Evolutionary mechanisms acting on proteinase inhibitor variability
J. T. Christeller, HortResearch, Private Bag 11030, Palmerston North, New Zealand
Tel: +64 6356 8080, ext 7760
The interaction of proteinase inhibitors produced, in most cases, by host organisms and the invasive proteinases of pathogens or parasites or the dietary proteinases of predators, results in an evolutionary ‘arms race’ of rapid and ongoing change in both interacting proteins. The importance of these interactions in pathogenicity and predation is indicated by the high level and diversity of observable evolutionary activity that has been found. At the initial level of evolutionary change, recruitment of other functional protein-folding families has occurred, with the more recent evolution of one class of proteinase inhibitor from another, using the same mechanism and proteinase contact residues. The combination of different inhibitor domains into a single molecule is also observed. The basis from which variation is possible is shown by the high rate of retention of gene duplication events and by the associated process of inhibitory domain multiplication. At this level of reorganization, mutually exclusive splicing is also observed. Finally, the major mechanism by which variation is achieved rapidly is hypervariation of contact residues, an almost ubiquitous feature of proteinase inhibitors. The diversity of evolutionary mechanisms in a single class of proteins is unlikely to be common, because few systems are under similar pressure to create variation. Proteinase inhibitors are therefore a potential model system in which to study basic evolutionary process such as functional diversification.
Proteinase inhibitors are a diverse group of proteins that share not only a common biochemical activity, but also the distinguishing feature of rapidly undergoing evolutionary variation. Currently, 59 distinct families of proteinase inhibitors have been recognized . I use the term ‘family’ in this review to denote these phylogenetic groupings and the term ‘class’ to denote inhibitors that interact with proteinases with mechanistic similarities, i.e. the serine-, cysteine-, aspartic and metallo-proteinases. Phylogenetic relationships among several of these inhibitor families have been analysed: including the serpin family [2–8], Bowman–Birk [9,10], cereal trypsin/α-amylase inhibitor , proteinase inhibitor I , proteinase inhibitor II  and cystatin [14,15]. Compared with the total number of families that are currently recognized, this represents a very small proportion, although phylogenetic trees for all families have been constructed (http://merops.sanger.ac.uk). These relationships are useful in developing an understanding of when and where the inhibitor class evolved; however, they do not provide information on the mechanisms driving gene evolution.
The focus of many reviews of proteinase inhibitors over the last 25 years has been on classification and structure–function relationships. These proteins have not been well recognized as a class of proteins with an interesting evolutionary history. The purpose of this review is to summarize, for the first time, information relevant to proteinase inhibitor evolution, much of it collected incidentally, with the express intention of stimulating possible interest in this area.
Proteinase inhibitors and their binding to proteinases have been extremely well characterized for more than 70 years and I focus only on those inhibitors that have the attributes of being involved in protein–protein interactions representing antagonistic interorganism interactions. In this review, I draw attention to the evidence that proteinase inhibitor evolution appears to occur by multiple and interacting mechanisms not currently identified for other coevolving molecules and that this feature may be indicative of both a high rate of evolutionary change and the role of protein–protein interactions. Both attributes appear necessary, the mechanisms are less apparent in proteinase inhibitors targeted at intraorganism target proteinases than for interorganism interactions, from which the majority of the examples are drawn.
Proteinase inhibitors, pathogens and pests
Evolutionary pressures of various kinds have often been hypothesized to cause active and rapid evolutionary change. Various lines of evidence suggest that a major function of proteinase inhibitors is to combat the proteinases of pests and pathogens [16–20]. The secreted proteinases of the latter organisms are key components of invasive cocktails, required for entry into the host and rapid utilization of its constituent proteins. In these situations, there is clearly evolutionary pressure for the host to respond by evolving new and effective inhibitors. This model is often termed the ‘evolutionary arms race’.
