Genetic and phenotypic effects of hybridization in independently introduced populations of the invasive maize pest Diabrotica virgifera virgifera in Europe

Origin and crossing of Diabrotica v. virgifera Experiments were conducted on six laboratory-reared D. v. virgifera populations, i.e. two parental and four hybrid populations. The parental populations were the Central and South-eastern European population (CSE European) and the Northwest Italian population (NW Italian), as defined by Miller et al (2005). Hybrid populations were the four possible crossings of both (Table 1). Adults of the CSE European population were collected from maize fields near Crvenka in Severnobacki County in northern Serbia and near Kondoros in Bekes County in southern Hungary (only 200 km distant from each other). Adults from the NW Italian population were collected from a maize field near Como in Lombardy County in northern Italy. Adults were shaken from the maize plants into a funnel with an attached gauze bag. In the laboratory, adults were reared in gauzed cages (45 x 45 x 60 cm) for egg laying under standardized conditions according to Krysan & Miller (1986), Singh & Moore (1999), and similar to conditions during hybridisation and for trait assessments as described in detail below. Adults of the first laboratory generation (G1) of the CSE European and NW Italian populations were crossed in the laboratory. Newly emerged adults from both populations were sexed according to antenna length (Kuhar & Youngman, 1995; Staetz et al., 1976) and tarsus characteristics (Hammack & French, 2007). For crossing, single CSE European females were kept with single NW Italian males (or one NW Italian female with one CSE European male) in small gauze-covered plastic cylinders (dia.: 120 mm, h: 140 mm) for five days to allow mating. Two soft, unripe kernels of organically-produced maize, one 13 x 13 x 13 mm piece of zucchini flesh, one 13 x 13 x 13 mm piece of pumpkin flesh (Li et al., 2014), and a 5 x 5 x 5 mm piece of artificial pollen diet (Branson & Jackson, 1988; Singh & Moore, 1999) were provided as food, as well as a 10 x 5 x 5 mm cube of 15% watery agar. After 5 days, female-male pairs were double-checked for correct sex identification to avoid mistakes in crossings. Then, about 100-200 crossed pairs were pooled and transferred into a larger gauze cage for oviposition. Importantly, the CSE European female X NW Italian male population was reared separately from the NW Italian female X CSE European male population. The parental populations were mass-reared until G2 and the hybrid populations until Fx1 for experimentation. Parental effects - the fact that the phenotype of an individual is affected by the environment of its parents (Johnston et al., 2004) - were reduced because each D. v. virgifera was reared under similar standardised laboratory conditions as per Li et al. (2014) for more than one generation. Those conditions were the same as for the trait assessments described below. Assessment of neutral genetic traits and their variability DNA extraction and genotyping All D. v. virgifera adults assessed for phenotypic traits, had been stored in 90–96% ethanol until DNA extraction. Individuals were then washed in 0.065% NaCl at least three times to remove ethanol from the tissues (Bermond et al. 2012). DNA was extracted from half of each body cut lengthwise, using the DNeasy tissue kit according to the manufacturer’s instructions (Qiagen, Hilden, Germany) with an elution volume of 100 ml (AE buffer, Qiagen, Hilden, Germany). Thirteen microsatellite loci of D. v. virgifera as described by Bermond et al. (2012), were amplified using three separate multiplex PCRs performed in a PTC-225 MJ Research thermocycler (GMI, Minnesota, USA). Family reconstruction In the rearing, mated females had been grouped per population for oviposition, so that the offspring from different females were mixed. This resulted in cages in which several families of various sizes were present. This family structure is expected to affect genetic variability estimates. Therefore, the microsatellite markers were used to reconstruct the kinship between individuals, which allowed reconstructing siblings and thus considering the family structure in the genetic analyses. Families were reconstructed using the COLONY v. 2.0.1.1 software of R (Wang, 2004) (R Development CoreTeam, 2012). COLONY implements a maximum likelihood method to group individuals with siblings on the basis of their multi-locus microsatellite marker - defined genotypes. The connection between most likely siblings was obtained through a simulated annealing algorithm (Wang 2004). This method takes different types of genotyping errors into account when working with markers (dropout, mutations, contamination, false alleles) and can be used with or without knowledge of the parental genotypes. In our case, we did not have access to the genotypes of each sibling`s parents. Therefore, we performed four independent runs for each population to