Harnessing the functional diversity of plant cystatins to design inhibitor variants highly active against herbivorous arthropod digestive proteases

Protein engineering approaches have been proposed to improve the inhibitory properties of plant cystatins against herbivorous arthropod digestive proteases. These approaches typically involve the site-directed mutagenesis of functionally relevant amino acids, the production and selection of improved inhibitory variants by molecular phage display procedures, or the design of bi/multifunctional translational fusions integrating one or several cystatin inhibitory domains. Here, we propose a new approach where the function-related structural elements of a cystatin are substituted by the corresponding elements of an alternative cystatin. Cys protease inhibitory assays were first performed with 20 representative plant cystatins and model Cys proteases, including herbivorous arthropod digestive proteases, to appreciate the extent of functional variability among plant cystatin protein family members. The most, and less, potent of these cystatins were then used as ‘donors’ of structural elements to create hybrids of tomato cystatin SlCYS8 used as a model ‘recipient’ inhibitor. Our data confirm the wide variety of cystatin protease inhibitory profiles among plant taxa. They also demonstrate the usefulness of these proteins as a pool of discrete structural elements for the design of cystatin variants with improved potency against herbivorous pest digestive Cys proteases.


INTRODUCTION
Protease inhibitors of the cystatin protein superfamily play various roles in plants, from the regulation of cysteine (Cys) proteases in seeds and senescent organs to the inhibition of exogenous digestive proteases upon arthropod herbivory or pathogenic infection (Arai et al. 2002;Benchabane et al. 2010). Cystatins act as reversible pseudosubstrate inhibitors to hinder the active site of target proteases and block their catalytic action on the peptide bonds of protein substrates (Turk and Bode 1991). The inhibitory function of these proteins relies on two structural elements, a central hairpin loop with the conserved pentapeptide motif Gln-X-Val-X-Gly (where X is any amino acid) and a second hairpin loop in the C-terminal region with a conserved Trp residue, which physically interact with amino acid residues in the active site cleft of the target enzyme. A third structural element is also involved, consisting of a flexible, N-terminal amino acid string presenting a conserved Gly-Gly dipeptide motif (Benchabane et al. 2010). This third element, referred to as the N-terminal trunk, interacts with surface residues on the target enzyme to strongly influence the inhibitory potency and specificity of the cystatin towards different protease isoforms (Rasoolizadeh et al. 2016a).
An increasing body of knowledge about the properties and physiological roles of plant cystatins has triggered the development of various biotechnological applications over the years (Tremblay et al. 2019). Studies have shown the potential of these proteins as ectopic regulators of endogenous Cys proteases to regulate storage protein deposition and degradation in reproductive organs (Hwang et al. 2009, Munger et al. 2015, to restore fertility in Cys protease-induced male sterile plants (Shukla et al. 2016) or to avoid the detrimental action of endogenous Cys proteases on heterologous proteins in plants used as bio-factories for clinically relevant proteins (Sainsbury et al. 2013, Grosse-Holz et al. 2018, Jutras et al. 2019. Other studies have shown their potential to implement drought, cold or salt tolerance in different crops (Chen et al. 2014, Quain et al. 2014, Tan et al., 2016, 2017a, associated with the induction of abiotic stress-related genes upon recombinant cystatin expression (Munger et al. 2012, Tan et al. 2017b). Most importantly, numerous studies have described the potential of plant cystatins to protect plants from microbial pathogens, root parasitic nematodes and phytophagous arthropods (reviewed in Schlüter et al. 2010, Lima et al. 2015, Macedo et al. 2015, Martinez et al. 2016. Cystatins inhibit digestive Cys proteases secreted in the extracellular milieu of microbial cells or digestive tract of herbivorous arthropods, to cause amino acid shortage, growth delays and eventual death of the pathogenic or herbivorous enemy (Broadway 2001, Lima et al. 2015.