Consistent with the role of proteinase inhibitors in resistance to invasion is the observation that massive accumulation of proteinase inhibitors occurs in certain tissues and organs that are likely sites of attack. First are those tissues whose high nutritional value presents to a pest or pathogen the best possible site for attack, for example, seeds , other plant storage organs such as plant tubers [23,24], and the eggs of birds . The reproductive strategies of these organisms require that the best possible nutrition be provided in these tissues and they are therefore clearly an attractive food sources for others. Proteinase inhibitors from these sources have long been extensively studied and often contain inhibitors of multiple families and classes. There appear to be few, if any, studies on the proteinase inhibitors in the eggs of other egg-laying organisms such as fish and insects, or on the proteinase inhibitors of organisms that retain their eggs internally. It is possible that the evolution of internal egg-bearing is related to the reduced pest, pathogen and parasite attack that comes with this strategy.
The second attractive site of attack by invasive organisms is fluids that permit transport of the pathogen throughout the host, for example, mammalian serum, invertebrate haemolymph and plant phloem. Indeed, these three fluids are, once again, a rich source of many and varied proteinase inhibitors and have been studied extensively. However, caution needs to be applied; the primary role of these inhibitors in blood appears to be regulation of the blood-clotting cascade. In insects, proteinase inhibitors in the haemolymph clearly play a part in the immune response, particularly in regulating the activation of prophenoloxidase in response to invasion by pathogens . The function of the diverse phloem proteinase inhibitors is unclear [27–30].
The third situation that can be identified in which extensive and varied proteinase inhibitor accumulation arises is the reverse situation, where proteinase inhibitors themselves are the pathogenic determinants. For example, the salivas of leeches and blood-sucking insects contain multiple inhibitors  that inactivate the proteinases of the blood-clotting cascade, thereby preventing blood-clotting and permitting the invader to feed freely.
Structural gene evolution
In this review, the evolutionary mechanisms used by organisms to enhance variation in the structural genes of proteinase inhibitors are discussed. This story is but one third of the equation; the other two parts, evolution of cognate proteinases and evolution of the proteinase inhibitor regulatory sequences are not discussed except en passant, partly because of the need to limit the breadth of the review and partly because the identification of the cognate proteinase is, in many cases, yet to be verified and because mechanisms of promoter evolution are a distinct and new topic entirely. The identification of cognate proteinases is often extremely challenging and, where achieved, has led to specific studies on coevolution [19,32,33]. The evolution of promoters has clearly occurred with equal rapidity to that of the structural genes. For example, orthologues of the proteinase inhibitor I family are known that have constitutive, tuber-specific , wounding signal-induced, leaf-specific , phloem-specific , fruit-specific , developmentally regulated , ethylene-induced  and cell-cycle-specific  promoters.
These processes may be most active in plants because many pests and pathogens use S1 superfamily serine proteinases as pathogenic determinants and these are not common in plants (The Arabidopsis Information Resource, TAIR, database at http://www.arabidopsis.org, lists 55 serine proteinases from 546 endopeptidases); thus the problem of isolating serine proteinase inhibitors from plant metabolic and regulatory processes is less critical, a consideration that might slow inhibitor evolution in other organisms. In mammals, serpins separate into secreted inhibitors involved in pathogen defence processes and intracellular inhibitors involved in cellular regulation; the former show high rates of evolution, whereas the latter do not .
In the following sections, each distinct evolutionary mechanism is discussed separately, yet it is clear from the examples given that they do not act in isolation. The first three mechanisms relate to the fundamental evolution of this group of molecules. The fourth to the sixth mechanisms appear to establish a higher level of diversity that forms the basis from which the seventh and final mechanism described is fundamental for generation of the large variation observed.
Recruitment of other protein-folding scaffolds to proteinase inhibitor function
It is apparent that recruitment of numerous protein-folding scaffolds to proteinase inhibitor function has occurred. This is clearly seen in the various folds of several proteinase inhibitors.
