SEARCH

SEARCH BY CITATION

Keywords:

  • clonal diversity;
  • clonal selection;
  • host plant;
  • microsatellites;
  • Myzus persicae;
  • parthenogenesis

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Parthenogenetic organisms often harbour substantial genotypic diversity. This diversity may be the result of recurrent formations of new clones, or it may be maintained by environmental heterogeneity acting on ecological differences among clones. In aphids, both processes may be important because obligate and cyclical parthenogens can form mixed populations. Using microsatellites, I analysed the temporal dynamics of clonal diversity in such a population of the aphid Myzus persicae over a 1-year period. The frequency distribution of clonal genotypes was very skewed, with many rare and few common clones. The relative frequencies of common clones underwent strong and rapid changes indicative of intense clonal selection. Differences in their host associations suggest that these shifts may partly be caused by changes in the abundance of annual host plants. Other selective factors of potential importance are also discussed. New, sexually produced genotypes made a minor contribution to clonal diversity, consistent with the observed heterozygote excess characteristic of predominantly asexual populations in M. persicae.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Populations of asexual organisms often harbour substantial genotypic diversity (Bell, 1982; Hughes, 1989). This raises the question of how this diversity is maintained. If the observed genotypic diversity were largely neutral, one would expect a continuous decline of diversity due to the stochastic loss of clones. If the observed diversity was not neutral, competitive exclusion may erode clonal diversity even more rapidly. However, clonal diversity can be maintained by selection if temporal and/or spatial variation in the environment leads to shifts in the relative fitnesses of clones (Vrijenhoek, 1979; Maynard Smith & Hoekstra, 1980; Weeks & Hoffmann, 1998), or if selection acts in a negative frequency-dependent manner, for example if common clones are affected disproportionately by parasites (e.g. Dybdahl & Lively, 1998). Clonal diversity will also be high if new clones continuously originate from otherwise sexual species (see Simon et al., 2003, for a review of the different mechanism leading to asexuality). Then clonal diversity may reflect the balance between the influx of clones and their elimination through drift and/or selection. That clonal diversity can erode quickly has been shown in cyclical parthenogens like rotifers or aphids, in which a single generation of sexual reproduction is followed by a large number of parthenogenetic generations before the next sexual generation. The burst of clonal variation immediately after a sexual generation is lost rapidly during the asexual phase of the life cycle (Rhomberg et al., 1985; Gómez & Carvalho, 2000). Typically, it is assumed that selection at least contributes to this loss, which is supported by studies demonstrating that clones of cyclical parthenogens that are distinguishable by molecular markers also differ significantly in ecologically relevant traits (Weider, 1993; Epp, 1996; Sunnucks et al., 1998; Turak et al., 1998; Vorburger et al., 2003b; Vorburger, 2004).

In many aphid species, the dynamics of clonal diversity are particularly complex because cyclical as well as obligate parthenogens occur in mixed populations. Well-documented cases include Rhopalosiphum padi L. in France (Simon et al., 1991, 1996; Delmotte et al., 2002), Sitobion avenae (Fabricius) in France and the UK (Simon et al., 1999; Dedryver et al., 2001; Llewellyn et al., 2003), or Myzus persicae (Sulzer) in Europe and Australia (Margaritopoulos et al., 2002; Guillemaud et al., 2003; Vorburger et al., 2003a). Several studies on aphid populations came to the conclusion that selection has a strong influence on their genotypic composition (e.g. Sunnucks et al., 1997; Guillemaud et al., 2003; Llewellyn et al., 2004; Zamoum et al., 2005). An important ecological difference between obligate and cyclical parthenogens in aphids is that only cyclical parthenogens can produce cold-resistant eggs during their sexual generation in autumn, while obligate parthenogens are viviparous throughout the year. As a consequence, cold winters can select for sex (Simon et al., 2002). At a global scale, this is evident from a decrease of obligate parthenogens towards extreme latitudes, as illustrated by Blackman (1974) for M. persicae. Typically, the tropics and subtropics harbour only obligate parthenogens and cold climates only cyclical parthenogens, while both reproductive modes occur in temperate climate zones because their coexistence is facilitated by temporal variation in winter severity (Rispe & Pierre, 1998; Rispe et al., 1998).

