjac, J., Jongsma, M. and Bosch, D. (1999) Aphagemid vector using the E. coli phage shock promoter facil-itates phage display of toxic proteins. Gene, 228, 23–31.

Beekwilder, J., Schipper, A., Bakker, P., Bosch, D. and Jongsma,M.A. (2000) Mutagenesis and recombinant expression of potatoproteinase inhibitor II. Eur. J. Biochem. 267, 1975–1984.

Bell, H.A., Fitches, E.C., Down, R.E., Ford, L., Marris, G.C., Edwards,J.P., Gatehouse, J.A. and Gatehouse, A.M. (2001) Effect of diet-ary cowpea trypsin inhibitor (CpTI) on the growth and develop-ment of the tomato moth Lacanobia oleracea (Lepidoptera:Noctuidae) and on the success of the gregarious ectoparasitoidEulophus pennicornis (Hymenoptera: Eulophidae). Pest Manag.Sci. 57, 57–65.

Brito, L.O., Lopes, A.R., Parra, J.R.P., Terra, W.R. and Silva, M.C.(2001) Adaptation of tobacco budworm Heliothis virescens toproteinase inhibitors may be mediated by the synthesis of newproteinases. Comp. Biochem. Physiol. B Biochem. Mol. Biol.128, 365–375.

Casaretto, J.A. and Corcuera, L.J. (1998) Proteinase inhibitor accu-mulation in aphid-infested barley leaves. Phytochemistry, 49,2279–2286.

Ceciliani, F., Bortolotti, F., Menegatti, E., Ronchi, S., Ascenzi, P.and Palmieri, S. (1994) Purification, inhibitory properties, aminoacid sequence and identification of the reactive site of a newserine proteinase from oil-rape (Brassica napus) seeds. FEBSLett. 342, 221–224.

Cowgill, S.E., Wright, C. and Atkinson, H.J. (2002) Transgenicpotatoes with enhanced levels of nematode resistance do nothave altered susceptibility to non-target aphids. Mol. Ecol. 11,821–827.

De Leo, F., Bonade-Bottino, M.A., Ceci, L.R., Gallerani, R. andJouanin, L. (1998) Opposite effects on Spodoptera littoralislarvae of high and low level of a trypsin proteinase inhibitorin transgenic plants. Plant Physiol. 118, 997–1004.

De Leo, F., Bonade-Bottino, M.A., Ceci, L.R., Gallerani, R. andJouanin, L. (2001) Effects of a mustard trypsin inhibitorexpressed in different plants on three lepidopteran pests. InsectBiochem. Mol. Biol. 31, 593–602.

De Leo, F., Volpicella, M., Licciulli, F., Liuni, S., Gallerani, R. andCeci, L.R. (2002) PLANT-PIs: a database for plant protease inhi-bitors and their genes. Nucl. Acids Res. 30, 347–348.

Delledonne, M., Allegro, G., Belenghi, B., Balestrazzi, A., Picco, F.,Levine, A., Zelasco, S., Calligari, P. and Confalonieri, M. (2001)Transformation of white poplar (Populus alba L.) with a novelArabidopsis thaliana cysteine proteinase inhibitor and analysisof insect pest resistance. Mol. Breed. 7, 35–42.

Fitches, E., Woodhouse, S.D., Edwards, J.P. and Gatehouse, J.A.(2001) In vitro and in vivo binding of snowdrop (Galan-thus nivalis agglutinin; GNA) and jackbean (Canavalia ensifor-mis; Con A) lectins within tomato moth (Lacanobia oleracea)larvae; mechanisms of insecticidal action. J. Insect Physiol. 47,777–787.

Griffiths, A.D., Williams, S.C., Hartley, O. et al. (1994) Isolation ofhigh affinity human antibodies directly from large syntheticrepertoires. EMBO J. 13, 3245–3260.