First, inhibitory serpins share 30% amino acid sequence homology with ovalbumin, the major storage protein of egg white, and share, with several other noninhibitory proteins, the same basic structure . Second, five members of the large cereal α-amylase inhibitor family have developed proteinase inhibitor function, and three of these have lost the former activity during evolution . Interestingly, these inhibitors exhibit the same novel backbone structure as is also observed in 2S seed storage proteins and in nonspecific lipid transfer proteins [11,42], suggesting even earlier recruitment of α-amylase inhibitor function. Third, equistatin [43,44], fish egg inhibitor , saxiphilin , testican  and p41 major histocompatibility complex fragment  are cysteine proteinase inhibitors based on the thyroglobulin fold. Fourth, soybean Kunitz inhibitors have homology to noninhibitory proteins sporamin A , stress-induced proteins , dehydroascorbate reductase  and miraculin . Finally, it has been suggested that both Bowman–Birk and cystatin inhibitors have evolved from an ancient ribonuclease-like gene .
Other major plant storage proteins, as well as the cereal α-amylase inhibitor and the soybean Kunitz inhibitor may have been recruited to additional function. For example, the Bowman–Birk inhibitors in seeds and proteinase inhibitors I and II in potato tubers are present at such high levels that they function as storage proteins . Possibly their ancestral types, lacking proteinase inhibitor function, have been lost due to the evolutionary advantage of producing dual-function proteins. Thus, although we have only identified five examples from 59 known inhibitor classes  it is possible that further examples will be identified in the future as additional proteinase inhibitor structures become available.
Knowledge of the inhibitory mechanism may assist understanding of how the recruitment processes evolved. Two of the above example inhibitors are serine proteinase inhibitors that use the ‘standard canonical’ or ‘Laskowski mechanism’. This involves the presence on the surface of the inhibitor of a stabilized loop that can mimic a substrate but which has long residency times in the proteinase active site pocket as a result of that conformational stability. The loop also has a protruding amino acid side chain that mimics the proteinase target specificity [54–56]. The combination of these features produces a rapid-binding, slow-release specific inhibitor rather than a substrate. The distinction may, in fact, not be clear cut, as substrate proteins may, under specific conditions, have inhibitory properties relative to small reporter substrates used to assay proteinases, for example, napins, legumins  (W. A. Laing, HortResearch, personal communication).
It seems entirely possible that any surface loop or even stretch of exposed amino acids could evolve inhibitory function by reducing flexibility through evolving intraprotein interactions, creating a loop by insertion of a very few amino acids and mutating a single amino acid to form a P1bait site residue. Evidence for this can be found by inspection of crystal structures of inhibitors and their putative ancestral proteins. This is clear for the cereal trypsin/amylase family in which the structure of the bifunctional ragi inhibitor complex with Tenebrio molitorα-amylase  showed that the proteinase binding loop adopts a canonical conformation at the opposite side of the molecule. This loop is absent from the α-amylase inhibitor . It also appears correct for serpins, because the intact active site loop adopts a distorted helical conformation in the P10–P3′ region, overlapping a type I β-turn in P2′–P5′ compared with an undistorted α-helix for ovalbumin . The modification in structure between inhibitory and noninhibitory forms is probably due to the presence of proline residues and a four amino acid insertion at this point in the inhibitor sequence .
This hypothesis requires that the rapid evolution of proteinase inhibitors began somewhat late in evolutionary history, when many of the major protein folds had already evolved, and that evolution has been from established essential functional proteins to new supplementary functions. If the stimulus for this to occur is the development of pests and pathogens it supposes that these interactions also were not a feature of early life and, so far as the discussion revolves around multicellular organisms as hosts, this is a reasonable assumption.
It should be noted, however, that the ‘Laskowski mechanism’ is not a feature of most families of inhibitors that inhibit proteinase classes, i.e. inhibitors of metalloproteinases, cysteine proteinases and aspartic proteinases even though these proteinases do possess characteristic residues at their active sites (cation, sulfydryl and aspartyl, respectively). Although many of these inhibitors are competitive, binding at the active site to prevent access to substrates, they do not mimic a substrate, are not cleaved reversibly and often utilize more than a single exposed loop in their direct active site contacts. Of these proteins, only one class, the thyropins, has been identified as recruited structures.