Vorburger et al. (2003a) investigated the clonal diversity and the distribution of reproductive modes in the aphid M. persicae in Victoria, Southeastern Australia. They found that even on a small geographic scale, the relative frequency of reproductive modes was affected by winter severity, but also by the availability of M. persicae's primary host, peach, that is required by cyclical parthenogens for egg laying. Cyclical parthenogenesis was more prevalent in areas with colder winters and a larger amount of commercial peach farming. Clonal diversity increased with the proportion of cyclical parthenogens in a sample, suggesting that the influx of new clones from sexual reproduction can make an important contribution to the observed clonal diversity at a given site. The same study also revealed the presence of two extremely common and widespread obligately parthenogenetic clones that occurred in the majority of all samples and together comprised more than 40% of the individuals collected (Vorburger et al., 2003a). Such ‘superclones’ have been detected in a number of aphid species (Sunnucks et al., 1996; Fenton et al., 1998; Fuller et al., 1999; Wilson et al., 1999; Haack et al., 2000; Llewellyn et al., 2003), yet the reasons for their success remain elusive. A problem with the study by Vorburger et al. (2003a) is that it only represents a snapshot in time, taken in early autumn 2002. Whether or not it was representative depends on how stable clonal frequencies are in time. The original survey did not reveal, for example, whether the dominant clones occurred at high frequency throughout the year or built up their dominance during the growth season. The high resolution of microsatellites and the increasing ease and cost-efficiency of their use make it possible to track the frequencies of individual clones through time. Thus I selected one of the sampling sites described in Vorburger et al. (2003a), and collected samples from secondary hosts at monthly intervals during the course of 1 year to describe the temporal dynamics of the clonal composition of M. persicae. I show that a skewed frequency distribution of clonal genotypes also characterizes this specific site, but that the relative frequencies of common clones undergo strong and rapid changes indicative of intense clonal selection. I provide evidence suggesting that these fluctuations may partly be caused by temporal changes in the abundance of annual host plants, yet other factors that may maintain clonal diversity via fluctuating selection are also discussed. Finally, the data suggest that although the study site is dominated by obligate parthenogens, new, sexually produced genotypes also make some (albeit minor) contribution to clonal diversity.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Study species

The peach-potato aphid, M. persicae, is presumed to be of Asian origin, but it is globally distributed today and represents an economically important pest (Blackman & Eastop, 2000). The earliest records of this species from Australia date back to late in the 19th century (Wilson et al., 2002). Cyclical parthenogens of M. persicae are host alternating. In autumn, they mate and lay eggs on their primary host, peach (Prunus persica), where the eggs undergo diapause over winter. In spring, viviparous parthenogenetic females hatch from these eggs and their clonal descendants disperse to secondary hosts, including plants from over 40 different families (Blackman & Eastop, 2000). There they produce many parthenogenetic generations before returning to peach in autumn. However, many genotypes in this species have lost the ability to reproduce sexually and live permanently on secondary hosts (obligate parthenogens).

Sampling and molecular methods

The sampling site was a vegetable farm producing Broccoli (Brassica oleracea), located in Bacchus Marsh, approximately 50 km west of Melbourne, Australia. Broccoli is produced throughout the year at this farm hence plants of all growth stages are available continuously, yet at changing locations. The farm is situated within a larger area of vegetable and fruit production, including some peach orchards. Despite the local availability of the primary host, a previous study found that the population of M. persicae at Bacchus Marsh consisted predominantly of obligate parthenogens (Vorburger et al., 2003a).

Sampling started in October 2002 (austral spring), i.e. about the time when sexually produced M. persicae migrate from their primary host peach to secondary hosts, and was repeated at monthly intervals until September 2003. Based on the development time of M. persicae under different temperatures (Liu & Meng, 1999), one month should correspond to 1–2 generations during the coldest and 4–5 generations during the warmest time of the year at Bacchus Marsh. The exact date and some descriptive statistics for each sample are given in Table 1. I sampled within an area of about 8 Ha by haphazardly screening plants for the presence of M. persicae and clipping infested leaves. The main crop (Broccoli) was examined as well as about a dozen species of weeds that are suitable secondary hosts for M. persicae, the most common being Chenopodium album, Hirschfeldia incana, Solanum physalifolium, and three closely related species of mallows (Malva neglecta, M. nicaeensis and M. parviflora, hereafter mallows). When an infested plant was found, no other plants were examined within a radius of 2 m from that plant. The number of samples collected per person hour was used as a rough estimate of the density of M. persicae for each sampling date.

