Introduction

The appearance of a novel, resistant plant genotype in the environment presents a challenge to its herbivores; insect populations utilising that host must be able to evolve resistance in order that they can maintain their ability to feed on that host. To predict how an insect species will respond to the emergence of a new, resistant host genotype, we need to know three things. First, is there additive genetic variation in the insect population for the ability to utilise the resistant host genotype? A number of studies have estimated levels of additive genetic variation in host use in herbivorous insects, the majority of which consider variation in the ability of a generalist herbivore to feed on different host species (Via, 1984a, 1984b; Ward et al, 1993; Carriere and Roitberg, 1994; Fox and Caldwell, 1994; Hawthorne and Via, 1994; Sheck and Gould, 1996; Thompson, 1996; Tucic et al, 1997; Ueno et al, 1997; Bossart, 1998; Hawthorne, 1998; Lazarevic et al, 1998; Gu et al, 2001; Poore and Steinberg, 2001).

Fewer studies examine genetic variation in the ability of a generalist to feed on novel resistant varieties of an existing host, although this is important in terms of the ability of a generalist to maintain its host range. One such study examined variation in the ability of the polyphagous leafminer, Liriomyza trifolii, to feed on a resistant genotype of an existing host, the chrysanthemum (Hawthorne, 1998). Genetic variation for performance on resistant chrysanthemums was identified and 10 generations of selection for resistance in the leafminer resulted in survival in the selected lines matching that found on susceptible chrysanthemum genotypes. Therefore, in this case, the leafminer showed the potential to respond to a novel, resistant genotype of an existing host, thus maintaining its host range.

The second thing we need to know is, if there is additive genetic variation present, at which life stage does this variation manifest itself? For many insect species, the age at which genetic variation in host use becomes apparent could be important. For example, in many lepidopteran species neonates tend to feed on the plant upon which they hatch and so are dependent on their mothers' choice of host, while older larvae are able to move to neighbouring plants, potentially encountering novel plant genotypes (Zalucki et al, 2002). This could result in differing selection pressures at different life stages. Moreover, levels of additive genetic variation can also vary with age or life stage as genes may be differentially expressed, both qualitatively and quantitatively during development (Zhu-Salzman et al, 2003). If the levels of additive genetic variation in performance differ between young and old larvae, then the population-level response to selection could depend more strongly on responses in one life stage over another.

Consequently, the third thing we need to know is whether there is a genetic correlation between offspring performance and adult oviposition preference. For example, if there is genetic variation in adult oviposition preference for the resistant host genotype and this is positively genetically correlated with larval performance on the plant, then we would expect a strong response to selection in both of these traits. However, if the adults do not recognise the new, resistant genotype as a host and do not oviposit on it, then the ability of the neonate larvae to develop on the plant may be irrelevant. A number of studies have examined the genetic correlation between preference and performance (Via, 1986; Fox, 1993; Ward et al, 1993; Nylin and Janz, 1996; Tucic et al, 1997; Gu et al, 2001), but to our knowledge no previous studies have considered age-related effects on additive genetic variation in host use.

Helicoverpa armigera (Hübner) is a highly polyphagous noctuid moth. It is found on a number of agriculturally important species including the chickpea, Cicer arietinum (L.). Chickpea is a self-fertilising species and as such occurs as highly inbred lines. Screening of chickpea germplasm has identified a number of varieties potentially resistant to H. armigera (Lateef, 1985). Using the H. armigera–chickpea, insect–host relationship as a model system, we address the question of how a polyphagous species might respond to a novel, resistant host genotype by investigating variation in larval performance on, and adult preference for a susceptible and a resistant plant genotype. This variation was then partitioned into additive genetic and residual variance and the genetic correlation across life stages and environments examined.

Methods

Plants

Two varieties of the chickpea C. arietinum were used in this study: Tyson and ICC506. ICC506 is a variety that has been shown to have high levels of resistance to H. armigera in both field and laboratory studies (Lateef, 1985). Early in December 2002, 40 seeds of each variety were sown in commercially available UC Riverside potting soil mix for use in the neonate feeding assays. After 1 week, and 1 month, a further 40 seeds of each variety were sown for use in the third-instar feeding assays and adult oviposition trials, respectively. Thus, a large number of 6-week-old, preflowering plants were available for each assay. All seeds were sacrified prior to sowing and inoculated with rhizobium.

H. armigera culture

The laboratory culture was founded from larvae collected from northern Western Australia and from several locations on the east coast of Australia. A previous study found that gene flow was high even between distant populations in Australia and that the effective population size was large (Daly and Gregg, 1985). It was on this basis that we decided to collect larvae from sites on the east and west coasts of Australia and outcross them, thus establishing a laboratory colony that encompassed the variation present in the Australian population. To reduce the risk of inbreeding, eggs were collected from more than 200 adults each generation. The colony had been kept in the laboratory for three generations at the beginning of the experiment and was reared at 26°C with natural light, a necessary condition for breeding in this species.

