Introduction

Dichlorodiphenyltrichloroethane (DDT) has gone from being a worldwide panacea of insect control and probably the most famous pesticide in modern history, to being a critical flashpoint of the modern environmental movement and becoming the most infamous pesticide of recent times. DDT was first used to control pest insect populations beginning in the 1940s, but instances of field resistance were observed among many species of insects, including Drosophila melanogaster (Drosophila) [1]. Subsequent deleterious side effects were observed in non-target mammalian and avian species, and were linked to the environmental persistence of this insecticide or its metabolites [2–5], ultimately leading to DDT being banned in most countries. However, DDT remains in industrial production due to its continued use for the control of mosquitoes that vector malaria; a niche where DDT is very effective [6].

DDT disrupts nervous system function in arthropods by affecting nerve cell plasma membrane permeability and causing paralysis [7]. Contrary to expectations, when selection pressures for DDT resistance were eliminated following bans on DDT use in many nations, the frequencies of resistance phenotypes remained high in many endemic pest insect populations. Persistence has been attributed to random genetic drift of alleles that have no fitness costs compared to susceptible counterparts [8]. Additionally, DDT resistance mechanisms can confer cross-resistance to pyrethroid [9, 10] and neonicotinoid insecticides [11], and may be a factor that contributes to maintenance of resistance alleles in the absence of a direct DDT selection [12].

The genetic basis of DDT resistance traits in the mosquito Anopheles gambiae involves the additive effects of two quantitative trait loci (QTL) [13]. Similarly, two QTL with major effects were mapped to the para sodium channel and the CCEunk7 esterase genes in Aedes aegypti, along with minor QTL that implicated the role of 20 other detoxification enzymes [14]. Additionally, a significant amount of research on DDT resistance in Drosophila has focused on metabolic resistance [11, 12, 15–17].

DDT resistance in Drosophila is not a uniform phenotype, with varying levels of resistance observed across different Drosophila strains, and resistance can roughly be categorized into low-, medium- and high-level resistance [18]. Initial work on DDT by Crow [19] demonstrated the polygenic nature of DDT resistance, however, subsequent research singled out one low-level DDT resistant phenotype involving the Rst(2)DDT locus on chromosome two [20]. The chromosome region of Rst(2)DDT contains two cytochrome monooxygenease (P450) genes, Cyp6g1 and Cyp12d1, that are over expressed in at least some DDT resistant strains [18, 21]. Transcription of Cyp6g1 in DDT resistant Drosophila was up-regulated by an upstream insertion of the Accord transposon [22], and all subsequently described field resistant strains similarly show an over expression of Cyp6g1 due to this Accord insertion [12].

In contrast, DDT resistance among laboratory strains involves over expression of multiple P450 genes in addition to Cyp6g1 and Cyp12d1 [23, 24]. Specifically, the laboratory selected strain 91-R showed significant increases in expression of multiple cytochrome P450s and dozens of other genes. Furthermore, over expression of Cyp6g1 in transgenic Drosophila with a susceptible genetic background failed to reconstitute high levels of DDT resistance [25] and reinforced the hypothesis that DDT resistance may be a multilocus trait in this species [17, 26]. Indeed, a recent toxicokinetic analysis of 91-R revealed that oxidative P450s likely causes little direct metabolic resistance, but reduced cuticular penetration, increased reductive dechlorination, and enhanced excretion have been shown to play dominant roles [27].

With the advent of next generation sequencing (NGS) technologies, full-genome re-sequencing has become logistically feasible, and allows for ultra-fine resolution to map the genome location of mutations [28]. It is also a tool for genome-wide association studies (GWAS), population genomic [29] and phylogenomic studies [30]. GWAS using NGS-based re-sequencing approaches has been effectively applied to estimate genome variation within and between populations, and has identified genome regions that are associated with the expression of various traits [31–33]. GWAS in Drosophila are feasible due to a high-quality genome sequence assembly, gene models and tools for genome-wide molecular analyses [5, 34, 35].