Table 1.  Summary statistics for the 12 samples of M. persicae collected at Bacchus Marsh, including collection date, sample size (n), the number of different MLGs (No. of genotypes) and FIS estimated from the complete samples and from samples reduced to one copy of each genotype.
DatenNo. of genotypesFIS (clonal copies included)FIS (clonal copies excluded)
  1. *Indicates significant deviations from Hardy–Weinberg equilibrium at the 5% level after sequential Bonferroni correction for multiple tests (Rice, 1989).

14.10.20022414−0.128*−0.068
13.11.20022810−0.244*−0.181
12.12.20023310−0.374*−0.243*
15.01.20032715−0.161*−0.072
17.02.20033822−0.114*0.018
13.03.20033614−0.279*−0.090
22.04.20033012−0.326*−0.063
19.05.2003359−0.315*−0.067
18.06.2003277−0.375*−0.103
18.07.2003247−0.392*−0.138
15.08.20032810−0.267*−0.101
15.09.2003358−0.417*−0.134

The aphids were taken back to the laboratory alive. If several adults were present in a sample (most samples contained only one adult), a single adult was chosen haphazardly for genotyping. After recording its morph (apterous or alate), the DNA of this adult was extracted following the salting-out protocol described in Sunnucks & Hales (1996), and resuspended in 200 μL of 0.1 × TE (0.1 mm EDTA, 1 mm Tris base, pH 7.5). Each individual was genotyped at seven microsatellite loci, M37, M40, M63, M86, M107, myz2 and myz9 (Sloane et al., 2001; Wilson et al., 2004). Locus M86 is X-linked; all other loci are autosomal (Sloane et al., 2001). Conditions for PCR and visualization of products were as described in Vorburger et al. (2003a).

Data analyses

Individuals with the same multilocus genotype (MLG) can be members of the same clone or else arise by chance from independent instances of sexual reproduction. I used the programme MLGSIM by Stenberg et al. (2003) to calculate the probability of the latter. In a first step, this programme calculates Psex for genotypes that are found more than once, i.e. the probability that these genotypes occur the observed number of times in a sexual population with the observed allele frequencies, assuming Hardy–Weinberg and linkage equilibria. In a second step, the programme uses Monte Carlo simulation to obtain the critical values of Psex for the desired significance level (here 0.05), thus identifying the genotypes that are significantly overrepresented and therefore likely to be members of the same clone (Stenberg et al., 2003). Unfortunately, MLGSIM can only handle samples of up to 200 individuals, less than the total sample size in this study. To still be able to use the method, I calculated the expected number of representatives (rounding to the nearest integer) of multicopy MLGs in a sample of 200 with the same relative frequencies as the original, larger sample. This yielded 16 MLGs with an expected frequency of two or more (in 200), for which Psex could be calculated.

To obtain an estimate of the total number of MLGs in the population from the samples, I used the programme ESTIMATES (Colwell, 2005) to calculate the Chao1 richness estimator along with its 95% confidence intervals (Chao, 1987).

The genetic composition of an aphid population on secondary hosts is determined by the gain and loss of genotypes through immigration and emigration/extinction, but also by the clonal amplification of genotypes. To determine whether these processes lead to significant temporal changes in the genetic structure of our study population, we calculated global and pairwise FST (Weir & Cockerham, 1984) across loci between the 12 samples, using FSTAT Version 2.9.3 (Goudet, 2001). The ISOLDE programme implemented in GENEPOP Version 3.4 (Raymond & Rousset, 1995) was used to perform a Mantel test of independence between the matrix of pairwise FST and the temporal distance matrix (in days) between samples to test for ‘isolation by time’ (Guillemaud et al., 2003). These analyses were performed twice, once with a dataset reduced to one copy per genotype and sample, testing for temporal changes from the gain and loss of genotypes, and once with the complete dataset, testing for the additional effect of clonal amplification.

Clonal reproduction with the amplification of identical genotypes is expected to lead to heterozygote excess (Balloux et al., 2003), but also when population samples are reduced to just a single copy per genotype, excess heterozygosity is often detected in predominantly asexual populations of aphids (Delmotte et al., 2002; Vorburger et al., 2003a), apparently because obligate parthenogens exhibit higher heterozygosity than cyclical parthenogens (Vorburger et al., 2003a). I therefore used the randomization procedure implemented in FSTAT to test for deviations from Hardy–Weinberg equilibrium within samples, again using the complete dataset as well as the one with clonal copies excluded.