Change of inhibitory class
Recruitment of one family of proteinase inhibitors to inhibit a second class of proteinases is an evolutionary mechanism that can be readily identified and for which several examples are known. It is a special case of recruitment of another protein fold where the proteinase inhibitory structure is also recruited, or is likely to have been recruited, and is likely to have occurred more recently. This is seen in (a) serine proteinase inhibitors of the serpin family recruited to cysteine proteinase inhibition , (b) serine proteinase inhibitor of the serpin family recruited to aspartic proteinase inhibition [3,62], (c) serine proteinase inhibitor of the seed Kunitz family recruited to cysteine proteinase inhibition , (d) serine proteinase inhibitor of the seed Kunitz family recruited to aspartic proteinase inhibition , (e) serine proteinase inhibitor of the Bowman–Birk class recruited to cysteine proteinase inhibition , (f) cysteine proteinase inhibitors of the cystatin family recruited to aspartic proteinase inhibition (W. A. Laing, unpublished observations), and (g) cysteine proteinase inhibitors of the thyropin class, recruited to aspartic proteinase inhibition . This recruitment mechanism is not uncommon, having been identified in six of 59 known inhibitor families to date.
It is necessary to discuss each case individually, rather than assume that this represents a special and recent example of recruitment of another functional protein as discussed above. The serine proteinase inhibitors of the serpin family do not operate via the ‘Laskowski mechanism’. However, they have evolved a very effective ‘suicide’ irreversible mechanism in which the inhibitor (following Michelis complex formation) partitions between a tetrahedral stable intermediate and a cleaved, inactive inhibitor [67,68] leading to covalent bond formation via acyl bond formation and a large conformation. This is often observed as an SDS-stable product that migrates more slowly during PAGE than the unreacted proteinase and inhibitor. It seems that this mechanism has been maintained during evolution to the cysteine inhibitory form because the structure with a reactive site loop remains [69,70], the corresponding thioacyl complex has been detected  and the cleavage site remains within the single reactive loop of several dual class inhibitors [72,73].
Because the serine proteinase inhibitor molecule is apparently found in a diverse range of organisms: eukarya, bacteria, archaea and viruses  it is clearly the ancestral form and appears to have evolved directly to the new class of cysteine proteinase inhibitors found, to date, in mammals and viruses . The situation with the aspartic proteinase inhibitors has not yet been resolved with no studies investigating the presence of a covalent bond being reported. Interestingly, both these examples of recruitment of function appear to be an old divergence, occurring at a similar time to mammalian divergence  although these altered inhibitors have, to date, been reported only from mammals. It is also noteworthy that whereas most serpins are secreted, one clade is intracellular and members function as regulators of ‘promiscuous’ proteinases , rather than being involved regulation of endogenous cascades or in protection from pathogens.
Despite extensive studies, we still do not know whether the aspartic proteinase inhibitors that have clearly been recruited from standard mechanism Kunitz seed serine inhibitors, have recruited the latter molecules' serine proteinase inhibitor loop to inhibit aspartic proteinases. At least one member of this small family restricted to Solanaceae  has the ability to inhibit both classes . Clearly, a structure of an inhibitor–aspartic protease complex would be of great interest because successful attempts to determine a cleavage site for this system have not been reported. Owing to the very different pH optima for complex formation of the two classes it is difficult to even show whether simultaneous binding of two proteinases is possible. Although the active site loop of serine proteinase inhibitors of the seed Kunitz appears to have been disrupted in the cysteine inhibitors of this class , there is no information on the mode of interaction of these inhibitors with cysteine proteinases although some members of these inhibitors retain weak antitrypsin and antichymotrypsin activity .
The cysteine proteinase inhibitor, bromelain inhibitor VI from pineapple, is a double-chain inhibitor that shares similar folding and disulfide bond connectivities with the Bowman–Birk trypsin/chymotrypsin inhibitor . The authors suggested that these inhibitors have evolved from a common ancestor and differentiated in function during a course of molecular evolution. However, the B-domain of bromelain inhibitor VI has weak antitryptic activity, suggesting that class conversion is a reasonable alternative.
