Functional proteomics-aided selection of protease inhibitors for herbivore insect control

Studies have reported the potential of protease inhibitors to engineer insect resistance in transgenic plants but the general usefulness of this approach in crop protection still remains to

be established. Insects have evolved strategies to cope with dietary protease inhibitors, such as the use of proteases recalcitrant to inhibition, that often make the selection of effective

inhibitors very challenging. Here, we used a functional proteomics approach for the ‘capture’ of Cys protease targets in crude protein extracts as a tool to identify promising cystatins for

plant improvement. Two cystatins found to differ in their efficiency to capture Cys proteases of the coleopteran pest Leptinotarsa decemlineata also differed in their usefulness to produce

transgenic potato lines resistant to this insect. Plants expressing the most potent cystatin at high level had a strong repressing effect on larval growth and leaf intake, while plants

expressing the weakest cystatin showed no effect on both two parameters compared to untransformed parental line used for genetic transformation. Our data underline the relevance of

considering the whole range of possible protease targets when selecting an inhibitor for plant pest control. They also confirm the feasibility of developing cystatin-expressing transgenics

resistant to a major pest of potato.

Three papers have described, almost 30 years ago, the potential of plant genetic transformation to implement insect resistance into crop genomes. Two of those papers, by Vaeck et al.1 and

Fischhoff et al.2, reported on the potential of Cry toxin-encoding genes from the soil bacterium Bacillus thuringiensis (Bt) to produce transgenic plant lines resistant to the tobacco

hornworm Manduca sexta. The third paper, by Hilder et al.3, discussed the potential of a trypsin inhibitor from cowpea to produce transgenic lines resistant to another lepidopteran pest, the

tobacco budworm Heliothis virescens. These three seminal papers were followed by hundreds of reports addressing various questions on insect-resistant transgenic crops, notably related to

the large-scale deployment and durable use of ‘Bt crops’ worldwide4,5,6 or to the basic reasons for the mitigated efficiency and still limited use of protease inhibitors in plant

protection7,8.

One explanation for the commercial success of Bt crops over protease inhibitor-expressing crops most likely lies in the different modes of action and pesticidal efficiency of the expressed

proteins in agricultural contexts9. A second explanation is the natural ability of herbivore pests to elude the effects of protease inhibitors shortly after consumption, as a result of a

long co-evolutionary history with their plant hosts that use these proteins as a defensive strategy against predation9,10,11. Whereas Cry toxins show strong toxicity towards relatively

specific pests and allow for an effective insecticide-like effect in the field, protease inhibitors interfere with dietary protein digestion and lead, in the most potent cases, to amino acid

shortage and a detrimental overexpression of digestive proteases causing growth delays and eventual death of the herbivore12. Most importantly, herbivorous insects have evolved a range of

strategies to cope with dietary protease inhibitors, typically involving the secretion of complex midgut protease complements, the overexpression of inhibitor-sensitive proteases to

outnumber the ingested inhibitors and the up-regulation of protease isoforms weakly sensitive to inhibition8. Nevertheless, and despite numerous unsuccessful attempts to use protease

inhibitors in pest control, a number of promising cases have been reported recently13,14,15,16,17,18,19,20 that still remind the importance of digestive proteases in herbivorous pests and

the possible relevance of these enzymes as effective targets for crop improvement21,22,23,24.

Protein engineering efforts have been made over the years to enhance the protective effects of plant protease inhibitors, notably involving fusion protein constructs to integrate

complementary inhibitor domains into single polypeptides or the rational design of inhibitor variants with improved activity towards animal or plant protease models25. A practical challenge

at present is to develop strategies for the selection of potent inhibitor candidates in such a way as to limit compensatory responses in the herbivore upon ingestion. Herbivorous arthropod

genomes encode large families of digestive protease genes26,27,28 that allow the herbivores to produce protease isoforms with a wide range of affinity spectra towards dietary protein

substrates and protease inhibitors11,29,30,31,32,33. Considering this, the most effective way to select useful inhibitors among a collection of available candidates may not be to test their

inhibitory potency against one or a few model proteases, but to compare their effective binding range against the whole complement of possible protease targets in the pest midgut. In this

study, we used a functional proteomics approach for the ‘capture’ and tandem mass spectrometry (MS/MS) analysis of protease inhibitor-susceptible proteases in crude biological extracts33 as

a decision tool for the rational selection of a protease inhibitor useful to engineer resistance to the major coleopteran pest Colorado potato beetle (Leptinotarsa decemlineata) in potato,

Solanum tuberosum. Unsuccessful attempts to implement resistance to this pest in potato varieties using recombinant protease inhibitors have been associated with the onset of multiple

adaptive responses including increased leaf consumption to counterbalance the loss of digestive functions and overexpression of inhibitor-insensitive proteases to sustain basic protein

digestion rates34,35,36. Here we show such compensatory responses to be compromised in larvae fed potato plants engineered to express a Cys protease inhibitor –or cystatin37– variant

efficient in capturing midgut Cys proteases of this insect under our proteomics setup.