To document the temporal changes in clonal diversity, the Shannon-Wiener (S-W) index was calculated for each date, as well as the G/n ratio (Llewellyn et al., 2003), i.e. the number of different MLGs divided by the sample size. For the most common MLGs, the proportions they comprised of each sample were plotted against sampling date. Exact binomial confidence intervals for the observed proportions were calculated using a Microsoft EXCEL macro written by John P. Pezzullo, available at http://members.aol.com/johnp71/confint.html. To test whether temporal changes in the frequencies of the most common clones were larger than expected by sampling error only, I used Fisher's exact tests to directly compare their frequencies vs. that of all other genotypes between subsequent peaks and valleys of their temporal trajectories. From the Australian Bureau of Meteorology I obtained climatic data for the sampling period, i.e. monthly mean maximal and minimal temperatures recorded at the closest automatic weather station (Sheoaks), as well as the monthly rainfall recorded at the closest rain station (Bacchus Marsh Golf Club). These data were compared with the overall clonal diversity and the relative frequencies of the most common MLGs using Pearson's product-moment correlation.

Finally, I assessed whether M. persicae collected from different hosts represent genetically differentiated subpopulations or one homogeneous population by pooling samples across dates and calculating multilocus FST among hosts. To test for differentiation among hosts I used randomizations of MLGs among hosts and the G-statistic as implemented in FSTAT and recommended by Goudet et al. (1996). Again, these analyses were performed once with a dataset reduced to one copy of each MLG per host and once with the complete dataset, in order to ascertain whether possible differentiation is mainly the result of different genotypes colonizing different hosts or of differences in the relative frequencies of genotypes on different hosts. However, due to temporal changes in their availability, the effect of host plant is confounded with temporal changes that may have occurred for other reasons. Fortunately, sample sizes were large enough for the two most common clones to test for differences in host association after the effect of date was controlled for (see Results for the details of this analysis).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Microsatellite variability

All microsatellite loci were polymorphic in all 12 samples. The total number of alleles per locus ranged between three (M37) and ten (M63 and myz9), theoretically allowing the distinction of 13.45 × 109 MLGs. Among the 365 individuals that were genotyped, a total of 72 different MLGs could be identified. Their frequency distribution was highly skewed. Forty-nine MLGs were collected only once, 16 MLGs between two and four times, five MLGs between 6 and 12 times, and the remaining two genotypes comprised nearly two-third of the entire sample. MLG 58 was collected 148 times (40.5%) and MLG 61 was collected 78 times (21.4%). The total number of MLGs in the population was estimated as 244 (95% C.I. = 142; 491) with the Chao1 estimator. Values of Psex obtained with MLGSIM ranged between 10−11 and 10−15 and were all highly significant, suggesting that individuals with the same MLG are indeed members of the same clone. All microsatellite genotypes and the exact composition of each sample are provided in the appendix.

Deviations from Hardy–Weinberg equilibrium with excess heterozygosity are expected in populations with predominantly clonal reproduction (Balloux et al., 2003). Accordingly, all samples exhibited significant heterozygote excess when all individuals were included (Table 1). When samples were reduced to once copy per MLG, all but one FIS value were still negative, but due to reduced sample sizes the deviation from Hardy–Weinberg equilibrium was only significant for the December sample (Table 1).

Temporal dynamics

The numbers of samples collected per person hour provide a rough estimate of the density of M. persicae for each sampling date (Fig. 1a). Numbers at Bacchus Marsh were high in spring when sampling started, but then declined during peak summer. By February (late summer), numbers had recovered and stayed high until May (late autumn), followed by a second decline during peak winter, from which the population had again recovered by the last collection date in September (early Spring). The proportions of alate individuals in the monthly samples ranged between 0 and 0.31 and followed a similar trajectory as the density, such that higher proportions of alates tended to be associated with higher densities (Fig. 1a), although this relationship was not statistically significant (r = 0.531, P =0.076). The exact dates and sample sizes for the 12 months are presented with some summary statistics in Table 1.

image

Figure 1. Temporal variation of (a) density estimated as the number of samples collected per person hour (solid circles) and the proportion of collected individuals that were winged (alates; open circles), and (b) clonal diversity estimated as the S-W index (solid circles) and the G/n ratio (open circles) for the study population of M. persicae. For sample sizes see Table 1. Panel (c) shows the mean monthly maximum (filled circles) and minimum (open circles) temperatures for the sampling period, recorded at the closest automatic weather station (Sheoaks), as well as the monthly rainfall (stars) recorded at the closest rain station (Bacchus Marsh Golf Club).

