Introduction

The development of resistance to pesticides by insects represents a modern-day example of evolution in action. Several attributes make resistance attractive to study as a model of molecular evolution including the speed of the onset of resistance, the involvement of major single genes, and the ability to identify these single genes and mutations therein that confer the trait. Such a model system allows empirical testing of some of the central tenets of micro-evolutionary theory concerning the origin and frequencies of beneficial mutations and the effects of their selection on the fate of adjacent polymorphisms (Nurminsky 2001). Interestingly, cases of resistance analyzed to date have revealed contrasting results with respect to the origin and molecular mechanisms of resistance. Single mutational events have been found to be responsible for DDT resistance in Drosophila (Daborn et al. 2002), while multiple resistance events have been found in the red flour beetle, Tribolium casteneum (Andreev et al. 1999), and organophosphate (OP)-resistant Culex pipiens (Raymond et al. 1998). Mutations that give rise to resistance range from point mutations in insecticidal target site encoding genes (ffrench-Constant et al. 1998) to amplification of esterase genes resulting in overexpression of the protein that then sequesters the insecticides (Hemingway 2000). This paper represents a comparable analysis of OP resistance in the sheep blowfly, Lucilia cuprina.

The history of OP usage to control L. cuprina has been well documented in Australia and New Zealand, where this species is an ectoparasite of sheep (Tillyard and Seddon 1933; Heath et al. 1991). Diethyl OPs, such as diazinon, have been extensively used on sheep to control the fly as both prophylactic and curative treatments. Diazinon was first introduced in Australia in 1958. Single gene diazinon resistance mapping to the Rop-1 locus was detected soon after, in 1965 (Shanahan and Hart 1966; Arnold and Whitten 1976), rising rapidly to high frequency in both Australia (97%; Hughes 1981; Hughes and MacKenzie 1987; Levot et al. 1995) and New Zealand (Gleeson et al. 1994; Wilson and Heath 1994). The dimethyl OP malathion has also been used on sheep, albeit to control lice rather than L. cuprina directly. Nevertheless, resistance was detected in L. cuprina (Hughes et al. 1984), which also maps to the Rop-1 region of the fourth chromosome (Rmal; Raftos 1986; Smyth et al. 1994). The frequency of malathion resistance in the field is lower than that for diazinon resistance (Hughes et al. 1984; Levot et al. 1995).

The Rop-1 and Rmal genes have been isolated from L. cuprina and shown to be synonymous with the LcαE7 gene, which encodes the esterase, E3 (Newcomb et al. 1997b; Campbell et al. 1998b). LcαE7 is a member of a cluster of α-esterase genes that is shared across the higher Diptera (Oakeshott et al. 1999; Claudianos et al. 2001), with orthologues identified in Drosophila melanogaster (Russell et al. 1996; Robin et al. 2000b), D. buzzatii (Robin et al. 2000a), and M. domestica (Claudianos et al. 1999). In L. cuprina resistance to OPs is encoded by distinct mutations in the LcαE7 gene. A Gly137Asp substitution in the oxyanion hole within the active site of the enzyme transforms the carboxylesterase to an OP hydrolase conferring resistance to OPs with a preference for diethyl OPs such as diazinon (Newcomb et al. 1997b). A second substitution, Trp251Leu, in the acyl pocket of the active site also confers resistance to OPs, this time preferring dimethyl compounds, particularly those containing a carboxylester linkage in their leaving group such as malathion (Campbell et al. 1998b).

Previously, approximately 40 strains of L. cuprina were made homozygous for the fourth chromosome and were characterized for OP resistance both toxicologically and biochemically (Campbell et al. 1997, 1998b; Smyth et al. 2000). Four classes of isogenic (IV) strains were described: a susceptible class with high levels of carboxylesterase activity, a malathion-resistant class with high levels of malathion carboxylesterase (MCE) activity, a diazinon-resistant class with low levels of carboxylesterase activity and detectable levels of OP hydrolyase activity, and a class highly resistant to both diazinon and malathion. Surveys to test whether resistance correlates with a particular amino acid substitution have detected some variation within resistance classes. Newcomb et al. (1997a) did not detect any replacement site substitutions in a 103-bp segment of genomic DNA around the Gly137Asp substitution in a survey of 15 isogenic (IV) strains. However, Campbell et al. (1998a) did detect sequence variation within Trp251Leu containing types of L. cuprina, indicating that at least two haplotypes exist among the four isogenic (IV) strains surveyed. Furthermore, a RFLP survey of 12 variable restriction sites across 8 kb including LcαE7 revealed three RFLP haplotypes associated with susceptibility (A,B,C), and two each associated with diethyl (D,G) and dimethyl OP resistance (E,F) segregating among 25 isogenic (IV) strains (Symth et al. 2000).

Here we analyze sequence variation across a region of the LcαE7 gene in these and other extant isogenic (IV) strains, plus some pinned specimens from earlier collections, in order to determine the association of the resistance mutations with the different classes of resistance and their inferred haplotypes. In doing so, we are able to examine the impact of selection on variation at the Rop-1 locus and infer the consequential effects on haplotype diversity over time. In addition we describe the sequence of haplotypes involved in gene duplication events associated with resistance-conferring alleles of LcαE7.