reconstruct the families. Siblings were formed after observing consensus groups of individuals in all the four different runs (Table 1). Only the individuals assigned to a genetic family were used for these genetic and phenotypic analyses that considered the family structure. Assessment of neutral genetic variation Neutral genetic variation was calculated after pooling individuals from the two experimental series per population (see below), because the pairwise fixation index (FST) between both series was not higher than 1%. The four hybrid populations were not pooled for genetic analyses because of a slight genetic differentiation detected with the microsatellite markers (see below). Neutral genetic variation within populations was evaluated by determining the mean number of alleles per locus (A) and the mean expected heterozygosity (He) (Nei, 1987) using GENECLASS version 2.0.h software (Piry, et al. 2004). We compared the mean number of alleles between populations by estimating allelic richness (AR) based on the smallest sample size using the rarefaction method of Petit et al. (1998) using Fstat version 2.9.3 software (Goudet, 2001). We calculated individual inbreeding coefficients (FIS) in each population and tested for deviations from the Hardy-Weinberg (HW) equilibrium using probability-test of “Genepop on the Web” -software (Raymond and Rousset, 1995a, b) similar to the exact Hardy-Weinberg tests of Weir & Cockerham (1990). Then, neutral genetic variation within populations was analysed a second time, this time accounting for family structures of the studied individuals, thus allowing the extraction of inter-family neutral genetic variation. From each population, 100 samples were randomly selected by sampling one individual per reconstructed family using the COLONY v. 2.0.1.1 software of R (R Development CoreTeam, 2012). From these, we generated one hundred values of DC, AR, He, FIS and probabilities associated with the Hardy Weinberg equilibrium test using Genepop 4.0 software (Raymond & Rousset 1995a, b). To calculate AR, seven families (the minimum number of families per population) were randomly drawn per population. This resampling procedure was used to correctly estimate mean values of DC, AR, He, FIS and their standard deviations when considering family structure. The 100 samples are not independent samples that could be used as independent observations to perform parametric tests. When analysing statistics of genetic diversity such as He and AR, the unit of observation was the locus (loci are independent in a coalescence framework, not individuals). Thus, the homogeneity of He and AR among populations was tested using non-parametric Friedman test (with the locus as repeat unit) taking and not taking the family structure into account. In case of detecting an overall heterogeneity of He and AR among samples, the hypothesis of an equal genetic variation between each hybrid and parental population was tested using the two-sided Wilcoxon signed rank tests. Due to multiple comparisons, the level of significance of each of the Wilcoxon test results was corrected with the false-discovery rate procedure of Benjamini & Hochberg (1995). When considering the family structure, the tests of Hardy-Weinberg equilibrium and homogeneity of He and AR among populations were performed 100 times, i.e. once for each of the 100 random draws (see above). The aim was to get a point estimate of the p-values via the resampling procedure and the averaging over the 100 values. We thus considered the means of the 100 p-values obtained from the random draw of 100 samples and their standard deviation. Assessment of phenotypic traits and their variability Ten phenotypic traits were assessed following Li et al. (2009; 2010) for each population (Table 1). These included six fitness traits (adult lifespan, fecundity, overwintering survival of eggs, survival from eggs to adulthood, overall fitness), two activity traits (proportion of adults flying, flight take-off response) and three morphometric traits (fresh body weight, elytra length, elytra width). Fitness traits To assess adult lifespan and fecundity, newly emerged male-female pairs were transferred into small bioassay containers (L: D, 24: 18 °C, light regime L: D, 14: 10). The containers consisted of two plastic urinalysis cups (diameter: 48 mm, height: 80 mm), stacked one inside the other providing 175 cm3 of space as per Toepfer et al. (2012). The upper cup had a 10 - mm hole in the bottom to give the female access to the lower, soil-filled cup for egg-laying. Abundant - for the insects unlimited - food was provided in each container. This was, two soft, unripe kernels of organically-produced maize, one 13 x 13 x 13 mm piece of zucchini flesh, one 13 x 13 x 13 mm piece of pumpkin flesh (Li et al., 2009), and a 5 x 5 x 5 mm piece of artificial pollen diet (Branson & Jackson, 1988). A 10 x 5 x 5 mm cube of 15% watery agar served as a water source for the adults. Food and agar were changed every 5 to 7 