Since transgenic crops expressing Bacillus thuringiensis (Bt) toxins were first released, resistance evolution leading to failure in control of pests populations has been observed in a number of species. Field resistance of the moth Busseola fusca was acknowledged 8 years after Bt maize was introduced in South Africa. Since then, field resistance of this corn borer has been observed at several locations, raising questions about the nature, distribution and dynamics of the resistance trait. Using genetic markers, our study identified four outlier loci clearly associated with resistance. In addition, genetic structure at neutral loci reflected extensive gene flow among populations. A realistically parameterised model suggests that resistance could travel in space at speed of several kilometres a year. Markers at outlier loci delineated a geographic region associated with resistance spread. This was an area of approximately 100 km radius, including the location where resistance was first reported. Controlled crosses corroborated these findings and showed significant differences of progeny survival on Bt plants depending on the origin of the resistant parent. Last, our study suggests diverse resistance mutations, which would explain the widespread occurrence of resistant larvae in Bt fields across the main area of maize production in South Africa.
Adaptation to human driven environmental pressures present insightful illustrations of the various facets of contemporary evolution and their implications. The evolution of resistance, the process by which a population of organisms acquires a genetically based decrease of susceptibility to a toxic compound (Tabashnik et al., 2009), may result in spectacularly fast responses to strong selective pressures (e.g., Mallet, 1989). Genetically modified crops, expressing Bacillus thuringiensis (Bt) toxins, are expected to result in elevated and continuous selection whenever Bt crops dominate the landscape (Tabashnik et al., 2013).
Evolution of resistance to Bt toxins expressed by crops is a direct threat to sustained control of pest populations, which may seriously impact agricultural yields. Even if a history of 10−20 years has revealed the first empirical trends (Tabashnik et al., 2013), the questions of how fast resistance to Bt crops toxins is likely to evolve and how quickly it may spread remain challenging. While Bt crops have been deployed for almost two decades, loss of control in pest populations, due to field-evolved resistance, arose in a few pest species only. However, decrease of susceptibility to transgenic Bt toxins (field-evolved resistance), which has not led to control failure, has been observed in many species (Tabashnik et al., 2013). Empirical knowledge about the spatial dynamics of resistance spread is scarce. Extensive gene flow is commonly observed in pest populations (e.g., Daly and Gregg, 1985; Korman et al. 1993; Bourguet et al., 2000; Endersby et al., 2007; Kim et al., 2009) and may promote rapid diffusion of resistance alleles across space (e.g., Peck et al., 1999; Bourguet et al., 2000).
The main option for delaying resistance evolution is the refuge strategy, which amounts to planting non-Bt plants within or surrounding Bt fields, allowing the survival of some susceptible larvae in a Bt-dominated environment. The proportion of refuge crop required to effectively delay resistance evolution is strongly dependent on the recessiveness of the resistance trait and, more generally, on the survival of heterozygote-resistant individuals at the dose of toxin expressed in plant tissues (e.g., see Vacher et al., 2003; Vacher et al., 2004).
The nature and uniformity of the genetic bases of resistance to Bt toxins are thus crucial aspects influencing resistance evolution and management (e.g., Tabashnik et al., 1998). Multiple recessive mutations that cause resistance to Cry1A toxins in Lepidoptera have been reported (see Heckel et al., 2007) in genes encoding Cadherin-like proteins (e.g., Morin et al., 2003) and ABC transporters (e.g., Baxter et al., 2011). In the field, diverse alleles conferring resistance were identified within Chinese populations of Helicoverpa armigera (Zhang et al., 2012), including recessive resistance alleles at the Cadherin locus as well as two non-recessive mutations, one of which was not linked to this locus.
Strong selection acting at a particular resistance locus is expected to affect not only the frequency of the selected mutation but also flanking polymorphisms. Alleles present on the ancestral chromosome on which a favourable mutation arose will tend to increase in frequency as the mutation invades the population, a consequence, which is known as the genetic hitchhiking effect (Kojima and Schaffer, 1967; Maynard-Smith and Haigh, 1974). Selective shifts may therefore give rise to strong linkage disequilibrium, i.e., some non-random allelic associations at the population level, between loci of the genomic region under selection.
Various population genetics approaches have aimed at taking advantage of such phenomena to capture signatures from positive selection (Jensen, 2014), notably in non-model organisms (e.g., de Villemereuil and Gaggiotti, 2015). The size of a hitchhiking effect depends on several factors, including: age and frequency of the mutation under selection; the intensity of selection; and recombination distance to the locus under selection. In particular, the initial frequency of the selected allele in the population before a selective shift, as well as the age of a mutation, has a crucial role (Maynard-Smith and Haigh, 1974; Jensen, 2014). As a consequence, the effectiveness of genome scans is lowered when e.g.: populations have undergone a particular demographic history (e.g., bottleneck); a recessive allele is involved; or selection acts on an ancient and previously neutral allele (Teshima et al., 2006; Jensen, 2014).
In South Africa, the pest Busseola fusca (Lepidoptera, Noctuidae) evolved resistance to the Bt toxin Cry1Ab expressed by transgenic maize within 8 years of the technology being released (Van Rensburg, 2007). Although this pest species is seldom hosted by wild grasses (Le Rü et al., 2006), it is responsible for substantial yield losses in maize, sorghum (Kfir et al., 2002) and more marginally sugarcane crops (Assefa et al., 2015). Being endemic to Sub-Saharan Africa and widely distributed throughout the region (Dupas et al., 2014), B. fusca is considered a pest of economic importance. To date, field resistance in B. fusca remains one of the few cases in the world where Bt crop failure has clearly been documented (see Tabashnik et al., 2013). Resistance evolution in this species was initially ascribed to an insufficient implementation of the refuge strategy (Kruger et al., 2012; Van den Berg et al., 2013). Nonetheless, the existence of a dominant resistance trait, such as found in this pest (Campagne et al., 2013), is expected to have greatly accelerated the spread of resistance and to have drastically reduced refuge efficiency. Furthermore, the presumed low survival of resistant individuals on non-Bt plants is unlikely to have mitigated the fitness benefits conferred by the resistance trait since no fitness cost associated with resistance was found in B. fusca (Kruger et al., 2014).