Materials and Methods

Strains

A total of 41 strains of field collected L. cuprina were made isogenic for their fourth chromosome through a series of crosses to a multiply inverted fourth chromosome balancer [In(4)6+8+12Sh gl/+ + +] (Smyth et al, 2000). All Australian isogenic (IV) strains have been described previously (Smyth et al. 1994, 2000) and the three isogenic (IV) strains from New Zealand were all collected from Wanganui in the North Island and made as above. Males of 19 of the Australian isogenic (IV) strains were screened with 0.2% (w/v) malathion during the isogenization process and are termed screened lines (Smyth et al. 2000). The other 22 isogenic (IV) strains are termed unscreened (Table 1). A strain of L. sericata was collected from Canberra, ACT, Australia.

Table 1 Summary of location, OP resistance status, carboxylesterase activity, OP hydrolyase activity, resistance residue type, and haplotypes deduced from the LcαE7 sequence of L. cuprina isogenic (IV) strains and L. sericata
Full size table

Larvae were raised on a combination of liver and meat meal media at 27°C, while adults were maintained on sugar and water and given a protein feed prior to oviposition (Weller et al. 1993). Adult flies that were used for DNA extractions were maintained on a diet of sugar and water. Toxicological and biochemical analysis of many of these strains has been described previously (Campbell et al. 1997, 1998b; Smyth et al. 2000).

Pinned material was obtained from the Australian National Insect Collection (CSIRO-Division of Entomology, Canberra, Australia).

Sequence Analysis

Genomic DNA from the nonpinned material was prepared either from individual pupae or flies using the DIGSOL (Newcomb et al. 1998) or CTAB methods (Crozier 1991), or from eggs using the method of Davis et al. (1986). An ∼1.2-kb region of the LcαE7 gene (coordinates 276–1105 in LcαE7 cDNA; Newcomb et al. 1997b) was amplified from genomic DNA using the Lc7pop1 (5′-agataagtcagtgcaagttgattt-3′) and Lc7pop2 (5′-attccttaacaagcataggcattt-3′) primers. This includes the last 85 bp of the second exon, through to the first 33 bp of exon 5. The corresponding region of the protein includes the active site serine (Ser218) and residues involved in the oxyanion hole (Gly136, Gly137, Ala219) and acyl pocket (Trp251, Val307, and Phe309). In particular, it covers the sites that encode the resistance-conferring Gly137Asp (diazinon) and Trp251Leu (malathion) substitutions. Amplification reactions (50 μl) consisted of 10 pmol of each primer, 10 mM Tris–Cl, pH 8.3, 1.5 mM MgCl, 50 mM KCl, and a 0.2 mM concentration of each dNTP. The addition of 2 units of Taq polymerase (Boehringer Mannheim) followed an initial step of 2-min denaturation at 94°C. The remaining cycling conditions consisted of denaturation at 94°C for 1 min, annealing at 55°C for 30 s, and extension at 72°C for 1 min 20 s for 35 cycles. A final step included a 10-min extension at 72°C.

PCR products were purified using the Qiagen PCR direct purification kit, following the procedure described by the manufacturer. Direct sequencing of purified products using Lc7pop1, Lc7pop2, and Lc7pop3 (5′aaagtgtgtggctcagaggattg-3′) and the primers described by Newcomb et al. (1997b) was achieved using the Prism Ready Reaction Dye Deoxy Terminator Cycle Sequencing Kit (Perkin Elmer) following the manufacturer’s instructions. Sequencing reactions were resolved on an Applied Biosystems Model 377 automated DNA sequencer.

For strains of L. cuprina showing resistance to both malathion and diazinon and the L. sericata strain, the PCR products were first cloned into pBluescript and then sequenced as above. Multiple clones were sequenced until identical representative clone sequences were recovered. Resulting sequences were compiled using Sequencher (GeneCodes Corporation) to generate sequences minus primer sites for further analysis.

For pinned material the method of Phillips and Simon (1995) was used to extract genomic DNA from whole flies. Resulting DNA was resuspended in 20 μl of dH2O or 0.5× TE buffer. Usual precautions were taken for the preparation of ancient material. Extractions were performed in a separate fly-naive room in a UV-treated laminar flow hood using separate reagents and equipment. DNA quantity was assessed using Sybr Green I fluorescent dye (Molecular Probes) and quality assessed on 0.7% agarose gels. In test PCRs some samples did not yield products even though DNA was present. Cleanup using the Concert PCR purification Kit (LifeTechnologies) often yielded amplifiable DNA, presumably by removing PCR-inhibiting contaminants. PCR was conducted as described above using the Expand High Fidelity PCR system (Roche) except 40 cycles of amplification were used and various concentrations of template DNA trialed. Reactions were set up in a UV-treated laminar flow hood. Attempts to amplify the entire 1.2-kb region as above were not successful from DNA extracted from pinned material due to the DNA being degraded to sizes between 200–500 bp (data not shown). Primer combinations of 7F1a><7R2 or Lc7pop1><7R2, and 7F3><7R4 (Newcomb et al. 1997b) were used to amplify fragments covering the Gly137Asp and Trp251Leu resistance substitutions, respectively. Resulting products were cloned using the pGEM-T Easy system (Promega) and sequenced as described above.

