Genetic variation in response to an indirect ecological effect

Abstract

Indirect ecological effects (IEEs) are widespread and often as strong as the phenotypic effects arising from direct interactions in natural communities. Indirect effects can influence competitive interactions, and are thought to be important selective forces. However, the extent that selection arising from IEEs results in long-term evolutionary change depends on genetic variation underlying the phenotypic response—that is, a genotype-by-IEE interaction. We provide the first data on genetic variation in the response of traits to an IEE, and illustrate how such genetic variation might be detected and analysed. We used a model tri-trophic system to investigate the effect of host plants on two populations of predatory ladybirds through a clonal aphid herbivore. A split-family experimental design allowed us to estimate the effects of aphid host plant on ladybird traits (IEE) and the extent of genetic variation in ladybird predators for response to these effects (genotype-by-indirect environmental effect interaction). We found significant genetic variation in the response of ladybird phenotypes to the indirect effect of host plant of their aphid prey, demonstrating the potential for evolutionary responses to selection arising from the prey host.

1. Introduction

Indirect ecological effects (IEEs) are interactions between two species that depend on the presence of a third (Wootton 1994; Abrams 1995). Examples of indirect effects are the herbivore-mediated interactions that occur between primary producers and predators in a trophic cascade. In natural food webs consisting of many interconnected food chains, there exists the possibility of numerous IEEs. These can be a significant cause of phenotypic variation, are sometimes stronger than direct effects (Brown et al. 2001), and may even act antagonistically to direct effects thereby leading to unexpected experimental results (Wootton 1994). As Miller & Travis (1996) highlight, given that IEEs have significant community effects they could also be an important selective force affecting adaptive evolution of the species involved. The importance of this idea has recently been reiterated by Neuhauser et al. (2003) in their review of community genetics (interaction between genes in one species and the populations of other species in the community). Theoretical work considering competitive interactions shows that IEEs can indeed be an important selective force (Connell 1980; Jeffries & Lawton 1984; Johnson & Seinen 2002). Furthermore, empirical (Agrawal 2000; Janssen et al. 2002) and theoretical (vanderMeijden 1996; Agrawal et al. 2002) studies of plant–herbivore–predator interactions show that IEEs may play an important role in the evolution of traits that affect interactions between the species involved.

The extent that selection results in adaptive evolutionary change depends on genetic variation. While genetic variation in the phenotypes affected by interactions would be unsurprising, genetic variation in the extent of the IEE response (i.e. a genotype by IEE interaction, or G×IEE) has never been shown. Like the evolution of phenotypic plasticity, it is the G×IEE that indicates the potential for evolution in response to selection arising from IEEs.

Studies that examine related aspects of genetic variation in the context of IEEs fall into two categories. One set of studies explores how genetic variation in one species affects the IEE transmitted in terms of phenotypic effects on a focal population or community (Rice & Wilde 1989; Birch et al. 1999; Holton et al. 2003). For example, Whitham et al. (1999) review how pure and hybrid plant species vary in the arthropod communities that they attract. The other type of study documents genetic variation in a species experiencing two varying direct effects (i.e. genetic variation in an intermediate species; (Pruter & Zebitz 1991; Cronin & Abrahamson 1999; Hufbauer & Via 1999). Neither asks if genetic variation exists for response to an IEE. Thus, experiments of this sort cannot answer questions about the population's ability to respond to selection. The existence of genetic variation in response to IEEs, although intuitively obvious, remains unexamined.

Quantifying a G×IEE requires a quantitative genetic design. Unlike typical G×E studies, there is a complication because an intermediate species forms an additional environment. Thus, the environment experienced by the focal species is partly determined by the genotype of the intermediate organism. Variation in the genotype of the intermediate organism must therefore either be accounted for (producing a genotype-by-genotype-by-environment, G×G×E, experimental design) or eliminated. The multiplicative nature of factors in such a design is such that to account for all the variation, the number of individuals to be examined quickly becomes prohibitively large unless genetic variation is eliminated in one of the species.