Four years after being first recorded in the field in 2006, resistance in B. fusca was observed at a distance of ~40 km from the location reported initially (Kruger et al., 2011). By 2011, Bt maize fields infested by B. fusca were observed in various places up to ~400 km distance from the initial location (see van den Berg et al., 2013). The latter sequence of observations raises a number of puzzling questions regarding both the initial distribution and spatial dynamics of resistance traits to Bt-maize in this species. Although gene flow may either swamp local resistance evolution or contribute to the spread of resistance alleles (Bourguet et al., 2000), little is known about the levels of migration and gene flow taking place in populations of B. fusca at sub-regional scales.
On the basis of empirical genetic data collected in the field, controlled crosses, and theoretical considerations, the purpose of this study was to investigate dynamics and diversity of resistance evolution in field populations of B. fusca. Our aims were to: (i) to evaluate the extent of gene flow in this pest within and between sites distributed across the major region of maize cultivation in South Africa; (ii) to identify markers associated with resistance and further investigate spatial distribution of resistance traits; and (iii) to explore the geographic variability of resistance inheritance in B. fusca within the study region.
Materials and methods
Study context and sampling
The first official observation of field resistance in B. fusca was recorded in the Christiana area (Van Rensburg, 2007; Kruger et al., 2011) which is located at the western extremity of the main maize production area (Van den Berg et al., 2013). Available data indicate, between 1998/1999 and 2010/2011, an average proportion of 60% Bt maize in the area where resistance was first reported (Kruger et al., 2012). The landscape is composed of 20% crops (the main one being maize) while most of the matrix consists of semi-natural or natural habitats (source: FAO GeoNetwork, www.fao.org). In this area, populations of B. fusca are bivoltine with a diapause stage between cropping seasons. Sampling was performed during the growing season 2010/2011, i.e., ~k≈26 generations after Bt maize was released, across the main region of maize production in South Africa (stretching across more than 600 km East-West).
To carry out population genetics analyses, 20–25 larvae per field were sampled in a total of 28 fields (14 non-Bt fields and 14 Bt fields, Table 1). Sampling was performed within maize fields at advanced stage of development (late vegetative stage or higher), and consisted in five collection points distributed along a line, separated by ~50 m. At each point, we collected five larvae originating from five different plants separated by a few metres from each other. Care was taken to sample larvae in their fourth instar at least, in order to minimise the number of susceptible larvae collected in Bt fields. Larvae originating from different plants were conserved separately in 90% ethanol. Finally, a crude assessment of field damage level was performed in most sampled fields by recording the number of plants with leaf damage in three separate rows chosen at random (at a distance>20 m from each other, depending on an arbitrary number of steps) of 100 plants each (see Table 1). In order to perform controlled crosses, an additional number of larvae was collected in Bt fields at sites where resistance had been reported.
Total DNA isolation of 20 mg of larvae thoraces was done using the DNeasy Tissue Kit (Qiagen, Hilden, Germany). DNA concentration was determined using a Nanodrop (Thermo Scientific, Wilmington, DE, USA). The amplified fragment length polymorphism (AFLP) reaction was performed as described by Vos et al. (1995) with slight modifications. A quantity of 300 ng of DNA was digested using the following restriction enzymes: EcoRI and PstI. Digestion was made overnight at 37 °C with five units of EcoRI (New England Biolabs, Ipswich, MA, USA), and five units of PstI (New England Biolabs) in a total volume of 25 μl. Ligation of double-stranded EcoRI and PstI adaptors to the ends of the restriction fragments was performed (for 8 h at 15 °C) by adding 10 pmol of EcoRI and PstI adaptors, five units T4 DNA ligase (Promega, Madison, WI, USA). A 5 μl volume of 10 × diluted ligation products were used as a template for pre-amplification, using 10 pmol of EcoRI and MseI primers, 0.4 mm dNTPs, 1.5 mm MgCl2 and 1 unit of GoTaq Flexi DNA Polymerase (Promega) in a final volume of 25 μl. The pre-amplification thermocycle profile was: 94 °C for 5 min, followed by 20 cycles at 94 °C for 30 s, 56 °C for 1 min, 72 °C for 1 min and 72 °C for 5 min. Selective amplification steps were performed using 5 pmol EcoRI (+NN) and PstI (+NN) primers with 2.5 μl of 20 × diluted preamplification product in a final volume of 12.5 μl. Each selective amplification reaction mixture contained 0.4 mm dNTPs, 1.5 mm MgCl2 and 0.5 unit of GoTaq Flexi DNA Polymerase (Promega). The selective amplification thermocycler profile was: 94 °C for 2 min, followed by 12 cycles at 94 °C for 30 s, 65 °C for 30 s, 72 °C for 1 min, followed by 23 cycles at 94 °C for 30 s, 56 °C for 30 s, 72 °C for 1 min and 72 °C for 5 min. The different primers combinations used were: EcoRI+AA/PstI+AG, EcoRI+AG/PstI+AG, EcoRI+GA/PstI+AT, EcoRI+AG/PstI+AT, EcoRI+GA/PstI+AG, EcoRI+CA/PstI+AA, EcoRI+CT/PstI+AT, EcoRI+TG/PstI+GA and EcoRI+GG/PstI+AC. AFLP markers were scored according to the absence/presence of peaks using Genemapper V4.1 (Invitrogen/ThermoScientific, Wilmington, DE, USA). To optimise the reliability of the markers, 65 individuals were processed twice. Loci characterised by more than three discrepancies (5%) were considered prone to genotyping errors and were discarded. A total of 629 polymorphic markers were retained for further statistical analysis. The overall genotyping error rate per locus (see Bonin et al. 2004), calculated as the ratio between observed number of differences and total number of comparisons, was 1.3%.
