Introduction
In many organisms, the environment experienced in previous generations has a profound impact on contemporary phenotypic expression; in particular, environmental conditions experienced by a mother can determine the phenotype of her offspring (Mousseau, Reference Mousseau1998). In such cases, this so-called maternal effect (Mousseau & Dingle, Reference Mousseau and Dingle1991a, Reference Mousseau, Dingle and Dudleyb; Fox et al., Reference Fox, Waddell and Mousseau1995; Mousseau & Fox, Reference Mousseau and Fox1998; Hunter, Reference Hunter2002), can evolve as adaptive transgenerational phenotypic plasticity (Mousseau & Dingle, Reference Mousseau and Dingle1991a; Mousseau, Reference Mousseau1998; Mousseau & Fox, Reference Mousseau and Fox1998; Hunter, Reference Hunter2002), whereby the mother can programme a developmental switch in her offspring which tunes them to a future environment predicted by cues such as humidity, temperature, food quality, presence of natural enemies, population density and photoperiod (Fox et al., Reference Fox, Waddell and Mousseau1995; Fox & Mousseau, Reference Fox, Mousseau and Mousseau1998; Hunter, Reference Hunter2002).
Diapause and wing polymorphism are among the best studied examples of maternal effects in insects and have been demonstrated to be adaptive and likely to be adaptive, respectively (reviewed by Tauber et al., Reference Tauber, Tauber and Masaki1986; Mousseau & Dingle, Reference Mousseau and Dingle1991a, Reference Mousseau, Dingle and Dudleyb; Fox & Mousseau, Reference Fox, Mousseau and Mousseau1998). Among insects, aphids are a group of particular interest when studying transgenerational adaptive maternal effects on account of their telescopic reproduction, where the parthenogenetic female carries within herself three generations of embryos (Dixon, Reference Dixon1998) and, thus, her transgenerational influence encompasses at least three generations (Mousseau & Dingle, Reference Mousseau and Dingle1991a). Maternal influence is among the mechanisms known to trigger the production of winged morphs on parthenogenetic females (reviewed by Braendle et al., Reference Braendle, Davis, Brisson and Stern2006); and both maternal host-plant condition and crowding may also induce progeny to enter diapause (reviewed by Tauber et al., Reference Tauber, Tauber and Masaki1986). The host plant, in particular, is an important environmental variable since it represents a spatially and temporally heterogeneous resource in terms of quality and availability (Fox et al., Reference Fox, Waddell and Mousseau1995); for example, both poor and high quality hosts may cause inhibition of wing induction (Müller et al., Reference Müller, Williams and Hardie2001). Previous laboratory studies on the performance of aphids which have experienced host transfer have shown that some aphid clones show a low initial performance on the newly colonized host which improves through generations (Mackenzie & Guldemond, Reference Mackenzie, Guldemond, Leather, Watt, Mills and Walters1994; Via, Reference Via1991; De Barro et al., Reference De Barro, Sherratt, David and Maclean1995; Douglas, Reference Douglas1997; Caballero et al., Reference Caballero, Ramírez and Niemeyer2001; Gorur et al., Reference Gorur, Lomonaco and Mackenzie2005). Some authors have suggested that the cause of this performance improvement is a weakening of maternal effects (influence of the rearing host) (De Barro et al., Reference De Barro, Sherratt, David and Maclean1995; Douglas, Reference Douglas1997; Caballero et al., Reference Caballero, Ramírez and Niemeyer2001), and some claim that genetic differences related to host adaptation play a major role (Via, Reference Via1991; Mackenzie & Guldemond, Reference Mackenzie, Guldemond, Leather, Watt, Mills and Walters1994; De Barro et al., Reference De Barro, Sherratt, David and Maclean1995).