Sib analysis

A full-sib/half-sib design was used to determine heritabilities of feeding performance and adult host preference (Lynch and Walsh, 1998). In all, 15 virgin males were each mated to two virgin females, resulting in 30 families in total. The mated females were then placed in individual containers with access to honey water, filter paper and nappy liner on which to lay eggs. Eggs from each female were collected and allowed to hatch in plastic tubs with access to artificial diet. Immediately upon hatching, 30 neonates from each family were assigned to the neonate assay. Approximately 5 days after hatching, larvae were transferred to individual 25 ml plastic cups. Upon reaching the third instar, 20 larvae per family were assigned to the third-instar assay. The remaining larvae were allowed to pupate in the cups and upon emergence moths were assigned to the adult choice assay.

Larval performance assays

Neonates

From each family, 15 neonates were randomly assigned to the Tyson and 15 to the ICC506 treatment groups. Larvae were then placed, in groups of five, into 200 ml plastic pots containing 10 ml of 10 g/l water agar into which the stems of five chickpea leaves were pushed. H. armigera moths lay eggs singly, but will lay several eggs on a single leaf (S Cotter, personal observation). As such, young larvae are likely to encounter each other during feeding. Rearing neonates in groups more closely mimics larval distribution in the field than would rearing in individual containers. Chickpea leaves were collected from Tyson and ICC506 plants immediately prior to testing and randomly distributed among the containers. Larvae were left to feed for 5 days, after which time each surviving larva was weighed. Containers were checked daily to ensure that sufficient leaf material remained. It was not necessary to replace leaf material during the experiment.

Third instars

From each family, 10 third-instar larvae were randomly assigned to the Tyson, and 10 to the ICC506 treatment groups. Larvae were starved for 2 h, weighed and then placed, individually, into 25 ml plastic pots containing 5 ml of 10 g/l water agar into which the stems of a single chickpea leaf was pushed. Older H. armigera larvae are more solitary and can be cannibalistic and so rearing in individual cups at this stage is necessary (S Cotter, personal observation). Again, chickpea leaves were collected immediately prior to testing and randomly distributed among the containers. After 24 h of feeding, larvae were weighed a second time to give an estimate of weight gain.

Adult choice assay

Five female moths from each family were mated and then placed in individual 600 ml containers with two branches of Tyson and two of ICC506 in agar. The branches were matched for size and arranged alternately in a circle around a central feeder. After 24 h, each container was rotated to avoid any positional effects on female choice. Females were left to lay eggs for 2 days, after which time the branches were removed and the eggs laid on each counted. Preliminary tests found that the repeatability of female choice using this experimental technique was high (r=0.78±0.13, MSamong females=0.193, MSwithin females=0.024, n=2 preference measurements per female).

Variance components analysis

Heritability estimates of each trait and genetic correlations between traits were estimated using a multivariate restricted estimate maximum-likelihood (REML) procedure (VCE version 4, Groeneveld and Kovac, 1990; see http://w3.tzv.fal.de/genetik/public_html/). This involved fitting an individual ‘animal model’ where the phenotype of each individual was separated into additive genetic components of variance plus other random and fixed effects, such that: y=X b+Z a+e, where y was a vector of phenotypic values, b and a were vectors of fixed and random effects, e was a vector of residual values and X and Z were the corresponding design matrices relating records to the appropriate fixed or random effects (Lynch and Walsh, 1998). The phenotypic variance of each trait, VP is thus described as VP=VA+VM+VR, where VA is the additive genetic variance, VM is the variance attributable to maternal effects and VR is the residual variance, which includes nonadditive sources of genetic variance such as dominance variance or epistatic effects, environmental effects and error variance. All of the estimates for variance due to maternal effects were nonsignificant and so were removed from the models. The effect of cage on the neonate estimates of performance was also nonsignificant and so was removed from the models.

The heritability of each trait was calculated as the ratio of additive genetic variance to phenotypic variance: h2=VA/VP. Genetic correlations between each pair of traits, rA, were estimated from the genetic covariance estimate between the two traits Cov[x,y], and the estimate of additive genetic variance for each trait VAx and VAy, where rA=Cov[x,y]/[(VAx)(VAy)]0.5. The VCE program returns SEs for all estimates, the significance of which could then be determined with t-tests. As the REML procedure assumes that the data are normally distributed, larval weight gain data were log-transformed and percentage egg-lay data were angular-transformed prior to analysis to conform to this assumption. The analysis was then repeated with the untransformed data in order to calculate estimates for the coefficients of additive genetic and residual variance (CVA and CVR, respectively), where CVA=100(VA)0.5/X and CVR=100(VR)0.5/X and X is the population mean.