Sequences were aligned initially using Sequencher (GeneCodes Corporation) and then corrected manually as required. Phylogenetic trees, tests of monophyly and bootstrap analysis (1000 replicates), were calculated using PAUP* (Swofford 1998). The set of evolutionary changes in DNA polyphorphism was traced onto the halpotype tree using MacClade 4.0 (Maddison and Maddison 2000). DNA polymorphism analysis, recombination analysis, linkage disequilibrium, coalescent simulations, and test statistics to detect departures from neutrality, including Tajima’s (1989) D, Fu and Li’s (1993) D and F, and the MacDonald Krietman (1991) tests, were performed used DnaSP Version 3.50 (Rozas and Rozas 1999). For simulations of haplotype diversity and numbers of haplotypes, 10,000 random simulations were performed under a no-recombination model with the number of segregating sites fixed at 100. Fay and Wu’s (2000) H statistic was calculated at the Web site serverhttp://www.genetics.wustl.edu/jflab/htest.html. Ten thousand simulations were run using a divergence of 0.217 and assuming no recombination. Sites segregating for or containing gaps in the outgroup were excluded from the analysis.

Results

We have determined the nucleotide sequence of a 1.2-kb region of LcαE7 from 41 isogenic (IV) strains of L. cuprina from Australia and New Zealand (GenBank accession nos. AY691501–AY691508; nucleotides 300–1105 in LcαE7 cDNA [Newcomb et al. 1997b]). The corresponding region of the orthologous gene from L. sericata was also determined for use as an outgroup. Seven L. cuprina strains screened with malathion during their construction were confirmed to carry a duplication of the LcαE7 gene (Smyth et al. 2000). For these strains PCR products of the LcαE7 gene were cloned and the same region as above sequenced. An alignment of sequences from the 34 strains carrying a single copy of the LcαE7 gene has a total length of 1194 nucleotides including gaps and 1151 excluding gaps, all of which are in introns (Fig. 1). Of the 86 polymorphic sites within L. cuprina sequences, 48 are in introns and 38 are in the coding region, 29 being silent and nine being replacement site polymorphisms (Table 2).

Figure 1
figure 1

Variable sites within the seven haplotypes of the LcαE7 and L. sericata αE7 sequence. The positions of OP resistance mutations conferring diethyl (e.g., diazinon) and dimethyl OP resistance (e.g., malathion) are marked with a D and M, respectively.

Full size image
Table 2 Sequence polymorphism within regions of the LcαE7 gene
Full size table

Despite all the polymorphisms no additional haplotypes were recorded in L. cuprina over the seven identified in the RFLP survey of similar material presented by Smyth et al. (2000) (Table 1). Each of the three susceptible strains has a different haplotype (haplotypes A, B, and C [Smyth et al. 2000]). Of the resistant strains that carry a single copy of the LcαE7 gene, 16 share a common Gly137Asp containing haplotype (haplotype D), 2 share a second Gly137Asp containing haplotype (haplotype G), 9 contain a Tr251Leu containing haplotype (haplotype E), and 4 have a second Tr251Leu containing haplotype (haplotype F). Of the seven strains carrying two copies of the LcαE7 gene, three contain a D and an E haplotype, three a D and an F haplotype, and one a G and an F haplotype. The L. sericata haplotype is unique (haplotype H). Of the 22 unscreened L. cuprina strains, i.e., those not selected with malathion during the isogenation process, 16 have the Gly137Asp containing D haplotype (72%), 2 have the rarer G haplotype (9%), 1 is a Trp251Leu containing haplotype (E; 4%), and 3 are susceptible haplotypes (A, B, and C; 5% each).

The inferred translation products of all sequences contain an active site serine (Ser218) plus all the conserved oxyanion hole and acyl pocket residues associated with a functional esterase enzyme (Oakeshott et al. 1999). The nature of the residues at positions 137 and 251 correlates well with existing biochemical data for the isogenic (IV) strains and their resistance phenotype (Table 1). The three strains carrying Gly137 and Trp251 (LS2, LBB101, Llandillo 104) are susceptible to diazinon and malathion and show high levels of carboxylesterase activity for the artificial substrates methyl butyrate and naphthylacetate (termed ali-esterase and E3 staining, respectively), intermediate levels of MCE activity, and low levels of OP hydrolase activity (as measured using the OP chlorfenvinphos as a substrate [Smyth et al. 2000]). In nonduplicated strains where resistance and biochemical activity have been determined, Asp137 is completely associated with reduced E3 staining, ali-esterase and MCE activities, high levels of OP-hydrolyzing ability, and resistance to diazinon, while Trp251 is associated with reduced ali-esterase and E3 activity, high levels of MCE activity, and malathion resistance (Smyth et al. 2000). No individual sequences of LcαE7 carrying both resistance substitutions Asp137 and Leu251 were detected. The seven strains containing both one form of LcαE7 with Asp137 and Trp251 and another with Gly137 and Leu251 are not heterozygous but contain duplicated copies of at least the LcαE7 locus (Smyth et al. 2000). Toxicology with diazinon and malathion has been conducted on four of these strains, revealing that they are resistant to both compounds, except for Tooronga 2.3, where resistance to diazinon was not detected (Table 1). This result either reflects an error in the discriminating toxicology or may be due to incomplete expression of the Asp137 duplicated copy in Tooronga 2.3.

