• Tidak ada hasil yang ditemukan

Department of Biology, University of California, Riverside, CA 92521, USA

© CAB International 2003. Quality Control and Production of Biological Control Agents:

Theory and Testing Procedures (ed. J.C. van Lenteren) 73

Abstract

The success of biological control, particularly augmentative biological control, depends upon the effec- tive mass rearing of natural enemies. However, developing the best rearing strategy is complicated by the ‘paradox of captive breeding’: increasing quantity generally decreases quality. Quantity is the num- ber of individuals produced per unit time, and it is easily measured. Quality is the ability of captive- reared individuals to function as intended in the field, and can be measured as field success relative to the success of individuals from a natural population. Such field measurements are almost always diffi- cult and expensive. Unfortunately, quantity and quality are usually negatively correlated, since genetic adaptation to the rearing environment often adversely affects adaptation to the field environment.

Here I review examples of adaptation to captive rearing and of the trade-off with field performance.

Given this trade-off, the optimum management strategy is a compromise that can only be defined after extensive field tests. However, we can identify some general factors that are likely to influence the out- come of a breeding programme. A large, genetically variable founding population from a geographical region climatically similar to the release site maximizes the chance of adaptation of the control agent to the release site. While a large founding population minimizes the immediate risk of inbreeding, it may include a few genotypes preadapted to the captive rearing conditions. The success of these few geno- types can result in a genetic bottleneck. The problem can be minimized by temporarily maintaining many small breeding units, so that the reproductive success of a lot of individuals is ensured during the initial phase of domestication. However, any genetically variable population will adapt to its new artificial conditions, and the breeding facility should be designed to minimize selection for characteris- tics known to reduce field performance. Even so, field-adapted genotypes must be incorporated on a regular basis, either by monitored addition to the captive population or by establishing a completely new population. Alternatively, adaptation to the captive environment can be avoided by maintaining a large number of inbred (isofemale) lines. This approach, combined with prerelease crosses, can be very effective at maintaining quality. These considerations highlight an important problem associated with using genetically manipulated stocks. We must be careful that the potential benefits of genetic engi- neering are not squandered by incorporating beneficial genetic changes into laboratory-adapted stocks that are ill suited to field release.

Introduction

Optimizing the mass rearing of arthropods for release into the field is an extremely com- plex problem. Much of this complexity arises from a fundamental evolutionary conflict implicit in the mass-rearing process: the ideal captive population is maximally adapted to both the rearing conditions and the field conditions into which it will be released. In most cases, this ideal cannot be achieved and the optimal solution is a com- promise between efficient rearing and good field performance. However, the details of the compromise will depend upon the genetic structure of the population. By adopting a population-genetic approach, we can avoid many of the pitfalls of establishing and managing captive populations and hopefully achieve close to the best solution.

There is an extensive literature on the evolutionary problem of how organisms in nature maintain simultaneous adaptation in two (or more) environments. This is the problem of adaptive plasticity (see Via et al., 1995) and the evolutionary solution depends upon the degree of genetic correlation between high fitness in the alternative envi- ronments. A high positive correlation means that genotypes with high fitness in one envi- ronment also have high fitness in the second.

However, a high negative correlation means that genotypes with high fitness in one envi- ronment have a low fitness in the other. It is the existence of negative genetic correlations that creates the evolutionary problem of adaptive plasticity and what I shall refer to as the ‘paradox of captive breeding’ – improving performance in the rearing facil- ity can result in decreased performance in the field.

The paradox of captive breeding means that the optimal solution is generally a com- promise between quantity and quality.

Quantity is the productivity of the rearing programme, measured by numbers reared per unit time. Quality is the ability of cap- tive-reared individuals to function as intended after field release (see Chapter 1), and can be measured relative to the field per- formance of individuals from the natural

population. The optimal solution aims to maximize the product (quantity) ×(quality), and I shall argue that, in general, this most effective strategy does not maximize either quantity or quality.

The Problem: the Trade-off Between Quantity and Quality Measuring the trade-off

The measurement of quantity, i.e. the success of captive rearing, is a straightforward count of the numbers of individuals produced under the rearing protocol. Quantity is expected to improve with domestication, i.e.

adaptation to the captive environment. This improvement can be evaluated experimen- tally by comparing the productivity of the established captive population to a recently wild-caught control population. The mea- surement of quality, i.e. field performance, is more complex and must be determined by the specifics of the release programme.