Results

Larval performance and adult oviposition on each chickpea variety

The effects of plant genotype on larval performance and adult oviposition preference were analysed with linear mixed models using REML in Genstat. We included sire as a random effect, and plant and the interaction between sire and plant as fixed effects. As expected, neonate performance, measured as weight gain over 5 days feeding on chickpea, was significantly higher on the susceptible chickpea Tyson than on the resistant variety ICC506 (Wald statistic χ2=106.07, df=1, P<0.001, Table 1), although there was no effect of chickpea variety on neonate survival (logistic regression, χ2=0.36, df=1, P=0.54).

Table 1 Trait means and heritability estimates
Full size table

Third-instar performance, measured as weight gain over 24 h, was also higher on Tyson, although the effect was much smaller (Wald statistic χ2=4.69, df=1, P<0.05; Table 1); there was no mortality in third instars over the course of the feeding test. Despite the fact that the suitability of Tyson for larval development seemed to be higher than that of ICC506, there was no significant difference between the numbers of eggs laid by adult moths on each variety (Wald statistic χ2=0.06, df=1, P=0.81; Table 1).

Heritability of larval performance

All of the heritability estimates for larval performance were highly significant (Table 1). In contrast, the heritability of innate adult host preference was not significant, although the CVA values calculated using the untransformed data suggest that there is additive genetic variation present in this trait.

The estimates for the heritability of neonate performance on each host plant were higher than the respective third-instar estimates, although this was marginally nonsignificant for heritability of performance on Tyson (Tyson, hneonate2=0.441±0.050, h3rd instar2=0.295±0.056, t28=1.95, P=0.061; ICC506, hneonate2=0.578±0.069,h3rd instar2=0.344±0.060, t28=2.56, P<0.05). There was a trend for the heritability estimates of performance on ICC506 to be higher than on Tyson, but this was not significant (Table 1). An examination of the CVA and CVR calculated for each trait show that there are similar levels of additive genetic variation present in all the measures of performance, but that the levels of residual variance are higher for third-instar performance (Table 1).

Genetic correlations across life stages

While there was a significant positive genetic correlation between neonate and third-instar performance on ICC506 (rA=0.517±0.122, t13=4.24, P<0.001; Table 2), there was no comparable correlation across life stages for larvae feeding on Tyson (rA=0.014±0.142, t13=0.10, P>0.05; Table 2).

Table 2 Genetic correlations between traits
Full size table

Trait variation and the environment

Trait variation across environments can be examined in two ways. Firstly, variable trait expression can be regarded as the trait itself and variation partitioned into that explained by the genotype and that explained by the environment. A significant genotype-by-environment interaction shows that genotypes perform relatively differently in each environment. The second approach considers trait expression in each environment as a different trait and examines the genetic covariance between them; a genetic correlation significantly lower than 1 indicates that the ranking of genotypes differs across environments (Lynch and Walsh, 1998).

Using the first approach, there was a significant genotype-by-environment interaction for neonate performance (Wald statistic χ2=68.87, df=28, P<0.001; Figure 1a) but not for third-instar performance (Wald statistic χ2=12.84, df=28, P=0.99; Figure 1b). The second approach confirms this result. The genetic correlation across environments for the neonates was significantly lower than 1 (rA=0.198+0.130, t13=−6.17, P<0.001; Table 2), whereas the correlation across environments for the third-instar larvae was not significantly different from 1 (rA=0.945+0.049, t13=−1.12, P=0.28; Table 2).

Figure 1
figure 1

Genotype-by-environment interactions for each life stage. Family means±SE using the transformed data are plotted for each chickpea variety for neonates (a) and for third instars (b). Nonparallel lines are an indication of a genotype-by-environment interaction. Neonate growth rate is measured as log (weight in mg after 5 days +1). Third-instar growth rate is measured as the residuals from the regression of weight after assay on weight before assay.

Full size image

Discussion

To predict how a generalist will respond to the emergence of a new, resistant host genotype in the environment, we need to know three things: is there additive genetic variation for the ability to utilise the resistant host genotype, at which life stage does this variation manifest itself and is there a genetic correlation between offspring performance and adult oviposition preference?

In this study, we found moderate levels of additive genetic variation in the ability to utilise a susceptible chickpea and a novel, resistant chickpea genotype in the H. armigera population. It has been argued that CVA rather than heritability should be used as an indicator of evolvability (Houle, 1992). Heritability is the measure of evolvability in the standard deviation (SD)-standardised version of the breeders equation (Lynch and Walsh, 1998); however, Hereford et al (2004) suggest that standardising by the SD is inappropriate as the standardisation factor is itself a function of the additive genetic variance. Alternatively, standardising by the trait mean results in the measure of evolvability being the CVA. As such, if the mean-standardised strength of selection is equal for all measured traits, the CVA would be the best predictor of the response to selection, whereas if the SD-standardised strength of selection is equal for all measured traits, the heritability would be the best predictor of the response to selection.

