Evolutionary Ecology

, Volume 26, Issue 6, pp 1311–1329

Intraspecific hybridization, life history strategies and potential invasion success in a parasitoid wasp

Authors

  • Chiara Benvenuto
    • Biology of Introduced PopulationsInstitute Sophia-Agrobiotech, (INRA–CNRS–UNS)
    • UCD School of Biology and Environmental ScienceUniversity College Dublin
  • Sandrine Cheyppe-Buchmann
    • Biology of Introduced PopulationsInstitute Sophia-Agrobiotech, (INRA–CNRS–UNS)
  • Gérald Bermond
    • Research and Development in Biological ControlInstitute Sophia-Agrobiotech, (INRA–CNRS–UNS)
    • Research and Development in Biological ControlInstitute Sophia-Agrobiotech, (INRA–CNRS–UNS)
    • Research and Development in Biological ControlCentre INRA PACA
  • Xavier Fauvergue
    • Biology of Introduced PopulationsInstitute Sophia-Agrobiotech, (INRA–CNRS–UNS)
Original Paper

DOI: 10.1007/s10682-011-9553-z

Cite this article as:
Benvenuto, C., Cheyppe-Buchmann, S., Bermond, G. et al. Evol Ecol (2012) 26: 1311. doi:10.1007/s10682-011-9553-z
  • 327 Views

Abstract

Classical biological control—the introduction of exotic species to permanently control pests—offers an applied framework to test ecological and evolutionary hypotheses derived from invasion biology. One such hypothesis is that intraspecific hybridization can facilitate invasions because hybrids express higher phenotypic mean and/or variance than their parents. We tested this hypothesis using the parasitoid wasp Psyttalia lounsburyi, a candidate biocontrol agent for the olive fly Bactrocera oleae. Under laboratory conditions, we found marked differentiations between two populations of wasps, from South Africa and Kenya, in terms of life history strategies. South African females were better reproducers than Kenyan females, but the opposite was observed for males. Reaction norms showed different effects of developmental temperature on fecundity depending on the genotype. However, neither heterosis nor hybrid breakdown were observed. Hence, in this system, sex-specific effects of hybridization and genotype-by-environment interactions jeopardize any straightforward prediction on the fitness of hybrids. Therefore, our paper contributes to tone down the hybrid advantage hypothesis in invasion biology.

Keywords

HybridizationClassical biological controlCyto-nuclear effectGenotype-by-environment interactionHeterosisVarianceReaction norms

Introduction

A fascinating question in invasion biology arises in the form of a paradox (Sax and Brown 2000): how can exotic species, despite their presumed low initial genetic diversity, invade remote areas, different from their native area, sometimes even outperforming locally adapted populations? Long-term invasion success can be explained by invaders characteristics, such as broad range of ecological tolerance or resistance to parasites and diseases, as well as extrinsic factors such as reduced predation rates and lack of competitors (e.g., Kolar and Lodge 2001; Marchetti et al. 2004). From an ecological-genetics point of view, the paradox can be solved if we consider the different conditions which can occur during invasion (Frankham 2005; Hufbauer 2008). These conditions result from a combination of genetic and demographic processes, which are in turn determined by a variety of factors, including gene flow and population dimension and structure (Lowe et al. 2004). At the beginning of biological invasions, initial propagule size can be large, with no significant reduction in the genetic variation of invaders. However, this is generally not true, and small initial population size generally results in severe genetic bottlenecks. In that case, successive invasions from one source population may progressively restore the genetic variability from the native area (e.g., Bousset et al. 2004). Another process that can resolve the paradox of biological invasions is the hybridization of different introduced populations (multiple introductions from different source populations); this admixture may increase genetic variation and create beneficial genetic innovations that might favour adaptation to novel environments (e.g., Roman 2006; Facon et al. 2008).

The concept of hybridization refers to the mixing of divergent genotypes (Anderson and Stebbins 1954; Stebbins 1959), either at the species level (interspecific hybridization) or at the population level (intraspecific hybridization). It is commonly observed that if the genetic divergences between the parental lineages are too marked, the intermixed offspring might not be successful or might present reduced fitness (hybrid inferiority; see review in Burke and Arnold 2001). But if hybrids are able to persist and reproduce successfully, inter- and intraspecific hybridization enhance genetic diversity, possibly increasing the adaptive potential of invaders (Ellstrand and Schierenbeck 2000; Schierenbeck and Ellstrand 2009). Indeed, highly successful hybrid invaders have been recorded in plants (Ellstrand and Schierenbeck 2000; Lavergne and Molofsky 2007; Rosenthal et al. 2008; Culley and Hardiman 2009; Keller and Taylor 2010; but see Wolfe et al. 2007) and, to a lesser extent, in animals (Kolbe et al. 2004; Facon et al. 2005, 2008).

Two main mechanisms, not mutually exclusive, have been proposed to explain the success of hybrids as invaders: (1) hybrid vigour (or heterosis) which results in higher phenotypic mean for some traits in hybrids compared to their parents; (2) high genetic and phenotypic variance which increases evolutionary potential (Rieseberg et al. 1999; Ellstrand and Schierenbeck 2000; Facon et al. 2005).