While the goal is to control the numbers of some specific pest, the specific agent used may be a parasitoid, predator or, in the case of the sterile-insect technique (SIT), a conspe- cific male. However, at a minimum, the mea- sure must integrate three components: first, the ability of individuals to disperse from the release site and find the target (hosts or prey in the case of natural enemies, females in the case of SIT); secondly, their ability to success- fully interact with the target (parasitize the hosts or eat the prey in the case of natural enemies, or mate in the case of SIT); and, thirdly, the ability of the released individu- als to survive in the field and continue to find their targets. Finally, if the project goal is to establish a self-perpetuating population of a natural enemy (classical biological con- trol), then the ability of individuals to repro- duce and to survive through unfavourable seasons defines a fourth essential component of quality.

It is important that the quality of the cap- tive-reared population is measured relative to a recently wild-caught control population.

However, in practice, it is unlikely that the field performance of a captive strain could 74 L. Nunney

be compared directly with that of a wild- caught population. Instead, the captive strain could be tested alone and its perfor- mance compared with some previously established ‘wild’ standard. To establish the standard, replicated trials (preferably run at different times and at different sites) must be conducted using a wild-caught popula- tion. These trials would establish the stan- dard in terms of some appropriate measure of quality. The measure should reflect the success of an individual over the whole of its useful life, and include the necessity for active searching beyond the immediate release site. For example, for a parasitoid, this standard could be per cent parasitism per parasitoid per unit area measured a set number of days after release and given some typical host density.

I have been unable to find any experi- ments that measure both quantity and qual- ity in strains with different degrees of domestication. Indeed, there are relatively few studies that document the consequences for biological control of a decline in quality associated with mass rearing (although see Ito, 1988; Calkins and Ashley, 1989). This is probably due to an understandable reluc- tance of those responsible for captive rearing to document any such decline, since the effort is likely to provoke criticism of their captive- rearing strategy. However, this attitude is misguided. The important question is not whether quality has declined, but whether the product of quality and quantity has been maximized. The theoretical expectation is that maximizing this product will almost inevitably involve some decline in quality.

Theoretical expectations

In general, quantity and quality are inter- changeable, i.e. a loss of quality can be com- pensated for by increased quantity (Nunney, 2002). This assumption leads to the conclu- sion that the optimum strategy maximizes the product of these two parameters. Only if there is an interaction between quantity and quality is it necessary to maximize a more complex function. For example, if density- dependent interference among individuals

caused individual field performance (qual- ity) to decline at high density, then field den- sity would need to be factored into the maximization. Except under these condi- tions, the effectiveness of mass-rearing pro- grammes can be evaluated along the two dimensions of the quantity produced and the quality of the individuals released.

Mathematically, the effectiveness (E) of a captive-rearing programme is defined by:

E= Pw (1)

where P and w are quantity (productivity) and quality (individual field performance), respectively. Furthermore, we expect that adaptation to the rearing environment (increasing P) will change (and generally decrease) waccording to some function f:

w= f(P) (2)

It follows from (1) and (2) that maximizing E requires:

dln f(P)

dln P =1 (3)

The interpretation of equations (2) and (3) is shown in Fig. 6.1. In both the upper and lower graphs, the solid curve defines f(P), the relationship defining how field quality (w) changes as an initially wild-caught popu- lation adapts to captive rearing. The popula- tion is expected to gradually shift from its initial state (‘wild population’) to a relatively stable ‘domesticated stock’. This transition is marked by some decrease in field perfor- mance. Since effectiveness is the product Pw (equation 1), points of equal effectiveness are linked on a log scale by a line of slope 1 (see Fig. 6.1, dashed lines). The maximum effectiveness is usually defined by equation (3), i.e. where one of these lines is tangential to f(P). This can be seen in the upper graph of Fig. 6.1. The ‘optimum strategy’ shown on the graph maintains a population that is par- tially domesticated – at this point the gain in quantity far outweighs the loss in quality.

The lower graph differs in the shape of f(P), a difference that alters the optimum strategy. In this graph, there is a local maxi- mum near to the point of full domestication;

however, the overall optimum strategy is to minimize domestication, because of the large Managing Captive Populations 75

loss in quality occurring during the initial stages of domestication. Such a pattern sug- gests investment in improving the rearing conditions so that the selection that results in an excessive loss of quality is reduced.