days. All pairs of D. v. virgifera were provided with the same amount of food, as D. v. virgifera fitness and activity is influenced by diet experience (Mabry et al., 2004). To assess adult lifespan, bioassay containers were daily checked for live and dead adults. The date of death was recorded, lifespan calculated, and dead adults removed. The experiment was stopped after 70 days because it was thought that this period was long enough to reliably reflect the total adult lifespan (Li et al., 2009). The proportion of females and males surviving until day 70 was calculated for each tested population. To assess fecundity, two teaspoons of moist, sterile black field soil (sieved at 0.15 mm mesh size; 25-35 w% moisture) were placed into the lower cup of the bioassay containers after 7 days maturation period (Branson & Johnson, 1973; Hill, 1975). Then, every 14 days after adult emergence, the lower cup of the bioassay container (containing soil and eggs) was removed and replaced with a new one. The soil with eggs was washed with tap water through a 0.25 mm sieve, and recovered eggs were counted. The experiment was stopped after 70 days (Li et al., 2009). The realised fecundity was calculated for each individual female. Eggs were then stored in sterile moist soil for pre-diapause at 24° C during two weeks (Branson, 1976; Krysan, 1972), and then for diapause at 6 to 8°C during another five months (Krysan, 1982). To determine the overwintering survival of eggs, the soil with eggs was washed through a sieve after five months diapause. Egg survival per individual parental female of each tested population was assessed under stereomicroscope according to Modic et al. (2005). To determine survival from eggs to adulthood, 200 successfully overwintered viable eggs were incubated at a temperature of L: D, 25: 21 °C for 14 to 21 days to initiate hatching. As not each female had laid enough eggs to obtain 200 eggs after overwintering, this experimental part was not conducted on an individual female base, but with pooled egg batches. About 2,000 to 3,000 organic, untreated maize seeds (var. Gavott UFA Semences, Bussigny, Switzerland) were soaked in water for 24h and germinated in a plastic tray (330 x 190 x110 mm) with a gauze lid. Three days after germination, ready-to-hatch eggs were transferred onto the seeds in the plastic trays and maintained at a temperature of L: D, 25: 21 °C, and light regime of L: D, 14: 10. Emerging larvae found unlimited maize roots. After 14 to 20 days, third instar larvae were transferred along with the maize seedlings into gauze-covered cylinders (diameter: 120 mm, h: 140 mm) containing sterilised field soil for pupation (sieved at < 5 mm mesh size, 25-35 w% moisture). The transferred maize still provided food for the larvae until entering the soil for pupation. Adults started to emerge about one week later. Adult emergence was recorded daily until no emergence was anymore recorded during four consecutive days. The total number of emerged adults was divided by the initial number of 200 incubated viable eggs. Survival from eggs via all developmental stages until adulthood was calculated by combining the data from the overwintering survival of eggs with the survival from larvae and pupae until adulthood. The overall fitness of each individual D. v. virgifera female equals the net reproductive rate R of a population (Begon et al., 1990) and was calculated according to Lombaert and al. (2008) assuming a 50 : 50 sex ratio (Spencer et al., 2009; Toepfer & Kuhlmann, 2006): Fitness = Larvae to adult survival * Adult sex ratio * Realised fecundity * Egg overwintering survival Activity traits Flight activity was assessed as a measure of dispersal capability using seven-day old adults (Naranjo, 1991). Activity measures on young adults better reflect activity differences between individuals and between populations than measures on mature adults which are influenced by nutritional status and egg load (Li et al., 2010). Flight activity was assessed by measuring the proportion of adults flying and the flight take-off response as described by Li et al.(2010) between 14:00 and 16:00 under laboratory conditions of 24 °C and 60% relative humidity. Flight stands consisted of a wooden pin (h: 40 mm, diameter 10 mm) fixed onto the end of an inverted white plastic funnel (h: 160 mm, diameter 135 mm at base). The base of the funnel was surrounded with water to prevent the adults from walking off the stand. An individual adult was gently transferred onto the base of the funnel using an aspirator device. Following adult release, the incidence of take-off and the time from release to take off were recorded. The trial ended when the adults flew off the stand or when 300 seconds had elapsed (Li et al., 2010). The proportion of adults flying and mean time until take-off was calculated for female and male adults (non-fliers were recorded as taking more than 300 seconds). Tested adults were returned to the bioassay containers. Morphometric traits Fresh body