The laboratory selected DDT resistant and control strains 91-R and 91-C, respectively, represent lines that have a common origin, have been kept side-by-side in the laboratory, yet while 91-C received no DDT selection, 91-R received intense DDT selection pressure for over 50 years. Using these resource populations, whole genome re-sequencing data were generated from pools of individuals from the Drosophila strains 91-R and 91-C, and applied to detect chromosome regions putatively affected by prolonged DDT selection. To the authors’ knowledge, analysis across the entire insect genome to identify regions influenced by selective sweeps in highly DDT resistant insects has not been previously performed, although there has been a study focused on insecticide resistant blow flies (Lucilia cuprina) examining selective sweeps around an individual gene [36] and some additional studies looking at signatures of selection around transposable elements [37, 38]. We tested the hypothesis that the expected usual gene suspects for DDT resistance would be detected, as opposed to the alternative that resistance is polygenic with many other genes impacting resistance. Elucidating these genetic and biochemical mechanisms associated with pesticide resistance evolution might lead to improved pest management strategies. Equally important, such information has the potential to further our understanding of how DDT impacts biological processes that are evolutionarily conserved between insects and mammals.

Results

Detection of selective sweeps in 91-R by mapping Pool-seq data

Mean nucleotide diversity (π) and Tajima's D (D) estimates were made among major chromosome arms for mapped read data from strains 91-R and 91-C (Table 1). Both estimates were generally higher for 91-R compared to 91-C and were also ~10-fold lower on the X chromosome compared to autosomes (Table 1). The variation within sliding window estimates for π and D were also greater for 91-R compared to 91-C. Use of an arbitrary cutoff of a ≥ 100-fold reduction of π in a given 500-kb window compared to the mean π across the same chromosome resulted in the identification of 13 genome intervals in the 91-R genome (Fig 1A). Strong evidence for directional selection was not found on chromosomes X or 3L. Reduced nucleotide diversity was used as a proxy for the identification of genome regions, which potentially show the effects of a selective sweep, where the size of these 13 genome regions ranged from 0.1 to 0.9-Mbp (Table 2). A second tier cutoff of a ≥ 90- but < 100-fold reduction of π in a given 500-kb window was used to identify putative genome regions affected to a lesser degree following chronic DDT exposure in 91-R, and resulted in the identification of three additional genome intervals showing “minor” selective sweeps (Fig 1A). By comparison to estimates obtained across the genome of 91-C, the effects of random genetic drift, as opposed to directional selection, could possibly be accounted for on the 91-R genome. Decreases in the estimates of D along chromosome arms were also interpreted as being derived from the effects of directional selection in those corresponding genome regions. The calculated values of D along 500-kb windows of genomic sequence from 91-R indicated the effects of negative section were detected in 2L and 3R, and that these genome regions corresponded to those also predicted to show significant reductions in π described above. Similar to the estimates of π described above, there were no intervals on chromosomes X or 3L that showed significant reductions in estimates of D in the 91-R or 91-C genome. In contrast, the estimates of D did appear to mirror those of π along 2L, 2R and 3R in 91-R but not 91-C (Fig 1B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Estimates of (A) nucleotide diversity (pi; π) and (B) Tajima's D (D) across the chromosome arms of Drosophila melanogaster from the DDT resistant 91-R strain (black line) and DDT susceptible 91-C strain (grey line).

Metrics obtained from 500-kb sliding windows with a step size of 100-kb. The genes identified to be associated with each of the major selection sweeps are as follows: (1) CG42329, (2) CG15394, (3) NinaC, (4) CG6453, (5) CG17568 and Ref(2)P, (6) RtGEF, (7) CG12050, CG8677, and Dtr, (8) CG31612, (9) Dscam1, (10) Sut1, (11) MDR49, (12) CG1041 and (13) CG31495. The genes identified to be associated with each of the minor selection sweeps are as follows: (A) Cyp4g1; (B) AlstR, Mnt and Fd3F; and (C) Kon.

https://doi.org/10.1371/journal.pone.0123066.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Mean nucleotide diversity (pi; π) and Tajima's D (D) among chromosome arms for DDT resistant 91-R and susceptible 91-C Drosophila melanogaster strains.

https://doi.org/10.1371/journal.pone.0123066.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Thirteen genomic intervals in Drosophila melanogaster strain 91-R (ID 1 to 13) with nucleotide diversity estimated reduced ≥ 100-fold compared to respective chromosome means, and indicate putative genome regions under directional selection for survival when exposed to DDT.