As well as the Gly137Asp and Trp251Leu substitutions, five of the other seven replacement site polymorphisms found here were also reported by Newcomb et al. (1997a), in their comparison between the LS2 (susceptible) and the Llandillo 103 (diazinon-resistant) strains. These five encode four amino acid polymorphisms, Ala267Val, Met283Leu, Thr335His, and Ile358Phe. (Two nucleotide substitutions are involved in the Thr335His change.) None of these other substitutions, however, contribute to OP-hydrolase activity, nor are they predicted to lie in the active site of the enzyme (Newcomb et al. 1997a). The remaining two replacement site polymorphisms involve an Asn122Thr and Pro317Ser in the OP-susceptible Llandillo104 sequence. Both residues are predicted to be over 12 Å away from the active site serine using the three-dimensional structure of acetylcholinesterase (lace) as a model (data not shown).

An unweighted parsimony tree was constructed from the nucleotide data to determine relationships among the observed haplotypes (Fig. 2; tree length = 224 with gaps scored as missing). Distance- and maximum likelihood-based trees had identical structures (data not shown). Bootstrap analysis (1000 replicates) showed reasonable support for all the groupings, with weakly supported groupings being the haplotypes B + C clade, with 58% support, and the haplotypes G + H clade, with 47% support. Using the L. sericata sequence as an outgroup, haplotype G (Gly137Asp) is predicted to be most ancestral. The ancestral assignment of a resistant haplotype raises the possibility that the resistance mutation predated the use of OPs, but it might also reflect recombination or a recent origin of the mutation in the otherwise ancestral background of the haplotype G. We examined these possibilities further in the analysis of pinned specimens below. The Trp251Leu containing haplotype E is closely related to the susceptible haplotype A and may be derived from a closely related common ancestor. Neither the two Gly137Asp containing haplotypes (D,G) nor the two Trp251Leu containing haplotypes (E,F) are closely related, indicating that they are unlikely to be derived from each other, but instead represent separate mutational events. Phylogenetic trees constructed with Gly137Asp constrained to have arisen once increased the tree length from 224 to 228, whereas a tree constrained to evolve Trp251Leu once increased the tree length from 224 to 233. The hypothesis that these mutations each arose once was also tested using the Kishino–Hasegawa (1989) and Templeton (1983) signed rank tests (Table 3). A single origin for the Leu251 mutation was rejected by both tests, however, the single origin hypothesis was not rejected for Asp137.

Figure 2
figure 2

Phylogenetic tree of the relationships among haplotypes constructed using parsimony scoring gaps as missing. Changes were calculated along branches using MacClade 4.0. Substitutions causing resistance-associated changes are in boldface (387, G>A is Gly137Asp; 787, G>T is Trp251) and ambiguous changes are in italics. All inferred changes along the branch to the outgroup are not shown. Haplotypes A–H are described in Table 1. Under haplotype names are resistance-causing substitutions, with resistance-associated Asp137 and Trp251 substitutions in boldface. Number in parentheses is the frequency at which this haplotype was observed. Duplicated lines have not been taken into consideration. Numbers alongside branches represent percentage bootstrap values calculated from 1000 replicates.

Full size image
Table 3 Kishino–Hasagewa and templeton tests of monophyly of the origins of resistance
Full size table

Recent recombination or gene conversion among the haplotypes to move the resistance mutations onto a different background would create overestimates of the numbers of origins of resistance mutations and compromise the inference of multiple origins for the two resistance mutations. Therefore we assessed the possibility that recombination has been involved in the history of these alleles. The recombination parameter (Hudson 1987) was calculated for the data (C = 0.0015), while the estimated minimum number of recombination events (Hudson and Kalpan 1985) detected for the data was eight, between sites (46,253) (323,384) (384,510) (589,640) (640,652) (766,783) (783,871) (1102,1142). Plotting all the predicted substitutions onto the L. cuprina portion of the tree in Fig. 2 allowed the visualization of any blocks of sites that appear to be convergent nucleotide substitutions around the resistance mutations at nucleotides 387 and 787. Such patterns would indicate recombination had moved the resistance mutations onto another allelic background rather than mutated de novo. While no such blocks can be observed in Fig. 2, recombination or conversion of the resistance mutation onto another background cannot be totally ruled out since there is evidence for recombination in the data.

A selective increase in frequency of resistant haplotypes in the population should reduce levels of variation at the locus through the removal of neutral variation present in susceptible haplotypes. Tests of significant deviations from neutral expectations should be able to detect the effects of the reduction of haplotype number or haplotype diversity at the locus where this is occurring. Such an analysis can be achieved by comparing observed haplotype data to those obtained from random coalescent simulations as implemented in DnaSP (Rozas and Rozas 1999). Data from only the unscreened lines were used in these tests, as they represent unbiased collections of field chromosomes (no. of lines = 22, sites = 1180, S = 82, π = 0.0161, θ = 0.0193). Haplotype number (Nei 1987) is less than expected under neutrality (Nhap = 17, actual = 6, p < 0.00001). Similarly, haplotype diversity is lower than expected (Hdiv = 0.967, actual = 0.476, p < 0.00001). When recombination is assumed, even greater deviations from neutral expectations for Nhap and Hdiv are achieved (data not show).