In a captive environment, we can expect the ‘domesticated’ stock (Fig. 6.1) to be rela- tively stable. However, over time, we would expect the accumulated effects of inbreeding to result in a slow general decline that reduces both quantity and quality (the lower part of the curves shown in Fig. 6.1).

Genetic Change due to Captive Rearing Most of the studies of genetic change occur- ring in captive-reared populations have been on the tephritid fruit flies used in SIT. These studies are useful because they illustrate the dramatic changes that occur when insects are intensively cultured to produce very large numbers. There is no reason to believe that the data from fruit flies are unusual. Genetic changes are inevitable whenever a genetically variable population is reared in a novel envi- ronment. Natural selection will act and adap- tation to the new environment will occur.

Mating behaviour

One very common adaptation to captive rearing is earlier mating and oviposition (e.g.

melon fly: Miyatake, 1998; medfly: Rössler, 1975; Wong and Nakahara, 1978; Vargas and Carey, 1989; oriental fruit fly: Foote and Carey, 1987; tobacco budworm: Raulston, 1975). The time scale of this adaptation is quite rapid. Using a wild-caught population of tobacco budworm, Raulston (1975) found that, after seven generations of captive rear- ing, the shift to the domesticated pattern of early mating was almost complete.

More complex changes in mating behav- iour may also occur. Haeger and O’Meara (1970) showed that the captive rearing of the mosquito Culex nigripalpus resulted from a shift in female behaviour. The mating success of colony females was about 70%, regardless of whether the males were from the colony or from the wild; however, under similar condi- 76 L. Nunney

log (quantity)

log (quantity)

log (quality)log (quality)

Wild population

Wild population and optimum strategy

Optimum strategy

Domesticated stock

Domesticated stock

Fig. 6.1. The expected trade-off between the numbers produced in mass rearing (quantity) and field performance (quality). The solid curve defines the trade-off. The natural (wild) population is arbitrarily placed at the origin and the position of a population adapted to the captive-rearing facility over many generations (domesticated stock) is shown. The region of the trade-off curve below the point of domestication defines decreasing quality and quantity, due to the effect of long-term inbreeding depression. The dashed lines link combinations of quality and quantity that are equally effective (as defined by equation 1). The upper graph shows a trade-off curve with an optimum strategy of partial domestication; the lower graph shows a curve with no such optimum – the best strategy is to minimize domestication.

tions only about 1% of wild-caught females mated. Changes in mating behaviour appar- ently due to captive rearing have also been observed in houseflies (Fye and LaBrecque, 1966). In a laboratory simulation of SIT, ster- ile males from a 20-year-old laboratory popu- lation competed poorly with males from a wild-caught population, whereas sterile males from a newly established laboratory population were much more successful. In another example, Fletcher et al. (1968) showed significant differences between two captive populations of screw-worm fly. Males from both populations produced the male pheromone, but only females from one of the populations responded to the chemical. The authors suggested that differences in the cap- tive rearing of the two populations may have selected for this difference.

Changes in mating behaviour are not only a problem for SIT. In classical biological con- trol, the aim is to establish a self-perpetuat- ing population of the natural enemy. If the mating system has been disrupted through domestication, the probability of establish- ment is inevitably reduced.

Life-history traits

Captive-rearing conditions almost inevitably select for faster development. This has been observed in medfly (Rössler, 1975; Wong and Nakahara, 1978; Vargas and Carey, 1989), ori- ental fruit fly (Foote and Carey, 1987), Caribbean fruit fly (Leppla et al., 1976) and melon fly (Miyatake, 1993; Miyatake and Yamagishi, 1999). In the melon fly, these changes occurred during the first nine gener- ations of captive rearing (Miyatake and Yamagishi, 1999).

Selection for faster development generally leads to a correlated decrease in adult size and lifetime female fecundity (Nunney, 1996);

however, the expected correlations can break down when a population is introduced into a new environment (Service and Rose, 1985).