The recruitment of a cystatin to aspartic proteinase inhibitor [28,76] is based on sequence homology of around 30% and similarity of around 50% between rice cystatin and squash aspartic proteinase inhibitor (SQAPI) and modelling the later onto the crystal structure of the former (W. A. Laing, unpublished observations). Inhibition appears to involve the same areas of interaction; mutation and removal of residues at two regions known to be involved in cystatin interactions  abolishing aspartic protein inhibition (P. Farley, unpublished data) and hypervariability within the small inhibitor family also occurs at these sites and at the third site known to be involved in cystatin interactions (J. T. Christeller, unpublished observations). The tentative conclusion is that this small family of aspartic proteinase inhibitors, restricted to members of the Cucurbitales (J. T. Christeller, unpublished data) has evolved directly from the much more widespread cystatin inhibitors.
Equistatin is a protein consisting of three thyroglobulin domains , the N-terminal domain inhibiting cysteine proteinases and the central and C-terminal domains inhibiting aspartic proteinases. Equistatin is therefore a member of the thyropin class of inhibitors, of which all other known members are cysteine proteinase inhibitors. The published structure of the thyropin p41 fragment shows a wedge shape and three-loop inhibitory structure similar to cystatins, thus suggestive of convergent evolution . There is no information on the mode of inhibition of the aspartic proteinase inhibitor variant domains of equistatin.
Proteinaceous aspartic proteinase inhibitors are very rare in nature and four of the six known families are recruited from other classes, with only two, the yeast inhibitor IA3  and the Ascaris PI-3 inhibitor  being uniquely aspartic inhibitor families. Both inhibitors are small and have quite idiosyncratic inhibitory mechanisms. Thus inhibitor class change appears to be the mechanism of choice for this class of proteinases. These observations, combined with their rarity, may indicate that the aspartic class of proteinases have evolved relatively recently, at least to fulfil the function of defence proteins. Given the relative rarity of metalloproteinase inhibitors, similar questions may be asked about the time of their evolution and their recruitment to a defence protein role.
Recently, two examples of multidomain proteins have been identified in which different domains with distinct inhibitor classes are fused into a single inhibitor. First, testican is a multidomain protein with three domains having homology to different proteinase inhibitors, an N-terminal domain with metalloproteinase inhibitory activity , a follistatin-like domain with similarity to Kazal serine proteinase inhibitors although no serine proteinase inhibition has yet been reported and a thyropin domain that inhibits cysteine proteinases . Second, the WFIKKN protein  has, based on homology, a whey acidic protein metalloproteinase inhibitor module, a follistatin/Kazal inhibitor module and two Kunitz-type modules. One of the latter domains has been shown to inhibit trypsin . Although this evolutionary mechanism seems to be uncommon, the existence of this type of domain-shuffling event may indicate the high evolutionary pressure that proteinase inhibitors experience, permitting coordinate expression of proteinase inhibitors against a cocktail of secreted proteinases.
Gene duplication is a very common feature of proteinase inhibitors. Many, if not most, inhibitors are present as small gene families with altered specificities among the paralogues. Two features of pathogenicity and predation may drive this process; first, the presence of multiple attacking organisms with variation in their proteinases and, second, the presence of multiple proteinases in single organisms, for example, secreted aspartic proteinases of Candida number more than 10, whereas the digestive proteinase genes of lepidopteran larvae number in the hundreds. Thus gene duplication, coupled with rapid adaptation processes (detailed below) form a mechanism to resist invaders. Because there is no evidence that these genes duplicate at higher rates than any others, their fixation in the genome is more likely to be due to rapid adaptation and the selective advantage obtained. Fryxell  considers that fixation of gene duplication is maintained by coevolution of functionally related gene families. His hypothesis seems appropriate for proteinases and their inhibitors. Habu et al.  studied the evolution of duplicated Kunitz inhibitor genes in winged bean and a complex series of gene inactivation and gene conversion events was inferred.