The hypothesis that hybridization plays a major role in biological invasions is weakened by methodological difficulties and observational biases. First, because introduced populations are small and most often undetectable before they establish and spread, it is not easy to investigate genetic hybridization during the initial phases of invasion (but see the approach of Ciosi et al. 2008), precisely when it is expected to have the strongest effect (Puth and Post 2005; Marsico et al. 2010). Second, because most studies on biological invasions rely on the observation of fortuitously introduced populations which have already established and spread, it is almost inevitable to conclude that hybridization has a positive effect on invasion success. In fact, if hybridization had a negative impact via hybrid breakdown during the establishment period, it would have less chances of being documented, while neutral effects of hybridization might be underestimated because of publication biases (since outcomes with non-significant effects are often not published). Third, the inference that hybridization is important simply because it is frequent fails to recognize that hybridization is likely to be more common in introduced ranges (where species and genotypes that are separated in their native ranges are brought into contact) than in the their original areas (Hufbauer 2008; but see Verhoeven et al. 2011).

An alternative approach to the study of intraspecific hybridization in natural conditions consists in performing manipulative experiments. Classical biological control introductions are introductions carefully planned and controlled by humans to permanently control pests. They represent examples of “experimental invasions” (Grevstad 1999; Lexer et al. 2003) for which it is possible to manipulate initial conditions (e.g., size and genetic composition of introduced populations) and follow invasion dynamics from the very first steps. Even though this approach is not novel and has been successfully employed (Grevstad 1999; Memmott et al. 2005; Fauvergue et al. 2007) it is still underused (Fauvergue and Hopper 2009; Marsico et al. 2010). Classical biological control projects can be extremely valuable in gaining information on the role of hybridization in invasion dynamics, especially during initial phases of the invasions where data are lacking.

We used a candidate biological control agent, the haplodiploid parasitoid wasp Psyttalia lounsburyi Silvestri (Hymenoptera: Braconidae), to test the hypothesis that hybrids exhibit superior performances than their parents. Our study organism is specialized on the olive fruit fly Bactrocera oleae (Rossi) (Diptera: Tephritidae), a major pest of olive trees (Olea europaea); it has therefore gained importance in classical biological control programmes (e.g., in California, see Daane et al. 2008; in France, see Malausa et al. 2010). Here we present laboratory experiments which we used to explore the potential invasive success of two different strains of P. lounsburyi and their hybrids. The actual efficiency of this wasp against the olive fly in nature will be further assessed upon its release in the field.

Two populations of P. lounsburyi (from Kenya and South Africa) were tested to assess whether hybrids exhibit some advantages over individuals from the pure parental populations. In particular, we were interested in assessing best-parent heterosis, i.e., superiority of hybrids compared to the best parent. Mid-parent heterosis (the superior performance of hybrids compared to the average value of the two parents) can be useful to reveal the presence of non-additive genetic mechanisms of hybrid fitness (Johansen-Morris and Latta 2006; Stupar et al. 2008) but best-parent heterosis has more practical value for population management.

In order to test our hypothesis, we compared potential and realized female fecundity as well as one component of male fitness (ability to produce daughters) among hybrid and parental lineages under three developmental temperatures. Contrary to the hypothesis proposed by Ellstrand and Schierenbeck (2000) that hybridization facilitates invasions, our experimental approach revealed hybridization effects much more complex than expected. Hence, our results highlight the difficulty of using hybridization as a major input when designing optimal release strategies for biological control, and more generally, our results indicate that hybridization does not always provide a solution to the paradox of biological invasions.

Materials and methods

Psyttalia lounsburyi strains and rearing

Psyttalia lounsburyi is a solitary parasitoid wasp which develops inside its host larvae without stopping their development (i.e., it is a koinobiont endoparasitoid). Like in other species of the order Hymenoptera, males are haploid (they develop from unfertilized eggs) and inherit all their genetic material from their mother, whereas females are diploid and get half of their genes from their mother and half from their father. Thus recombination is absent in F1 males but occurs in females (Hedrick and Parker 1997) at a rate similar to that observed in diploid organisms (Niehuis et al. 2010). Recombined males can nonetheless be obtained from back-crosses. Because of this lag between the sexes, it has been suggested that haplodiploidy helps to maintain a higher proportion of parental genotypes in the population, favouring additional genetic admixture (Legner 1988). Overall expectations about the effect of hybridization derived from diploid species should hold in solitary hymenopterans (Breeuwer and Werren 1995; Antolin 1999; Niehuis et al. 2010).

We used two outbred strains of P. lounsburyi, one originating from South Africa (SA) and the other from Kenya (KE). The two populations are known to be genetically distinct (FST = 0.42, which represents a highly significant pairwise genetic differentiation between the two populations; Cheyppe-Buchmann et al. 2011; see also Bon et al. 2008; Rugman-Jones et al. 2009; Malausa et al. 2010). Initial parasitoid wasps were obtained from laboratory cultures maintained in quarantine at the USDA-ARS European Biological Control Laboratory in Montpellier, France. They were transferred to the INRA laboratory in Sophia-Antipolis, France, in 2007. Initially, the population from South Africa was found to be less polymorphic than the one from Kenya in terms of number of microsatellite alleles (Cheyppe-Buchmann et al. 2011). At least 25 founders were available to start the rearing of the two strains, which were kept more than 20 generations in the laboratory prior to the beginning of the experiment. The two strains were naturally infected by Wolbachia but the bacterial infection was removed at the beginning of 2008 by treating the wasps with the antibiotic rifampicin for two consecutive generations. Aposymbiotic individuals were thus used in our experiments.

