The G119S ace‐1 mutation confers adaptive organophosphate resistance in a nontarget amphipod

Abstract Organophosphate (OP) and carbamate (CM) insecticides are widely used in the United States and share the same mode of toxic action. Both classes are frequently documented in aquatic ecosystems, sometimes at levels that exceed aquatic life benchmarks. We previously identified a population of the nontarget amphipod, Hyalella azteca, thriving in an agricultural creek with high sediment levels of the OP chlorpyrifos, suggesting the population may have acquired genetic resistance to the pesticide. In the present study, we surveyed 17 populations of H. azteca in California to screen for phenotypic resistance to chlorpyrifos as well as genetic signatures of resistance in the acetylcholinesterase (ace‐1) gene. We found no phenotypic chlorpyrifos resistance in populations from areas with little or no pesticide use. However, there was ~3‐ to 1,000‐fold resistance in H. azteca populations from agricultural and/or urban areas, with resistance levels in agriculture being far higher than urban areas due to greater ongoing use of OP and CM pesticides. In every case of resistance in H. azteca, we identified a glycine‐to‐serine amino acid substitution (G119S) that has been shown to confer OP and CM resistance in mosquitoes and has been associated with resistance in other insects. We found that the G119S mutation was always present in a heterozygous state. Further, we provide tentative evidence of an ace‐1 gene duplication in H. azteca that may play a role in chlorpyrifos resistance in some populations. The detection of a genetically based, adaptive OP and CM resistance in some of the same populations of H. azteca previously shown to harbor a genetically based adaptive pyrethroid resistance indicates that these nontarget amphipod populations have become resistant to many of the insecticides now in common use. The terrestrial application of pesticides has provided strong selective pressures to drive evolution in a nontarget, aquatic species.