Thus, although the adult size of melon fly decreased in response to selection for a shorter developmental period, lifetime fecundity did not, and captive melon-fly populations gener- ally have a higher fecundity than wild-caught

flies (Miyatake, 1998). Similarly, in both the Caribbean fruit fly (Leppla et al., 1976) and the oriental fruit fly (Foote and Carey, 1987), the shorter development time of domesticated populations was associated with higher fecun- dity, relative to recently wild-caught flies.

Correlated responses can affect traits that we may not a priori expect to be influenced.

Miyatake (1998) notes that selection for faster development in the melon fly results in indi- viduals that have a shortened circadian period and that mate earlier in the day. These responses were not arbitrary; they were due to the pleiotropic effects of a single gene (Shimizu et al., 1997). This result is an excel- lent illustration of how adaptation to the rearing facility (faster development) could have an unexpected negative effect on mat- ing success in the field (due to flies attempt- ing to mate at the wrong time of day).

General

We do not know which genetic loci are involved in the adaptation to a captive envi- ronment. The rapidity of adaptation is sugges- tive that relatively few loci are responsible for most of the change. For example, in tobacco budworm, it took only four generations for the oviposition pattern of a wild-caught popula- tion to converge on that of a laboratory culture (Raulston, 1975). More typically, significant adaptive change seems to occur over the first 6–10 generations (Raulston, 1975; Loukas et al., 1985; Miyatake and Yamagishi, 1999).

In the screw-worm fly, Bush and Neck (1976) identified a candidate gene, the - glycerophosphate dehydrogenase (-GDH) locus. They found that one allele, rare in nat- ural Texas populations, was very common in each of four large ‘factory’ populations. They argued that this was an adaptive change in response to the novel rearing conditions (constant high temperature, combined with selection for rapid development and reduced flight). Similarly, Loukas et al.(1985) found rapid changes at several allozyme loci when a population of the olive fly was reared in the laboratory. In only five generations, the commonest allele at the 6-phosphogluconate dehydrogenase (6-PGD) and alcohol dehy- Managing Captive Populations 77

drogenase (ADH) loci declined from an ini- tial frequency of 0.6–0.7 to close to 0.2.

Strong selection for adaptation to the cap- tive-rearing environment can be expected to reduce the genetic variability of populations.

Bush and Neck (1976) and Loukas et al.(1985) noted that the genetic (allozyme) variability of captive colonies decreased over time, and Miyatake and Yamagishi (1999) found that the heritability of larval development time in captive melon fly declined over time until it was not statistically different from zero.

Changes in Field Performance There is no question that captive rearing results in a cascade of genetic changes as a population adapts to its new environment.

But this leaves open the question of the extent to which these changes reduce field performance. As noted earlier, measuring field performance is almost always difficult and understandably many researchers have used simple laboratory tests to infer effective- ness in the field (see, for example, Cohen, 2000). However, data from the Japanese melon-fly SIT programme argue strongly against this approach.

The melon fly was successfully eradi- cated from the Japanese islands of Kume- zima in 1977 and Miyako-zima in 1987. This success came after several failed attempts in other parts of the world to use SIT to eradi- cate fruit flies. Ito (1988) reviewed the Japanese project and concluded that, con- trary to prevailing practice, a relatively low sterile : wild-fly ratio is sufficient to achieve success provided that the harmful effects of domestication can be avoided. In particular, he emphasized that, although the negative effects of irradiation have always been a con- cern in SIT, mass rearing can cause a much greater reduction in the mating competitive- ness of released males. This is a very impor- tant point. It suggests that the prevailing emphasis on quantity is misplaced. A failure to maintain quality can drive up the cost of biological control and can make successful control unlikely. Calkins and Ashley (1989) stressed this same point for medfly SIT.

Using estimates from three medfly stocks, they calculated the dramatic increase in costs incurred when quality is compromised.

A decline in the field mating competitive- ness of mass-reared melon flies became apparent after about 15 generations. No such decline was apparent under laboratory con- ditions (Fig. 6.2). At generation 18, mating 78 L. Nunney

Laboratory

Field

5 10 15 20

Generations of mass rearing 0

0.5 1.0

Competitiveness

Fig. 6.2. The decline in the field mating competitiveness of sterile, captive-reared male melon flies, as a function of their time in mass production. Also shown are the results of laboratory trials at generations 16 and 17. The release programme on Kume-zima, Japan, started after generation 5. (Figure from Ito, 1988.)