Parental strains were maintained in a controlled environment (temperature: 22 ± 1°C; relative humidity: 55%; photoperiod: 16L:8D) on a substitution host, the Mediterranean fruit fly Ceratitis capitata Wiedemann (Diptera: Tephritidae). Fruit flies are easier to rear in the laboratory than olive flies and their larvae have been successfully used as hosts for P. lounsburyi (Thaon et al. 2009; Malausa et al. 2010). The rearing proceeded as follows: third instar C. capitata larvae were placed in a layer of nutritive medium, which was gently spread on Parafilm and wrapped around ping pong balls. These “infestation units” were then exposed to adult parasitoids in cylindrical infestation chambers (20 cm length × 30 cm diameter) for eight hours (Thaon et al. 2009). At the end of this period, medium and larvae (parasitized and unparasitized) were removed and placed in new boxes with a supply of fresh medium. Under such conditions, flies emerge in about 15 days from unparasitized pupae. The first male parasitoids emerge 3–4 days later, followed by females.

Crosses

The experiment was designed to generate simultaneously the eight types of crosses detailed in Fig. 1. This required two consecutive generations of crosses. Host pupae, which had been exposed to one of the two parental parasitoid strains, were individually isolated. Emerged virgin adults were sexed and reallocated for seven days into plastic containers with food and water for mating (30 couples per container for each type of cross). In this first generation, females from each parental strain (SA and KE) were either crossed with males from the same strain (SA and KE respectively) to obtain pure parental offspring or with males from the other strain (KE and SA respectively) to obtain hybrid offspring (F1 hybrids). Once mated, females were allowed to oviposit on infestation units for five consecutive days. Parasitized pupae were isolated again. With emerged adults from the second generation, we again performed the same four types of above-mentioned crosses with 30 couples for each cross; in addition, the two types of F1 hybrid females were backcrossed against each of the two parental strains to obtain the four types of back-cross hybrids. Host larvae were exposed to these mated females for seven days and they were subsequently randomly assigned to three developmental temperatures (18, 22, or 26°C) in climatic developmental chambers (Sanyo MLR-351H). Other environmental parameters were maintained constant (16L:8D with light intensity of 2,230 Lux and 75% RH). The whole crossing design was replicated five times between October 2008 and July 2009, each replicate being treated as a block in the statistical analysis. At the end of the second generation of crosses we could thus simultaneously analyze eight cross types (two parental cross types, two F1 hybrids and four back-cross hybrids) across three developmental temperatures for a total of 24 combinations of treatment. Thirty couples were used for each combination and block, for a total of 3,600 individuals tested.
https://static-content.springer.com/image/art%3A10.1007%2Fs10682-011-9553-z/MediaObjects/10682_2011_9553_Fig1_HTML.gif
Fig. 1

Crossing design. The boxon the left contains crosses characterized by South African cytoplasm (individuals originated by South African mothers); the box on the right (light grey) contains crosses characterized by Kenyan cytoplasm (individuals originated by Kenyan mothers). Cytoplasm is also specified in squared brackets for F1. Continuous lines indicate crosses internal to each parental lineage (P); dotted lines indicate crosses between the two parental lineages (F1); dashed lines indicate back-crosses (BC). Nuclear genome is reported at the bottom of the figure as mean expected proportion of the South African genome. The percentage of nuclear genome is different between daughters (diploid) and sons (haploid)

Biological traits

For each block, cross type and developmental temperature, we estimated the number of females (“daughters”) and males (“sons”). It is assumed that these two traits depend primarily on the genotype of the parents. In fact, the number of produced offspring results from a combination of mother’s fecundity and the mating compatibility between the two parents; moreover, in parasitoid wasps, offsprings’ survival is mainly triggered by compounds (such as venom proteins, VLP, and teratocytes) injected by the mother during oviposition (Godfray 1994; Quicke 1997). The genotypes of the parents that we could analyse were the two parental strains and the two F1 generations (see Fig. 5). The third investigated biological trait was the number of mature oocytes in the ovaries for a subset of the daughters. For this latter measure, daughters were maintained in isolation cages at 22°C and provided with honey and water during one week to allow for egg development. In three sampling days (every other day from day 7) five females from each cross type and each temperature were collected randomly and stored at −18°C. These females (N = 1,669) were dissected under a binocular microscope at 25× magnification and mature oocytes were counted. Because ovigeny takes place in the late part of the pre-imaginal development (Godfray 1994; Quicke 1997), it is assumed that this trait is determined by the individual’s genotype rather than those of the parents. Thus, we could analyse this trait for all the crosses: the two parental strains, the two F1 generations and the four back-crosses (see Fig. 1).

The number of emerged progeny (daughters and sons) will henceforward be referred to as the realized fecundity while the number of oocytes in daughters will be referred to as the potential fecundity. We present the results starting from potential fecundity, despite the fact that it was measured later in our design, because potential fecundity chronologically precedes realized fecundity in the individual’s life.

