Introduction
The evolution of insecticide resistance provides many classic examples of rapid evolutionary change in populations (Georghiou, Reference Georghiou1972; Tabashnik, Reference Tabashnik1994). Because resistance is often based on a simple genetic change involving one or a few loci (Roush & McKenzie, Reference Roush and McKenzie1987), the genes and processes involved in resistance evolution can be followed relatively easily. Yet the pattern of spread of resistance remains poorly documented. In most cases, resistance has been identified from a limited number of populations or strains, and both longitudinal and spatial data on changes in resistance are relatively rare.
Longitudinal data on resistance patterns are available in some situations where resistance management strategies have been implemented for pest species. For example, Forrester et al. (Reference Forrester, Cahill, Bird and Layland1993) conducted an intensive, long-term monitoring program to evaluate the impact of a pyrethroid resistance management strategy in cotton for Helicoverpa armigera (Hübner) (Lepidoptera: Noctuidae) in Australia. They observed a steady increase in pyrethroid resistance over time in both sprayed cotton areas and an unsprayed refugium. Similarly, in the housefly, Musca domestica L., Denholm et al. (Reference Denholm, Sawicki and Farnham1985) tracked pyrethroid resistance alleles selected in treated populations in enclosed farm buildings to untreated populations outdoors and on neighbouring farms. These studies suggest that resistance can evolve locally but also spread to adjacent areas through gene flow. In the case of the mosquito, Culex pipiens L., gene flow appears to have been very widespread, in that a resistance allele derived from a single mutation event now has a worldwide distribution (Labbe et al., Reference Labbe, Lenormand and Raymond2005).
When gene flow spreads resistance alleles from areas where selection occurs, resistance is expected to be high in areas adjacent to those where chemicals are applied, as well as in areas where selection is taking place. Eventually, resistance levels should increase in all areas unless there is selection against the resistance alleles in the absence of chemical applications. However, very high levels of gene flow may disrupt this spatial structure and lead to high levels of resistance only in areas directly affected by spraying and an increase in resistance levels over time in all areas where a pest occurs (Caprio & Tabashnik, Reference Caprio and Tabashnik1992). Conversely, very low levels of gene flow could produce sharp changes in resistance levels between adjacent areas dictated by local selection pressures and only a very slow overall increase in resistance in unsprayed areas.
In this paper, we consider patterns of resistance in multiple populations of the diamondback moth, Plutella xylostella (L.), an insect renowned for developing resistance to insecticides (Talekar & Shelton, Reference Talekar and Shelton1993), which has damaged Brassica vegetable crops throughout the world. We address the development of resistance to pyrethroids. Previous research on P. xylostella has indicated that resistance to insecticides in this species can vary over short distances (<10 km) within islands (Tabashnik et al., Reference Tabashnik, Cushing and Johnson1987) and also among mainland populations (Shelton et al., Reference Shelton, Wyman, Cushing, Apfelbeck, Dennehy, Mahr and Eigenbrode1993). Microsatellite data pertaining to P. xylostella suggest high levels of gene flow in this species (Endersby et al., Reference Endersby, McKechnie, Ridland and Weeks2006), and there is also direct evidence of long distance movement (Chapman et al., Reference Chapman, Reynolds, Smith, Riley, Pedgley and Woiwood2002). This suggests the potential for resistance to spread relatively rapidly and to affect region-specific levels of resistance.
Our sampling extends over a 12-year period in southern Australia, where vegetable growers started using low cost, synthetic pyrethroid insecticides in the 1980s. In many horticultural districts, synthetic pyrethroids, particularly permethrin, became the sole means of controlling Lepidoptera in Brassica vegetables. Growers first experienced difficulty in controlling P. xylostella in the early 1980s; and, by 1985, insecticide control failures and ploughing of damaged crops were common. Resistance to a range of insecticide chemical groups, including the synthetic pyrethroids, was documented at this time (Wilcox, Reference Wilcox1986; Altmann, Reference Altmann1988; Hargreaves, Reference Hargreaves1996). The initial response by the growers to control failure of synthetic pyrethroids was to increase the rates and frequency of insecticide applications and revert to use of organophosphate insecticides.