Rearing experience has been shown to affect selection abilities of insects (Szentesi & Jermy, Reference Szentesi, Jermy and Bernays1990). For example, data has been presented in support of generalist and specialist aphid taxa differing in their host selection speed and efficiency, with specialists taking faster and more efficient decisions than generalists, when reared and tested on their respective host plants (Bernays & Funk, Reference Bernays and Funk1999; Funk & Bernays, Reference Funk and Bernays2001; Vargas et al., Reference Vargas, Troncoso, Tapia, Olivares-Donoso and Niemeyer2005). However, this faster selection by specialists did not occur when aphids were evaluated on hosts different from their rearing hosts (Tosh et al., Reference Tosh, Powell and Hardie2003; Troncoso et al., Reference Troncoso, Vargas, Tapia, Olivares-Donoso and Niemeyer2005). These results suggest that the extent of maternal effects as well as their transgenerational influence is related to the degree of host specialization. In this work, we evaluated whether the performance and developmental instability of a generalist and a specialist aphid, reared on optimal and suboptimal hosts (as sources of different maternal background, from now on referred to as rearing hosts) and then transferred to optimal hosts (as sources of different degrees of stress), experienced variations along four successive generations. The aphids chosen for this study were the green peach aphid, Myzus persicae s.s. (Blackman & Eastop) (Hemiptera: Aphididae), a generalist aphid that uses peach, Prunus persica L. (Rosaceae) as its primary host and more than 400 plant species from different families as secondary hosts (Blackman & Eastop, Reference Blackman and Eastop2000), and a form of M. persicae s.s. associated to tobacco (Brain, Reference Brain1940), with distinctive biological and genetic traits which allow it to be treated as the subspecies, M. persicae nicotianae (Blackman & Eastop) (Blackman & Eastop, Reference Blackman, Eastop, van Emden and Harrington2006). We expect that: (i) the performance of both taxa on optimal hosts will be lower when transferred from suboptimal host than from optimal host, and the recovery of performance levels will be different in the generalist than in the specialist; and (ii) developmental instability will be greater when aphids are transferred from the suboptimal rearing host to the optimal test host, will be greater in the specialist than in the generalist, and will decrease along the four successive generations.
Materials and methods
Insects and plants
Aphids of both taxa were collected from parthenogenetic populations in agricultural fields in the neighbourhood of Talca, Chile (M. persicae s.s. from sugarbeet Beta vulgaris L., and M. persicae nicotianae from tobacco Nicotiana tabacum L., cv. BY 64). Aphid taxa were separated on the basis of morphometrical multivariate analysis; clonal characterization of the colonies was achieved with microsatellite markers (Fuentes-Contreras et al., Reference Fuentes-Contreras, Figueroa, Reyes, Briones and Niemeyer2004). A colony of the only clone of M. persicae nicotianae present in Chile (Fuentes-Contreras et al., Reference Fuentes-Contreras, Figueroa, Reyes, Briones and Niemeyer2004), and a polyclonal colony of M. persicae s.s. were established in the laboratory. Optimal and suboptimal hosts were chosen after a study of the performance (r m) of the two taxa on different hosts. Cabbage was designated as the suboptimal host for both the generalist and the specialist, and sweet pepper and tobacco the optimal hosts for these taxa, respectively. Colonies were maintained at 20±2°C, and 16L:8D photoperiod for more than ten parthenogenetic generations before the beginning of the experiments. Plants of cabbage, sweet pepper and tobacco were grown under the same conditions as aphids.