Download figure to PowerPoint

The temporal changes in clonal diversity expressed as the S-W index and as the G/n ratio are illustrated in Fig. 1b. The two measures provide a very similar pattern. Clonal diversity was high in the very first sample (October 2002), but lower in the next 2 months that were followed by another increase peaking in February 2003. After that, clonal diversity declined almost continuously to reach the lowest levels toward the end of the sampling period, i.e. in late winter/early spring (Fig. 1b). Over the course of the whole year, there was a significant decline in clonal diversity for both estimators (S-W index: r2 = 0.345, F1,10 = 5.268, P = 0.045; G/n ratio: r2 = 0.393, F1,10 = 6.467, P = 0.029). Clonal diversity was unrelated to the proportion of alate individuals in the monthly samples (r = 0.142, P = 0.66, using the S-W index).

Climatic data for the weather stations closest to the sampling site are depicted in Fig. 1c. The trajectories of monthly mean maximal and minimal temperatures are very similar, reaching the highest values in January/February and the lowest in July/August. Rainfall was generally very low during the sampling period, but highest in winter (Fig. 1c). Reflecting this seasonal pattern, rainfall was negatively correlated with temperature (r = −0.808, P < 0.001, using the mean maxima). Consistent with the overall decline in clonal diversity over the year, the S-W index of diversity was positively correlated with temperature (r = 0.624, P = 0.030, again using the mean maxima).

If and how changes in genotypic composition lead to temporal genetic differentiation between samples was assessed by testing for ‘isolation by time’. With each sample reduced to one copy per genotype, global FST was very low (FST = 0.006, P = 0.15), and ‘isolation by time’ was weak, although marginally significant (Fig. 2). With clonal copies included in the samples, global FST was higher and statistically significant (FST = 0.019, P =0.001), and ‘isolation by time’ stronger (Fig. 2), suggesting that changes in the relative frequencies of genotypes have a stronger effect on temporal differentiation than their gain or loss.

image

Figure 2. Isolation by time’ between 12 monthly samples of M. persicae from Bacchus Marsh for samples reduced to one copy per genotype [open circles; FST/(1 − FST) = −0.00085 + 0.00005 × days; Mantel test with 10 000 permutations: P = 0.034], and for the complete samples [solid circles; FST/(1 − FST) = 0.00526 + 0.00012 × days; Mantel test with 10 000 permutations: P = 0.007].

Download figure to PowerPoint

Temporal shifts in the frequencies of the most common clones were indeed strong and rapid. Figure 3a shows the trajectories of the six most common clones across the 12 months. Four of these clones, the four most common ones, had already been detected at the same site in an earlier sample taken in March 2002 for a different study (Vorburger et al., 2003a). Their frequencies in that sample are also depicted in Fig. 3a. These clones are known to reproduce by obligate parthenogenesis because their reproductive modes have previously been determined experimentally (Vorburger et al., 2003a).

image

Figure 3. Twelve-month temporal trajectories of the relative frequencies of (a) the six most common clones of M. persicae at Bacchus Marsh, including the frequencies of four of these clones in an earlier sample collected at the same site, and (b–d) the trajectories of the three most common clones with 95% C.I.