The inbreeding coefficient FIS was estimated at both population and individual levels, for each sampled field using the software I4A, which is designed to deal with dominant genetic markers (Chybicki et al., 2011). After testing the convergence of different run lengths, data were analysed using 10 000 burn-in and 50 000 sampling iterations considering flat priors for the inbreeding coefficients. Estimates were provided with 95%-credibility intervals.
Rapid resistance evolution is expected to have left specific signatures in the genome of B. fusca as directional selection imposed by Bt toxins may result in increased differentiation between populations at hitchhiking loci. Using Bayescan (see Foll and Gaggiotti, 2008), we scanned the genome of this pest for loci at which genetic differentiation between populations (FST) is significantly higher than the genome baseline (outlier loci). Two sets of markers were defined and analysed separately: non-outlier loci, expected to reflect neutral processes such as genetic drift and gene flow between populations; and outlier loci, further tested for being associated with resistance.
To analyse the set of non-outlier loci, Euclidean genetic distances between individuals were computed (Peakall and Smouse, 2006) and genetic differentiation (ΦST) among fields was estimated with an analysis of molecular variance (Excoffier et al., 1992; Peakall, Smouse 2006). Statistical significance was tested using 999 permutations. A between-field analysis was performed on Principal Coordinate Analysis (PCO) axes (ade4 package, R Core Team, 2012). In addition, pairwise kinship coefficients (Fij) were estimated using the software SPAGeDi (Hardy and Vekemans, 2002). Estimation of kinship coefficients based on dominant markers requires an externally supplied estimate of the inbreeding coefficient. As the vast majority of individuals did not exhibit evidence of inbreeding (see Results section) we used F=0. Average kinship was then calculated within and between fields. The genetic neighbourhood size (Nb*), a measure of the local extent of breeding, was estimated following Hardy's (2003) procedure: Nb*≈(F0–1)/b, where F0 is the average kinship for adjacent individuals (within a field) and b is the slope of the linear decay between kinship and the natural logarithm of distance. Pairs of geographic distances higher than 100 km were excluded (see Hardy, 2003) and a confidence interval was calculated using 999 bootstrap iterations. Nb* is an empirical estimate of the theoretical demographic parameter Nb=4πσ2ρe, where ρe denotes the effective population density (e.g., see Rousset 1997; Hardy, 2003) and σ2 (km2 per generation) is the axial variance of an isotropic Gaussian dispersal distribution (2σ2 being the mean squared parent–offspring-displacement).
Association between field resistance and the genetic structure at outlier loci was then investigated. Analyses of population clustering were performed exclusively on individuals which were sampled in Bt fields. Ward's method for hierarchical clustering based on pairwise genetic distances (Peakall and Smouse, 2006) and the software STRUCTURE 2.3.4 (Pritchard et al., 2000) were used to identify genetic clusters. The latter analysis was executed under the admixture model and correlated allele frequencies, with 10 000 burn-in iterations and 50 000 main iterations. The optimal number of genetic clusters (K) was determined following the procedure described in Evanno et al. (2005), with five replicates for each K-value. For the sake of comparison, the same analysis was performed using the set of non-outlier loci.
The difference in marker frequency at outlier loci between Bt and non-Bt fields was tested using generalised linear mixed models (Venables and Ripley, 2002) with spatially correlated random effects, assuming a binomial distribution. Crop type (i.e., Bt vs. non-Bt) was treated as a fixed effect and the geographic coordinates of the fields were considered as random components of the analysis. Correlations between marker frequencies at outlier loci were evaluated with Pearson's correlation coefficient (R Core Team, 2012). Linkage disequilibrium between outlier loci was tested following Hill's procedure (Hill, 1974), which provides a parametrically explicit maximum-likelihood computation method for dominant markers. Both a standardised estimate of linkage disequilibrium (r) and its associated log10-likelihood ratio (LR) were calculated.
Heritability of resistance traits provides insights into the genetic bases of resistance. To explore the geographic variations in larvae survival on Bt maize, we performed single-pair crosses between resistant parents and a strain of a priori susceptible larvae originating from Tanzania (Tz), where no Bt crops are grown. Three sites, where resistance had been reported, were selected: Valhaarts (Va), an irrigation scheme close (~50 km) to the location, where resistance was first reported and in which resistance has been previously studied (Kruger et al., 2011); Lichtenburg (Li) at 180–200 km distance from the initial location; and Bethleem (Be), at ~300 km from the initial location (see also Figure 1). Controlled crosses were performed between adults originating from Bt fields and susceptible individuals: Va × Tz; Li × Tz; and Be × Tz. Progeny survival was assessed in a greenhouse on Bt and non-Bt maize planted in pots. Maize plants were inoculated with 20–30 neonates, which were gently transferred with a paintbrush within the whorl of young maize plants (i.e., from the 5th leaf collar stage). Plants were contained in fine-meshed nets to avoid larvae escaping. The number of individuals surviving on a plant was recorded after a period of 21 days. Preliminary observations indicate that survival after 2 weeks on Bt maize of individuals originating from Tanzania, is typically ⩽2%. A fraction of the offspring were reared on non-Bt maize as a control. About 1700 larvae were used in the experiment. Survival of the progeny on maize was compared using Fisher’s exact tests (R Core Team, 2012). Contingent upon emergence of F1 adults and various experimental constraints, sib-crosses were undertaken in order to test whether homozygosity could reinforce resistance in F2 offspring of selected parents. One sib-cross was successfully performed among individuals originating from the Li × Tz cross and which survived on Bt maize.