Brassica vegetables account for a very small proportion of the host plants available for P. xylostella in southern Australia but receive the most intensive use of insecticide. Vast areas of brassicaceous weeds occur throughout southern Australia, particularly in Western Australia, where wild radish, Raphanus raphanistrum, has developed resistance to multiple herbicide modes of action (Rieger et al., Reference Rieger, Preston and Powles1999; Walsh et al., Reference Walsh, Powles, Beard, Parkin and Porter2004) and in New South Wales, where wild turnip, Brassica tournefortii, and turnip weed, Rapistrum rugosum, are particularly widespread. In high rainfall areas of southern Victoria and Tasmania, about 70% of dairy farmers grow large areas of forage brassicas such as turnips (Brassica rapa) in late spring to early summer to fill a gap in summer feed (Moate et al., Reference Moate, Dalley, Roche, Grainger, Hannah and Martin1999). Moreover, recent expansion of canola growing areas in southern Australia has provided a new host for P. xylostella and insecticide application to control outbreaks of the pest on this crop has started relatively recently.
Samples across these hosts since 1993 provide a resource for investigating spatial and longitudinal changes in resistance patterns. We consider four issues. First, to what extent are patterns of resistance stable over time and differences among populations maintained? We examine several populations where repeated sampling has taken place. Second, is resistance associated with different types of host plant? We address the association between resistance and host type, focusing particularly on whether resistance in samples from weeds has remained lower than in other crops and whether resistance on samples from canola has changed over time as pesticide applications have started to be applied to this crop. Third, is there evidence of a change in spatial resistance patterns, as gene flow has spread resistance alleles to populations adjacent to sprayed areas? We test for a spatial pattern across all hosts and on specific host types using Mantel tests. Finally, is there an association between genetic variation and population differences in resistance patterns? We use genetic variation at microsatellite loci to compare patterns of variation and also consider levels of genetic and phenotypic differentiation among populations. Answers to these questions highlight the importance of local selection for resistance in determining patterns across populations, despite the spread of resistance alleles to unsprayed hosts. They also show a recent history of selection in canola increasing levels of resistance. Implications for the management of P. xylostella resistance across regional areas are discussed.
Materials and methods
Insect collection
One hundred and fifty samples of eggs, larvae or pupae of P. xylostella were collected from 104 locations in southern Australia from April 1993 to September 2005 (fig. 1, table 1). Larvae were reared on seedling leaves of cabbage (Brassica oleracea var. capitata cv. Green Coronet) at 25°C (16 h:8 h, L:D) at ambient relative humidity for one to nine generations. A population (Waite) of P. xylostella that is susceptible to synthetic pyrethroid insecticides was collected in an organic vegetable garden ca. 1987 (Baker & Kovaliski, Reference Baker and Kovaliski1999) and has been maintained in the laboratory for use as a reference in each assay since 1996.
Fig. 1. Sampling locations of Plutella xylostella (L.) in southern Australia. Numbers within circles signify the number of unique sites within the circled region from which a sample was tested for resistance to permethrin.
Table 1. Resistance ratio (RR), LC50 and 95% confidence intervals (C.I.) for permethrin tested on populations of Plutella xylostella (L.) collected on Brassica vegetables, canola, weeds and forage Brassica crops from southern Australia. RR was computed from comparison with bioassay data from the standard laboratory population (Waite) and tested on the same date as the field population.
Date, field collection date; Gen, generation tested since field collection; *, RR calculated at LC50 due to nonparallel slopes; n, number of individuals exposed to permethrin; s.e., standard error of slope.