Statistical analyses

The eight types of crosses were categorized based on the bi-parental source (mother × father; i.e., SA × KE means that South African females are crossed with Kenyan males). Mothers determine the nuclear and cytoplasmic components of their offspring, while fathers’ influence is limited to the nuclear component in daughters. For each cross type, we also assigned specific genotypic values (based on the mean expected proportion of the South African genome) combined with the origin of the cytoplasm (Fig. 1). Overall, genotypes were categorized as parental, F1 generation and back-crosses. F1 were further categorised has having South African or Kenyan cytoplasm (in squared bracket). Statistical analyses relied mostly on generalized linear models (GLM) (Crawley 1993), using the Genmod procedure in the 9.1.3 SAS statistical package (SAS Institute Inc. 1999). We used the model deviance divided by its degree of freedom and diagnostic plots (scattered plots of standardized deviance residuals vs. predicted values and observed vs. predicted values) to check the adequacy of the models (deviance/df = 1; see Montgomery et al. 2001). Preliminary explorative analyses (not shown) revealed that our data were overdispersed (deviance/df ≫ 1). We thus applied GLMs with a negative binomial error distribution and log link function, which fitted our data better (deviance/df ≈ 1; see model fit in Table 1).
Table 1

Most parsimonious generalized linear models with negative binomial error distribution and log link function for (1) potential fecundity of mature fed daughters; (2) total offspring (all sexes combined); (3) number of emerged daughters; (4) number of emerged sons

 

Deviance/df

Model fit

Source

df

χ2

P value

1. Female potential fecundity

a) By type of cross

1,683.1/1,615

1.04

Block

4

196.0

<0.0001

Temperature

2

372.7

<0.0001

Cross

7

225.6

<0.0001

Cross*Temp

14

50.6

<0.0001

b) By female’s characteristics

1,682.9/1,631

1.03

Block

4

201.2

<0.0001

Temperature

2

46.9

<0.0001

Nuclear genotype

1

135.5

<0.0001

Cytoplasm

1

2.2

0.1363

Nuc*Cyto

1

7.8

0.0054

Nuc*Temp

2

24.4

<0.0001

2. Total offspring

a) By type of cross

123.1/106

1.16

Block

4

67.2

<0.0001

Temperature

2

25.0

<0.0001

Cross

7

19.0

0.0083

c) By parental genotype

123.1/110

1.12

Block

4

66.1

<0.0001

Temperature

2

24.9

<0.0001

Mother’s genotype

3

16.5

0.0009

3. Female offspring

a) By type of cross

123.1/108

1.14

Block

4

83.8

<0.0001

Cross

7

25.8

0.0006

c) By parental genotype

124.3/111

1.12

Block

4

81.3

<0.0001

Mother’s genotype

3

13.7

0.0034

Father’s genotype

1

9.5

0.0021

d) SA & KE mothers only

59.7/54

1.11

Block

4

44.3

<0.0001

Mother’s genotype

1

9.8

0.0017

e) F1[SA]& F1[KE]mothers only

63.4/54

1.17

Block

4

41.5

<0.0001

Father’s genotype

1

9.7

0.0018

4. Male offspring

a) By type of cross

123.9/106

1.17

Block

4

45.9

<0.0001

Temperature

2

32.1

<0.0001

Cross

7

16.7

0.0193

c) By parental genotype

123.8/110

1.13

Block

4

45.5

<0.0001

Temperature

2

31.8

<0.0001

Mother’s genotype

3

14.0

0.0029

For each analysis: a) cross type is coded as a categorical variable; b) cross type is replaced by the female own characteristics (nuclear genotype and cytoplasm); c) cross type is replaced by mother’s and father’s genotype; d) Mother’s genotype = South Africa (SA) and Kenya (KE) only; e) Mother’s genotype = F1[SA] and F1[KE] only. Model deviance and the corresponding degrees of freedom are reported. χ2 values are obtained from type-3 likelihood ratio tests

Our goal was to compare potential and realized fecundity among parental and hybrid lines across three developmental temperatures and five blocks. We thus systematically used cross type, developmental temperature and block as predictors (full model explicatory variables; see Table 1). However, offspring from each cross type can be characterized either from their parents (i.e., mother’s and father’s genotypes) or based on their own characteristics (nuclear genotype, as a proportion of SA genome, and cytoplasm). Thus, depending on the phenotypic traits analyzed, we substituted cross type either with the offspring’s own genotype and cytoplasm (Table 1.1b) or with the parental genotypes (Table 1.2c, 3c, 4c). Interactions between factors were only investigated for the number of mature oocytes, for which replications were available. A manual backward model selection procedure was performed by removing non-significant (at the P = 0.05 level) explanatory variables and interactions from full models in order to obtain most parsimonious models (Crawley 1993). To test the significance of the explanatory variables, we used type-3 likelihood ratio tests (which are not affected by the order in which the variables are specified in the model).

Data were graphed as exponential back-transformed least square means or expected values of the parameters of interest from most parsimonious GLM together with their 95% confidence intervals (CI). Where required, post hoc tests for multiple comparisons were performed and P values estimated after sequential Bonferroni correction (Holm 1979).

Potential female fecundity

In order to test for differences among parental and hybrid lines in potential fecundity, we analyzed oocyte load using the daughters’ own characteristics as predictors. Both the influence of the cytoplasmic background (SA or KE) and the nuclear genotype, expressed as the mean expected proportion of South African genome, were taken into account to investigate more precisely the genetic determinism of potential fecundity. Because the expectation from our working hypothesis was that hybrids perform better than parents, nuclear genotype was initially considered as a qualitative variable. Since we did not find evidence of departure from linearity, we then repeated the analysis considering nuclear genotype as a continuous variable. A limited number of females (26 out of 1,669 females: 1.56%) presented no eggs in their ovaries and were excluded from the analysis.