Domain replication and circular permutation
Domain replication serves a similar function to gene duplication, both providing a base from which variability can be established. Instead of complete gene duplication, including promoter and terminator sequences and possible reintegration at a distinct locus, there is duplication of the inhibitory domain sequence with the domains remaining fused. This evolutionary mechanism is very common in many inhibitor gene families and has been well reviewed for some time [1,55,86]. Again, it serves the function of not only coordinated expression, but also increased levels of expression. The most extreme examples are genes in the proteinase inhibitor II and cystatin families. Proteinase inhibitor II family members have been characterized with varying numbers of domains, from one [13,87], two , three , four , six  through to eight . Many of these proteinase inhibitor II polyproteins are processed proteolytically, displaying the highly unusual phenomenon of cleavage within the domain , with the final molecule having N- and C-terminal sections being circularly permutated. The final conformation adopted by these domain hybrid molecules is identical to that adopted by single domain versions , likely to represent the putative ancestral sequence order . A mechanism involving unequal gene cross-over has been proposed to account for this variation by Barta et al.  who also noted that this may be a scenario to enhance functional diversity against pathogens. These multiplication and circular-permutation events have been followed by rapid divergence within single genes to target diverse proteinases Members of the proteinase inhibitor II family have been reported to inhibit chymotrypsin, elastase, oryzin, pronase E, subtilisin and trypsin [95–97]. Circular permutation has been observed in other proteins, including proteinases [98,99].
The second extreme example occurs in potato tubers where a single 85 kDa polypeptide, potato multicystatin comprises eight tandem cystatin domains, with 53–89% identity of residues, linked by proteolytically sensitive junctions [100,101]. Potato multicystatin comprises a family of four to six genes in potato and the pattern of gene expression, as well as the properties of the protein suggest that potato multicystatin has a role in plant defence . Although single domain cystatins are most common, a three-domain multicystatin has been isolated from sunflower seeds  and kininogens are also three-domain cystatins .
Mutually exclusive splicing
A single serpin gene of Manduca sexta expressed in the haemolymph comprises 10 exons, with the ninth, containing the reactive site loop, existing as 12 variants all positioned between the eighth and tenth exons [103–105]. All 12 variants, each possessing a single ninth exon, are found in a M. sexta cDNA library, indicating that mutually exclusive exon splicing is occurring. The mechanism occurs in many other other genes, being first reported in tropomyosin  and includes serpin genes from Bombyx mori, Ctenocephalides felis, Drosophila melanogaster and Caenorhabditis elegans; albeit with smaller numbers of exclusive exons. The only other proteinase inhibitors reported to use alternative splicing are mammalian calpastatins  in which the variant exons include the initiation codons. It is likely that the evolution of the system occurs by uneven crossover of chromosomes .
Hypervariability in proteins, which may be defined as enhanced variation among orthologous and paralogous genes at the contact residues, is often proposed as an example of positive Darwinian evolution although this idea remains controversial . Hypervariability is virtually a defining feature among proteinase inhibitors and the key mechanism in creating variation. Hypervariability in proteinase inhibitors occurs when nucleotides encoding proteinase contact residues within the active site loops in ‘Laskowski mechanism’ inhibitors such as the ovomucoids [112,113] and the aprotinin family  and in ‘non-Laskowski (nonstandard)’ inhibitors such as the serpins [17,114,115] mutate and are fixed in the genome at a much higher rate as amino acid variants compared with residues elsewhere in the molecule , or, at a nucleotide level, at a rate higher than silent mutations . That hypervariation results in functional diversity, although long the predicted outcome, has been demonstrated by Barbour et al. . The importance of the three preceding mechanisms that act to produce the gene multiplicity needed to generate functional diversity cannot be overemphasized. Ohta  also concluded that ‘that positive selection operated after duplication to increase functional diversity’ and concluded that mechanistically ‘hypervariability of amino acids at the reactive center is generated by an interaction among natural selection, random genetic drift, point mutation, and gene conversion’.
Proteinase inhibitors from classes other than serine proteinase inhibitors have also shown hypervariability, such as cystatins  and squash aspartic proteinase inhibitor (SQAPI; J. T. Christeller, unpublished data). Hypervariability is extremely common among proteinase inhibitors and proteinases [32, 33,118], with only two examples of diversification of proteinase inhibitors without evidence of positive selection have been reported [12,119]. Hypervariability appears to be a feature of other pairs of interacting molecules such as resistance genes , surface antigens , thionins  and conotoxins [111,123], but is otherwise rare, in line with the neutral model of evolution . These interactions also seem to be examples of the operation of a coevolutionary ‘arms race’. Whether these proteins will also show a similar range of evolutionary mechanisms as proteinase inhibitors in addition to hypervariability is not yet known, although as avirulence genes are being identified, some information on their evolution is being published [125–128].