Bioassay
Third instar larvae of P. xylostella were tested for susceptibility to permethrin using a leaf dip bioassay (after Tabashnik & Cushing (Reference Tabashnik and Cushing1987)). Cabbage leaf discs of 4.5 cm diameter were dipped for 5 s in distilled water solutions of formulated insecticide (Ambush® Emulsifiable Concentrate Insecticide–Crop Care Australasia Pty Ltd) and hung vertically to dry in a fume hood for 2 h. Control discs were dipped in distilled water. No wetting agents were used. Discs were placed into Gelman® 50 mm diameter×9 mm plastic Petri dishes. Ten larvae were placed on each disc and four replicates of seven or eight concentrations were set up for the field populations tested and for the susceptible laboratory population. From 1996 onwards, the susceptible population was tested at the same time as every field population. Mortality was assessed after 48 h at 28°C. Larvae were considered dead if they did not move when touched with a paintbrush. In all, we tested 41 populations of P. xylostella from canola, 23 from weeds, 10 from forage crops and 76 from vegetables.
Analysis
Concentration-mortality data were analysed using the probit analysis program, POLO-PC (LeOra Software) (Russell et al., Reference Russell, Robertson and Savin1977). The lethal concentration expected to cause 50% mortality (LC50) of each insecticide for each sample of P. xylostella, the 95% confidence intervals for these concentrations and the slope (+standard error) of the probit line was estimated. χ2 tests for goodness-of-fit of the data to the probit model were run using POLO-PC. If the probit model did not fit (χ2 test), the LC50 value for the particular sample may not have been reliably estimated and was adjusted by POLO-PC with the heterogeneity factor (χ2/df). The index of significance for potency estimation (g) was used to calculate 95% confidence intervals for potency (relative potency is equivalent to resistance ratio) and if parallel slopes could not be fitted for a particular assay, then the ratio was calculated at LC50 (Robertson & Preisler, Reference Robertson and Preisler1992).
Log transformed values of LC50 were used throughout the analyses to ensure normality. To test for differences between 11 locations sampled repeatedly over time for both the LC50s and the slopes, a univariate analysis of covariance (ANCOVA), using the Generalised Linear Model in SPSS (Version 13.0), was undertaken that included location as a random factor and time since sampling started as a covariate. The LC50 and slope of the laboratory susceptible population (Waite) was also included in this analysis as a covariate but did not significantly influence values from the field populations. The variance component due to location was estimated with restricted maximum likelihood.
The effects of host, time since start of sampling and host by time interactions on LC50 and slope of the probit line were assessed using ANCOVA on one sample taken randomly from each location from the entire dataset, with LC50 of the laboratory susceptible population and generation in the laboratory at time of testing as covariates. An association was found with the LC50 of the covariate (see below), so subsequent analyses looking at separate hosts over time were made using unstandardized residuals derived from the regression.
Mantel tests (Mantel, Reference Mantel1967) in PopTools (Hood, Reference Hood2002) with 10,000 iterations were used to look for spatial structure within the data by comparing population pairwise matrices of geographic distance (km) and difference between LC50. Mantel tests were made on a host by host basis (canola, vegetables, weeds and forage) to exclude any confounding host effect. Effects of time were not controlled. The Mantel tests were conducted on differences between the residuals from the regression of geographic distance and LC50 after controlling for variation in LC50 of the susceptible laboratory population used as a comparison. To look for effects of potential laboratory adaptation, Mantel tests were repeated using samples that had only been reared through one generation prior to bioassay.
Thirteen of the samples tested for insecticide resistance (table 2) were also screened for the six microsatellite loci described by Endersby et al. (Reference Endersby, McKechnie, Vogel, Gahan, Baxter, Ridland and Weeks2005). Nei's measure of genetic distance (D) (Nei, Reference Nei1978) for the microsatellite data was estimated with GDA (Lewis & Zaykin, Reference Lewis and Zaykin2001) and compared with distance measured in terms of the residuals from the regression with LC50. Mantel tests (10,000 permutations in PopTools) were used to determine the significance of associations between these variables.
Table 2. Australian samples of Plutella xylostella screened for both insecticide resistance and microsatellite loci.
n=number screened with microsatellites.
Comparing variation among populations
To compare levels of genetic variation among populations for the microsatellite markers with variation in resistance, we undertook a comparison of molecular variation and quantitative variation within and among populations. For the molecular comparison, we undertook an AMOVA with Arlequin (Schneider et al., Reference Schneider, Roessli and Excoffier2000) on the microsatellite data and computed the variance among populations. We then compared the proportion of variance accounted for by the population term for the molecular variation relative to the quantitative variation. F ST values were also computed with Arlequin.