Phenotypic variance (number of oocytes)

To estimate potential female fecundity, approximately 14 females were dissected for each of the eight cross types, the three temperatures and the five blocks. We were thus able to test differences in phenotypic variance for potential fecundity among crosses. In particular, our goal was to check for an increase of phenotypic variance in F1 hybrids and back-crosses (which could result in increased adaptive potential). We used the methodology described by O’Brien (1979, 1981): individual data were r-transformed in such a way that the average of the transformed values for each cross was equal to the initial variance of that cross (see also Connallon and Knowles 2006). We then performed an analysis of variance (ANOVA) on this new data set, nesting the eight cross types within three overall categories: parents (KE, SA), F1 hybrids (F1[SA], F1[KE]) and back-cross hybrids (F1[SA] × SA, F1[KE] × SA, F1[SA] × KE, F1[KE] × KE). R-transformed data were not normally distributed, but this method is robust to violations of normality assumptions (O’Brien 1981). Since data were unbalanced we used weighted Welch’s one-way ANOVAs to confirm our results (as suggested by O’Brien 1981).

Realized fecundity

To test for differences among parental and hybrid lines in realized fecundity we analyzed total offspring production using a GLM. Following Krackow et al. (2002) on the analysis of sex ratio data in arrhenotokous haplodiploids (where males develop from unfertilized eggs), we also analyzed daughters and sons separately. Such an analysis is useful to disentangle different roles of factors affecting each sex. We then focused on the production of daughters, because daughters, not sons, are influenced by both maternal and paternal genotypes and should therefore best reveal effects of hybridization. In parasitoid wasps, offspring production is determined mainly by behavioural and physiological features of the parents (Godfray 1994; Quicke 1997). We therefore used the genotypes of the mothers (SA, KE, F1[SA], F1[KE]) and fathers (SA, KE) as explanatory variables in this analysis.

To check if the number of C. capitata larvae was a limiting factor for offspring production, we tested the correlation between the number of emerged parasitoids and the total number of emerged insects (flies and wasps), the latter indicating initial host availability. In order to do this, we used only the subset of individuals which had developed at 22°C (standard rearing temperature in the lab) wherein we can assume similar mortality across blocks for parasitoids and hosts.

Results

Potential female fecundity

We analyzed potential female fecundity from the number of mature oocytes in the gonads of fed daughters. Fecundity was influenced by block, developmental temperature, cross type and the interaction between cross type and temperature (Table 1.1a). To shed more light on the effect of cross type we repeated the analysis using daughter’s characteristics, i.e., nuclear genotype (mean expected proportion of South African genome) and cytoplasm, as explanatory variables. Figure 2 (panels a, b) shows the raw data and predictions from the GLM performed using nuclear genotype as a qualitative variable (model details not shown). From the graph we detected no departures from linearity, hybrids showing intermediate fecundities in-between parental lineages. We thus used nuclear genotype as a continuous variable (Table 1.1b, Fig. 2c). This highlighted a differential response of nuclear genes depending on the cytoplasmic background. Potential fecundity increased with increasing proportion of the South African genotype but the fecundity advantage was more pronounced in the South African cytoplasm than in the Kenyan cytoplasm (gray lines vs. black lines in Fig. 2c). The analysis also revealed significant genotype-by-environment interactions. The relative advantage given by the South African genome was indeed maximal at 22°C but was less pronounced at 18 and 26°C. The various genotypes (designated by combination of mean proportion of South African genome and Kenyan or South African cytoplasm) expressed different potential fecundity depending on the developmental temperature.
https://static-content.springer.com/image/art%3A10.1007%2Fs10682-011-9553-z/MediaObjects/10682_2011_9553_Fig2_HTML.gif
Fig. 2

a, b Effect of nuclear genotype (expressed as mean expected percentage of the South African genotype and considered as a qualitative variable) on potential female fecundity (number of oocytes), in different cytoplasmic environments (a South African; b Kenyan) at three different developmental temperatures (18, 22, and 26°C, from left to right). Lines represent predictions from generalized linear model (±95% confidence intervals) fitted to actual data with cross and temperature as explanatory variables (model details not reported; block effect was not included in the graphs for ease of interpretation). c Combined graph with the predictions from the two generalized linear models (gray line response in the South African cytoplasm; black line response in Kenyan cytoplasm) considering nuclear genotype as a continuous variable (see Table 1.1b for model details)

Phenotypic variance

Because recombined genotypes may induce an increase in variance within hybrid populations (particularly in back-crosses), we tested the heterogeneity of variances for potential female fecundity according to the category of cross types (parents, F1 hybrids and back-cross hybrids), developmental temperature and block (Table 2). Results of the global test indicated that both “block” and “temperature” were significant, as well as their interaction. However, contrary to the expectations, hybrids and parental lineages had similar phenotypic variance.
Table 2

Crossed-nested ANOVA on r-transformed data (see text for details) to analyse the effect of the various explanatory variables on the phenotypic variance for potential female fecundity

Source

df

SS

MS

F

P value

Block

4

15,294,299.09

3,823,574.77

23.20

<0.0001

Temperature

2

6,167,681.34

3,083,840.67

18.71

<0.0001

Temp*block

8

8,041,597.60

1,005,199.70

6.10

<0.0001

Cross category

2

912,702.77

456,351.39

2.77

0.0630

Cross [cross category]