Although the above description completes the specific evolutionary mechanisms currently known to occur in proteinase inhibitors, there are two additional related areas that are relevant to proteinase inhibitor evolution.
Interaction of proteinase inhibitors with inactive proteinases
An emerging story is the possibility of proteinase inhibitor inactivation by inactive variants of proteinases as antagonists. Mathialagan and Hansen  suggest that the pregnancy-associated glycoproteins that are inactive aspartic proteinases  may be the cognate proteinases of uterine serpins. The inactive proteinases would act to bind a proportion of the inhibitors, leaving some active proteinases to fulfil the desired function, in a situation where overexpression of active proteinases is itself undesirable. A similar role has been suggested for the multigene families of inactivated serine and cysteine proteases in Sarcoptes scabiei[129,130], i.e. antagonists of host proteinase inhibitors. Both stories may represent an adaptation to the parasitic interactions involved in pregnancy and scabies infection. This system may operate elsewhere. For example, there are inactive proteinase cDNAs in insect midgut, both induced  and constitutively expressed . If these transcripts are translated and are active in binding proteinase inhibitors, then they may have a role in insect resistance  and in explaining the patterns of adaptation observed when insects are fed diets containing proteinase inhibitors [133–135]. This inundative strategy, a biochemical ‘male sterile technique’, may be relatively uncommon because it requires additional resources in terms of protein synthesis. However, it also presents a new challenge for proteinase inhibitor evolution because the mechanisms described above operate on the evolving proteinase specificity and structural changes rather than the cryptic changes occurring in proteinase inactivation.
Proteinase inhibitors and parasitism
Our discussion so far has described the various evolutionary mechanisms that have been observed in proteinase inhibitors in their coevolutionary variation with cognate proteinases. The examples used exclusively illustrate the concept of a causal relationship between this process and successful plant predation and pathogenesis. It is reasonable to interpret parasitism as a less extreme form of pathogenesis or predation in which the host is maintained in a live state. The literature on the determinants for successfully establishing a parasitic relationship includes many examples in which proteinases have been implicated. These include: (a) stress-induced ClpP of Listeria monocytogenes and its crucial role in intracellular survival of this pathogen , (b) inactivation of serine proteases in the Scabies mite , (c) mycoparasitism of Agaricus bisporus, (d) phytoplasma virulence , (e) serine proteinase inhibitors may play a role in the tick larvae fixation and feeding processes , (f) Trypanosoma cruzi infection , and (g) hookworm adaptation . It is therefore possible that in these situations the same causal relationship exists between proteinase–proteinase inhibitor evolution and parasitism. However, the existence of proteinase inhibitors, let alone variation and adaptation, has not been demonstrated in all these parasitic relationships. We can speculate that this relationship has more far-reaching implications. Parasitism has been implicated as a driving force in the development of sex [141–143]. If this is true then perhaps proteinase inhibitor evolution is even more important they previously recognized.
The evolutionary pressure surrounding the interaction of proteinases and their inhibitors in an antagonist environment seems to be immense. The impression left by the survey presented here is that inhibitors are using virtually every trick in the evolutionary book, and sometimes in combination, to create variation and that the various mechanisms occurring do so in a random fashion. Whether there are new inhibitors and adaptive mechanisms yet to be discovered does not diminish this already impressive list. It is probable that proteinase inhibitors are one of the most actively evolving proteins and that they deserve further consideration as model systems to study important evolutionary phenomena.
I would like to thank W. A. Laing (HortResearch, Auckland, New Zealand) for reading the manuscript. The project was supported by funding from the Public Good Science Fund, administered by the New Zealand Foundation for Research, Science and Technology (Contract C06X0207).