For the quantitative analysis, we generated a distribution of concentrations at which individuals would have died based on the raw data and for the same number of individuals as scored in the microsatellite analysis. These data consist of a series of concentrations and survival of individual larvae at each concentration (four groups of larvae tested in groups of ten) and were used to generate a distribution of concentrations at which the individual larvae died. For instance, if 10 larvae died at 56.2 ppm and 12 larvae died at 100 ppm, we gave 10 larvae a score of 56.2 and 2 larvae a score of 100 and so on. Where a few larvae survived the highest concentration tested, we gave these larvae the value of the next highest concentration. We then undertook an analysis of covariance on log transforms of concentrations to compare the populations, with LC50s for control values as a covariate.
Results
Resistance to the synthetic pyrethroid, permethrin, was confirmed in laboratory-reared P. xylostella from a range of host plants in Australia. For the overall study, LC50 values ranged from 2 to 1113 ppm and the resistance ratio varied from 0.1 to 69.5 in the field samples. However, 14 samples of P. xylostella out of 150 tested were equivalent in susceptibility to the laboratory population used as a control and two were more susceptible (table 3). Between 1995 and 1999, six of these samples were collected from weeds, six from vegetables, three from canola and one from a forage crop. LC50 values of the susceptible populations ranged from 1.5 to 36.0 ppm.
Table 3. Australian populations of Plutella xylostella (L.) susceptible to permethrin. Populations were classified as susceptible according to the method of Robertson & Preisler (Reference Robertson and Preisler1992), in which the 95% confidence intervals of the resistance ratio with a designated susceptible population include the value 1.0.
RR, resistance ratio; *, calculated at LC50.
Location differences over time
The study of P. xylostella from 11 locations over time showed that resistance levels (measured as LC50) varied between locations (F9,23=4.37, P=0.002) (fig. 2), but changes over time within a site were not observed (F1,23=0.71, P=0.410). There was a significant interaction between location and time (F9,23=2.55, P=0.034). LC50 was not significantly affected by the number of generations the population had been in the laboratory at the time of testing (F1,23=2.57, P=0.122). Resistance ratios in samples of P. xylostella collected from Gatton, Queensland, on four occasions between 2001 and 2004 have remained consistently low (mean=6.7). In contrast, at Tenthill, 25 km from Gatton, resistance ratios have been consistently very high (mean=65.6) each time a sample from this location has been tested (1996, 1999, 2004). The slope of the probit line also varied between locations (F9,23=2.50, P=0.037), but not over time (F1,23=0.61, P=0.443), and showed an interaction between time and location (F9,23=2.53, P=0.035), indicating that the shape of the dose-response relationship changed over time in some locations. The number of generations a population had been in the laboratory at time of testing showed no significant effect on slope (F1,23=0.85, P=0.366).
Fig. 2. Mean LC50 of permethrin (ppm) and standard deviation of Plutella xylostella (L.) sampled on multiple occasions from 11 locations in Australia.
Resistance levels across regions, time and host
When considering all regions and using one sample per site, LC50 values were affected significantly by host type and there was a sampling time effect, an interaction between host and time and an interaction between location and time (table 4). The LC50 values for controls which were entered as a covariate in the analysis were also significant. To examine the effect of host on resistance further, we carried out an ANOVA on residuals after accounting for LC50 values for controls in a linear regression and then undertook post hoc tests. These indicated that the mean residuals for weeds were significantly lower by a post hoc test (Tukey b) than for the other hosts, as evident when the residuals for host are plotted (fig. 3a). The host by sampling time interaction reflected the fact that the effect of sampling time was significant for canola (P<0.001) but not for any of the other hosts. The positive regression coefficient (0.00074±0.00017) indicates that resistance on canola increased linearly at later sampling times (fig. 3b).
Fig. 3. (a) Mean residuals (±standard error) of regression of ln LC50 of permethrin against time (after correction for LC50 of control) of Plutella xylostella (L.) sampled from four host plant types in Australia. (b) Plot of residuals from regression of ln LC50 of permethrin against time (after correction for LC50 of control) for samples of P. xylostella collected from canola.