Determination of performance along successive generations
Aphids from the established colonies (M. persicae s.s. reared on cabbage and sweet pepper, M. persicae nicotianae reared on cabbage and tobacco) were transferred to their respective optimal host and studied over four generations. Four host transfer treatments were established: (i) transfer of M. persicae s.s. from optimal to optimal host (Mp O→O); (ii) transfer of M. persicae s.s. from suboptimal to optimal host (Mp SO→O); (iii) transfer of M. persicae nicotianae from optimal to optimal host (Mpn O→O); and (iv) transfer of M. persicae nicotianae from suboptimal to optimal host (Mpn SO→O). For each treatment, three adult apterae were placed on plants at the 4- to 6-leaf stage in tobacco or at the 14- to 16-leaf stage in sweet pepper. The aphids were allowed to reproduce for 24 h and then adults were removed. Five of the newly born nymphs on each plant (first generation, G 1) were left undisturbed for a fixed number of days for each taxon (four days for M. persicae s.s. and five days for M. persicae nicotianae) and the rest removed. After this time, one nymph per plant was haphazardly selected to be monitored and the other four were removed. Survival and fecundity were evaluated daily. The number of days from birth to first reproduction (T) and fecundity of individual females in the subsequent T days were used to calculate the intrinsic rate of natural increase (r m) according to Wyatt & White (Reference Wyatt and White1977). Starting the day in which each adult aphid began to reproduce, the progeny was counted and removed daily with the exception of nymphs produced on the third reproductive day, which were not removed. Five of the nymphs born in the third reproductive day (G 2) were left undisturbed for a fixed number of days for each taxon (see above) and the rest removed. After this time, one nymph per plant was haphazardly selected to be monitored and the other four were removed. Survival and fecundity were evaluated daily. Starting the day in which each adult aphid began to reproduce, the progeny was counted and removed daily with the exception of nymphs produced on the third reproductive day, which were not removed. They constituted the third generation (G 3) and were reared using the protocol described above. The same protocol was used for G 4. For all four generations the same plant was used in order to avoid the stress produced to the nymphs by the change of plant to measure their performance. The experiment began with 30 replicates for each treatment, but the number decreased as the generations progressed.
Two-way ANOVA with repeated measures was used to evaluate the effects on r m of aphid taxon (generalist vs. specialist), rearing host (optimal vs. suboptimal), and generation (first to fourth, corresponding to the repeated measures). Post-hoc comparisons were performed using Tukey HSD test modified for unbalanced samples (StatSoft, Reference StatSoft2006). Student t-tests were used to compare the r m of the first generation of aphids transferred from suboptimal to optimal hosts with the respective r m on the rearing suboptimal host. Pearson product moment correlations were used to evaluate the significance of eventual relationships between r m and generation. Data passed the tests of parametric analysis assumptions for each procedure and all analyses were performed with STATISTICA 6 (StatSoft, Reference StatSoft2001). To analyse differences in aphid mortality during the period of study, the total number of remaining replicates within treatments were counted after each generation, and rates of loss of replicates were compared using a 2×2 factorial ANCOVA (Lowry, Reference Lowry2006).
Assessment of developmental stability
Adults (age=2T) whose performance had been evaluated were individually stored in 70% ethanol and later mounted on slides as described by Blackman & Eastop (Reference Blackman and Eastop2000). The length of the third antennal segments and the siphunculi (abdominal tubes) were measured according to Ilharco & van Harten (Reference Ilharco, van Harten, Minks and Harrewijn1987), with a binocular microscope (magnification 40×). These traits were chosen based on previous studies in which they showed sensitivity to stressing conditions, as expressed in significant fluctuating asymmetry values (Gorur, Reference Gorur2004; Liu et al., Reference Liu, Zhai, Zhang and Zong2005). In particular, Liu et al. (Reference Liu, Zhai, Zhang and Zong2005) indicated that the third antennal segment is one of the parameters most sensitive to host variations and hence an ideal trait, likely due to its function in host-plant volatile detection. Both traits were measured three times and then evaluated for measurement error, antisymmetry, directional asymmetry, and trait-size dependence of fluctuating asymmetry values, following procedures suggested by Palmer (Reference Palmer and Markow1994).
Fluctuating asymmetry values (FA) were estimated using the index FA5, 
, where L and R represent the length of the left and right traits, respectively (Palmer & Strobeck, Reference Palmer and Strobeck1986; Palmer, Reference Palmer and Markow1994). Two-way ANOVA with repeated measures was used to compare differences in FA values of each taxon on different host transfers through generations.
Morphological abnormalities observed were quantified in each treatment. Frequencies of abnormalities in generalist and specialist aphids were compared with a Fisher exact test (Zar, Reference Zar1996).