5

4,153,673.68

830,734.74

5.04

0.0001

Error

1,621

267,187,225.0

164,828.6

  

The eight experimental crosses were nested (squared brackets) in three general categories (cross category: Parental, F1, and Back-crosses)

Realized fecundity

Total offspring

A preliminary exploratory analysis (daughter and sons combined, N = 24 treatment levels and five blocks) revealed the influences of block and developmental temperature (environmental factors) and cross type on the total number of adult offspring (Table 1.2a). The block effect was not surprising because of the difficulty to control precisely the number of hosts in the infestation units. For example, the number of hosts was estimated four times greater in block 1 than in block 4. However, we found no correlation between number of emerged parasitoids and host availability (N = 32, Pearson correlation coefficient r = 0.25; P = 0.17 all blocks combined; no significant correlations were found within each block). Hence, the number of hosts did not limit parasitoid oviposition. Developmental temperature affected realized fecundity: there were fewer offspring at 26°C (average number of emerged offspring and 95% CI: 89.67 [80.01–100.48]) compared to 18°C (137.22 [122.67–153.48]) and 22°C (120.92 [108.08–135.27]), (18°C vs. 26°C: df = 1, χ2 = 26.9, P < 0.0001; 22°C vs. 26°C: df = 1, χ2 = 13.3, P = 0.0003; 18°C vs. 22°C: df = 1, χ2 = 2.43, P = 0.12). The type of cross was also a significant predictor (Table 1.2a; see also Fig. 3). To investigate its effect in detail, we replaced all cross types with the corresponding maternal and paternal genotypes. In this way we were able to reveal a significant effect of mother’s genotypes on the production of offspring by each cross (Table 1.2c). South African maternal genotype produced more offspring (125.12 [109.67–142.74) than the Kenyan maternal genotype (93.19 [81.66–106.34]; df = 1, χ2 = 9.52, P = 0.002; 25.5% more offspring were produced by SA genotype; see colour code by maternal genotype in Fig. 3). We found no superiority of hybrids: F1[SA] (the best reproducer among the two hybrid lines) produced more offspring (135.07 [118.46–154.01]) than the pure Kenyan line (df = 1, χ2 = 15.29, P < 0.0001) but it produced no more offspring than the pure South African line (df = 1, χ2 = 0.65, P = 0.42); F1[KE] did not produce less offspring (108.43 [95.059–123.69]) than F1[SA] (df = 1, χ2 = 5.32, P = 0.02, non significant after sequential Bonferroni correction) and showed no significant difference from the pure parental Kenyan (df = 1, χ2 = 2.55, P = 0.11) or South African (df = 1, χ2 = 2.26, P = 0.13) lines (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs10682-011-9553-z/MediaObjects/10682_2011_9553_Fig3_HTML.gif
Fig. 3

Realized fecundity (number of emerged offspring) for each of the eight cross types (categorized by cross categories). Histograms represent least square means obtained after fitting the most parsimonious generalized linear model to the observed data (see Table 1.2a for model details). Error bars represent 95% confidence intervals. Maternal genotypes (significant predictor for realized fecundity) are designated with different colours

Offspring by sex

These global patterns were slightly modified when daughters and sons were analyzed separately (see Tables 1.3 and 1.4 for female and male offspring respectively). In fact, developmental temperature affected the number of males (Table 1.4a; Fig. 4) but not that of females (from full model: χ2 = 3.08; P = 0.21; effect removed in the most parsimonious model reported in Table 1.3a). More precisely, the increase in developmental temperature resulted in a decrease of the number of emerged males (average and 95% CI at 18°C: 98.05 [86.16–111.59]; at 22°C: 80.96 [71.09–92.18]; at 26°C: 55.31 [48.45–63.14]), presumably through a reduced pre-imaginal male survival (Table 1.4; Fig. 4; 18°C vs. 26°C: df = 1, χ2 = 36.42, P < 0.0001; 22°C vs. 26°C: df = 1, χ2 = 16.13, P < 0.0001; 18°C vs. 22°C: df = 1, χ2 = 4.2, P = 0.04, non significant after sequential Bonferroni correction). Figure 4 represents the overall effect of temperature on the production of offspring, regardless the cross under consideration. We did not find any interaction between the temperature and the cross: temperature is affecting generally all genotypes in a similar way (as the temperature increases, the production of male offspring decreases). Parental genotypes also affected differently the production of male and female offspring. The number of daughters was influenced by both paternal and maternal genotypes (Table 1.3c). Both F1 females (F1[SA] and F1[KE]) produced more daughters (40.50 [34.83–47.08] and 39.97 [34.40–46.45] respectively) than pure Kenyan females (28.24 [24.18–32.98]; F1[SA] > KE: df = 1, χ2 = 10.65, P = 0.0011; F1[KE] > KE: df = 1, χ2 = 10.01, P = 0.0016); between the pure parental lines pure South African mothers (average number of daughters and 95% CI 39.97 [34.37–46.48]) performed better than pure Kenyan mother (SA > KE: df = 1, χ2 = 9.84, P = 0.0017), while the opposite was true for fathers (average number of daughter and CI for South African males: 32.54 [29.19–36.27]; for Kenyan males: 41.54 [37.35–46.20]; KE > SA: df = 1, χ2 = 9.85, P = 0.0017). Instead, the number of sons was influenced only by the genotype of the mothers (Table 1.4c) with F1[SA] females producing more sons (92.96 [79.85–108.23]) than pure Kenyan (64.35 [55.23–74.99]; df = 1, χ2 = 11.19, P < 0.0008) and F1[KE] (67.66 [58.04–78.88]; df = 1, χ2 = 8.23, P < 0.0041) but not than pure South African females (83.09 [71.23–96.92]; df = 1, χ2 = 1.02, P = 0.31). Again, pure South African mothers produces more sons than pure Kenyan mothers (df = 1, χ2 = 5.26, P = 0.0219).
https://static-content.springer.com/image/art%3A10.1007%2Fs10682-011-9553-z/MediaObjects/10682_2011_9553_Fig4_HTML.gif
Fig. 4