Download figure to PowerPoint

The temporal changes are described in more detail for the three most common clones (Fig. 3b–d): Comprising 29% of the individuals, MLG 45 was the most common clone in the first sample (October). This clone was present at the site already during an earlier collection in March 2002 (Fig. 3a) and at that time was found to be the second most common and widespread clone throughout Victoria (Vorburger et al., 2003a). At Bacchus Marsh, it declined after October 2002 and was never collected again after January 2003 (Fig. 3b). Comparing the October and February samples, this change is significant (Fisher's exact test, P < 0.001). The most common clone averaged across the 12 months was MLG 58. It had already been the most common clone at the site in the earlier sample of March 2002 and was then found to be the most common and widespread clone of M. persicae throughout Victoria (Vorburger et al., 2003a). However, it only comprised about 12% of the first sample collected in October 2003 (Fig. 3c), followed by a steep increase to make up more than half of the following two samples (October vs. December, P = 0.002). The peak in December was followed by a decline to a low of about 17% in April (December vs. April, P = 0.004), after which it increased again to reach 49% of the last sample in September (April vs. September, P = 0.009). The relative frequency of MLG 58 did not correlate significantly with either temperature or rainfall (mean maximum temperature: r = 0.010, P = 0.976; rainfall: r = −0.093, P = 0.774). The second most common clone at Bacchus Marsh was MLG 61. Also this clone had already been present at the site in March 2002 (Fig. 3a). It was present in all 12 of the monthly samples, but only at a low frequency of between 3 and 6% during the first 4 months (Fig. 3d). After January, MLG 61 started to increase and reached a peak of 47% in the April sample (January vs. April, P < 0.001). After April it decreased again slightly but retained substantial frequencies of between 20 and 40% until the end of the sampling period (Fig. 3d). Comparing the April and September samples, this decline was not statistically significant (P = 0.307). The relative frequency of MLG 61 was negatively correlated with the mean maximum temperature (r =−0.629, P = 0.028) and positively correlated with rainfall (r = 0.643, P = 0.024), consistent with its higher abundance during the autumn and winter months.

Although the estimated frequencies are associated with wide confidence intervals (Fig. 3) due to the limited size of the monthly samples, some of the changes are too strong to be caused by drift alone. For example, the strong increase of MLG 61 from January to April 2003 (Fig. 3d) required only about 10 generations in a population that must have been very large even during periods of low abundance. Density estimates for M. persicae in potato fields ranged between 0.5 and 591 individuals per 70 × 70 cm sampling unit (Ro & Long, 1999), corresponding to between about 20 000 and 12 million individuals per hectare, which would only represent a small fraction of the total population size within a horticultural area like that of Bacchus Marsh (approximately 250 Ha). To obtain an appreciation of the effect of clonal drift over the relevant time scale in reasonably large populations, I used the programme EASYPOP (Version 1.8) (Balloux, 2001) to run 100 simulations of 10 (nonoverlapping) generations in populations of size 10 000 and a starting composition of 10 clonal genotypes at a frequency of 10% each. The minimal and maximal frequencies to which a single clone drifted in any of the simulations were 7.1 and 13.1%, respectively, demonstrating that clonal drift would have had very limited power even during peak summer and peak winter, when population density was lowest.

Host plant relations

When samples were pooled across the 12 months, sample sizes for five host plants were large enough to test for genetic differentiation among hosts (Broccoli: 115 individuals/24 MLGs; C. album 23/12: H. incana: 43/10; mallows: 69/21; S. physalifolium: 92/33). With the data set reduced to one copy of each MLG per host, there was no significant genetic differentiation between hosts with a multilocus FST (±SE) of −0.001 ± 0.003 (P = 0.639). When clonal copies were included in the data set, genetic differentiation was still very low but significant (FST = 0.004 ± 0.002, P = 0.001), suggesting that allele frequencies of MLGs found on different hosts do not differ, but that some MLGs are disproportionately represented on specific hosts. Comprising nearly two-third of the population, the two most common clones MLG 58 and MLG 61 must be mainly responsible for this effect. Indeed, compared to its total frequency of 21.4%, MLG 61 was overrepresented on broccoli (33.9%) and H. incana (32.6%), underrepresented on C. album (8.7%) and S. physalifolium (5.4%), and about proportionately represented on mallows (20.3%). MLG 58 (total frequency 40.5%) was only slightly overrepresented on H. incana (44.2%), mallows (44.9%) and S. physalifolium (44.6%) and slightly underrepresented on broccoli (38.3%) and C. album (39.1%). However, these figures are severely confounded by the temporal changes in clonal frequencies and host availability (Fig. 4). For example, the annual weed S. physalifolium appeared to be the most attractive of all hosts for M. persicae and thus yielded many samples, but was only available from spring until late autumn, when MLG 58 was generally much more frequent than MLG 61 (Fig. 4). To disentangle these effects I reduced the data set to just the two most common clones and their four most common hosts (broccoli, H. incana, mallows and S. physalifolium, sample sizes for C. album were too small), and used maximum likelihood logistic regression in PROC GENMOD (SAS Institute, 1996) to model the probability that any given individual was MLG 58, as a function of the month in which the sample was collected and the host plant it came from. Month was fitted first and was of course highly significant (inline image = 34.34, P < 0.001). After month was controlled for, the effect of host plant was fitted and also found to be significant (inline image = 11.63, P = 0.009), indicating that the two most common clones indeed exhibit some differences in host plant association. The month × host interaction was not significant (inline image = 25.25, P = 0.237). Based on inspection of Fig. 4, the main difference appears to be that MLG 58 was more common than expected on S. physalifolium, while MLG 61 was more common on broccoli.