Results
The results revealed significant overall effects of taxon, rearing host, and generation on the intrinsic rate of natural increase (r m). Most importantly, these factors showed a significant interaction (table 1), indicating that the performance of the generalist and specialist aphids after host transfer depends on their original provenance (optimal or suboptimal hosts), and that the maternal background operates differentially along generations (fig. 1). The only biologically meaningful post-hoc comparison which was significant occurred in the first generation of M. persicae s.s. between host transfer treatments (P=0.013, fig. 1). Additionally, when comparing the r m value of the first generation of the generalist and specialist aphids from suboptimal rearing host transferred to optimal hosts, with r m values on the respective suboptimal hosts, the generalist showed a clearly significant increase (t=−5.49, P<0.001) while the specialist did not (t=−2.02, P=0.056). The latter P value was considered non-significant because no further tendency to increased performance was observed along generations.
Fig. 1. Intrinsic rates of natural increase (r m) of Myzus persicae s.s. reared on the optimal (●) and suboptimal (○) hosts, and of M. persicae nicotianae reared on the optimal (▲) and suboptimal (△) hosts, and then tested on the optimal host over four generations. The letters indicate significant differences between host transfer treatments of M. persicae s.s. at the first generation.
Table 1. Summary of two-way ANOVA with repeated measures performed to evaluate the effect of maternal effect on the intrinsic rate of increase between taxa and across generations.
The values of r m for the generalist aphid from the suboptimal rearing host increased significantly along generations (r=0.209, d.f.=91, P=0.04); however, there was a significant decrease of r m in the generalist aphid from the optimal rearing host (r=−0.384, d.f.=99, P<0.01). No significant increase or decrease was found for the specialist either from the optimal or suboptimal rearing hosts (r=0.028, d.f.=63, P=0.83, and r=−0.101, d.f.=51, P=0.48, respectively).
Rates of loss of replicates along generations were significantly different between taxa as well as between host transfer treatments (table 2, fig. 2).
Fig. 2. Rate of replicate loss by the end of each generation of Myzus persicae s.s. reared on the optimal (●) and suboptimal (○) hosts, and of M. persicae nicotianae reared on the optimal (▲) and suboptimal (△) hosts, and then tested on the optimal host over four generations. Bold and thin regression lines correspond to ‘optimal to optimal’ and ‘suboptimal to optimal’ host transfer treatments, respectively; continuous and dashed regression lines correspond to the generalist and specialist taxon, respectively. Regression lines followed by a different letter differ at P<0.05 (table 2).
Table 2. Summary of 2×2 factorial ANCOVA performed to compare the rates of replicate loss between treatments accross generations (fig. 2).
No significant effects were found within any of the treatments for measurement error, antisymmetry, directional asymmetry, and trait-size dependence of fluctuating asymmetry values. Taxon but not host transfer nor generation significantly affected fluctuating asymmetry values of siphunculi (table 3). FA values were significantly higher for M. persicae nicotianae (table 4). On the other hand, none of the factors studied significantly affected the FA values for third antennal segments (table 3). Finally, body abnormalities consisting of atrophied structures, antennal protuberances, and reduction of body size (dwarfism) occurred with low frequencies and without significant differences between generalist and specialist (Fisher exact test, P=0.30).
Table 3. Summary of results of two-way ANOVA with repeated measures for fluctuating asymmetry values (FA5).
Table 4. Values of fluctuating asymmetry for trait, aphid taxon (Myzus persicae s.s., Mp; M. persicae nicotianae, Mpn), and rearing host (O, optimal; SO, suboptimal) through generations G1 thorough G4. Fluctuating asymmetry values (FA, in mm2) were calculated using the formula: (see Materials and methods).
Discussion
In the present study, the effects of host transfer treatments on aphid performance differed between the generalist and specialist aphids. In the optimal to optimal host transfer, r m decreased along generations in the generalist on sweet pepper but not in the specialist on tobacco. The decreased performance of the generalist may be attributed to the accumulation of induced chemical defences in the plant arising from its interaction with the aphid; for instance, feeding by M. persicae s.s. has been shown to give rise to induced defences in sweet pepper (Tapia, Reference Tapia2006), and by M. persicae nicotianae to induced chemical defences in wild tobacco (Heidel & Baldwin, Reference Heidel and Baldwin2004). The insensitivity of the performance of M. persicae nicotianae to induced defences in tobacco may be attributed to the enzymatic detoxifying system it possesses (Cabrera et al., Reference Cabrera, Fuentes-Contreras and Figueroa2005).