Interestingly, the heritability of performance was higher for neonates than third-instar larvae, suggesting that the SD-standardised response to selection for performance on a plant should be stronger in neonates than in older larvae. In contrast, the CVA values calculated from the untransformed data were very similar across all groups, suggesting that the mean-standardised response to selection would be similar at both life stages. It seems likely that the probability an individual will survive to reproduce in the field would be associated with its absolute growth rate rather than with its growth rate relative to other individuals. As such, it is not possible to predict which form of standardised selection would be more comparable to that acting in the field. Therefore, in order to predict the response to selection in the field, further information regarding the nature of selection at each life stage would be required.

On the resistant chickpea genotype, ICC506, there was a strong positive genetic correlation in larval performance across the life stages. If the basis of this correlation is pleiotropy rather than linkage disequilibrium, then the genes controlling the trait at each life stage were largely the same. Conversely, on the susceptible genotype, Tyson, there was no genetic correlation in performance across life stages. It is interesting to note that there was also a strong positive genetic correlation between neonate performance on ICC506 and third-instar performance on Tyson. This poses interesting questions regarding the genes controlling larval performance on each host.

A possible explanation for the pattern of genetic correlations in larval performance could be due to constitutive versus induced resistance to plant defence mechanisms at the different life stages. Plants use a number of resistance mechanisms that can affect insect feeding, including physical factors such as leaf toughness or trichome density, or chemical factors such as toxic allelochemicals and proteinase inhibitors. When encountering chemical defences, insects can respond in kind. For example, insects can detoxify allelochemicals via the inducible cytochrome P450 monoxygenase system (Berenbaum, 1991; Berenbaum et al, 1992; Rose et al, 1992; Hung et al, 1995; Scott et al, 1998; Harrison et al, 2001; Li et al, 2002), or produce or upregulate alternative proteases that are not susceptible to inhibition, or that can digest the proteinase inhibitors present in the diet (Broadway, 1996, 1997; Wu et al, 1997; Patankar et al, 1999, 2001; Moon et al, 2004).

The specific mechanism of resistance responsible for the differences between Tyson and ICC506 is unknown; however, the evidence to date suggests that resistance in chickpeas is primarily due to acid exudates on the leaf surface (Lateef, 1985) or isoflavonoids (Simmonds and Stevenson, 2001); it is therefore likely to involve some kind of secondary compound to which the feeding larva is forced to respond. Neonates have few fat reserves and establishment on a plant is critical to survival. In these circumstances, it would be beneficial for a detoxification system to be an induced response, switched on only when necessary. In other words, neonates placed on the resistant host, ICC506, faced with secondary compounds would respond by switching on genes responsible for the production of detoxifying enzymes or insensitive proteases, while neonates on Tyson, the more palatable host, would not.

Third-instar larvae with greater fat reserves may have these genes switched on as a form of constitutive resistance, or alternately, the threshold level of gut function disruption at which these genes are switched on may be much lower than for neonates. This scenario could result in genes being switched on in all larvae tested, except for the neonates on Tyson. Further studies are necessary to determine the secondary compounds responsible for resistance in ICC506 and the mechanisms of detoxification used by H. armigera in response to these compounds in the diet.

In contrast to performance, adult oviposition preference was not heritable and there was no overall preference for the susceptible chickpea, Tyson (at least under the experimental protocol employed here). It may be that the females were unable to discriminate between the two varieties of chickpea. Both lines appear to be identical, and while there may be differences in secondary compounds between the lines, these may not be detectable prior to ingestion. As such, there may be no chemical cues available to a female to indicate that one line is less suitable for her offspring than another. A lack of genetic correlation between oviposition preference and offspring performance, while counterintuitive, is not unusual even at the level of species, (Thompson, 1988 and references therein; Fox, 1993; Nylin and Janz, 1996; Gu et al, 2001), and as such, it may be that the lack of genetic correlation in this case is accurate.

Previous studies have suggested that adult host preference is strongly influenced by factors such as plant abundance and experience with a particular host (Papaj and Rausher, 1987; Cunningham et al, 1998, 1999; Cunningham and West, 2001). Whether the lack of preference for either genotype of chickpea found here is due to the inability to discriminate between the two or due to a lack of experience with either host, it seems likely that females encountering a large patch of host plants, as would occur in an agricultural situation, would oviposit regardless of the genotype present. Under these circumstances, it seems likely that there would be strong selection acting on neonates for improved performance and that the response to such selection would be rapid. Future studies should focus on the strength of selection at each life stage to further examine the potential for insect adaptation to resistant hosts in the field.