Effect of developmental temperature on realized fecundity, by sex. Open circles (daughters) and solid circles (sons) represent least square means obtained after fitting the most parsimonious generalized linear model to the observed data. Error bars represent 95% confidence intervals

When focusing on the production of daughters, we found contrasting life history strategies between the South African and Kenyan populations. Comparing only pure parental genotypes (Table 1.3d; Fig. 5a), South African females produced more daughters (39.73 [34.54–45.70]) than Kenyan females (28.37 [24.54–32.79]; SA > KE: df = 1, χ2 = 10.61, P = 0.0011). This result held when South African females were crossed with South African males (contrast: SA × SA vs. KE × SA: χ2 = 4.69; P = 0.03) or with Kenyan males (SA × KE vs. KE × KE: χ2 = 5.09 P = 0.024). Hybrids showed neither absolute superiority over both parental lineages nor hybrid breakdown. The number of emerged daughters produced by F1 females when crossed with South African males (F1[SA] × SA and F1[KE] × SA) was not significantly different from the South African (contrast: χ2 = 0.82; P = 0.36; Fig. 5a) or Kenyan parental lineages (contrast: χ2 = 0.68; P = 0.41; Fig. 5a). The number of emerged daughters in the two back-crosses introgressed with Kenyan paternal genes (F1[SA] × KE and F1[KE] × KE) was significantly higher than the number of daughters of the Kenyan pure line (contrast: χ2 = 12.08; P = 0.0005; Fig. 5a) but not higher than the number of daughters of the South African pure line (contrast: χ2 = 3.52; P = 0.06; Fig. 5a). When the two father’s genotypes were compared using “neutral” mother’s genotypes (i.e., F1 genotypes, to avoid the influence of homogamy on the production of daughters), we detected a strong influence of the paternal genotype (Table 1.3e; Fig. 5b). In this case, Kenyan males produced more daughters than South African males (KE > SA: df = 1, χ2 = 10.36, P = 0.0013) either when crossed with F1[SA] females (contrast: F1[SA] × SA vs. F1[SA] × KE: χ2 = 5.17; P = 0.023) or with F1[KE] females (F1[KE] × SA vs. F1[KE] × KE: χ2 = 6.44 P = 0.011). The comparison of hybrid males with parental males was not possible with our design.
https://static-content.springer.com/image/art%3A10.1007%2Fs10682-011-9553-z/MediaObjects/10682_2011_9553_Fig5_HTML.gif
Fig. 5

Realized fecundity (number of emerged females) for each of the eight possible cross types. a Maternal genotypes crossed with South African males (left) or Kenyan males (right); b Paternal genotypes crossed with F1[SA] females (left) or F1[KE] females (right). Histograms represent least square means obtained after fitting the most parsimonious generalized linear model to the observed data (see Table 1.3a for model details). Error bars represent 95% confidence intervals

Discussion

In the framework of a biological control programme against the olive fruit fly Bactrocera oleae, we analyzed the potential and realized fecundity of parental and hybrid lines of the parasitoid wasp Psyttalia lounsburyi. The two populations under study, from Kenya and South Africa, were previously molecularly characterized and resulted genetically distinct from each other (Cheyppe-Buchmann et al. 2011). The molecular characterization of candidate biocontrol agent is a fundamental step to assess the genetic variability of geographically distant populations and thus recognize their potential for hybridization. Our results show that the two populations have deeply diverged also in terms of one fundamental life history trait: fecundity. The relative fitness of the two genotypes varies according to the sex considered; South African females have more offspring than Kenyan females while Kenyan males sire more daughters than South African males. This apparent between-sex trade-off probably indicates that some genetic constraints prevent the simultaneous optimization of physiological and behavioural traits involved in offspring production and that different life history strategies may have been selected according to local conditions (Roff 1992; Stearns 1992). Unfortunately, as often in biological control programmes (Hufbauer and Roderick 2005), we do not have precise ecological data from the native areas for these two populations. Nevertheless, the genetic and phenotypic differentiation between South African and Kenyan populations provides a suitable case to test the effect of hybridization and discuss its potential consequences for biological invasions.