https://doi.org/10.1371/journal.pone.0123066.t002

Discussion

In the current study we used a whole genome approach to detect nucleotide signatures of directional selection in the 91-R population that resulted following chronic DDT exposure. Specifically, the effects of DDT selection on localized regions of the 91-R genome were measured by reductions in nucleotide diversity and corresponding estimates of directional selection using Tajima's D. The implication of individual genes in DDT resistance was obscured using this approach since genetic hitchhiking of flanking genes and genome regions occurs during selection due to limits on recombination to reduce the size of haplotype blocks [39, 40]. Thus, selective sweeps in 91-R have resulted in the near fixation of nucleotide sequence of the causal genetic factor(s) that are directly involved in the genes giving rise to this DDT resistant phenotype, as well as genes in proximity. It is important to note that some gene(s) located within some of the selective sweeps may appear associated with the resistant phenotype in 91-R, but may have been carried to fixation due to physical genetic linkage to gene(s) with major effect due to genetic hitchhiking.

Overall, genes within the 13 major (Fig 1 and S6 Table) and three minor effect regions (Fig 1 and S6 Table) identified in 91-R by selective sweeps do not correspond to the genes that have been previously associated with DDT resistance using gene expression analyses or identified in pesticide resistant Drosophila populations. The only exception is gene NinaC, which has predicted kinase activity related to sensory transduction/vision, and is over-expressed in DDT resistant strains [26]. Similarly, four of the genes located in genome regions of 91-R affected by selective sweeps, Dscam1, Dtr, RtGEF, and CG6453, have previously been implicated in synaptic development and/or function [41–44]. Additionally, the candidate genes of 91-R affected by selective sweeps may also be involved in regulation of cellular growth (CG6453) [44], as well as genes that are involved in cellular communication and signal transduction cascades (S4 Table). Although implication of these mutant alleles in DDT resistance can be rationalized, additional functional studies are required to deduce individual roles as well as the effects of the non-synonymous changes in 91-R on subsequent protein function and resulting phenotype. Additionally, the nucleotide diversity (pi; π) and D in the sliding window figure indicated that the control (91-C) strain showed no evidence for reduced variation or coding sequences (CDS) selection, in that sliding windows are fairly uniform across the chromosome—with exception of certain regions likely representing those near centromeres.

The genes CG17568, ref(2)P, CG8677, and CG31612 located in selective sweeps five, seven and eight, respectively, contain Zn-finger DNA binding motifs, which could suggest the resultant proteins have potential roles as transcription factors. Since trans-regulatory control of transcription by soluble transcription factors often occurs at cis-promoter or enhancer elements by way of Zn-finger mediated protein-DNA interactions, mutations in transcription factors that affect DNA-protein or protein-protein interactions at the promoter or with enhancer elements can cause changes in expression at physically unlinked genes. Therefore, genes in an interconnected gene regulatory pathway may show a coordinated response to transcription factor mutations. It is conceivable (but speculative) that the mutant transcription factor alleles in 91-R might be involved in the gene regulatory networks which lead to up-regulation of the transcripts in 91-R described by Pedra et al. [17], a hypothesis that remains to be tested. In a broader context, the basal cause of apparent incongruent results obtained from gene expression and genetic mapping/phylogenomic studies may be rooted in the effect that genes in QTL intervals/selective sweeps have upon gene networks. This hypothesis might also suggest that system approaches may yield greater insight into the genetic and genomic basis of insecticide resistance traits.

In the three genome regions showing less major effects of selection at a ≥ 90-fold reduction in nucleotide diversity cutoff, (A, B and C in Fig 1 and S2 Table), the candidate genes were associated with the nervous system and only one was a P450, Cyp4g1. The Cyp4g1 protein is known to be associated with hydrocarbon production, converting long-chained aldehydes to long-chained hydrocarbons in oenocytes in the epidermis of Drosophila, that are then in turn transported to the waxy layer of the epicuticle [45]. Additionally, Strycharz et al. [27] recently demonstrated that 91-R had higher quantities of cuticular hydrocarbons, visible changes in the cuticle (via electron microscopy), and that reduced penetration is an important component of DDT resistance in 91-R. Although work by Waters et al. [46] suggested no difference in expression levels of Cyp4g1 between 91-C and 91-R, it is not currently known if differential expression of the Cyp4g1 protein might be localized near the cuticle of the insect or if structural changes in Cyp4g1 may play some role in resistance.