Tomato cystatin SlCYS838 and single functional variants of this protein39 bearing an isoleucine (P2I), a leucine (P2L) or a valine (P2V) in place of the original proline at position 2, or an

arginine (T6R) in place of the original threonine at position 6, were considered as possible candidates for potato transformation (Fig. 1a). L. decemlineata digestive Cys proteases in

theory sensitive to cystatin inhibition, the so-called ‘intestains’40, are categorized into six functional families referred to, respectively, as intestains A (IntA), IntB, IntC, IntD, IntE

and IntF33,40. Our proteomic approach consists of capturing cystatin-sensitive intestains in midgut extracts after binding biotinylated versions of the cystatins produced in Escherichia coli

to an avidin-linked matrix for affinity enrichment33. After recovery, the captured intestains are detected as three bands of 25, 27 and 30 kDa on Coomassie blue-stained gels following

SDS-PAGE, which include isoforms of the different functional families representing the complement of proteases bound by the cystatin variant. The three bands are excised, digested with

trypsin, and the resulting peptides submitted to MS/MS for intestain identification and quantitation. The abundance of inhibitor variant-sensitive intestains in source extracts may be

inferred by the counting of MS/MS unique peptide spectra, assuming a positive correlation between the number of captured peptides and the inhibitory range of the cystatin variant against

specific intestain families or the whole range of intestains33,41.

Spectral counts for intestain unique peptides captured with biotinylated versions of wild-type SlCYS8 and single functional variants P2I, P2L, P2V and T6R in midgut extracts of potato-fed L.

decemlineata larvae.

(a) Tertiary structure model for SlCYS8 (GenBank accession no. AF198390) showing the approximate positions of residues Pro-2 (P2) and Thr-6 (T6) targeted for mutagenesis in the N-terminal

region, relative to the central (Loop 1) and C-terminal (Loop 2) inhibitory loops of the protein. The in silico model was generated with Modeller, v. 9.758 using the NMR solution structure

coordinates of oryzacystatin59 as a template (Protein Data Bank accession no. 1EQK). (b) Intestain unique peptides counted for the five SlCYS8 variants, as inferred from MS/MS datasets of

refs 11 and 42. Data are expressed relative to total spectra counted for wild-type SlCYS8 (mean value adjusted to 1; see Supplementary Table S1 for details on unique peptide counts for each

cystatin variant. Each bar is the mean of three independent (insect replicate) values ± se.

We used this approach recently to address basic questions on the evolution and structure/function determinants of intestain–cystatin interactions in L. decemlineata11,42. An MS/MS peptides

dataset generated during these studies was here reassessed to compare the ability of our SlCYS8 variant candidates to bind the insect intestains, assimilating the complement of captured

proteases to the binding range of each variant. Thirty unique intestain peptides were detected overall following MS/MS, that could each be assigned to a single functional family

(Supplementary Table S1). IntB-, IntC- and IntD-specific peptides were detected for all cystatin variants while no IntA-, IntE- or IntF-specific peptides were detected, presumably due to low

numbers of intestains from these families in midgut extracts or to the reported insensitivity of some intestains to cystatin inhibition40. As inferred from total spectral counts, mean

numbers of captured peptides differed from one variant to another, giving for instance mean numbers with P2V– and P2L–biotin six to seven times the peptide numbers obtained with the original

SlCYS8 inhibitor (Fig. 1b). By comparison, T6R–biotin captured a smaller number of peptides, similar to the number captured with SlCYS8. These observations pointed overall to the usefulness

of our functional proteomics approach to discriminate cystatin variants based on their effective binding range towards the whole complement of possible target Cys proteases in midgut

extracts. They suggested, in practice, the potential of this approach to identify potent inhibitors, such as P2V in the present case, to implement insect resistance in planta.