Theoretical resistance spread
One approach for understanding the geographic distribution of resistance is to model the spatial spread of resistance alleles. In our survey, larvae were found to survive on Bt maize at various locations which also correspond to sites where resistance was reported (Van den Berg et al., 2014), translating into a vast expanse of approximately 200 km × 300 km. Assuming relatively simple genetic bases of resistance, the most elementary scenarios consistent with these observations could include: (i) a rapid propagation of a unique mutation conferring resistance from an initial location; or (ii) ancient and widespread resistance mutation(s), before the selective shift. We explored the feasibility of a dominant mutation spreading across space, from an initially confined distribution, under the selective pressure of Bt maize. Resistance was considered to involve a single locus, with a resistance allele (R) of frequency q, and a susceptible allele (S) of frequency 1 – q. For modelling purposes, the spatial distribution of Bt and non-Bt plants is considered continuous and random. The proportion of Bt crop in the landscape (ω) is assumed to determine the survival probability (e.g., see Tabashnik et al., 2008; Jin et al., 2015) of the three genotypes (RR, RS and SS): WSS=1 – ω, WRS=(1 – ω)+h × ω and WRR=1, where h denotes the dominance of the resistance allele. Spatial spread of a resistance allele under selection was considered under a demogenetic system of deterministic reaction-diffusion equations proposed by Tyutyunov et al. (2008) and slightly modified to our purpose (see Supplementary Appendix S1). The reaction term of the model accounts for local increase of resistance allele frequency, in relation to ω, h, WSS, WRR, q and the birth rate β per individual and per generation. The diffusion term captures the effect of individual dispersal under diffusive Brownian motion (diffusion constant D), scaling to 2σ2=4D after one generation.
Genetic structure at non-outlier loci
The five primer combinations yielded 629 polymorphic and reproducible fragments in B. fusca, for the entire set of 693 individuals. All individuals had different AFLP profiles, differentiated by at least eight markers. Estimates of population inbreeding coefficient were low, ranging between 0.007 and 0.058 in the 28 sampled fields (Table 1), indicating no strong heterozygote deficiency. Similarly, estimates of median individual inbreeding were <0.01 in ~97% of individuals. The outlier analysis carried out with all 28 fields detected 26 loci characterised by FST>0.07 (logarithm of the posterior odds, log10 PO>1) while the non-outliers were characterised by a lower differentiation FST=0.027 (s.d.=0.008). On this basis, loci were split into two groups: ‘outlier’ and ‘non-outlier’ loci and were further analyzed separately.
Evidence for extensive gene flow between populations was found when investigating the genetic structure of non-outlier loci. Overall genetic differentiation among fields was low but non-null (ΦST=0.03, P<0.001). Even at long distances (>400 km), average pairwise genetic differentiation between sampled fields was weak (ΦST≈0.04). In spite of a lack of genetic differentiation at broad spatial scales, a smooth pattern of isolation by distance was found. Firstly, pairwise ΦST were associated with geographic distances (R=0.47; P<0.001). Similarly, kinship coefficients were significantly correlated with log-transformed distances (R=0.52; P<0.001; Figure 2a). Parameters of the latter regression yielded the estimate for the neighbourhood size, Nb*≈(F0–1)/b=706 with a 95% bootstrap interval (526, 1010), which translates into ρe.σ2≈55 (i.e., with Nb*≈4πσ2ρe), which is well above the theoretical threshold of ρe.σ2>1 at which a population distributed in a continuous space behaves nearly as a panmictic population (see Maruyama, 1972). Secondly, a between-class (i.e., fields) PCO (Figure 2b) revealed congruent mapping between B-PCO axes and the geographical origin of the field-collected populations (see Figure 1 for the locations); e.g., clusters of fields (5, 6, 7), (2, 3, 4) and (54, 64, 68, 71, 75). The small amount of variation captured by the analysis (6.7% of the total inertia; 1.7% onto axes 1 and 2) consistently reflected low levels of genetic structure.
Outlier loci and resistance
The genetic structure of Bt fields at outlier loci revealed two spatially segregated groups of Bt fields, henceforth A1 and A2 (Figure 3). The optimal number of clusters obtained using a Bayesian clustering method (STRUCTURE) was K=2 in both sets of loci; higher values of K did not result in better fits of the model (see Supplementary Appendix S2). Although essentially no genetic structure appeared in non-outliers, a clustering based on outlier loci brought to light a clearer pattern which was further supported by the Ward's clustering method (Figure 3). One cluster was composed of five Bt fields (3, 5, 7, 43 and 65) which were spatially aggregated (area A1, Figure 1) in the surroundings of the location where resistance was first reported. While the extent of A1 could be sketched as an area of ~100 km radius, A2 was defined as an outgroup which included the remaining nine Bt fields.
A second round of outlier analysis (using Bayescan) was performed within each group A1 and A2, including the four closest non-Bt fields (1, 2, 4 and 6) in A1 and the remaining ones in A2. These additional analyses did not yield outliers which were not detected previously (Table 2). However, some outliers returned in the first round were not detected as significant (log10 PO<1) in the second round, presumably due to a reduction in statistical power.
The association of marker frequency at outlier loci with Bt and non-Bt fields was then examined. Mixed generalised linear model with spatially correlated random effects showed significant differences in frequency at four outlier markers (L192, L204, L299 and L369) in A1 (P<0.005, Table 2) whereas no significant association was found in A2.
These four loci were therefore considered strong candidates associated with resistance to Bt toxins. Spatial distribution of these putative hitchhikers within A1 was highly consistent (Figure 4): marker frequency in the Bt fields was either systematically lower (L192, L204 and 299) or systematically higher (L369) than in the non-Bt fields. It is worth noticing that these contrasts were stronger in the two most Southern Bt fields of A1 (5 and 7), suggesting a higher resistance allele frequency in these fields. Overall, allelic frequencies at the four outlier markers (in the 28 fields) were significantly inter-correlated (see also Supplementary Appendix S3). Pearson's correlation coefficients were: 0.6<R<0.7 (P<0.001) in the pairs (L192-L204), (L192-299) and (L204-299); R=-0.51 (P=0.005) in the pair (L204-L369); R=-0.44 (P=0.017) in the pair (L192-L369); and not significant in one pair of loci only, R=−0.30 (P=0.115).