image

Figure 4. Host association of the two most common clones of M. persicae at Bacchus Marsh.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The parthenogenetic population of M. persicae on secondary hosts at the Bacchus Marsh study site is characterized by a high clonal diversity, a very skewed distribution of clonal frequencies (few common and many rare clones) and strong and rapid changes in the relative frequencies of common clones during the course of a year. Drift or selection can cause shifts in the relative frequencies of clones. However, in the large populations typical of aphids, drift only plays a negligible role, arguing for the observed shifts being the result of intense clonal selection. These shifts are mainly responsible for the genetic differentiation among samples taken at different times from the same site, as evident from the observed ‘isolation by time’.

Strong fitness differences among clones and shifts in their relative fitness in response to environmental variation (i.e. genotype-by-environment interactions) are the conditions required for the maintenance of diversity via environmental heterogeneity (Vrijenhoek, 1979; Maynard Smith & Hoekstra, 1980). The actual environmental factors responsible for these fitness differences are currently unknown. It is likely that seasonal changes in environmental conditions lead to shifts in the relative fitness of different clones. For example, Niklasson et al. (2004) report temporal fluctuations of four clones of the thelytokous fly Dipsa bifurcata over an 8-year period that are synchronous among populations and correlated with seasonal changes in the environment. Similarly, Carvalho & Crisp (1987) observed a seasonal succession of abundant clones of the waterflea Daphnia magna over the course of 2 years, and Carvalho (1987) found a close correspondence between their optimal temperature ranges and the water temperatures during the seasons at which they were most abundant. In the present study, the increased abundance of MLG 61 during autumn and winter would be consistent with such a seasonal pattern, but with data from 1 year only it is not possible to pinpoint the relevant selective factors. Nevertheless, the finding that the two most common clones at the Bacchus Marsh study site exhibit different host plant associations hints at one such factor. Although M. persicae is one of the most polyphagous aphids known, genetic variation among clones in their performance on different hosts has repeatedly been demonstrated (Weber, 1985,1986; Edwards, 2001; Vorburger et al., 2003b). As host availability varies in space as well as in time (e.g. annual species), selection for or against clones adapted to particular hosts may vary accordingly. For example, its overrepresentation on S. physalifolium suggests that genotype 58 is well adapted to that host and may be under positive selection during the months of high abundance of this annual weed, but lose this advantage during the months when S. physalifolium is not available.

Other selective factors that are likely to exhibit temporal or spatial variation include temperature, natural enemies (parasitoids or predators), and insecticides.

Temperature has long been identified as one of the most important environmental variables in aphid ecology (Hales et al., 1997), and genetic variation among clones in their temperature tolerance has been detected in M. persicae (Vorburger, 2004). It is thus feasible that the relative fitnesses of different clones undergo shifts with seasonal changes in temperature, similar to those observed in D. magna by Carvalho (1987).

Aphids are attacked by a large number of natural enemies, and it has even been proposed that parasitoids are at least partly responsible for the mid-season population crash that is typical for many aphid species (Karley et al., 2004). Among their natural enemies, hymenopteran parasitoids of the family Aphidiidae probably have the strongest impact on aphid populations (Schmidt et al., 2003). In one aphid species, the pea aphid (Acyrthosiphon pisum), genetic variation for resistance to aphidiidine parasitoids has been demonstrated (Henter & Via, 1995; Hufbauer & Via, 1999; Ferrari et al., 2001; Stacey & Fellowes, 2002), so it is possible that more resistant clones are under positive selection when parasitoids are abundant, but – if resistance incurs a cost – under negative selection when parasitoids are scarce.