Variation closely linked to the resistance mutation should also be swept along as the allele increases in frequency in the population (Maynard Smith and Haigh 1974; Kaplan et al. 1989; Stephan et al. 1992). We searched for evidence of such hitchhiking by analysing linkage disequilibrium across the 22 unscreened isogenic (IV) strains using Fisher’s exact test. One thousand one hundred and forty-one of 2145 pairwise comparisons are individually significant (p < 0.05), of which 66 remain significant with Bonferroni’s correction. A plot of the pairwise comparisons across the LcαE7 gene (Fig. 3) reveals two sets of sites behaving as discrete blocks of disequilibrium. First, there is strong disequilibrium among most sites at both the 5′ (4–248) and the 3′ (835–1176) ends of the sequenced region. Second, there is a central block of disequilibrium which is strongest among sites in a region within the second intron (256–302) but which continues to a lesser degree into the adjacent exon (256–770). The resistance mutations (at 387 and 787) both show disequilibrium with the distal blocks, although physically the Gly137Asp substitution at nucleotide position 387 lies within the central block.

Figure 3
figure 3

Analysis of nonrandom associations among pairs of twofold degenerate sites within the LcαE7 (Rop-1) gene. White squares represent p > 0.05, light-gray squares represent 0.01 < p < 0.05, medium-gray squares represent 0.001 < p < 0.01, dark-gray squares represent p < 0.001, and black squares represent p < 0.001 and significant by the Bonferroni procedure.

Full size image
Figure 4
figure 4

Variable sites within LcαE7 from extant (haplotypes A–G) and from pinned L. cuprina. Variable sites are numbered along the top of the alignment, with D and M indicating the sites of the Gly137Asp and Trp251Leu mutations, respectively. These sites are then shaded gray in the alignment below. In the data from pinned L. cuprina, sites identical to haplotype A are marked with a dot. Polymorphisms that uniquely contribute to the formation of a novel haplotype are shown in boldface. Haplotype affinity is given at the right, with similarity to existing extant haplotypes noted. Four novel haplotypes are described. Resistance status predicted from the sequence is given in parentheses after the haplotype affinity. DR, diethyl OP resistant; DS, diethyl OP susceptible; MR, dimethyl OP resistant; MS, dimethyl OP susceptible.

Full size image

In addition, tests were applied to the LcαE7 data that can detect departures from neutral expectations with respect to the frequency distribution of variants at segregating sites. Tajima’s (1989) and Fu’s and Li’s (1993) tests detect hitchhiking by comparing estimates of θ, where θ = 4Nμ (N = effective population size, μ = mutation rate within a population). Neither Tajima’s D (D = −0.618, p > 0.10) nor the Fu and Li tests conducted either without an outgroup (D* = 0.590, p > 0.10; F* = 0.240, p > 0.10) or with an outgroup (D = 0.844, p > 0.10; F = 0.375, p > 0.10) revealed significant departures from neutrality. Fay and Wu’s (2000) H statistic also compares estimators of θ, namely θπ and θ H , where θ H is an estimator of θ weighted by the homozygosity of derived variants as opposed to the ancestral variants. H is significant (H = 32.45, pH = 0.029), due to the higher than expected number of sites with the derived state at high frequency. This result remains highly significant when a rapidly evolving intron is removed from the analysis (nucleotides 329–328), where assignment of the ancestral state may be effected by potential alignment inaccuracies and back mutations (H = 28.40, pH = 0.027).

The Macdonald–Kreitman test was also used to detect departures from neutrality in a data set containing LcαE7 sequences from the unscreened strains of L. cuprina and the L. sericata sequence as an outgroup. Under neutral expectations the ratio of nonsynonymous-to-synonymous substitutions should be equivalent in comparisons of intraspecific polymorphism and fixed difference data between species (McDonald and Krietman 1991). Deviation from neutral expectations was detected using a G test (p = 0.02030) and Fisher’s exact test (p = 0.046; Table 4). The ratio of nonsynonymous-to-synonymous variation is greater within species (0.321) than between (0.059).

Table 4 McDonald and Kreitman test
Full size table

PCR products of the LcαE7 gene were recovered from DNA extracted from seven pinned specimens dating from 1968 back to 1953. Amplification using primers 7F1a and 7R2 yielded a 145-bp product, while Lc7pop1 and 7R2 gave a 509-bp product, and 7F3 and 7R4 a 267-bp product. The first and second products covered regions of the LcαE7 gene that included the Gly137Asp mutation, while the third covered a region that included the Trp251Leu conferring mutation. Sequences from the resulting cloned products were compared to sequences from the eight haplotypes recovered from extant material (Fig. 1, Table 5). Representatives carrying the Asp137 mutation were recovered from three of the seven pinned flies sampled, including two flies collected from diverse locations within Australia in 1963 before the first reports of resistance. In all cases the haplotypes of the Asp137 alleles recovered from pinned flies could not be distinguished from either of the extant Asp137 haplotypes, D or G. Susceptible haplotypes from regions covering Gly137 were the same as extant susceptible haplotypes A, B, and C except for Moruya (2), which carried a new susceptible haplotype which looks like a recombinant between haplotypes D and B or C. No Leu251 mutations were detected from the pinned flies, but three novel susceptible Trp251 haplotypes were observed. One is a single substitution away from a D-like haplotype, one is two mutations away from a D-like haplotype, and the third is one mutation away from a G-like haplotype. Thus, in total, four new haplotypes were observed from this sample of seven pinned flies, all associated with susceptible substitutions. Potentially the number of novel haplotypes may actually be higher if new combinations of haplotypes between the regions amplified are real. For example, from the Cunnamulla fly a new allele combining the 7F1a><7R2 region with the haplotype D version of the 7F3><7R2 fragment would create a novel susceptible haplotype. Therefore from this sample of seven pinned flies (14 chromosomes), at least eight haplotypes were observed, not counting possible recombinant haplotypes.