Although ATP-binding cassette (ABC) transporters have previously been associated with DDT resistance in Drosophila, the selective sweeps analysis has shed new light on an additional candidate DDT resistance gene known as MDR49. Strycharz et al. [27] previously compared the transcription levels of the ATP-binding cassette transporters MDR49, MDR50 and MDR65 and MRP1 in 91-R versus Canton-S strains. Interestingly, MDR50, MDR65 and MRP1 were over-expressed in 91-R whereas MDR49 was not. RNAi knockdown of MDR50, MDR65 and MRP1 in DDT resistant flies results in increased sensitivity to DDT, however, knockdown of MDR49 had no effect [47]. Such aforementioned experimental approaches would only detect the putative role of differential transcription in DDT resistance and not structural changes in the protein that may lead to DDT resistance. Thus, the amino changes that we observed in MDR49 that may play a role in DDT resistance, if any, remain to be determined.

The other genes co-occurring with selective sweeps included: Dscam1, NinaC, CG6453, CG17568, Ref(2)P, RtGEF, CG12050, CG8677, Dtr, CG31612, Sut1, CG1041, and CG31495. Several of these aforementioned genes located within the identified selective sweeps contain known or suspected orthologs in other animals, including mammals, which could be useful in further investigation to better understand their potential links to DDT resistance. For example, some of these genes show plausible linkages to phenotypic resistance to DDT, however, based on their roles in mammals in biological processes known to be impacted by DDT exposure. For example, Dscam1 is associated with psychomotor retardation, and DDT (or more specifically the DDE byproduct) has been linked with retarded psychomotor development in humans exposed in the first trimester [48, 49]. NinaC is a retinal specific gene that codes for two photoreceptor cell specific proteins in Drosophila [50, 51], and mutations in NinaC were shown to cause light- and age-dependent retinal degeneration in Drosophila [51]. In human studies, Kamel et al. [52] found a dose-response relationship between exposure to organochlorides, such as DDT, and the risk of retinal degeneration. Male sterility in Drosophila results when Ref(2)P gene expression is absent in the testes, which suggested that the Ref(2)P gene expression is required for successful reproduction [53]. Although a somewhat controversial topic in the literature, there have been studies indicating that environmental pollutants (such as DDT) have links to male infertility [54, 55] in both humans [54] and rats [55]. The role these genes may play in high level DDT resistance remains to be determined. Additionally, it remains to be determined if some of these evolutionarily conserved, between insects and mammals, candidate genes may also provide insights into the impact of DDT exposure in mammalian systems, as has been done previously using Drosophila and human diseases [56–60].

Interestingly, we did not observe many of the genes typically associated with pesticide resistance [19, 61–65, 68–76]. Many initial publications reported that moderate to high level DDT resistance is thought to be polygenic with multiple genome regions contributing to this phenotype [19, 63, 64]. Previous researchers have identified loci on the second chromosome involved in DDT resistance [20, 65], whereas chromosomes X and three are thought to have some slight impact on the DDT resistance phenotype [64, 65]. The uniform reduction in estimates of π across the X-chromosome of 91-C and 91-C, were not surprising due to the effective 3/4 population size of the X-chromosome as compared to autosomes [66], which affects the rate at which chromosomal loci may become fixed by random genetic drift or influenced by selection [77]. The affects of random fixation by random genetic drift on the X-chromosome may also have been exacerbated by the relatively small number of individuals in 91-C and 91-R laboratory populations [67]. An increasing amount of research has focused on the single Rst(2)DDT locus [20, 68–71], whereby over-transcription of the cytochrome P450 Cyp6g1 located within the Rst(2)DDT region was suggested to be both necessary and sufficient for DDT resistance [25]. Daborn et al. [25] essentially proposed that resistance to DDT was monogenic, at least in Drosophila strains with low-level DDT resistance. The Rst(2)DDT locus maps to the second chromosome between the genes cinnabar (cn; location 2R:3,670,302–3,672,711) and vestigial (vg; 2R:8,772,137–8,786,890) [11, 21, 72, 73]. Although our current dataset is not from a low-level resistant strain, this genome interval in 91-R where Cyp6g1 exists is located between, and not within, the 91-R selective sweeps labeled 9 and 10 (Fig 1). Interestingly Cyp6g1 is over-expressed in the 91-R strain [17, 21, 78]. Thus, our analysis suggests that Rst(2)DDT is not a major factor, or potentially even involved, in the DDT resistance phenotype in 91-R, and agrees with prior results which showed that DDT resistance could be maintained in Drosophila strains that did not show high Cyp6g1 transcript levels derived from at the Rst(2)DDT locus [74]. The current results are also in agreement with the conclusion by Strycharz et al. [27] that metabolic resistance, particularly P450-based resistance, plays a negligible role in the overall DDT resistance phenotype in 91-R. In addition to Cyp6g1, expression of Cyp12d1 was implicated in being differentially expressed in DDT-resistant fly strains [18] as were non-synonymous coding sequence mutations Cyp6a2 [75] and para [76]. However, none of these genes occurred in any of the selective sweeps identified in our current experiments.