Heath, R., McDonald, G., Christeller, J.T., Lee, M., Bateman, K. andWest, J. (1997) Proteinase inhibitors from Nicotiana alataenhance plant resistance to insect pests. J. Insect Physiol. 43,833–842.

Jongsma, M.A., Bakker, P.L., Peters, J., Bosch, D. and Stiekema,W.J. (1995a) Adaptation of Spodoptera exigua larvae toplant proteinase inhibitors by induction of gut proteinase activ-ity insensitive to inhibition. Proc. Natl. Acad. Sci. USA, 92,8041–8045.

Jongsma, M.A., Bakker, P.L., Stiekema, W.J. and Bosch, D. (1995b)Phage display of a double-headed proteinase inhibitor: analysisof the binding domains of potato proteinase inhibitor II. Mol.Breed. 1, 181–191.

Jongsma, M.A. and Bolter, C. (1997) The adaptation of insects toplant protease inhibitors. J. Insect Physiol. 43, 885–895.

Jouanin, L., Bonade-Bottino, M., Girard, C., Morrot, G. and Giband,M. (1998) Transgenic plants for insect resistance. Plant Sci. 131,1–11.

Koiwa, H., Paino D’Urzo, M., Assfalg-Machleidt, I., Zhu-Salzman,K., Shade, R.E., An, H., Murdock, L.L., Machleidt, W., Bressan,R.A. and Hasegawa, P.M. (2001) Phage display selection ofhairpin loop soyacystatin variants that mediate high affinityinhibition of a cysteine proteinase. Plant J. 27, 383–391.

Koiwa, H., Shade, R.E., Zhu-Salzman, K., Subramanian, L., Mur-dock, L.L., Nielsen, S.S., Bressan, R.A. and Hasegawa, P.M.(1998) Phage display selection can differentiate insecticidalactivity of soybean cystatins. Plant J. 14, 371–379.



Koscielska, K., Kiczak, L., Kasztura, M., Wesolowska, O. andOtlewski, J. (1998) Phage display of proteins. Acta Biochim.Pol. 45, 705–720.

Laskowski, M. and Kato, I. (1980) Protein inhibitors of proteinases.Ann. Rev. Biochem. 49, 593–626.

Lee, I.S., Lee, S.-H., Koo, J.C., Chun, H.J., Lim, C.O., Mun, J.H.,Song, Y.H. and Cho, M.J. (1999) Soybean Kunitz trypsin inhibitor(SKTI) confers resistance to the brown planthopper (Nirvatalugens Stal) in transgenic rice. J. Mol. Breed. 5, 1–9.

Markland, W., Ley, A.C., Lee, S.W. and Ladner, R.C. (1996) Iterativeoptimisation of high-affinity protease inhibitors using phagedisplay. Part 1. Plasmin. Biochem. 35, 8045–8057.

Mazumdar-Leighton, S. and Broadway, R.M. (2001) Identificationof six chymotrypsin cDNAs from larval midguts of Helicoverpazea and Agrotis ipsilon feeding on the soybean (Kunitz) trypsininhibitor. Insect Biochem. Mol. Biol. 31, 633–644.

McBride, J.D. and Leatherbarrow, R.J. (2001) Synthetic peptidemimics of the Bowman-Birk inhibitor protein. Curr. Med. Chem.8, 909–917.

Menegatti, E., Tedeschi, G., Ronchi, S., Bortolotti, F., Ascenzi, P.,Thomas, R.M., Bolognesi, M. and Palmieri, S. (1992) Purification,inhibitory properties and amino acid sequence of a new serineproteinase inhibitor from white mustard (Sinapis alba L.) seeds.FEBS Lett. 301, 10–14.

Paulillo, L.C.M.S., Lopes, A.R., Cristofoletti, P.T., Parra, J.R.P.,Terra, W.R. and Silva-Filho, M.C. (2000) Changes in midgutendopeptidase activity of Spodoptera frugiperda (Lepidoptera:Noctuidae) are responsible for adaptation to soybean proteinaseinhibitors. J. Econom. Entomol. 93, 892–896.