Discussion

We have shown that the correlations of diethyl OP resistance with the Gly137Asp substitution and of dimethyl OP resistance with the Trp251Leu substitution extend across the 41 isogenic (IV) strains of L. cuprina analyzed with the exception of the Tooronga 2.3 strain. Of the 260 codons surveyed, the only consistent difference between predicted proteins from susceptible strains and the diethyl OP-resistant proteins is the Asp137 substitution. Similarly for the dimethyl OP-resistant proteins the Leu251 substitution is the only consistent difference. These findings support previous collative studies on smaller numbers of lines (Newcomb et al. 1997a; 15 isogenic lines). With respect to the diethyl OP resistance, these data largely corroborate previous experimental studies with artificial chimerics of LcαE7, which showed that the Gly137Asp substitution is sufficient to encode the ability to hydrolyze the diethyl OP chlorfenvinphos and simultaneously result in reduced carboxylesterase activity (Newcomb et al. 1997a).

Studies of OP resistance in other insects have found similar associations with variation in homologues of LcαE7. Molecular studies of the M. domestica homologue of LcαE7, MdαE7 found the same Gly137Asp substitution in diazinion-resistant housefly strains (Claudianos et al. 1999). In the parasitic wasp, Anisopteromalus calandrae, malathion resistance is associated with a Trp220Gly substitution in the LcαE7 homologue and increased levels of the expression of the gene’s transcript (Zhu et al. 1999). This Trp220Gly substitution is equivalent to Trp251 in LcαE7, suggesting that substitutions to other small amino acids at this position may as be able to confer increased OP hydrolase activity and resistance. In the hornfly, Haematobia irritans, the HiαE7 gene is expressed at five times higher levels in diazinon-resistant strains compared with susceptible strains, albeit there is no evidence for mutations at sites equivalent to Gly137 and Trp251 (Guerrero 2000). The hornfly data aside, these studies of OP resistance in a range of insects suggest that mutations at the 137 and 251 amino acid positions in orthologues of LcαE7 may be common mechanisms of OP resistance in insects. Intriguingly, resistance to diazinon in Drosophila melanogaster has never been mapped to the DmαE7 gene even though the equivalent substitutions are possible by single point mutations.

The evolution of OP resistance in L. cuprina has resulted in a high frequency of resistance-conferring alleles of Rop-1 in the field. From this survey of fourth chromosomes 18 of the 22 unscreened isogenic (IV) strains (82%) carry the Asp137 substitution. A far lower frequency carries the Leu251 mutation (4%) and susceptible alleles are also not common (14%). Although the sample size is small these results are similar to estimates of the levels of resistance in Australia as measured using direct toxicological assays (Wilson and Heath 1994; Gleeson et al. 1994), suggesting that the isogenization process has not highly biased the sampling of chromosomes from the wild. The high frequency of OP resistance in the field is due in part to the evolution of a modifier gene that neutralized the negative fitness effects that Asp137 and Trp251 have in noninsecticide environments. The modifier gene was detected in Australia in 1979 (McKenzie et al. 1982) and has also been detected in New Zealand flies (Gleeson 1995a).

Previous toxicological and biochemical studies have indicated a very strong negative association between diazinon resistance and malathion resistance (Smyth et al. 2000). At least in part, this will reflect the fact that the two mutations that confer their resistance are only 400 bp apart, giving recombination little chance of creating the double mutant. However, even if an Asp137 and Leu251 containing allele arose, it is unlikely to confer resistance to malathion as well as diazinon. This is because malathion resistance is thought to require the enzyme to maintain its carboxylesterase activity, while the Asp137 substitution abolishes this activity (Campbell et al. 1998a). The double mutant allele of LcαE7has been constructed and expressed in vitro, showing that the enzyme has low malathion carboxylesterase activity similar to the Asp137 mutant alone (Heidari et al. 2004). However, the blowfly has found another way to evolve double resistance—gene duplication (see below).

The Gly137Asp and Trp251Leu mutational classes are each represented by two distinct haplotypes. Two distinct groupings are resolved for each class of resistance in our analysis of the sequence relationships among the haplotypes. In the case of diazinion resistance, however, a single origin could not be rejected using tests of monophyly using phylogentic-based tests. Furthermore, Gly137Asp falls within a block of linkage disequilibrium found with the LcαE7 gene (sites 268–770), raising the possibility that the Gly137Asp mutation arose once and was transferred by recombination or gene conversion since low levels of recombination are predicted within the region.