Table 4. ANCOVA for effects of host, time since sampling started, host by time interaction, LC50 for control samples and generation in laboratory at time of testing (labtime) on LC50 of samples of Plutella xylostella from 104 locations in Australia. LC50 values were log transformed before analysis.
In contrast to the LC50 results, there was no significant effect of host, time or interaction between these factors on the slope of the probit lines (table 4). Note that control values for slope were not included in this analysis as a covariate because the slope of the controls and samples were not correlated.
Spatial structure in resistance levels
No spatial structure was observed for resistance measured as LC50 residuals after accounting for control values (table 5) when considered on a host by host basis against geographic distance (km) using Mantel tests. Using data from samples that had only been reared through one generation prior to bioassay had little effect on the P-values for the Mantel tests (table 5). There was no pattern regardless of whether data were considered separately for eastern and western Australia or considered across the entire continent. Locations that were close together were, therefore, no more likely to show similar resistance levels than those far apart from each other, despite the clustered nature of the sampling sites (fig.1). There was also no spatial structure for slope values of the probit lines (results not presented).
Table 5. Statistics for Mantel tests for spatial structure in permethrin resistance levels in Plutella xylostella (L.) from Australia from four categories of host plant. Samples from eastern and Western Australia were tested both joint and separately with 10,000 iterations.
Matrix variable 1=difference in residuals of ln LC50. Matrix variable 2=geographic distance (km). P(F0–F1)=P-values for Mantel tests on samples that had been through a maximum of one generation in the laboratory before bioassay.
Genetic distance versus difference in resistance
A comparison of Nei's genetic distance (Nei, Reference Nei1978) between samples with differences in level of resistance measured in terms of the residuals of the regression showed no association (Mantel r=0.17, P=0.1434). Genetic distance among the populations, therefore, did not associate with pesticide resistance. The AMOVA to compare levels of molecular variation measured by microsatellites relative to the quantitative variation of insecticide resistance indicated that 0.25% of the variance was accounted for by molecular variation among the 13 population samples. The F ST computed among all samples (table 2) across loci was 0.002, which did not differ significantly from a random value by permutation (P=0.12). These values are similar to those reported by Endersby et al. (Reference Endersby, McKechnie, Ridland and Weeks2006) for a wider range of populations of P. xylostella in Australia. In contrast, the analysis of covariance indicated a significant effect of population on the concentrations at which larvae died (F(12,503)=7.29, P<0.001), and differences among population samples accounted for 13.9% of the variance in concentration at death. Population differences in resistance are, therefore, much larger than differences in molecular variation.
Discussion
Patterns of resistance in populations
The results indicate that resistance in P. xylostella is localized and associated with the host crops from which the samples were obtained. There was no evidence of spatial structure for resistance even though our sampling involved clusters of sites where broadacre crops and vegetables were grown. Locations either had persistently high or low resistance levels over several years, likely to be related to continued selection for resistance because of pyrethroid applications. Although P. xylostella may be transported on vegetable transplants leading to resistance problems in other localities (Shelton et al., Reference Shelton, Kroening, Eigenbrode, Petzoldt, Hoffmann, Wyman, Wilsey, Cooley and Pedersen1996), no obvious patterns attributable to this phenomenon were observed in this study, though the origin of vegetable seedlings at particular locations was not traced in detail.
The extent of gene flow in P. xylostella in southern Australia is high enough to homogenise microsatellite allele frequencies (Endersby et al., Reference Endersby, McKechnie, Ridland and Weeks2006), but localized selection for insecticide resistance is strong enough to generate differences among populations without being reflected in neutral microsatellite allele frequency distributions. In a direct test of the association between resistance and genetic distance, we found no association between these variables. Phenotypic differentiation among populations was substantial in the absence of genetic population structure confirming that the neutral markers are not linked with resistance traits. Therefore, local patterns of selection are unrelated to genetic distance and are important in determining resistance patterns of P. xylostella across locations in Australia.