| INTRODUC TI ON
Organophosphate (OP) and carbamate (CM) pesticides have been widely used in the United States since the early 1970s when the U.S. Environmental Protection Agency (EPA) banned the organochlorine dichlorodiphenyltrichloroethane (DDT). Both OP and CM pesticides have the same mode of action, targeting acetylcholinesterase (AChE; EC 3.1.1.7). They elicit toxicity in insects by binding to AChE and preventing the breakdown of acetylcholine in the neuronal synapses, thus leading to over excitation of the nervous system, paralysis, and death (Aldridge, 1950). Human toxicity concerns associated with the OPs have led to the banning of some of these chemicals by the EPA (e.g., azinphos-methyl, ethyl parathion) and severe restrictions placed on the use of others (methyl parathion).
Among those restricted were the OPs chlorpyrifos and diazinon, for which all products intended for residential use were withdrawn from the marketplace in the early 2000s, though their agricultural uses continue. As a result of the restrictions on many OP products and general regulatory pressures to reduce their use, the annual use of OPs and CMs by professional pesticide applicators in California declined from a peak of nearly 8 million kg in 1995 to ~2 million kg in 2016 (CDPR, 1995(CDPR, , 1997(CDPR, , 2016 These quantities do not include residential home and garden use by nonprofessionals, which is not tracked by the state. The widespread use of OPs and CMs poses a risk to aquatic ecosystems. Both pesticide classes are frequently detected in U.S. streams and rivers, sometimes at levels that exceed the established benchmarks for aquatic life and contribute significantly to urban stream impairment (Stone, Gilliom, & Ryberg, ). Historically, monitoring in California has found both urban and agricultural runoff to transport toxic concentrations of OPs into aquatic ecosystems (Bailey et al., 2000;Kuivila & Foe, 1995;Kuivila & Foe, 1995;Kuivila & Foe, 1995). In more recent years, OP concentrations in urban runoff have declined dramatically due to the cessation of residential use of diazinon and chlorpyrifos, and urban environmental concentrations of these compounds are now quite low (Weston, Holmes, & Lydy, 2009;Weston & Lydy, 2010). Toxicity due to OPs in California agricultural runoff is observed less frequently and is less widespread than it was in the 1990s, yet since that time, OPs from agricultural use have been implicated in acute toxicity to aquatic invertebrates including daphnids (Ceriodaphnia dubia), mayflies (Procleon sp), midges (Chironomus dilutus), and amphipods (Hyalella azteca) (Anderson et al., 2006). In addition, legacy OPs sequestered in stream sediments contribute to macroinvertebrate pesticide exposure as well (Rasmussen et al., 2015).
Pest insects can serve as models for understanding the effects of pesticides on aquatic organisms, especially given that insecticides do not discriminate between target and nontarget organisms in their mode of toxicity. In fact, the evolution of adaptive pesticide resistance caused by pesticide selective pressure is common among target pest insects (Feyereisen, Dermauw, & Van Leeuwen, 2015) and has been documented in some populations of the nontarget aquatic invertebrate H. azteca (Major, Weston, Lydy, Wellborn, & Poynton, 2018;Weston et al., 2013). In H. azteca, resistance to pyrethroid insecticides is both genetically based and predictable, with resistance occurring exclusively in waterways near land uses associated with pyrethroid applications . In pyrethroid-resistant H. azteca from urban or agricultural areas, resistance was explained by the presence of any of several mutations in the gene coding for the pyrethroid target site, the voltage-gated sodium channel (vgsc) Weston et al., 2013). In one agricultural site with resident pyrethroid-resistant H. azteca near Salinas, California (Chualar Creek), we found the OP chlorpyrifos at acutely toxic levels in the sediment at 13 times the 10-day LC 50 of sensitive laboratory H. azteca populations (Weston et al., 2013). Since pyrethroids and organophosphates have different modes of toxic action, the pyrethroid-related mutations would be unlikely to confer resistance to chlorpyrifos. Thus, the persistence of the Chualar Creek H. azteca population in an environment with acutely toxic levels of chlorpyrifos suggests the population may have an additional adaptive mechanism providing OP resistance. However, the mechanism of that resistance has not yet been characterized.
Resistance to pesticides in some populations of H. azteca indicates a substantial pesticide presence in the environment capable of eliminating sensitive taxa. Pesticides acting as strong selective pressures may reduce genetic diversity via population bottlenecks and "genetic erosion" (Van Straalen & Timmermans, 2002). In pyrethroid-resistant H. azteca, fitness costs including decreased thermal tolerance and greater sensitivity to other chemicals have been associated with resistance to pyrethroids (Heim et al., 2018). Further, increased bioaccumulation potential in pyrethroid-resistant H. azteca and greater trophic transfer of pyrethroid residues to their fish predators have also been documented (Muggelberg et al., 2017). If some H. azteca have also evolved resistance to OP (and/or CM) pesticides, similar ecological and evolutionary costs could occur. Determining the extent of and mechanism behind the chlorpyrifos resistance in H. azteca as suggested by our Chualar Creek, observations must first be explored before we can fully comprehend the eco-evolutionary impacts of pesticide use on these nontarget organisms.
In the present study, we use many of the populations from our previous work throughout California  to screen H. azteca populations from agricultural and urban areas for chlorpyrifos resistance. Then, we investigate the mechanism of chlorpyrifos resistance in H. azteca by focusing on the gene that codes for the target site of OP and CM insecticides, acetylcholinesterase (ace-1).