From a physiological standpoint, the actual ability of a cystatin to protect the plant from herbivory is determined by its relative abundance compared to Cys proteases in the target herbivore midgut, by its inhibitory range towards these enzymes, and by any compensatory response induced in the herbivore after ingestion (Rasoolizadeh et al. 2016b). Herbivorous insects have developed effective strategies to avoid the negative effects of dietary protease inhibitors, including the secretion of digestive proteases from different functional classes, the overexpression of proteases following inhibitor uptake, and the production of protease isoforms weakly sensitive to inhibition (Zhu-Salzman and Zeng 2015). A well-documented example is the coleopteran insect pest Colorado potato beetle (Leptinotarsa decemlineata), which uses an array of positively selected digestive Cys protease isoforms to process leaf proteins (Vorster et al. 2015). Divergent, if not contradictory, effects have been reported for transgenic potato lines engineered to express cystatins, ranging from major developmental delays and mortality (Lecardonnel et al. 1999, Cingel et al. 2015, Rasoolizadeh et al. 2016b to compensatory growth and hypertrophic behavior sustained by Cys protease overexpression (Cloutier et al. 1999, Cingel al. 2015. Possible explanations for such discrepancy among studies include differential expression levels of the recombinant cystatin in leaf tissue, varying stability of this protein in different potato cultivars, distinct inhibitory ranges towards the insect Cys proteases and experimental biases influencing insect fitness.
Together, these observations stress the need for a better understanding of complex interactions between wound-inducible cystatins and Cys proteases in plant-insect systems.
They also underline the relevance of rational strategies for the molecular improvement of recombinant cystatin inhibitory profiles towards Cys proteases.
Three main approaches are generally adopted for the molecular improvement of plant cystatins (Sainsbury et al. 2012a, van Wyk et al. 2016, Tremblay et al. 2019). The first approach involves site-directed substitution at functionally relevant amino acid sites, the second approach the generation of improved inhibitory variants using phage display or DNA shuffling artificial evolution procedures, and the third approach the design of bi-or multifunctional translational fusions integrating one or more cystatin inhibitory domains. In this study, we explored the potential of a fourth approach based on the substitution of one or more function-related structural elements (SE's) of the cystatin by the corresponding element(s) of an alternative cystatin. Our goal was to assess the usefulness of potent cystatins from different plant taxa as SE "donors" to generate functional variability among the structural hybrids of a "recipient" cystatin. The idea was to translate the concept of 'loop replacement design' (LRD), as described for the engineering of multimeric mammalian antibodies (Clark et al., 2009), to the improvement of single-domain cystatins. Amino acid substitutions in the Nterminal trunk or the inhibitory loops of plant cystatins have proved useful to enhance the inhibitory potency or change the affinity profile of these proteins towards insect or nematode Cys proteases (Urwin et al. 1995, Kiggundu et al. 2006, Goulet et al. 2008, Rasoolizadeh et al. 2016a. We hypothesized that an LRD-like scheme by which the function-related elements of a protein are changed for the corresponding elements of a related protein would represent a welcome complement to site-directed mutagenesis as it would allow conformational changes in the cystatin on a length scale beyond that accessible to single mutations (Clark et al. 2009).
Cystatins are well suited to protein engineering and polypeptide grafting, as illustrated by their stability in fusion with different protein partners (Jutras et al. 2018, Tremblay et al. 2019, the structural stability of model cystatin tomato SlCYS8 bearing a poly-His tag for protein purification in a non-inhibitory loop of the protein scaffold (Sainsbury et al. 2016), the ability of SlCYS8 to stabilize a human protein translational fusion partner in planta (Sainsbury et al. 2013), the structural assessment of natural cystatins as a guide for de novo protein design (Marcos et al. 2017), and the use of a consensus plant cystatin scaffold to design Affimer binding proteins for a variety of imaging, diagnostic and therapeutic purposes (Tiede et al. 2014, Kyle 2018. Here, we confirm the usefulness of plant cystatins as a reservoir of discrete structural elements for cystatin engineering, and the potential of SE substitutions to create cystatins with improved inhibitory potency against arthropod Cys proteases.