In a simulation model of a finite population of P. xylostella presented by Caprio and Tabashnik (Reference Caprio and Tabashnik1992), moderate levels of gene flow promoted local adaptation or evolution of resistance when genetic variation (i.e. presence of resistance alleles) was constrained in some fields but not in others. In contrast, if the background frequency of resistance alleles is low, in an infinite population, then high gene flow can impede the development of resistance (Georghiou, Reference Georghiou, Georghiou and Saito1983; Lenormand & Raymond, Reference Lenormand and Raymond1998) but concurrently allow resistance alleles to spread to untreated areas, resulting in homogenisation of resistance (Caprio & Tabashnik, Reference Caprio and Tabashnik1992).
Our study suggests that the evolution of resistance is not constrained by gene flow, but that gene flow does result in the spread of resistance alleles in Australian production districts to initiate and maintain resistance to permethrin in P. xylostella. The high LC50 values in some weed crops reflect movement of resistance alleles into these unsprayed hosts. Moreover, the gradual linear increase in resistance in P. xylostella in canola over several years is consistent with the fact that these crops were not sprayed for control of this pest prior to 2000 but have experienced increased selection pressure for pyrethroid resistance in recent years as the pest status of P. xylostella in canola has increased (Endersby et al., Reference Endersby, Ridland, Zhang, Endersby and Ridland2004). With the expansion in use of pyrethroids throughout the vast area of canola plantings, there may eventually be an increase in the background frequency of resistance alleles so that resistance homogenisation occurs across locations. In contrast to the situation in canola, the lack of increase in LC50 over time in vegetable crops may reflect a general reduction in use of synthetic pyrethroid insecticides in vegetable production regions that occurred after widespread control failures, when insecticides with new modes of action were registered for use.
Local selection also appears to be important in evolution of insecticide resistance in the pear psylla, Psylla pyricola Foerster, in orchards (Tabashnik et al., Reference Tabashnik, Croft and Rosenheim1990), where a number of pyrethroid treatments explained a significant proportion of the variance in resistance over sites within regions. In this species, gene flow among populations (mean F ST=0.08), estimated using allozymes (Unruh, Reference Unruh1990), is thought to be too low to influence the response to selection (Tabashnik et al., Reference Tabashnik, Croft and Rosenheim1990), and population differentiation was greater at a local level than between regions. Moreover, management of insecticide resistance within an individual orchard may achieve a decrease in the rate of development of resistance, even if neighbouring orchards are sprayed more frequently (Tabashnik et al., Reference Tabashnik, Croft and Rosenheim1990).
Managing resistance
Strategies to delay evolution of resistance in P. xylostella, in which use of particular insecticide groups is restricted to particular times of the year, have been implemented to varying degrees in Australian vegetable crops (Deuter, Reference Deuter1989; Vickers et al., Reference Vickers, Endersby and Ridland2001). Within these restrictions on timing as well as a restriction on the number of applications made per planting, the strategies allow some flexibility in choice of insecticide mode of action at the level of the individual farmer. Despite the fact that the strategies were designed prior to studies of population structure in Australian P. xylostella, management on individual farms would seem to be appropriate when local selection is the major determinant of evolution of resistance (Tabashnik et al., Reference Tabashnik, Croft and Rosenheim1990).
However, resistance to spinosad in P. xylostella developed in Hawaii within 2.5 years of use, despite general adherence to resistance management guidelines (restriction on number of applications) on individual properties by individual farmers (Zhao et al., Reference Zhao, Li, Collins, Gusukuma-Minuto, Mau, Thompson and Shelton2002). This phenomenon occurred within an intensive Brassica production region with contiguous farms practising continuous, sequential planting of cabbage, which resulted in a common moth population being exposed to around 50 applications of spinosad in one year (Zhao et al., Reference Zhao, Li, Collins, Gusukuma-Minuto, Mau, Thompson and Shelton2002). Intensive systems, such as this, are common in the production of vegetables and need to be taken into consideration in strategies to mitigate or delay resistance to insecticides. Regionally focused resistance management of new chemistries subsequently was implemented in Hawaii and use of spinosad was withdrawn until susceptibility was restored (Mau & Gusukuma-Minuto, Reference Mau, Gusukuma-Minuto, Endersby and Ridland2004). Unfortunately, resistance then developed rapidly to indoxacarb (Zhao et al., Reference Zhao, Collins, Li, Mau, Thompson, Hertlein, Andaloro, Boykin and Shelton2006), one of the two new compounds that were used in rotation as replacements for spinosad.