We present evidence for genetic variation underlying a response to an IEE in a ladybird species, Harmonia axyridis in a simplified tri-trophic (plant–aphid–ladybird) system where the intermediate species is clonal. Indirect effects of prey host plant affect several predatory ladybirds including H. axyridis (Nishida & Fukami 1989; Ferry et al. 2003). However, these studies did not examine G×IEE. Demonstrating the existence of G×IEE is an important step towards integrating the community context of a population into models that aim to predict selection and subsequent evolution.

We tested for three components to demonstrate the potential evolutionary consequences of IEE in our model system. First, we tested for an IEE from the plant to the ladybird influencing traits related to fitness (i.e. the potential for selection from an IEE). Second, we tested for genetic variation underlying the traits of interest within each environment (i.e. the potential for evolutionary change). Finally, we tested for genetic variation among ladybirds in response to the IEE, revealed as an interaction between variation in ladybird genotype and variation in IEE (G×IEE, i.e. the potential for adaptive evolution in direct response to selection imposed by an IEE).

2. Methods

(a) Ladybird population and experimental design

We exposed H. axyridis larvae from a full-sib, split-family breeding design to different dietary treatments to induce IEEs. The first of our two diet treatments consisted of a single clone of pea aphid, Acyrthosiphon pisum, raised exclusively on sugar snap pea plants (Pisum sativum var. ‘macrocarpon Dwarf sweet green’; hereafter pea). The second treatment consisted of the same clone of A. pisum raised exclusively on fava bean plants (Vicia fava var. ‘Windsor White’; hereafter bean). This design allowed us to eliminate the effects of aphid genetic variation and to isolate the effect of host plant species on phenotypic variation in the ladybird. We collected aphids from plants and froze them at −18 °C in a sealed container to ensure sufficient numbers of aphids for experimentation. To reduce the unknown effects of within-species plant variation, we thoroughly mixed different collections of frozen aphid stocks from a given plant species.

Our ladybird population consisted of H. axyridis individuals obtained from a commercially cultured population (Koppert Biological Systems, The Netherlands) sold as a bio-control agent. This population was raised in the laboratory under common-garden conditions for at least one generation before use in our experiments. Both the stock population and experimental individuals were maintained at a temperature of 21±1 °C and on a 15 : 9 photoperiod in an incubator.

Because we used a full-sib breeding design the quantitative genetic parameters and genetic inferences from our analyses are potentially biased by dominance and maternal effects (Falconer & Makay 1997). Given the complexity of a half-sib G×E design that involves three species, and the sample sizes required, this design is not practical. Further, each ladybird requires an enormous number of aphids during development (Koch 2003). We, therefore, traded off the potential limitation of a full-sib study with the practicalities imposed by insect husbandry.

We randomly collected virgin males and females from mass colonies and placed them together in a clean 10 cm petri dish where they were allowed to mate. Males were removed after mating was complete. Females were left in the petri dishes and allowed to lay eggs. Dishes were checked every 24 h for eggs. All eggs laid within a single day were considered a single clutch. To minimize cannibalism and any post-hatching common environment effects, newly hatched larvae were transferred to individual petri dishes. Larvae were randomly assigned to the different treatments (aphids from pea or aphids from bean) and allowed to feed ad libitum. Fresh aphids were added to petri dishes every day. We collected data from 23 full-sib families. Male and female H. axyridis are sexually dimorphic for many of the traits we considered (Preziosi et al. unpublished data). We therefore analysed male and female data separately. There were no differences in the patterns for males and females, and so for brevity we present only the results from females. Family sizes are around ten and vary slightly due to missing data for some families for some traits.

(b) Focal traits

All the measurements were made on the full-sib offspring of the matings. We photographed individual ladybirds using a digital camera on a dissecting microscope and measured morphology (abdomen length, thorax length and pronotum width) from these images using NIH image analysis software v. 1.61 (available at http://rsb.info.nih.gov/nih-image/). We measured adult mass to the nearest 0.1 mg within 24 h of eclosion and before being given food or water. We recorded hatching time and development time to the nearest day. We analysed hatching time to check that the two halves of each family were equivalent at birth and that differences detected later were due to IEEs occurring after hatching, where the only treatment effects were the host plants of the prey.