In addition, linkage disequilibrium was detected among putative hitchhikers within the smaller and more homogeneous area A1. Highly significant pairwise linkage disequilibrium values 0.42<| r |<0.47 (log10-likelihood ratios LR⩾3) were obtained for the pairs of loci (L192-L369), (L204-L369) and (L204-L299), while the remaining three pairs were characterised by r-values>0.25: r=0.38 for the pair (L192-L369) (LR=2.4), 0.25<| r |<0.32 (LR>1.13) for (L299-L369) and (L192-L299). Consistently, genetic differentiation at these four outlier loci was elevated between Bt and non-Bt fields of A1 (ΦST=0.24; P<0.001) as well as between Bt fields of A1 and A2 (ΦST=0.21; P<0.001). By contrast, it was low (ΦST<0.05) within either type of field (Bt or non-Bt) and within each area (A1 and A2).
The site (Va) was located in the area A1 in which genetic contrasts were observed, while (Li) and (Be) were located farther apart in A2 (see Figure 1). Overall survival was 0.67 when offspring was reared on non-Bt maize, significantly higher (P<0.001) than in offspring reared on Bt maize (Figure 5). When reared on Bt maize, survival of Va × Tz offspring (0.32) was clearly higher than both Li × Tz and Be × Tz offspring (P<0.001). Given substantial heterogeneity within the two latter crosses, the highest survival in each type of offspring (0.11, 0.13) was compared with that of Va × Tz offspring reared on Bt; it was unambiguously confirmed that progeny obtained with individuals from Valhaarts outperformed progeny of the other crosses on Bt maize (P<0.001; Figure 5). Although survival in a Li × Tz progeny reared on Bt maize was particularly low (6/383), a F2 sib-cross between two out of the six survivors resulted in significantly increased survival on Bt maize (51/251, P<0.001), suggesting the existence of a recessive resistance trait in (Li).
Theoretical resistance spread
On the basis of a demogenetic system of deterministic reaction-diffusion equations, we show that the lower propagation speed (V) of resistance is a constant which takes the form (Supplementary Appendix S1, see also Tyutyunov et al., 2008):
The time required for a concentric wave of resistance to travel a distance d (km) may simply be approximated by T≈d / V generations.
From controlled crosses performed in the laboratory we estimated that a female of B. fusca produces on average ~250 viable neonates (see also Khadioli et al., 2014). In laboratory experiments, survival (egg to adult stage) of B. fusca has been estimated to be, respectively, of order 40% and 15–20% in optimal and in sub-optimal temperature conditions (Khadioli et al., 2014). In a different species of Noctuidae (Helicoverpa zea) the estimate of survival to adulthood in the field was close to 5% (Vargas and Nishida, 1980). Assuming a 1:1 sex ratio, we determined a conservative estimate of β=0.05 × 250/2≈6.
A resistance allele being propagated under constant selective pressure typically results in fast rapid spread. On the basis of equation (1), we explored the scale at which a dominant resistance allele may spread within ~26 generations (i.e., 13 years, and assuming h=1, ω=0.6, β=6) from an initially local distribution. Selection due to a moderate fraction of Bt crop in the landscape (ω≈0.6) and operating on a fully dominant resistance allele (h=1) yields speeds of several kilometres per year, e.g., 4.6 km per year<V<17 km per year for 1<σ2<10. Over 26 generations, the spread of resistance would attain a radius of several tens of kilometres (i.e., ~60–220 km) (Figure 6). Under the same parameter values, a high axial dispersal variance (σ2≈25 km2 per generation) would be required for the resistance wave to travel from the location where resistance was first reported to the farthest Bt fields in which individuals were found to survive (Figure 6).
Being one of the few instances of field-evolved resistance leading to Bt crop failure (Tabashnik et al., 2009; Tabashnik et al., 2013), Bt-resistance in B. fusca may be considered an insightful study case in the context of Bt crop deployment worldwide. Our results confirmed that a sharp selective shift, as indicated by the extensive uptake of Bt maize (Kruger et al., 2012), resulted in rapid evolution of resistance. Beyond the detection of outlier loci, our results provide further understanding of both gene flow and field resistance in this pest, notably in terms of spatial distribution, dynamics and diversity of resistance in the field.
Gene flow and resistance spread
In line with what was shown in several other moth species (e.g., Korman et al., 1993; Nibouche et al., 1998; Bourguet et al., 2000; Han and Caprio, 2002), our empirical results reflect extensive gene flow and low levels of population inbreeding in B. fusca. To a first approximation, the whole population seems close to panmixia at coarse scales (700 km × 400 km). It has been suggested that B. fusca could be characterised by limited dispersal capacities (Sezonlin et al., 2006). However, consistent with Dupas et al. (2014), our study did not reveal strong barriers to gene flow. Estimate of neighbourhood size in B. fusca (Nb*≈700 individuals) had the same order of magnitude as in the European corn borer Ostrinia nubilalis (Nb*≈500 on maize, see Martel et al., 2003). Classical methods for estimating gene flow (see Rousset, 1997; Hardy, 2003) cannot yield a separate estimation of the parameters σ2 and ρe, such that the product of these has to be considered, i.e., ρe.σ2≈55 in our study. Assuming an effective density of 14<ρe<55 individuals per km2 therefore amounts to assuming a dispersal parameter 1<σ2<4 km2. Similar effective population densities, i.e., 5<ρe<50 individuals per km2, have been proposed in other species of wild moths (Saccheri et al., 2008) and moth pests (Martel et al., 2003). Mean flight distance of several kilometres was observed in males of the moth pest Spodoptera litura, using a release-recapture experiment. In B. fusca, observations of dispersing females crossing an inter-field of 1.6 km have been reported (see Harris and Nwanze, 1992). Although dispersal distances and timing may differ between males and females, these observations are roughly compatible with a dispersal parameter σ2 scaling to a few km2 per generation.