The inability to find susceptible alleles within the extant collection of haplotypes that are the possible ancestors of the resistant haplotypes is probably due to the current low frequency of susceptible alleles; many susceptible alleles have probably been lost in the last 40 years. To test whether susceptible alleles were once more abundant, harboring more haplotype diversity before the selective sweep, haplotypes were sampled from pinned flies from the Australian National Insect collection. Certainly haplotype diversity is higher in the collection of haplotypes from pinned specimens (minimally 8/14) than in the extant collection (6/22). In addition four of the eight haplotypes from the pinned specimens were not seen in the larger extant collection. All four novel haplotypes in the pinned material are associated with susceptible genotypes. We should point out that there is a possibility that some of these novel haplotypes may be the result of cloned PCR errors even though a low error Taq polymerase was used. Two of the four novel haplotypes seen in the pinned material (Geraldton; Coorong) are a single mutational step from haplotypes seen previously. The other two novel haplotypes require either a recombination event (Moruya) or two mutations (Cunnamulla). Comparing the pinned data set with the extant haplotype collection shows a drop in haplotype diversity consistent with the hypothesis that a selective sweep for a few beneficial mutations has reduced variation at the locus. Unfortunately the sample size and fragmented nature of the data from the pinned flies make it difficult to perform statistical tests of these assertions.

Interestingly we may have uncovered an example of the D haplotype class from a fly caught in 1953, before OP insecticides were introduced. This haplotype was uncovered from sequencing over the region of the Trp251Leu mutation in a fly from Cunnamulla (NSW, Australia) and, unfortunately, not over the Gly137Asp mutation. This haplotype may either represent the progenitor of the common D haplotype class seen today or the susceptible version of the allele, in which the Gly137Asp mutation occurred. Further sampling of pinned specimens to increase the region sequenced and sample size will be essential if we are to be able to test whether significantly more susceptible alleles were indeed present before the selective sweep came through the population when OPs were introduced or whether resistant haplotypes can be detected from flies sampled before the introduction of OP insecticides.

The hitchhiking model of molecular evolution predicts that variation in and around a locus under selection will be reduced as the beneficial allele(s) increases in frequency (Maynard Smith and Haigh 1974; Kaplan et al. 1989; Stephan et al. 1992). At the Rop-1 (LcαE7) locus a majority of the field sampled alleles are of the haplotype class D (16/22), all having identical sequence. In contrast, the three susceptible haplotypes from the extant strains sequenced are all harboring unique, presumably neutral variation. Consistent with the hitchhiking model, levels of linkage disequilibrium within the LcαE7 gene are high. Also, tests detecting deviations from expected number of haplotypes and haplotype diversity detected an extreme underrepresentation of both in the LcαE7 sequence data.

The H statistic tests for an overabundance of high-frequency variants in a population. Such a test should detect recent episodes of genetic hitchhiking where one or a few beneficial alleles have been driven to high frequency in a population (Fay and Wu 2000). A significant H statistic was recovered for the L. cuprina data, suggestive that a recent selective sweep has or is occurring. While such a result is not unexpected, it would have been a serious failure of the test if it was unable to detect this very intensive selection and abundance of highly selected alleles. In comparison, Tajima’s D statistic detects deviations from neutral expectations through the occurrence of low-frequency variants. With our data this test was not significant, likely due to a balance of low-frequency variants contributed from the few susceptible alleles balanced against the lack of low-frequency variants in the four resistant haplotypes. During a selective sweep Tajima’s D can take on any value. It is only once an excess of low-frequency variants arises that this statistic becomes indicative of a past selective sweep. However, in this example there has been insufficient time (40 years, or not more than 200 generations) for these mutations to have arisen. On the other hand, the H statistic tests for an overabundance of high-frequency variants to detect hitchhiking. This only requires the sweep to have occurred, not subsequent mutations in the sweep allele(s), and so is able to detect very recent or ongoing sweeps (Prezeworski 2002).

The McDonald and Krietman test (1991) detects deviations from the neutral expectation that the ratio of nonsynonymous to synonymous changes is equivalent in interspecific versus intraspecific comparisons. Significant results with this test are usually associated with an overabundance of fixed replacement sites as evidence of positive selection. Using L. sericata for the interspecific comparisons, a significant result is obtained using Fisher’s exact test (although only just so; p = 0.046). There appears to be an overabundance of polymorphic replacement sites. Two of the replacement sites are in exon 3 and include the Gly137Asp, while seven are in exon 4 and include Trp251Leu. Assuming that the silent site variation is neutral, this result could be explained by either unusually high levels of intraspecific polymorphism, perhaps due to selection for different classes of OP resistance, or unusually low levels of interspecific protein divergence (Verrelli and Eanes 2000).

As well as being high in the LcαE7 gene, we would expect significant levels of linkage disequilibrium to extend well beyond the gene, since the selection pressure from OP use has been high and the sweep has occurred recently. RFLP analysis (Smyth et al. 2000) using the LcαE7 gene as a probe and further Southern analysis using other members of the αE7 cluster in L. cuprina (R.D.N., unpublished data) do suggest that the hitchhiking effect extends out across the α-esterase gene cluster. While the size of the cluster is not known in L. cuprina, in D. melanogaster it is about 60 kb (Robin et al. 2000a). This α-esterase cluster apparently contains a significant proportion of the detoxifying carboxylesterase genes in the dipteran genome (Oakeshott et al. 1999; Claudianos et al. 2002), so significant loss of diversity at this cluster may have implications to the species’ resilience to other ester xenobiotics. It will also be of interest to determine how much further the linkage disequilibrium extends beyond the α-esterase cluster in L. cuprina. It is possible that the selective sweep has reduced diversity at many linked loci on chromosome four. A reduction in haplotype diversity at this and linked loci would have a significant impact on the ability of flies to respond to selection pressures in the future.