Variable contributions of the N-terminal trunk and two inhibitory loops to the protease binding strength of plant cystatins
Docking simulations were performed in silico with three protease models to gain some preliminary insight about the functional variability of plant cystatins and the relative contributions of their function-related SE's to the enzyme-inhibitor complex. Five plant cystatins and the three model Cys proteases papain, human cathepsin L and Colorado potato beetle intestain D4 (IntD4) (Vorster et al. 2015) were selected for the simulations, for a total of 15 protease-cystatin complexes and 45 protease-SE interactions ( Table 1). Structure models were first built for the proteases and the cystatins by homology modelling with the solved structures of human cathepsin L (Ljunggren et al. 2007) and oryzacystatin I (OsCYS1) (Nagata et al. 2000), respectively. Protease-cystatin interactions were then simulated using the Z-Dock algorithm of Chen et al. (2003), by homology to the solved structure of papain in complex with human stefin B (Protein Data Bank Accession No. 1STF). In line with variable sequences in the functional regions of both the cystatins and the proteases, amino acid residues predicted to contribute to the binding process differed from one cystatin to another for a given protease, and from one protease to another for a given cystatin (Supplementary Figure 1). Accordingly, total binding energies differed for the 15 protease-cystatin complexes, from an inferred total energy value of -544 kcal/mol for maize cystatin ZmCYS1 interacting with papain to an energy value of -1234 kcal/mol indicating a stronger interaction between the same cystatin and cathepsin L ( In support to previously described models indicating variable contributions of the Nterminal trunk and two inhibitory loops to the protease binding process (Vorster et al. 2010), binding energies assigned to the three structural elements differed depending on the cystatin or the protease considered ( Table 1). For instance, a binding energy value of -334 kcal/mol accounting for 33% of the complex total binding energy was inferred for the N-terminal trunk of soybean cystatin GmCYS2 interacting with papain, compared to weaker energy values and relative contributions of less than 25% for the N-terminal trunks of tomato cystatins SlCYS8 and SlCYS9, OsCYS1 and ZmCYS1 interacting with the same enzyme (Figure 1). Likewise, an energy value of -402 kcal/mol accounting for 56% of the total was calculated for the second inhibitory loop of SlCYS9 interacting with papain, compared to weaker binding energies (and smaller relative contributions) of -275 kcal/mol (28%) and -272 kcal/mol (41%) for the same cystatin interacting with cathepsin L and IntD4, respectively ( Table 1).

Functional variability among plant cystatin protein family members
Protease inhibitory assays were conducted with cystatins of different plant taxa to empirically support our in silico assumptions suggesting functional variability among plant cystatins, to confirm the potential of these proteins as a source of SE's for cystatin improvement, and to identify potent cystatin donors for the SE substitution experiments. A multiple sequence alignment was generated with 262 cystatin primary sequences available in the National Center for Biotechnology Information (NCBI) database, using tomato SlCYS8 as a reference (Goulet et al. 2008). Double-stranded DNA fragments, or 'g-blocks', were then In agreement with our in silico models indicating variable binding energies for SlCYS8 and other plant cystatins interacting with papain (Table 1), inhibition constant (Ki) values against papain (Ki (papain)) differed by more than one order of magnitude from one cystatin to another, from 1.2 nM for Physcomitrella patens C-tailed cystatin domain PpCYS or 2.7 nM for tomato multicystatin domain SlCYS7, to 38.2 nM for soybean GmCYSB or even no measurable inhibitory activity for Arabidopsis AtCYS6B (Table 2) showing no activity against papain and L. decemlineata cathepsin L-like enzymes but easily measurable activity against T. urticae cathepsin L-like enzymes, or with Glycine soja GsCYS efficiently inhibiting papain and T. urticae proteases but showing weaker activity against the L. decemlineata enzymes.

A generic scheme for plant cystatin SE substitutions
An SE substitution strategy was designed to determine whether the variable inhibitory effects of plant cystatins against Cys proteases could be formally associated with the primary structures of their N-terminal trunk and inhibitory loops as suggested by our docking inferences, and whether plant cystatins as a group would represent a useful pool of discrete structural elements for the design of cystatin variants with improved potency against herbivorous pest digestive Cys proteases. Tomato SlCYS8 was used as a recipient protein model for hybrid design given its reported suitability for protein engineering (Goulet et al. 2008, Sainsbury et al. 2013, Jutras et al. 2018) and moderate activity against Cys proteases compared to other plant cystatins (Table 2 and Figure 2). P. patens PpCYS and potato StCYS5 were used as donors given their strong inhibitory potency against L. decemlineata proteases, and hence the expected potential of their function-related structural elements for SlCYS8 improvement. Cucumber CsCYS, weakly active against the arthropod cathepsin L-like enzymes (Figure 2), was selected as a 'flawed', negative control donor to further confirm the potential of inherently efficient inhibitors such as PpCYS and StCYS5 as loop donors to generate potent cystatins.
The SE hybrids were designed in silico by substituting the sequence(s) of SlCYS8 Nterminal trunk, first inhibitory loop (L1) and/or second inhibitory loop (L2) by the corresponding element(s) of PpCYS, StcYS5 or CsCYS (Figure 3). The N-and C-terminal boundaries of each structural element were defined based on their distance relative to conserved amino acid motifs essential for activity in the transferred element, in such a way as to also include all amino acids assumed to physically interact with amino acid residues of the target enzyme (Vorster et al. 2015) (Figure 3). More specifically, the N-terminal trunks were devised based on the Gly-Gly (-GG-) motif characteristic of the N-terminal region of plant cystatins, the first inhibitory loops based on the conserved pentapeptide motif Gln-X-Val-X-Gly (-QxVxG-) interacting with specific residues in the active site of the target protease, and the second inhibitory loops based on the conserved Trp residue also interacting with specific residues in the active site cleft (Benchabane et al. 2010). Twenty-one cystatin variants were designed overall, including all seven structural element combinations possible for each of the three cystatin donors ( Table 3 and Supplementary Table 2). DNA g-blocks were produced for the 21 hybrids and used as templates for bacterial expression and affinity purification using the GST gene fusion. As for the original cystatins above, some hybrids could not be properly expressed under our experimental conditions, likely due to deficient stability in a foreign cellular environment during heterologous expression. Overall, 16 hybrids were produced in a form suitable for protease inhibitory assays with papain and the arthropod proteases (Table 3), a large enough number of variants to draw conclusive trends about the potential of SE substitutions for cystatin engineering.