Should tactics to delay resistance be implemented at the level of the individual farm, throughout an intensive production district, within an industry (i.e. vegetables vs. canola) or as an area-wide management system? Although we sampled widely in space and time and compared sites across distances ranging from one to 3800 km, there will still be unsampled locations for which the resistance status of P. xylostella is unknown, making it difficult to determine the boundaries of localized units of selection. In cases such as that described in Hawaii (Zhao et al., Reference Zhao, Li, Collins, Gusukuma-Minuto, Mau, Thompson and Shelton2002), it would appear that resistance management should occur across the whole intensive production area, though a different approach from monthly rotations of two to three modes of action seems to be required. Similar tactics may apply wherever intensive production of vegetables takes place and where movement of moths between sequential plantings (Mo et al., Reference Mo, Baker, Keller and Roush2003; Schellhorn et al., Reference Schellhorn, Siekmann, Paull, Furness and Baker2004) is prevalent.
It will be important to consider all host plant types and the consequences of different patterns of insecticide use that occur within the vegetable, canola and dairy industries for control of P. xylostella. In particular, how do our findings relate to future use of insecticides in canola and vegetables? The theoretical frequency of any allele before selection in its favour is estimated to range from 10−2 to 10−13 (Roush & McKenzie, Reference Roush and McKenzie1987). With local selection influencing evolution of resistance to insecticides, it may be that compounds with new modes of action and, therefore, a low frequency of resistance alleles, could be used in canola during sporadic outbreaks of P. xylostella (for example, once every three years on average) without exacerbating resistance levels in the pest in vegetable crops.
One approach unlikely to be useful in the management of resistance in P. xylostella involves the use of neutral genetic markers. It has been proposed that such markers could be useful in identifying regions for area-wide control of resistance in Helicoverpa armigera (Hübner) (Scott et al., Reference Scott, Wilkinson, Merritt, Scott, Lange, Schutze, Kent, Merritt, Grundy and Graham2003, Reference Scott, Wilkinson, Lawrence, Lange, Scott, Merritt, Lowe and Graham2005). However, given that Australian populations of P. xylostella are only weakly differentiated genetically and given the lack of association between genetic distance and resistance variation, this approach would appear to be of little use in resistance management.
Because resistance is an evolutionary response to an environmental stress, the best that management strategies can hope to achieve is to delay the process (Hoy, Reference Hoy1998). The longitudinal data presented here indicate pyrethroid resistance develops locally on hosts that are sprayed regularly. Resistance then spreads to other areas, but local selection pressures, rather than gene flow, dictate levels of resistance.
Acknowledgements
Thanks to Jingye Zhang for technical assistance. Thanks to the Department of Agriculture, Western Australia; SARDI; Department of Primary Industries, Water and Environment, Tasmania; Agriculture NSW; Department of Primary Industries Queensland and others for collecting populations of P. xylostella from canola and vegetable crops. We are grateful to Mike Keller for supplying the Waite susceptible population of P. xylostella. Thanks to Steve McKechnie for comments on an earlier version of the manuscript. We also thank the agrochemical companies who were involved in supporting a national insecticide resistance testing program for P. xylostella: Aventis CropScience, BASF Australia Ltd, CropCare Australasia, Dow AgroSciences Australia Ltd, DuPont (Australia) Ltd, NuFarm Ltd, Sumitomo Chemical, Syngenta Crop Protection Pty Limited and AIRAC (AVCARE's Insecticide Resistance Action Committee). The study was funded by Horticulture Australia Limited, an Australian Research Council Strategic Partnership with Industry Research & Training Grant, Department of Primary Industries Victoria (Industry Partner), the Grains Research and Development Corporation and the Australian Research Council via their Special Research Centre program.