Our results show no drastic effect of hybridization on P. lounsburyi, be it negative (hybrid breakdown) or positive (heterosis). The phenotypes of hybrids are located in-between the phenotypes of their parents for both potential and realized fecundity. However, the exact position of the hybrid’s phenotypes differs for the two traits considered and, as explained below, it is determined by different genetic mechanisms. For potential fecundity, measured as the number of mature oocytes, the comparison among F1 hybrids, back-cross hybrids and parents resulted in a linear trend, well predicted by the proportion of parental nuclear genome. Hence, for potential fecundity, most of the genetic variability is transmitted additively. A significant part of this variability is nonetheless influenced by extra-nuclear components which interact with the nuclear components synergistically. Documented cases of cyto-nuclear interactions in animals focus mainly on incompatibilities (Breeuwer and Werren 1995; Ellison et al. 2008; Niehuis et al. 2008; Ellison and Burton 2010) and, for insects, on the role that Wolbachia plays on incompatibilities (Bordenstein et al. 2001; Bordenstein and Werren 2007). Cyto-nuclear coevolution may nonetheless promote local adaptations (reviewed by Rand et al. 2004; Dowling et al. 2008). For realized fecundity, which integrates the availability of mature eggs (potential fecundity) with many other physiological and behavioural traits (locomotion, host finding ability, quality of venoms…), back-crossed hybrids do not differ from the best parental population (from South Africa), suggesting some dominance effects. Hence, one important conclusion from our study is that the consequences of hybridization are trait-specific, even when traits are similar, as it is the case for potential and realized fecundity.

With reaction norms, we included developmental temperature in our experimental approach to test more thoroughly the effect of inter-population crosses on fitness. The effect of temperature can be particularly important when addressing the development of an organism in conditions different from its natural conditions (which can be the case for invading populations in a new environment). Developmental temperature affected realized fecundity. In particular, we recorded a decrease of male offspring emergence at 26°C, indicating sex-specific mortality in the haploid sex at high temperature. Moreover, environmental factors such as temperature have been reported to affect differentially the expression of various genotypes in parasitoids (Ris et al. 2004 and references therein). Here, when we analyzed potential female fecundity, we detected differential responses of the various genotypes depending on their rearing temperature. These thermal reaction norms reflect temperature-specific effects, which act in combination with the previously mentioned cyto-nuclear interactions, complicating predictions on hybrids’ fitness. Rarely do studies analyze the combined effect of genotype-by-temperature and cyto-nuclear interactions on fitness (but see Dowling et al. 2007). Developmental temperature, affecting differentially the fitness of different genotypes, can influence the success of parental and hybrid lines in thermally heterogeneous environments. In particular genotype-by-environment interactions may accentuate or decrease differences between genotypes and thus locally modify the response to selection. Significant genotype-by-environment interactions, which affect hybrids fitness, have also been reported in plants (e.g., Campbell and Waser 2001).

Our results are in line with a body of research which considers the complexity of genetic admixture and the difficulty of predicting its effects on fitness, in particular in the long term and in heterogeneous environments (Arnold and Hodges 1995; Barton 2001; Burke and Arnold 2001; Edmands 2002; Arnold and Martin 2010). In fact, despite the absence of hybrid vigour in our experiment, we cannot rule out a role of hybridization on invasion success in the long run. The results obtained with P. lounsburyi in the laboratory suggest that there is an adaptive diversification in the parental lineages. An admixed population composed of parents and hybrids would gather a pool of genotypes with different life-history traits, a situation which could benefit a population introduced into a novel environment.

Early stages of biological invasions are poorly studied, given the inherent difficulties of detecting directly small initial invasive populations or reconstructing a posteriori the initial phases of invasions (Guillemaud et al. 2010). Nevertheless, assessing the factors affecting the initial steps of a biological invasion is necessary to understand the overall invasion dynamic and possibly to allow preventive conservation actions aimed at restraining the spread of unwanted species (Kolar and Lodge 2001, 2002) or, conversely, to improve the establishment of beneficial ones (Ehler 1998; Fagan et al. 2002). Biological control programmes offer the rare opportunity to follow the progression of invasions, from their initial stage to their successful or unsuccessful establishment (Fauvergue et al. 2007; Marsico et al. 2010).

The debate on the actual role of hybridization has motivated this experimental study, performed in parallel to the introduction of P. lounsburyi in France, to explore the potential invasive success of intraspecific hybrids of this parasitoid wasp. We detected variability in life history traits in the two parental wasp populations but no hybrid superiority. We realize that our results are based on crosses performed using only two strains. However, considering the scarcity of data documenting the initial phases of an invasion process, we feel that our results can provide a glimpse at the complex interactions that occur when mixing divergent genotypes. Such interactions will be further investigated in the field, where these same genotypes are being released.

The results obtained in the laboratory underline that, even though hybridization is predicted to favour invasion success, its impact on fitness is difficult to determine because of (1) trait-specific responses, (2) between-traits variation in genetic architecture and (3) genotype-by-environment interactions. Although seducing and theoretically sound, the idea that hybridization acts in a straightforward and easily predictable way might be overly simplistic and might fail to capture the complexity of the process and the consequences for biological invasions. In the case of P. lounsburyi as in other biological invasions, the analysis of post-release populations, either based on neutral (molecular) markers or on quantitative traits (fitness components), is necessary to understand the evolutionary trajectories of hybrids in nature and their actual invasive potential.

Acknowledgments

We would like to thank Benoît Facon, Thomas Guillemaud, and Ruth Hufbauer for helpful comments on the manuscript as well as Marcel Thaon for providing biological material. This study was partially funded by the Agence Nationale de la Recherche (project BioInv4I, ANR-06-BDIV-008-01).

Copyright information

© Springer Science+Business Media B.V. 2012