Have geography and patterns of gene flow influenced the spread of resistance genotypes? Moderate levels of gene flow have been reported within Australian populations of L. cuprina using isozymes (Gleeson 1995b), while in New Zealand there are higher levels of population structuring, reflecting the prominence of significant barriers to gene flow such as mountain ranges and cold climate regions in the south (Gleeson 1995b). Such structuring can effect tests of neutrality such as Tajima’s D and the H statistic, however, most of our samples were from Australia, where levels of structuring are not great (Gleeson 1995b). Looking at haplotypes across geographic range there is no visible pattern of association (see Smyth et al. [2000] for a collection site map). The common OP-resistant D haplotype is found all over sheep farming areas of Australia and in New Zealand. The other diethyl-preferring OP-resistant haplotype, G, is found in both Australia and New Zealand. These data also lend weight to the argument that L. cuprina populations in New Zealand are derived from immigrants from Australia (Gleeson and Sarre 1997), bringing OP resistance with them.

Seven fourth chromosomes conferring resistance to both diethyl and dimethyl OPs were found in the isogenic (IV) strains made while screening for malathion resistance (Smyth et al. 2000). These seven strains carry two copies of the LcαE7 gene, one containing an Asp137 allele and the other a Leu251 allele. Three different haplotype combinations were detected among the seven strains: D/E, D/F and G/F (Table 1). These data are supported by RFLP analysis of two of these strains (Smyth et al. 2000). Tooronga 2.2 contains RFLP haplotypes D and E and 60NE3.1 haplotypes F and G (Smyth et al. 2000). These strains are not segregating for these alleles as revealed by Southern analysis of multiple individuals (Smyth et al. 2000). The added benefit of having both resistant alleles expressed, likely resulting in twice the amount of enzyme and of both resistant forms, results in a superior resistance phenotype for the fly (Campbell et al. 1998b). A similar duplication of an insecticide resistance locus has been inferred in mosquitoes (Bourguet et al. 1996). In insecticide resistant C. pipiens from the Caribbean, the acetylcholinesterase locus ACE.1 has become duplicated, with one copy being a susceptible allele and a second carrying mutations that confer insensitivity to the insecticide. While in this system only one of the copies is a resistant form, perhaps having the ancestral susceptible acetylcholinesterase gene present allows the organism to overcome any supposed fitness costs of using a mutated resistant acetylcholinesterase. In another case a metabolic esterase gene, estβ1, has been duplicated before being amplified multiple times to confer resistance to OPs in the TEMR strain of C. pipiens. The nonamplified copy is distinct from the amplified version of the gene indicating that there may be some differences in their ability to confer resistance (Raymond et al. 1998).

Gene duplication is one of the most important processes involved in the evolution of new genes and functions (Ohno 1970). Classically gene duplication is considered to take place before functional diversification. Because the gene duplicates are thought to be redundant in function when they are created, it is predicted that one of the duplicates will become silenced by mutation. Other models have suggested that the ancestral gene may possess multiple functions or sites of expression and that both copies of the gene would be required to maintain both roles or sites of expressions (Hughes 1994; Force et al. 1999; Lynch and Force 2000; Massingham et al. 2001). In particular, Hughes (1994) predicted that a period of “gene sharing” may predate the evolution of gene duplication, with multiple alleles encoding multiple functions at a single locus, rather than another function evolving in the duplicated gene after the duplication event. In this context the duplication of the LcαE7 gene can be seen as a way of capturing the benefits of both resistance alleles that previously only heterozygotes might enjoy. Certainly it is most parsimonious to infer that the duplication events postdate the origin of the two resistance alleles. The chances of creating identical haplotypes on the duplicated chromosomes as seen on the nonduplicated versions seems unlikely. It is interesting to note that none of the observed duplicated chromosomes involve a susceptible allele but all involve one Asp137 and one Trp251. This begs the question of the relevance of the susceptible allele in L. cuprina. We believe that this example of OP resistance conferred by the LcαE7 gene represents a case of gene sharing preceding gene duplication. We also note that such a process may have driven at least some of the other α-esterase gene duplication events to form the cluster within which LcαE7 now resides.

As a model of rapid evolution, OP resistance in L. cuprina shows many additional properties beyond the previous studies of resistance in insects. In particular, there are two allelic amino acid replacements, each with a distinct effect on function and presumptive selective advantage. Notwithstanding the multiple mutations, there is extensive linkage disequilibrium and an overabundance of high-frequency (resistant) alleles, highly indicative of a selective sweep. In addition, we see rapid exploitation of duplication events that combine the two classes of resistance alleles onto the same chromosome, with multiple independent duplication events involving different haplotypes. Thus the species response to the new selection pressure of widespread insecticide use appears to be exploiting both replacement site changes and chromosomal rearrangements, which, if the selection pressure continues, could significantly alter the constitution of the α-esterase cluster and surrounding genes in L. cuprina.