The generalist aphid reared on the suboptimal host was capable of progressively achieving higher performances after being transferred to the optimal host, r m eventually becoming indistinguishable from that of an aphid which had been reared and transferred to the optimal host. In fact, an increase was already apparent when the performance of the first generation after transfer from the suboptimal to the optimal host was compared to the performance on the suboptimal rearing host, thus showing a fast recovery of the generalist after host transfer. On the other hand, the specialist reared on the suboptimal host was incapable of increasing its performance upon transfer to the optimal host, at least within four generations after host transfer. Furthermore, loss of replicates – equivalent to aphid mortality – along generations was higher in the specialist than in the generalist, and loss was highest in the suboptimal to optimal host transfers. These phenomena may arise from differential prevalence of maternal effects and phenotypic plasticity in the generalist and in the specialist. Thus, the generalist achieved higher levels of performance by progressively overcoming its maternal effects, while the specialist seemed unable to do it, which could also be taken as a commitment to its rearing host, since it lasted for more than three generations. The maintenance of host commitment normally depends on a variety of traits, such as behavioural, physiological and/or developmental/morphological which show different levels of plasticity (Pigliucci, Reference Pigliucci2001), and also different evolutionary rates (Gittleman et al., Reference Gittleman, Anderson, Kot, Luh and Martins1996) In the present case, the steady maintenance of performance along generations in the specialist, suggest a commitment to the rearing host (supported by its maternal effect), and not to an ecological specialization history. The commitment of the specialist to its rearing host (in our case, cabbage) may be the result of the loss or inactivation of certain costly traits which were once adaptive to its life on tobacco, but were no longer necessary for thriving on cabbage. Recent evidence points to the importance of esterases in the detoxification of tobacco allelochemicals by M. persicae nicotianae (Cabrera et al., Reference Cabrera, Fuentes-Contreras and Figueroa2005). The re-installment of this detoxification capacity through re-allocation of the energy budget may require more than three generations.
Developmental instability was taxon and trait dependent, in agreement with correlational studies by Clarke (Reference Clarke1998). Thus, the specialist showed higher developmental instability of siphunculi (but not of third antennal segment) than the generalist. This, and the fact that the specialist showed lower performance (fig. 1) and higher mortality than the generalist (fig. 2) is not in disagreement with the proposal that fitness is lower in developmentally instable individuals (Møller, Reference Møller1997; Leamy & Klingenberg, Reference Leamy and Klingenberg2005). Since the effect of host transfer was negligible, such developmental instability can not be related to stress, as previously suggested by other authors (Palmer, Reference Palmer and Markow1994; Crespi & Vanderkist, Reference Crespi and Vanderkist1997).
The predictability of local environmental quality is an important criterion in the evolution of adaptive parental effects (Rossiter, Reference Rossiter1996, Reference Rossiter and Mousseau1998). Given the restricted host range of a specialist aphid, the adaptive value of maternal effects when colonizing highly predictable hosts is to be expected. When offspring face an environment different from the one experienced by the mother (as in our experimental manipulation), maternal effects would lose their adaptive value through a teleonomic phenotypic influence on such offspring and become a ballast for further performance improvement. On the other hand, a generalist aphid is capable of thriving on a wide range of hosts and thus may be adapted to unpredictable environments; hence, the adaptive value of maternal effects on host preference is questionable. When offspring face an environment different from the one experienced by the mother, performance will be mainly a result of its phenotypic flexibility (sensuPiersma & Drent, Reference Piersma and Drent2003) offsetting the eventual maternal effect. These considerations suggest that maternal effects should be discussed within the context of ecological specialization.
Acknowledgements
The authors wish to thank Ms Marcela Cordero for technical assistance and Francisca Zepeda and Christian C. Figueroa for genotyping the aphid colonies. This work was supported by FONDECYT (grant 1020561), and PBCT (Anillo ACT38). The authors are indebted to two anonymous reviewers whose comments and suggestions provided an opportunity for greatly improving the original manuscript.