Extensive gene flow in pest populations may affect the pace of resistance evolution by mitigating local increase of resistance with susceptible alleles originating from non-Bt fields, or by contributing to an accelerated spatial spread of resistance pockets. Besides gene dispersal (σ), the speed of resistance propagation crucially depends on the inheritance of resistance (i.e., h, the dominance parameter in our model) and the population ecology of a pest, notably its birth rate (β). These parameters would clearly deserve further consideration in order to better comprehend the spatial dynamics of resistance. Nonetheless, realistic numerical applications suggested a travelling speed of a few kilometres per generations. It seems plausible that a resistance pocket, initially small, may have quickly expanded into a bigger area of radius several tens of kilometres or more. Within 13 years (k=26 generations), one might expect the radius of such a pocket to increase by ~100 km (provided ω≈0.6, h=1, β=6, σ2≈2.1 km2, ρe≈26.8), an order of distance which is commensurate with the size of A1. However, our model suggested that only high dispersal variance could result in a propagation radius that would encompass all apparently resistant populations, which might lie at the margin of what can be reasonably assumed.
Outlier loci and resistance
While larvae surviving in Bt fields were observed across a vast area, a genetic signature associated with field-evolved resistance could be attributed to a genetic hitchhiking process within a smaller region only. The existence of such a pattern is also suggestive of a recent geographical spread of resistance. Indeed, due to genetic recombination, associations between resistance allele and alleles at linked loci are not expected to persist in time, unless recombination distance is very low (see Maynard-Smith and Haigh, 1974). Correlated allele frequencies and linkage disequilibrium are not necessarily indicative of physical linkage as they may also result from various stochastic processes at population level. Nevertheless, given an observed lack of neutral genetic structure, even at coarse spatial scale, we expect neither population differentiation, demographic history nor genetic drift to strongly affect linkage disequilibrium patterns in the wild.
In line with genetic hitchhiking arising due to resistance evolution, consistent differences in outlier marker frequencies were observed between Bt and non-Bt fields. The size of these differences could reflect an intermediate frequency of resistance alleles in A1 (see also Supplementary Appendix S4). On the one hand, a review study (Tabashnik et al., 2013) evaluated the proportion of resistant B. fusca individuals to be≈q2+2hq(1 − q)⩾50%, in the core resistance region, which translates into a resistance allele frequency q>0.25 assuming a strictly dominant (h=1) monogenic trait. On the other hand further assuming a resistance allele exclusively associated with null alleles (a0) at either outlier loci L192, L204 or L299, one would expect the frequency of homozygotes a0/a0 (corresponding to an absence of marker) to be close to P(a0/a0)=q2/(q2+2hpq) in Bt fields, i.e., leading to q=2 × P(a0/a0)/[1+P(a0/a0)]. Given P(a0/a0)≈0.3 at these loci in Bt fields (see Table 2), we attain a crude estimate q≈0.46. Finally, relaxing the stringent assumptions of full linkage and strict dominance, this exercise suggests a resistance allele frequency 0.25<q<0.5 (see also Supplementary Appendix S4).
Note that some outlier loci could not be associated with resistance, which suggests that these loci may not be related to resistance at all, or reflect a lack of power in some analyses, notably due to the use of dominant genetic markers. However, it cannot be excluded that a fraction of these outliers, loosely linked to the resistance trait, underwent quick hitchhiking but did not shift significantly during the time for which associations persisted, contributing to some levels of genetic differentiation between areas (see also Supplementary Appendix S5).
Uniformity of resistance
Since the first observations of field resistance in B. fusca were reported, in the Western part of the maize cutivation region, most conspicuous failures to control populations of B. fusca in Bt fields (e.g., see Van Rensburg, 2007; Kruger et al., 2011) have been observed in the surroundings of this location. Moreover, the resistance trait associated with these failures appeared to be dominant (Campagne et al., 2013). Our results showing differential progeny survival on Bt plants and genetic contrasts (between A1 and A2) support the idea of a dominant resistance trait being confined within A1. Furthermore, it is coherent with the chronology of resistance development in South Africa (Van den Berg et al., 2013) and with the levels of infestation in Bt fields perceived by farmers at different locations: perceived field infestations >10% mostly occurred within 100−150 km from the site where resistance was initially reported (Kruger et al., 2012; Van den Berg and Campagne, 2014).
By contrast, information about resistance found within A2 is rather scarce (Van den Berg et al., 2013). Even if survival of some susceptible individuals on plant tissues expressing a low dose of toxins might partly explain the development of larvae on Bt fields within A2, our results would advocate for independent evolution of resistance. Indeed, increased survival on Bt maize obtained in a F2 sib-cross clearly suggested the existence of at least one recessive resistance trait in A2. Similarly to what has already been reported in another pest (e.g., Zhang et al., 2012; Jin et al., 2013), diverse resistance mutations characterised by different levels of recessiveness might therefore be responsible for the observed pattern of widespread field resistance to Bt maize in South Africa.
Finally, our results draw a valuable picture of resistance distribution and dynamics for one of the few instances of resistance evolution leading to control failure in the field. Evidence suggests resistance being a nonuniform trait across the extended region of maize production. The distribution of the main dominant resistance trait seems confined to a smaller region (100 km radius). The scale at which the distribution of this trait takes place could correspond to a recent geographical spread as confirmed by theoretical considerations showing a fast propagation of a dominant resistance trait across space. This study illustrates both potential challenges and insights of a strong evolutionary response scaling, in space and time, with the selective pressure of drastic environmental change.
AFLP data are available from Dryad doi:10.5061/dryad.n3r35.
Assefa Y, Conlong DE, Van den Berg J, Martin LA . (2015). Ecological genetics and host range expansion by Busseola fusca (Lepidoptera: Noctuidae). Environ Entomol 44: 1265–1274.
Baxter SW, Badenes-Pérez FR, Morrison A, Vogel H, Crickmore N, Kain W et al. (2011). Parallel evolution of Bacillus thuringiensis toxin resistance in Lepidoptera. Genetics 189: 675–679.
Bonin A, Bellemain E, Bronken Eidesen P, Pompanon F, Brochmann C, Taberlet P . (2004). How to track and assess genotyping errors in population genetics studies. Mol Ecol 13: 3261–3273.