SE substitutions for the molecular improvement of tomato SlCYS8
Papain inhibitory assays were conducted to measure the impact of N-terminal trunk and inhibitory loop substitutions on the inhibitory activity of SlCYS8. In line with the variable efficiencies of SlCYS8 and donor cystatins against papain, Ki values for this enzyme differed from one hybrid to another (Figure 4). The most potent hybrids were hybrids P-1 and P-N1, both including the first loop of PpCYS, that showed Ki values for papain almost seven times smaller than the Ki value determined for SlCYS8. The less potent hybrid was hybrid C-N12, with the three functional elements of CsCYS, that showed no measurable activity against papain. Overall, most substitutions involving the structural elements of StCYS5 and PpCYS, both more potent than SlCYS8 and CsCYS against papain (see Table 2 to SlCYS8 (Hybrid S-N1) increased its inhibitory potency by more than 20 times against L. decemlineata cathepsin B-like enzymes ( Figure 5, lower panel). This improved inhibitory rate was more than three times the inhibitory rate observed for StCYS5 used at low concentration, suggesting the potential of SE substitutions not only to improve the inhibitory potency of a cystatin against its natural protease targets but also to broaden its inhibitory range to other Cys proteases.
Ki (papain) distribution maps were drawn to compare the overall impact of our SE substitutions strategy on SlCYS8 inhibitory activities with the impact of our site-directed mutagenesis approach involving single substitutions at functionally relevant, positively selected amino acid sites (Kiggundu et al. 2006) (Figure 6). An overall coefficient of variation (CV) of 53% was calculated for the relative Ki (papain) values of a previously described collection of 24 SlCYS8 single mutants bearing an alternative amino acid at positively selected sites Pro-2 (P2) or Thr-6 (T6) in the N-terminal trunk (Goulet et al. 2008). By comparison, CV values of 104% and 73% were calculated for the here tested 20 original (natural) cystatins and 16 SE hybrids, respectively. More specifically, Ki (papain) values for the single mutants were improved by 31% overall relative to wild-type SlCYS8, smaller than the average improvement rate of 154% observed for the SE hybrids (post-ANOVA Fisher's LSD test, P=0.02). Our underlying assumption for this work was that substituting the N-terminal trunk and/or inhibitory loops of a cystatin would represent a useful complement to current protein engineering approaches by allowing conformational changes on a length scale beyond that accessible to single mutations (Clark et al. 2009). Supporting this, SE substitutions as here implemented had a much greater impact on SlCYS8 inhibitory activities than our previously described approach involving single substitutions at positively selected amino acid sites ( Figure 6). Much interestingly, SE hybrids showed average Ki (papain) values improved by 150% overall relative to wild-type SlCYS8, five times higher than the overall improvement rate observed with the site-directed mutagenesis approach. Likewise, potent single mutants such as P2L, P2M and P2F produced earlier exhibited inhibitory activity rates increased by-i.e. IC50