Finally, M. persicae has evolved several resistance mechanisms to commonly used insecticides which entail costs in terms of a reduced reproductive success, increased mortality under adverse weather conditions, or maladaptive behaviour (reviewed by Foster et al., 2000). This suggests that spatial or temporal variation in the application of insecticides also has the potential to maintain clones with different levels of resistance in the population. In the present case, however, it is unlikely that the dramatic shifts in clonal frequencies are due to this cause because the main crop at the study site, broccoli, is grown and protected by spraying all around the year. But in an aphid with such a broad spectrum of secondary hosts, the permanent availability of suitable hosts at unsprayed sites is certainly important in preventing the fixation of clones with a high resistance to insecticides. Indeed, the average levels of insecticide resistance in M. persicae from Southeastern Australia have been found to be comparatively low (James Anstead, IACR Rothamstead, personal communication).

Apart from the different mechanisms by which environmental heterogeneity may maintain clonal diversity, the immigration of new, sexually produced clones from nearby peach orchards in spring appears to also contribute to the observed diversity at Bacchus Marsh. Sampling was started during the time when the descendants of sexually produced females migrate from their primary host, peach, to their secondary hosts. During the 11 months thereafter, the population on secondary hosts experienced a decline in clonal diversity. This decline was very unsteady, though, and I do not have a good explanation for the low levels of clonal diversity observed in November and December, followed by the intermittent increase in January and February (Fig. 1). Periods of increased migration and colonization by alates, commonly associated with high population density (e.g. McVean et al., 1999), could potentially increase clonal diversity, but there is little evidence for this in the present data set. The intermittent increase in diversity followed a period during which only few alate individuals were collected, and there was no relationship between clonal diversity and the proportion of alates (Fig. 1). Nevertheless, an overall decline is consistent with the influx of some sexually produced genotypes in spring and their disappearance towards the end of the growth season. However, this decline was rather shallow and the fact that many of the more common clones were collected during all seasons (Fig. 3), with the four most common clones already present at the site during the growth season preceding this study, support the previous finding that M. persicae at Bacchus Marsh primarily comprises obligate parthenogens that overwinter parthenogenetically (Vorburger et al., 2003a). This is further supported by the FIS values that were consistently negative even when samples were reduced to a single copy of each genotype. In M. persicae, this appears to be a characteristic of predominantly asexual populations due to the increased heterozygosity of obligate parthenogens (Vorburger et al., 2003a). A similar finding was made for the bird cherry-oat aphid R. padi (Delmotte et al., 2002).

An unresolved problem is why the frequency distribution of clones is so skewed. What selective advantage makes the ‘superclones’ so dominant? Two experimental studies compared the variance in fitness across different hosts (Vorburger et al., 2003b), as well as cold tolerance among a number of different clones of M. persicae (Vorburger, 2004), including genotypes 45 and 58, which are the most abundant clones across Victoria and were also found at Bacchus Marsh (Fig. 3). However, these experiments provided no evidence that their success could be explained by a particularly broad tolerance to host plant variation or by an increased cold tolerance, which would provide a selective advantage during parthenogenetic overwintering. Insecticide resistance may play a role, and this was suggested by a very recent study on M. persicae in France (Zamoum et al., 2005). Another hint stems from a common-garden experiment with 19 different clones collected during the study reported here, which was designed to estimate heritabilities of life-history traits and their genetic correlations (Vorburger, 2005). This experiment revealed positive genetic correlations among major components of fitness and thus strong fitness differences among clones, as well as a positive (albeit weak) relationship between the clones’ fitness estimates from the laboratory and their abundance in the field. This suggests that under benign conditions as experienced in the laboratory (which may temporarily be fulfilled in the field), there are differences in ‘general vigor’ among clones that allow certain clones to increase more rapidly than others.

Whatever the selective pressures are that cause the ecological success of the dominant clones, they appear to be strong yet unconstant, based on the fast and marked shifts in the relative frequencies of these clones. This study shows that a single sample would disregard the dynamic nature of aphid populations and could thus only provide an incomplete picture of their clonal composition.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

I thank V. Kellerman and C. Sands for their help in the field and F. Bacchin for letting us work on his farm. Meteorological data were provided by the Australian Bureau of Meteorology. Molecular work was conducted in P. Sunnucks’ lab at La Trobe University, supported by a Small Grant (104084) from La Trobe University's Faculty of Science, Technology and Engineering, using equipment funded by a DEST Department Systemic Infrastructure Initiative Grant. M. Arioli, P. Sunnucks and two anonymous reviewers provided helpful comments on a draft of this manuscript. I was supported by a fellowship (81ZH-066832) from the Swiss National Science Foundation and a research Grant (57202802) from the University of Zürich.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References