Bourguet D, Bethenod MT, Pasteur N, Viard F . (2000). Gene flow in the European corn borer Ostrinia nubilalis: implications for the sustainability of transgenic insecticidal maize. Proc Biol Sci 267: 117–122.
Campagne P, Kruger M, Pasquet R, Le Ru B, Van den Berg J . (2013). Dominant inheritance of field-evolved resistance to Bt corn in Busseola fusca. PLoS One 8: e69675.
Chybicki IJ, Oleksa A, Burczyk J . (2011). Increased inbreeding and strong kinship structure in Taxus baccata estimated from both AFLP and SSR data. Heredity 107: 589–600.
Daly JC, Gregg P . (1985). Genetic variation in Heliothis in Australia: species identification and gene flow in the two pest species H. armigera (Hübner) and H. punctigera Wallengren (Lepidoptera: Noctuidae). Bull Entomol Res 75: 169–184.
de Villemereuil P, Gaggiotti OE . (2015). A new F ST-based method to uncover local adaptation using environmental variables. Methods Ecol Evol 6: 1248–1258.
Dupas S, Le Rü B, Branca A, Faure N, Gigot G, Campagne P et al. (2014). Phylogeography in continuous space: coupling species distribution models and circuit theory to assess the effect of contiguous migration at different climatic periods on genetic differentiation in Busseola fusca (Lepidoptera: Noctuidae). Mol Ecol 23: 2313–2325.
Endersby NM, Hoffman AA, McKechnie SW, Weeks AR . (2007). Is there genetic structure in populations of Helicoverpa armigera from Australia? Entomol Exp Appl 122: 253–263.
Evanno G, Regnaut S, Goudet J . (2005). Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol Ecol 14: 2611–2620.
Excoffier L, Smouse PE, Quattro JM . (1992). Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131: 479–491.
Foll M, Gaggiotti OE . (2008). A genome scan method to identify selected loci appropriate for both dominant and codominant markers: a Bayesian perspective. Genetics 180: 977–993.
Han Q, Caprio MA . (2002). Temporal and spatial patterns of allelic frequencies in cotton bollworm (Lepidoptera: Noctuidae). Environ Entomol 31: 462–468.
Hardy OJ . (2003). Estimation of pairwise relatedness between individuals and characterization of isolation‐by‐distance processes using dominant genetic markers. Mol Ecol 12: 1577–1588.
Hardy OJ, Vekemans X . (2002). SPAGeDi: a versatile computer program to analyse spatial genetic structure at the individual or population levels. Mol Ecol Notes 2: 618–620.
Harris KM, Nwanze KE . (1992) Busseola fusca (Fuller), the African Maize Stalk Borer: A Handbook of Information. Manual. International Crops Research Institute for the Semi-Arid Tropics: Wallingford, UK.
Heckel DG, Gahan LJ, Baxter SW, Zhao JZ, Shelton AM, Gould F et al. (2007). The diversity of Bt resistance genes in species of Lepidoptera. J Invert Pathol 95: 192–197.
Hill WG . (1974). Estimation of linkage disequilibrium in randomly mating populations. Heredity 33: 229–239.
Jensen JD . (2014). On the unfounded enthusiasm for soft selective sweeps. Nat Commun 5: 5281.
Jin L, Zhang H, Lu Y, Yang Y, Wu K, Tabashnik BE et al. (2015). Large‐scale test of the natural refuge strategy for delaying insect resistance to transgenic Bt crops. Nat Biotechnol 33: 169–174.
Jin L, Wei Y, Zhang L, Yang Y, Tabashnik BE, Wu Y . (2013). Dominant resistance to Bt cotton and minor cross-resistance to Bt toxin Cry2Ab in cotton bollworm from China. Evol Appl 6: 1222–1235.
Kfir R, Overholt WA, Khan ZR, Polaszek A . (2002). Biology and management of economically important lepidopteran cereal stem borers in Africa. Annu Rev Entomol 47: 701–731.
Khadioli N, Tonnang ZEH, Ong'amo G, Achia T, Kipchirchir I, Kroschel J et al. (2014). Effect of temperature on the life history parameters of noctuid lepidopteran stem borers, Busseola fusca and Sesamia calamistis. Ann Appl Biol 165: 373–386.
Kim KS, Bagley MJ, Coates BS, Hellmich RL, Sappington TW . (2009). Spatial and temporal genetic analyses show high gene flow among European corn borer (Lepidoptera: Crambidae) populations across the central US corn belt. Environ Entomol 38: 1312–1323.
Kojima KI, Schaffer HE . (1967). Survival process of linked mutant gene. Evolution 21: 518–531.
Korman AK, Mallet J, Goodenough JL, Graves JB, Hayes JL, Hendricks DE et al. (1993). Population structure in Heliothis virescens (Lepidoptera: Noctuidae): an estimate of gene flow. Ann Entomol Soc Am 86: 182–188.
Kruger M, Van Rensburg JBJ, Van den Berg J . (2011). Resistance to Bt maize in Busseola fusca (Lepidoptera: Noctuidae) from Vaalharts, South Africa. Environ Entomol 40: 477–483.
Kruger M, Van Rensburg JBJ, Van den Berg J . (2012). Transgenic Bt maize: farmers’ perceptions, refuge compliance and reports of stem borer resistance in South Africa. J Appl Entomol 136: 38–50.
Kruger M, Van Rensburg JBJ, Van den Berg J . (2014). No fitness costs associated with resistance of Busseola fusca (Lepidoptera: Noctuidae) to genetically modified Bt maize. Crop Prot 55: 1–6.
Le Rü BP, Ong'amo GO, Moyal P, Muchugu E, Ngala L et al. (2006). Geographic distribution and host plant ranges of East African noctuid stem borers. Ann Soc Entomol Fr 42: 353–361.
Mallet J . (1989). The evolution of insecticide resistance: have insects won? Trends Ecol Evol 4: 336–340.
Martel C, Réjasse A, Rousset F, Bethenod MT, Bourguet D . (2003). Host-plant-associated genetic differentiation in Northern French populations of the European corn borer. Heredity 90: 141–149.