(c) Statistical analysis—heritabilities and coefficients of genetic variation

We estimated full-sib heritabilities and associated standard errors within each indirect environment using one-way ANOVA (Becker 1992). In addition to heritabilities, we calculated the coefficient of genetic variation (CV_G; Houle 1992). CV_G provides a measure of how much phenotypic change may occur given a unit of selection (‘evolvability’) and is presented as a percentage of the trait mean. For example a unit of selection on a trait with a CV_G of 10% produces a change in the population mean twice as large as a unit of selection acting on a trait with a CV_G of 5%.

(d) Statistical analysis—indirect ecological effect

The effect of indirect environment on individual traits was tested with a mixed-model two-way ANOVA using type III sums of squares in S-Plus (Insightful 2001). Indirect environment was analysed as a fixed effect and genotype as a random effect. The indirect environment factor in this model is a measure of the significance of IEEs, controlling for the effects of variation among families.

In addition to tests of significance, we estimated the proportion of the total variation seen in our ladybird populations due to each factor. We derived the following variance components:

$$V_{\text{P}}=V_{\text{G}}+V_{\text{IEE}}+V_{\text{G}\times \text{IEE}}+V_{\text{R}},$$

where V_P is the overall phenotypic variation, V_G is the variance component due to overall genotype effects, V_IEE is the variance component due to indirect environment, V_G×IEE is the variance component due to genotype-by-indirect environmental effect interaction and V_R is the residual variance component. We calculated the proportion of variance attributable to these factors as the ratio of the variance component due to the factor of interest to the overall phenotypic variance (obtained by summing all variance components).

(e) Statistical analysis—genetic variation for IEE

To test for significant genetic variation in response to the IEE we used the two-way ANOVA to estimate the variation in the response of different genotypes to indirect environments. The ratio of the interaction mean squares over the residual mean squares in this model provides a test of the significance of G×IEE interaction. In the same way that a significant G×E indicates variation among genotypes in their response to different environments (Via 1984), a significant G×IEE interaction effect indicates that there is variation among genotypes for response to the indirect environmental factors. We estimated the proportion of the total phenotypic variation attributable to G×IEE interaction as above.

3. Results

(a) Heritabilities and coefficients of genetic variation

High heritabilities for all traits in both indirect environments (table 1) indicates significant differences between families. CV_Gs were generally similar for traits between treatments although adult mass, hatching time and development time had higher CV_Gs than other traits (table 1).

Table 1.

Mean, heritability, and coefficient of variation for traits in different treatments.

	bean					pea
trait	$\overline{X}$ (SE)	H2 (SE)	CVG	NF	NI	$\overline{X}$ (SE)	H2 (SE)	CVG	NF	NI
abdomen length (mm)	2.540 (0.014)	1.029 (0.203)	4.306	22	113	2.479 (0.012)	1.079 (0.204)	3.602	22	102
thorax length (mm)	1.545 (0.010)	1.056 (0.199)	5.066	22	117	1.519 (0.009)	0.941 (0.211)	4.382	22	107
pronotum width (mm)	3.165 (0.017)	0.894 (0.208)	3.966	22	117	3.142 (0.017)	1.136 (0.196)	4.334	22	107
adult mass (g)	0.033 (0.001)	1.059 (0.200)	11.067	22	115	0.030 (0.001)	0.590 (0.221)	8.464	21	100
hatching time (days)	3.882 (0.043)	1.506 (0.132)	10.751	23	119	3.794 (0.049)	1.362 (0.164)	11.214	22	107
development time (days)	22.000 (0.193)	1.190 (0.183)	7.473	23	119	23.645 (0.218)	1.388 (0.160)	8.058	22	107

X¯, mean; SE, standard error; H², full-sib heritability; CV_G, coefficient of genetic variation; N_F, number of families; N_I, number of individuals.

(b) Indirect ecological effect

Controlling for within-family variation, ladybirds showed a strong response to the IEE for half of the traits we examined (table 1). Individuals generally had longer abdomens (F_1,158=5.400, p=0.021), greater mass (F_1,162=18.043, p<0.001), and shorter total development time (F_1,171=25.932, p<0.001) when reared on A. pisum from the bean treatment than individuals from the pea treatment. By contrast thorax length and pronotum width did not show a significant difference between treatments (thorax, F_1,167=2.035, p=0.156; pronotum, F_1,167<0.001, p=0.979). As expected hatching time, which occurs before the treatment is imposed, did not differ significantly between the two treatments (F_1,171=0.690, p=0.407).