weight, elytra length and elytra width were measured on each individual D. v. virgifera within 24 h following adult emergence (= initial morphometric traits without feeding). This is, because measures on young adults are less influenced by nutritional status and egg load than older adults (Li et al., 2014). The tested young adults were assumed to no longer be teneral, as they were fully coloured and did not have the light grey soft body typical of adults just after emergence. Fresh body weight was measured by transferring adults into a small plastic container and weighting them on a 0.1 mg to 160 g precision scale (Fox & Scheibly, 2006). Individuals were then returned to the bioassay containers. Elytra length and width (i.e., single measurement of both elytrons together across the dorsum) were measured according to Li et al. (2009;) and Mabry et al. (2004). Elytra length and width were chosen among other morphometric traits, such as hind tibia length or head capsule width because measures of elytra characters seem to well-reflect fecundity, life span and activity of D. v. virgifera (Li et al., 2009; Li et al., 2010). Adults were placed on a cool but not frozen pad (Icepack Migros, Delémont, Switzerland) to limit their activity during the measurements with a micrometre scale to the nearest 0.06 mm under a stereomicroscope (16 × magnification). Analyses of phenotypic trait differences among populations The “Family structure” was considered as a random factor in the applied linear mixed models (LMM) for normal distributed errors, generalized linear mixed models (GLMM) for other error distributions, or mixed Cox models for time-series data. The environmental variance was kept minimal because all animals were reared in the same laboratory conditions. Under these conditions, the inter-family variance of a trait reflects its genetic variance. This allowed us to compare the mean values and genetic variances of each phenotypic trait between populations. Our models allowed testing whether hybridization had an impact on the means and variances of phenotypic traits. The same models were used for each trait with the independent variables “Population type” (with the modalities NW Italian, CSE European (Serbia), each of four hybrids), “Sex” (female and male) and “Experimental series” (series 1 and 2) as fixed effects. We determined for each trait, whether the inter-family variance (the genetic variation proxy) varied or not between the “Population types”. For each trait, we compared a model with a nested random structure “Population type | Family” with a model with a simple random structure “1 | Family” (Table 4). The best model was chosen on the basis of the smallest value of the Akaike information criterion (AIC) (Burnham and Anderson, 2004). For some traits, only one or both sexes together were concerned in the phenotypic measures (fecundity, overwintering survival of eggs, survival of larvae and pupae until adulthood). Consequently, the two sub-types of generic models used to cover all the studied traits and factors were as follows: Phenotypic trait ~ Population type * Sex + Experimental series + Random factor Phenotypic trait ~ Population type + Experimental series + Random factor From these pre-established models we (i) determined the best random structure and its significance in the model; (ii) estimated the significance of the fixed effects; and (iii) tested the equality of traits for all comparisons among parental and hybrid populations. For that, multiple comparisons of the means of the phenotypic traits were conducted using the R package “multcomp” (R Development CoreTeam, 2012) and through defining a matrix of contrasts. The obtained p-values allowed us to determine which comparisons of mean phenotypic traits were significantly different between population types (see parental and hybrid populations in Table 4). All ten phenotypic traits (fecundity, egg overwintering survival, survival from eggs to adulthood, fitness index, adult lifespan, proportion of adults flying, flight take-off response, fresh body weight, elytra length, elytra width,) were analysed. Most traits were analysed by using LMM (with a normal error distribution). For that, data have been transformed to improve normality with a square-root transformation for fecundity and survival from eggs to adulthood, and with a logarithmic transformation for the fitness index. The proportion of adults flying (a binomial distribution of the error) was analysed by using a GLMM. The adult lifespan and flight take-off response were analysed by using a mixed Cox models (lifespan time-series censored to 70days, flight take of response to 300 sec). Normality of data had been visually confirmed for untransformed data as well as after-transformation- data before applying any model. The over-dispersion of binomial data has been tested for the proportion of adults flying by computing the over-dispersion parameter Φ, the ratio of the residual deviance of the model over the residual degrees of freedom.