| Site selection
Wild H. azteca were obtained from 17 sites throughout California, mostly between October 2014 and November 2015 as recorded in Major et al. (2018), but with the collection of one population in January 2018 (Table S1). All sites are herein designated with a threeletter code derived from the site name (e.g., Mosher Slough = MSH).
The sites were a priori placed into one of three categories: (a) little or no OP or CM exposure expected, usually due to lack of development in the watershed (referred to as LowOCU sites); (b) extensive residential and commercial development in surrounding lands (Urban sites); and (c) intensive irrigated agriculture in surrounding lands (Agricultural sites). Some of the agricultural sites also had large population centers in the watershed, so would have some urban influence as well.
The rationale for these three groups can best be understood in light of current and historical OP and CM use in California. Table 1 compares their use from the present day (2016 most recent data available) to their use in 1995, representative of a period of high OP and CM use and prior to many regulatory restrictions since placed upon the OPs. In the 1990s, OP use was far greater than it is currently, and it was prevalent in both agricultural and urban environments. Much of the diazinon was used for residential purposes. Chlorpyrifos was primarily an agricultural pesticide, but since the total amount used annually in California was 1,500 metric tons, even the 18% applied in nonagricultural environments represented a substantial quantity.
Malathion also had significant nonagricultural use. The withdrawal of We define nonagricultural use to include landscape maintenance, structural pest control, protection of public health, regulatory pest control, treatment of rights-of-way, and application to golf courses. Agricultural use comprises use in the growing and processing of crops. a The "na" indicates the percentage of nonagricultural use is not applicable since the total annual use is zero. b Subsequent to their original publication, the California Department of Pesticide Regulation adjusted the 1995 use data to remove suspected erroneous entries. The 1995 use totals shown are the revised amounts (DPR 1997) rather than those originally published (DPR 1995). However, the percentages of nonagricultural use could only be calculated using the original data. For nearly all compounds listed, the adjustments were trivial and inconsequential to this analysis. However, for carbaryl and diazinon the adjustments resulted in reducing the annual use by nearly half. Thus, the percentage of nonagricultural use shown for these two compounds may not be accurate.  Weston et al. (2013) and Major et al. (2018).

| Analysis of chlorpyrifos in sediment and water
For most sites, with the exception of Outlet Creek (OTL), surficial sediment (0-2 cm) was collected and sent to Southern Illinois University for chlorpyrifos analysis (Table S1). Water samples from chlorpyrifos toxicity tests (see below) were also sent for analysis.

| Chlorpyrifos toxicity testing
Field-collected and laboratory-cultured H. azteca were challenged with the OP chlorpyrifos in 96-hr water-only exposures. Organisms were size-fractionated prior to testing, and preference was given to juveniles that were able to pass through a 600-µm screen, but were retained on a 500-µm screen. For populations in which sufficient juveniles within that size range were unavailable, larger animals retained on a 1,000-µm screen, but passing through a 2,000-µm screen were used. Rostrum-to-telson body length measurements were recorded for approximately 30 individuals per population (Table S2).
Chlorpyrifos sensitivity was assessed using the 96-hr water-only acute toxicity methods as described in Major et al. (2018). Briefly, three replicate beakers, each containing 80 ml of media, 10 H. azteca and a 1-cm 2 piece of nylon mesh substrate were prepared for each treatment concentration. Because of difficulty with gender identification in H. azteca before euthanizing animals, gender was not scored before the toxicity tests were initiated. Base media consisted of Milli-Q deionized water reconstituted with salts and bromide (Borgmann, 1996;Smith, Lazorchak, Herrin, Brewer-Swartz, & Thoney, 1997 given that the testing was done using wild-collected individuals from diverse habitats throughout California, and the animals transported back to the laboratory, some relaxation of the survivorship threshold is reasonable. We provide control survivorship data for every test (Table S2) and include all tests exceeding 70% survival.
Testing of some populations yielded a small number of individuals that appeared healthy and exhibited normal behavior at concentrations far above those that had caused mortality to the majority of the test population. These were termed "survivors" and were set aside in ethanol for later analysis to determine the genetic basis for their insensitivity to chlorpyrifos toxicity. They are designated herein with an "S" suffix after the site name of the population (e.g.,

Mosher Slough survivors = MSH_S).
For each toxicity test, water was analyzed for chlorpyrifos from one concentration in the mid-point of the testing range. Each sample for water analysis included media from test initiation composited with the mid-test water change. Actual concentrations ranged from 69% to 120% of nominal concentrations. Reported LC 50 s were adjusted based on the actual, measured concentration of chlorpyrifos in that specific test, rather than nominal concentration.