Pouvreau, L., Chobert, J.M., Briand, L., Quillien, L., Tran, V.,Gueguen, J. and Haertle, T. (1998) Effect of pea and bovinetrypsin inhibitors on wild-type and modified trypsins. FEBS Lett.423, 167–172.

Quillien, L., Rahbe, Y. and Gueguen, J. (1998) Anti-aphid effect oftrypsin inhibitors from pea seeds. In Proceedings of 3rd Eur-opean Conference on Grain Legumes. Opportunities for HighQuality, Healthy and Added-Value Crops to Meet EuropeanDemands. Valladolid, Spain, 14–19 November 1998.

Rahbe, Y., Deraison, C., Bonade-Bottino, M., Girard, C., Nardon, C.and Jouanin, L. (2002a) Effects of the cysteine protease inhibitororyzacystatin (OC-I) on different aphids and reduced perfor-

mance of Myzus persicae on OC-I expressing transgenic oilseedrape. Plant Sci. (in press).

Rahbe, Y. and Febvay, G. (1993) Protein toxicity to aphids – anin vitro test on Acyrthosiphon pisum. Entomol. Exp. Appl. 67,149–160.

Rahbe, Y., Ferrasson, E., Rabesona, H. and Quillien, L. (2002b)Growth inhibition of the pea aphid Acyrthosiphon pisum byisoforms and fragments of the Bowman-Birk protease inhibitorsfrom pea seeds. Insect Biochem. Mol. Biol. (in press).

Rahbe, Y., Sauvion, N., Febvay, G., Peumans, W.J. and Gatehouse,A.M.R. (1995) Toxicity of lectins and processing of ingestedproteins in the pea aphid Acyrthosiphon pisum. Entomol.Exp. Appl. 76, 143–155.

Ruoppolo, M., Amoresano, A., Pucci, P., Pascarella, S., Polticelli, F.,Trovato, M., Menegatti, E. and Ascenzi, P. (2000) Characterisa-tion of five new low-molecular mass trypsin inhibitors fromwhite mustard (Sinapis alba L.) seed. Eur. J. Biochem. 267,6486–6492.

Ryan, C.A. (1990) Protease inhibitors in plants: genes for improvingdefenses against insects and pathogens. Annu. Rev. Phyto-pathol. 28, 425–449.

Scott, J.K. and Smith, G.P. (1990) Searching for peptide ligandswith epitope library. Science, 249, 386–390.

Tran, P., Cheesbrough, T.M. and Keickhefer, R.W. (1997) Plantproteinase inhibitors are potential anticereal aphid compounds.J. Econ. Entom. 90, 1672–1677.

Volpicella, M., Ceci, L.R., Gallerani, R., Jongsma, M.A. and Beek-wilder, J. (2001) Functional expression on bacteriophage of themustard trypsin inhibitor MTI-2. Biochem. Biophys. Res. Comm.280, 813–817.

Volpicella, M., Cordewener, J., America, T., Ceci, L.R., Gallerani, R.,Jongsma, M.A. and Beekwilder, J. (2002) Properties of digestiveenzymes from Helicoverpa zea adapted to plant defense. Eur. J.Biochem. (in press).

Volpicella, M., Shipper, A., Jongsma, M.A., Spoto, N., Gallerani, R.and Ceci, L.R. (2000) Characterisation of recombinant mustardtrypsin inhibitor 2 (MTI-2) expressed in Pichia pastoris. FEBSLett. 468, 137–141.

Witt, K.A., Gillespie, T.J., Huber, J.D., Egleton, R.D. and Davis, T.P.(2001) Peptide drug modifications to enhance bioavailability andblood–brain barrier permeability. Peptides, 22, 2329–2343.



selection by phage display of a variant mustard trypsin inhibitor toxic against aphids

Documents