DDT resistance, however, is not a single phenotype, but varies among strains from low- (e.g., Hikone-R), to moderate- (e.g., Wisconsin), to high-levels (e.g., 91-R) [18]. Based on microarray analysis, moderate- to high-level resistant phenotypes appeared to result from the effect of multiple differentially-regulated genes. Specifically, using microarray data, Pedra et al. [17] observed that numerous genes were over-transcribed in the 91-R strain including cytochrome P450s, glutahione S transferases, and a set of additional genes. Comparative analysis of microarray data from the Wisconsin and 91-R strains showed that multiple genes were differentially expressed, and that these genes were more numerous in the more highly resistant strain than the moderately resistant strain [17, 79]. A proteomic analysis also revealed that proteins associated with energy metabolism were differentially expressed in two DDT resistant as compared with a susceptible strain [26]. These combined observations suggest that at least moderate- to high-level DDT resistance may involve complex molecular interactions, and this might be consistent with a resistant phenotype that results from the effects of multiple genes. These results also suggest that several different genetic mechanisms may result in DDT resistant phenotypes, and that increasing levels of DDT resistance may be additive with an increasing number of genes involved, a hypothesis that remains to be tested. However, these results do not rule out the fact that low-level DDT resistant strains, taken directly from the field, may have fewer, i.e., monogenic, molecular mechanisms of resistance such as overly transcribed Cyp6g1 [78].

Certainly, the current data and analysis could be of potential importance to those insecticides where DDT resistance has been shown to confer cross-resistance to other types of insecticides, such as imidacloprid in Drosophila and pyrethroids in A. aegypti [11, 80]. Of greater practical importance, however, this general GWAS approach could be applied to other insect species currently being controlled by other pesticides, in order to understand the evolution of resistance in “real time” (i.e., follow field populations through generations of selection). A number of other insect genomes have been sequenced since the Drosophila genome was published in 2000 and has allowed for the study of insecticide resistance at the molecular level for a variety of species, such as in Anopheles gambiae [34, 81]. Although the 91-C and 91-R fly lines provide a unique system where selection has occurred for over half a century, there exist multiple Aedes aegypti laboratory strains, including strains selected for insecticide resistance to permethrin, where similar studies could be performed to identify structural mutations across the genome [82].

The discoveries of novel resistance mechanisms from such studies could help lead to new target genes and the development of novel control methods for these resistant species [83, 84]. This work also highlights that selection with DDT may result in the selection for novel mutations, potentially with some or many of these being associated with or directly involved in DDT resistance. Additional studies are required to validate the role of genes in each predicted selective sweep in DDT resistance by verifying the functional consequence of amino acid changes on protein structures and potential impact on the 91-R resistant phenotype. Of greatest importance, this study highlights the need for selective sweep analyses in pesticide resistant insect populations in order to identify potential candidate resistance traits. Further molecular examination of individual genes and a more detailed analysis of the specific effects of the structural changes within the insects are crucial to better understanding resistance, something beyond the scope of the current project.

Materials and Methods

Supporting Information

Acknowledgments

The authors wish to thank Dr. Ranjan Ganguly at the University of Tennessee-Knoxville for providing strains 91-R and 91-C material.