P2V and T6R were used as candidate inhibitors for potato transformation (Fig. 2) to confirm a possible relationship between protease capture efficiency under our proteomics setup, insect

resistance or susceptibility of transgenic plant lines expressing these inhibitors, and the relative ability of the target herbivore to mount an effective compensatory response upon leaf

consumption. P2V was selected as an ‘effective inhibitor’ candidate based on the broad binding range of P2V–biotin towards intestains (see Fig. 1b) and previously reported data suggesting a

strong inhibitory potency for this variant against L. decemlineata Cys proteases39. T6R was selected as a ‘weak inhibitor’ (negative control) candidate based on a narrow intestain binding

range of T6R–biotin similar to the binding range of SlCYS8–biotin (Fig. 1b), despite previously reported inhibitory data with synthetic peptide substrates suggesting a greater potency of the

single variant in vitro39. Potato lines engineered to express either inhibitors along with the antibiotic selection marker neomycin phosphotransferase II were regenerated on

kanamycin-containing growth medium following genetic transformation with the appropriate gene constructs (Fig. 2a). In vitro clones produced from independent calli were acclimated in

greenhouse and PCR-tested for the selection marker transgene in genomic DNA using appropriate DNA primers. A ~500-base-long ‘nptii’ amplicon was amplified from the DNA of all tested plants,

confirming that all clones regenerated on kanamycin had been genetically transformed by the Agrobacterium transgene vector.

Recombinant cystatin content and relative amount of pathogen-inducible ß-glucanases (PR-2 proteins) in transgenic lines of potato expressing SlCYS8 variants T6R or P2V.

(a) Gene construct elements for the cytosol-targeted expression of T6R and P2V in transgenic potato lines. Constructs included the basic coding sequence of either inhibitor39, a duplicated

version of the Cauliflower mosaic virus (CaMV) 35 S promoter (2XCaMV35S) for constitutive expression in leaves, a tobacco etch virus enhancer sequence (ES) in upstream position of the

cystatin coding sequence and a CaMV 35 S terminator sequence (TS) in downstream position. (b) Relative amounts of SlCYS8 variant in control (untransformed) and cystatin-expressing potato

lines. Data are expressed relative to low-expressing line P2V3 (arbitrary value of 1.0). (c) Relative amounts of ß-glucanases in control and SlCYS8-expressing lines. Data are expressed

relative to control line K (arbitrary value of 1.0). Numbers on the immunoblots indicate the position of commercial molecular mass markers (kDa). Each bar on panels b and c is the mean of

three independent (plant replicate) values ± se. Bars with the same letter on panel b are not significantly different (post-anova Tukey’s HSD; α = 0.05).

Immunodetections were performed with polyclonal IgG raised in rabbit against SlCYS8 to compare the relative amounts of recombinant P2V or T6R in transgenic leaves. As expected given the

random insertion of transgenes in the genome of Agrobacterium-infected cells, immunoblot signals differed from one clone to another, from weak to moderate in leaves of low-expressing clones

(such as P2V3) to strong or very strong in leaves of clones (such as T6R1, T6R3 or P2V10) expressing the cystatin at mean levels approximately seven times the mean level in line P2V3 (anova;

F(3,8) = 35.0, P = 0.001) (Fig. 2b). In agreement with studies reporting stress-related pleiotropic effects for recombinant cystatins in different plant species7 and the constitutive

expression of naturally inducible pathogenesis-related (PR) proteins in potato lines expressing corn cystatin II43, ß-glucanases of the PR-2 family were up-regulated in leaves of

cystatin-expressing clones compared to their steady-state levels in healthy leaves of parental line K (Fig. 2c). These observations confirmed the successful expression of T6R and P2V under

an active form inducing pleiotropy in leaf tissue. They also pointed, given the accumulation of ß-glucanases at comparable rates among transgenic clones (see Fig. 2c), to the possible

usefulness of T6R- and P2V-expressing lines as models to differenciate the expected protease inhibitory-mediated effects of P2V on L. decemlineata from eventual indirect effects via

pleiotropic alterations of the endogenous defense system altering leaf tissue composition.

A feeding assay was conducted to compare the impacts of ectopically expressed P2V and T6R on growth and leaf consumption rates of L. decemlineata larvae (Fig. 3). Potato plants expressing

either inhibitors were provided to 4th instars over 72 h, a period of time sufficient for the larvae to adjust their protease complement to dietary protease inhibitors36. Short-term effects

on foliage intake after 24 h were observed among larvae fed the different lines (anova; F(4,25) = 78.0, P