Maruyama T . (1972). Rate of decrease of genetic variability in a two-dimensional continuous population of finite size. Genetics 70: 639–651.
Maynard-Smith J, Haigh J . (1974). The hitch-hiking effect of a favourable gene. Genet Res 23: 23–35.
Morin S, Biggs RW, Sisterson MS, Shriver L, Ellers-Kirk C, Higginson D et al. (2003). Three cadherin alleles associated with resistance to Bacillus thuringiensis in pink bollworm. Proc Natl Acad Sci USA 100: 5004–5009.
Nibouche S, Buès R, Toubon JF, Poitout S . (1998). Allozyme polymorphism in the cotton bollworm Helicoverpa armigera (Lepidoptera: Noctuidae): comparison of African and European populations. Heredity 80: 438–445.
Peakall R, Smouse PE . (2006). Genalex 6: genetic analysis in Excel. Population genetic software for teaching and research. Mol Ecol Notes 6: 288–295.
Peck SL, Gould F, Ellner SP . (1999). Spread of Resistance in spatially extended regions of transgenic cotton: implications for management of Heliothis virescens (Lepidoptera: Noctuidae). J Econ Entomol 92: 1–16.
Pritchard JK, Stephens M, Donelly P . (2000). Inference of population structure using multilocus genotype data. Genetics 155: 945–959.
R Core Team. (2012) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing: Vienna, Austria.
Rousset F . (1997). Genetic differentiation and estimation of gene flow from F-statistics under isolation by distance. Genetics 145: 1219–1228.
Saccheri IJ, Rousset F, Watts PC, Brakefield PM, Cook LM . (2008). Selection and gene flow on a diminishing cline of melanic peppered moths. Proc Natl Acad Sci USA 105: 16212–16217.
Sezonlin M, Dupas S, Le Rü B, Le Gall P, Moyal P, Calatayud PA et al. (2006). Phylogeography and population genetics of the maize stalk borer Busseola fusca (Lepidoptera, Noctuidae) in sub‐Saharan Africa. Mol Ecol 15: 407–420.
Tabashnik BE, Brévault T, Carrière Y . (2013). Insect resistance to Bt crops: lessons from the first billion acres. Nat Biotechnol 31: 510–521.
Tabashnik BE, Gassmann AJ, Crowder DW, Carriere Y . (2008). Insect resistance to Bt crops: evidence versus theory. Nat Biotechnol 26: 199–202.
Tabashnik BE, Liu YB, Malvar T, Heckel DG, Masson L et al. (1998). Insect resistance to Bacillus thuringiensis: uniform or diverse? Philos Trans R Soc Lond B 353: 1751–1756.
Tabashnik BE, Van Rensburg JBJ, Carriere Y . (2009). Field-evolved insect resistance to Bt crops: definition, theory, and data. J Econ Entomol 102: 2011–2025.
Teshima KM, Coop G, Przeworski M . (2006). How reliable are empirical genomic scans for selective sweeps? Genome Res 16: 702–712.
Tyutyunov Y, Zhadanovskaya E, Bourguet D, Arditi R . (2008). Landscape refuges delay resistance of the European corn borer to Bt-maize: a demo-genetic dynamic model. Theor Popul Biol 74: 138–146.
Vacher C, Bourguet D, Rousset F, Chevillon C, Hochberg ME . (2003). Modelling the spatial configuration of refuges for a sustainable control of pests: a case study of Bt cotton. J Evol Biol 16: 378–387.
Vacher C, Bourguet D, Rousset F, Chevillon C, Hochberg ME . (2004). High dose refuge strategies and genetically modified crops – reply to Tabashnik et al. J Evol Biol 17: 913–918.
Van den Berg J, Campagne P . (2014). Resistance of Busseola fusca to Cry1Ab Bt maize plants in South Africa and challenges to insect resistance management in Africa. In: Soberón M, Gao Y, Bravo A (eds), Bt Resistance: Characterization and Strategies for GM Crops Expressing Bacillus thuringiensis Toxins. CAB International: Wallingford, UK. pp 36–48.
Van den Berg J, Hilbeck A, Bhøn T . (2013). Pest resistance to Cry1Ab Bt maize: field resistance, contributing factors and lessons from South Africa. Crop Prot 54: 154–160.
Van Rensburg JBJ . (2007). First report of field resistance by the stem borer, Busseola fusca (Fuller) to Bt-transgenic maize. S. Afr. J Plant Soil 24: 147–151.
Vargas R, Nishida T . (1980). Life table of the corn earworm, Heliothis zea (Boddie), in sweet corn in Hawaii. Proc Hawaiian Entomol Soc 23: 301–307.
Venables WN, Ripley BD . (2002) Modern Applied Statistics with S. 4th edn. Springer: New York, NY, USA.
Vos P, Hogers R, Bleeker M, Reijans M, van De Lee T, Hornes M et al. (1995). AFLP: a new technique for DNA fingerprinting. Nucleic Acids Res 23: 4407–4414.
Zhang H, Tian W, Zhao J, Jin L, Yang J, Wu S et al. (2012). Diverse genetic basis of field-evolved resistance to Bt cotton in cotton bollworm from China. Proc Natl Acad Sci USA 109: 10275–10280.
We thank four anonymous reviewers for their constructive comments about earlier versions of the mansucript as well as Peter E Smouse and Stéphane Dupas for helpful discussions. PC was supported by IRD and Natural Environment Research Council grant NE/J022993/1; RP, J-FS and BLR were funded by IRD; and JVdB was supported by Biosafety South Africa (Grant 08-001).
The authors declare no conflict of interest.
Supplementary Information accompanies this paper on Heredity website
About this article
Cite this article
Campagne, P., Capdevielle-Dulac, C., Pasquet, R. et al. Genetic hitchhiking and resistance evolution to transgenic Bt toxins: insights from the African stalk borer Busseola fusca (Noctuidae). Heredity 118, 330–339 (2017). https://doi.org/10.1038/hdy.2016.104