DISCUSSION
values decreased by-two-to threefold against L. decemlineata cathepsin L-like enzymes compared to wild-type SlCYS8 (Goulet et al. 2008). By comparison, most SE variants bearing one or two structural elements of StCYS5 or PpCYS here showed five-to tenfold inhibitory rate increases against these enzymes in non-saturating conditions (Figure 5). Overall, these observations point to the potential of SE substitutions as an effective way to derive functionally diverse cystatin variants from a plant cystatin template, and hence the potential of this approach as a valuable complement to current protein engineering strategies for cystatin molecular improvement.
An unsolved question at this point is the actual relevance of our new approach in practice considering the functional variability already observed among plant cystatin family members and the high inhibitory efficiency here measured for some of them. For instance, the inhibitory effect of potent SE hybrids like P-N1 or P-1 against papain was much stronger than the anti-papain activity of wild-type SlCYS8 but comparable to the inhibitory activity of PpCYS used as a donor cystatin for the two variants (Figure 4, Table 2). Similarly, StCYS5 and PpCYS both showed inhibitory values 10 times higher than SlCYS8 against L. decemlineata cathepsin L enzymes, comparable to the inhibitory values observed for the two most potent SE hybrids, S-12 and P-12, bearing the inhibitory loops of these donor cystatins (Figure 2, Figure 5). On the other hand, hybrid S-N1, with the N-terminal trunk and first inhibitory loop of StCYS5, exhibited very strong inhibitory activity against L. decemlineata cathepsin B-like enzymes, more than 15 times higher than, and approximately three times higher than, the inhibitory activities of SlCYS8 and StCYS5, respectively ( Figure 5). These data, while leaving open the question of a comparative plus-value for the SE substitutions strategy to produce potent inhibitors of cathepsin L-like enzymes, suggest the potential of this approach to generate broad-spectrum cystatins that also inhibit alternative proteases naturally recalcitrant to cystatin inhibition (Zhu-Salzman and Zeng 2015), such as those upregulated in L. decemlineata to sustain leaf consumption and larval growth (Rasoolizadeh et al. 2016b).
Additional studies will now be welcome to compare the plant protective effects of potent SE variants like hybrids S-N1, S-12 and P-12 with the protective effects of potent natural cystatins like StCYS5 and PpCYS or those of SlCYS8 single variant P2V SlCYS8, an improved but still moderately efficient inhibitor (see Figure 6) shown to negatively alter leaf consumption and growth of L. decemlineata larvae (Rasoolizadeh et al. 2016b). Studies will also be welcome to assess the inhibitory potential of SE hybrids integrating N-terminal trunk and inhibitory loop(s) of different cystatin donors, given the specific contributions of these structural elements to the Cys protease-cystatin complex. Studies will be welcome, finally, to measure the impact of a fourth 'structural element', the central fold supporting the N-terminal trunk and two inhibitory loops, on the inhibitory activity of plant cystatins. The strong inhibitory potency of hybrid S-N1 against L. decemlineata cathepsin B-like enzymes compared to wildtype StCYS5 used as a donor, or the occurrence of positively selected amino acids presumably influencing protease inhibition in the a-helix and inter-loop region of plant cystatins (Kiggundu et al. 2006), point to a possible impact of this structural element on the inhibitory efficiency of the protein. The central fold of cystatins is not directly involved in protease inhibition but still might represent a valuable target for plant cystatin engineering given its possible effects on the spatial orientation and stability of the three functional elements.

Representative plant cystatins
Phylogenetic inferences were performed with the MEGA6 software, v.6.06 (Tamura et al. 2013), using plant cystatin sequences available in the NCBI protein database (http://ncbi.nlm.nih.gov/). Non-redundant cystatin sequences were retrieved from the Viridiplantae domain of the database using the in-built Protein Blast tool, with the sequence of tomato SlCYS8 as a query sequence (GenBank Accession No. AF198390.1; gi|6671196).
Amino acid sequences including at least 90% of a full cystatin, here corresponding to 262 nonredundant NCBI accessions, were used as a starting point. A multiple sequence alignment was generated using the MUSCLE algorithm (Edgar 2004), from which a maximum likelihood tree was calculated based on the "JTT" amino acid substitution model (Jones et al. 1992) (Supplementary Figure 2). Fifty-seven 'representative' cystatins were identified by restricting the selection (i) to one sequence among highly similar sequences (>95% identity), and (ii) to one cystatin per branch of the phylogenetic tree. A subset of 30 cystatins was taken from this sample for the functional analyses, in such a way as to maximize sequence variability among the selected cystatins at amino acid positions expected to physically interact with the target proteases (Vorster et al. 2015).