(c) Genetic variation for IEE

We found significant or near significant interactions between ladybird genotype and IEE for most traits (abdomen length, F_19,158=1.989, p=0.011; pronotum width, F_19,167=3.466, p<0.001; adult mass, F_19,162=2.173, p=0.005; development time, F_20,171=1.556, p=0.069). The G×IEE for thorax length (F_19,167=1.058, p<0.399) and hatching time (F_20,171=0.953, p=0.521) was not significant. Significant G×IEE reflects non-parallel genotype reaction norms (figure 1). The amount of phenotypic variation explained by the G×IEE interaction could be substantial, ranging from 0% for thorax length and hatching time to 25% for pronotum width (table 2).

Figure 1.

(a)–(f) G×IEE reaction norms for Harmonia axyridis. Panels a–d illustrate reaction norms for morphological traits (a, abdomen length, b, thorax length, c, pronotum width, d, adult mass) and panels e and f illustrate reaction norms for life historical traits e, - hatching time, f, development time). Each line represents the mean reaction norm for a single genotype. The slope of the line is a graphical representation of the degree of phenotypic plasticity shown by that genotype. Indirect environments are plotted on the x-axis.

Table 2.

Percent of total phenotypic variance across both indirect environments explained by components derived from factors in our analysis.

trait	VG	VIEE	VG×IEE	VR
abdomen length	42.054	1.616	9.379	46.951
thorax length	8.779	0.086	0	91.134
pronotum width	22.212	0	25.104	52.684
adult mass	26.441	8.318	12.300	52.940
hatching time	71.133	0	0	28.867
larval development	60.226	6.801	3.260	29.713

Variance components presented as follows: V_G, genotypic; V_IEE, indirect environmental; V_G×IEE, genotype-by-indirect environmental; V_R, residual.

4. Discussion

(a) Indirect ecological effect

Half of all the traits we examined showed a highly significant phenotypic response to the IEE. These results agree with the only previous study that tested for IEE of host plant on H. axyridis (Ferry et al. 2003). Wootton (1994) argues that strong IEEs such as these could increase the effect that a single species has on the entire community. Such effects determine important community properties such as the overall strength of interactions and the level of connectedness in a community.

(b) Heritabilities and genetic variation for IEE

Our results suggest ample genetic variation in the life-historical and morphological traits we measured. Indeed, heritabilities were sometimes >1. This indicates that our estimates contain error around a high real heritability (Palmer 2000). This is not unusual for regression or sibling estimations of heritability and does not violate theoretical expectations (Weigensberg & Roff 1996).

As well as this heritable variation within indirect environments, we found significant genetic variation for response to an IEE (i.e. significant G×IEE interaction from the two-way ANOVAs). This suggests that it is not only the direct environment that interacts with genotype to affect phenotype (shown as G×E in other studies on H. axyridis; Grill et al. 1997; Ueno 2003) but also the indirect environment.

The genetic variation we detected underlying the response to the IEE explained up to 25% of the total phenotypic variation, which demonstrates the large effect that genetic variation in response to the IEE had in our population. The action of selection stemming from IEEs on this genetic variation could influence plasticity in response to the IEE (i.e. influence the level of specialization to prey from a particular host plant). Previous studies have found genetic variation influences the transmission and magnitude of IEEs (Rice & Wilde 1989; Whitham et al. 1999; Birch et al. 1999; Holton et al. 2003), or detected the effects of genetic variation in species that act as the intermediate species (Pruter & Zebitz 1991; Cronin & Abrahamson 1999; Hufbauer & Via 1999). Our study documents the first such evidence that individuals of a species subjected to IEEs may vary in phenotypic responses depending on their specific genotype. This supports the position that IEEs can play an important role in the evolution of species (Miller & Travis 1996; Neuhauser et al. 2003) and that the community in which a population exists can have an important impact on the evolution of that population.