Recombinant cystatins
All cystatins were produced in E. coli, strain BL21 as described previously (Goulet et al. 2008), using the GST gene fusion system for heterologous expression and affinity purification (GE Healthcare). DNA templates for the cystatins were synthesized as g-blocks (IDT) including GoldenGate BSAI cloning sites on both sides of the cystatin coding region (Sainsbury et al. 2012b). DNA coding sequences for the original cystatins corresponded to those sequences reported in GenBank (as listed in Table 2). DNA sequences for the SE hybrids were designed as described in the Results (Figure 3), with structural elements from donor cystatins potato StCYS5, P. patens PpCYS or cucumber CsCYS replacing the N-terminal trunk (amino acids 1-13, SlCYS8-numbering), first inhibitory loop (amino acids 44-55) and/or second inhibitory loop (amino acid 70-84) of tomato SlCYS8 (Supplementary Table 2). G-blocks were inserted in a modified version of the pGEX-3X expression vector (GE Healthcare) using the Golden Gate DNA shuffling method of Engler et al. (2009), downstream of a 'GST-factor Xa cleavage site' coding sequence (Sainsbury et al. 2012b). The GST tag was removed by cleavage with bovine factor Xa, according to the manufacturer's specifications (Novagen). Cystatin products showing fragmentation, as assessed by 15% (w/v) SDS-PAGE, were not considered further for the functional analyses. The purified cystatins were quantified by densitometric analysis of Coomassie blue-stained polyacrylamide slab gels following 12% (w/v) SDS-PAGE, using three technical replicates and bovine serum albumin (Sigma-Aldrich) as a protein standard.

Test proteases
Papain (E.C.3.4.22.2, from papaya latex) was purchased from Sigma-Aldrich. Colorado potato beetle (L. decemlineata) proteases were extracted from the midgut of fourth instars reared on greenhouse-grown potato plants, cv. Norland, as described previously (Goulet et al. 2008). Two-spotted spider mite (T. urticae) proteases were obtained from a laboratory colony reared in greenhouse on common bean (Phaseolus vulgaris). Whole mites were ground in liquid nitrogen, the resulting powder kept on ice for 10 min after resuspension in 50 mM Tris-HCl buffer, pH 7.0, and the whole mixture centrifuged at 4°C for 10 min at 20 000 g. The pellet was discarded, and the supernatant used as a source of digestive proteases for the protease inhibitory assays. Papain and soluble proteins in the arthropod crude extracts were assayed according to Bradford (1976), with bovine serum albumin as a protein standard.

Ki (papain) value determinations
Ki (papain) values for the cystatin variants were determined by the monitoring of substrate hydrolysis progress curves (Salvesen and Nagase 1989), based on the linear equation of Henderson (1972). Papain activity was monitored in 50 mM Tris-HCl, pH 7.0, using the synthetic peptide substrate Z-Phe-Arg-methylcoumarin (MCA) (Sigma-Aldrich). Hydrolysis was allowed to proceed at 25°C in reduced conditions (10 mM L-cysteine) with the substrate in large excess, after adding (or not) recombinant cystatins dissolved in a minimal volume of reaction buffer. Papain activity was monitored using a Synergy H1 fluorimeter (BioTek), using

Arthropod protease assays
Cys cathepsin activities in the arthropod protein extracts were assayed in 0.2 M NaH2PO4/0.2 M Na2HPO4 phosphate buffer, pH 6.5, using the synthetic peptide substrates Z-Phe-Arg-MCA for cathepsin L-like activities and Z-Arg-Arg-MCA for cathepsin B-like activities.
Hydrolysis was allowed to proceed in reduced conditions (10 mM L-cysteine) for 10 min at 25°C, with ~5-6 ng of arthropod protein per µl in the reaction mixture and the peptide substrate added in large excess. Cystatins dissolved in a minimal volume of reaction buffer were added to the reaction mixture for the inhibitory assays. Proteolytic activity was     Figure 1 for the three model proteases.
• • • 1 S, P and C stand for StCYS5, PpCYS and CsCYS, respectively; N, 1 and 2 for the N-terminal trunk, first inhibitory loop and second inhibitory loop of the donor cystatin. 2 Unsuccessfully produced in E. coli and not considered further for the functional assays.