| Genomic DNA extraction
Genomic DNA (gDNA) was extracted from individual H. azteca stored in ethanol. Preference was given to extracting males whenever possible to reduce the likelihood that offspring from a gravid female would contribute to the gDNA profile of an individual. When it was not possible to use only males, care was taken to prevent eggs from being transferred into the extractions. All extractions followed the Qiagen DNeasy® Blood & Tissue Kit (Qiagen) protocol with slight modifications documented in Major et al. (2018), including a steel-bead maceration step after initial buffer and proteinase K addition and an overnight incubation at 56°C. After extraction, gDNA was measured for purity (260/280 ratio) and nucleic acid concentration with a spectrophotometer (NanoDrop 2000; Thermo Scientific).

| COI genotyping
Hyalella azteca is known to be a species complex; thus, a 670-bp segment of the mitochondrial cytochrome c oxidase I (COI) gene was genotyped for a subset of individuals (5-10) from laboratory and wild populations to determine species identity within the complex.
For most animals in the present study, species identity had previously been determined and can be found in Major et al. (2018). Some animals, such as the chlorpyrifos toxicity test survivors as well as the ULC wild population collected in January 2018, had not been previously characterized. For those individuals, COI genotyping methods followed those in Major et al. (2018). Briefly, the COI segment was PCR-amplified using primer pairs IV, V, or VI (Table S3) and 5 min at 72°C. PCR products were gel-purified and sequenced with primer VII (Table S3) using an ABI 3730 automated sequencer.

| H. azteca species determination using two gene segments
COI sequences were primarily used for species determination in the present study, but in cases where COI sequences were not available for all individuals, a segment of a nuclear marker, the vgsc, was used to infer species affiliation as established and validated in Major et al. (2018). For most of the animals from wild populations in the present study, vgsc sequences came directly from the Major et al. (2018) study, and when they did not (ULC population and select chlorpyrifos toxicity test survivors), the same methods were employed to obtain http://www.atgc-montp ellier.fr/phyml/ ) was used to generate a maximum likelihood (ML) tree. The substitution model (HKY85 + G, gamma shape parameter = 0.264) was chosen automatically based on Akaike information criterion (AIC = 2,620). Branch supports of greater than 90% (10,000 bootstrap replicates) were retained and displayed on branches ( Figure S1) of an unrooted cladogram. Both COI and vgsc data were available for some individuals in the present study. In those cases, species determinations based on analysis of COI segments were overlaid onto the vgsc cladogram ( Figure S1).
Based on these distinctions, the highly supported branches of the vgsc trees were used to infer species affiliation for those individuals that were not sequenced at COI.
A minority of individuals from the present study, with available, but insufficient vgsc sequence quality or length (28), were excluded from the vgsc ML analysis, and thus, no species determinations could be directly inferred for those individuals. However, for sites at which only one species was identified, we assumed the remaining individuals at that site belonged to the same species group. For sites with more than one species present, COI and/or vgsc evidence was used to make species-level distinctions for all individuals.

| Target site gene cloning and resistance allele discrimination
To identify possible mutations associated with chlorpyrifos resistance in wild H. azteca, a segment of the acetylcholinesterase (ace-

| Ace-1 genotyping assay
An ace-1, direct genotyping assay was created based on the identification of a candidate resistance allele at amino acid position G119 (T. californica numbering) in select wild populations of cloned H. azteca (MSH, MSH_S, CLG, and CLG_S). As with amplicon cloning, primer pair I (Table S3) was used to PCR amplify the ace-1 segment using a high fidelity polymerase. PCRs had the same reagents, volumes, and settings as those listed for amplification of the vgsc segment (above), with the exception of the use of primer pair I instead of primer pair VIII (Table S3). After bands were confirmed on an agarose gel, they were cleaned with the QIAquick PCR Purification Kit (Qiagen) with a 40 µl elution volume. Between 200 and 300 ng of each cleaned PCR product was sent to the Massachusetts General Hospital DNA Core (Cambridge, MA) for sequencing with one of two internal primers. Internal sequencing primers were designed based on highly conserved ace-1 regions across the individuals using Primer3 v. 0.4.0 (Untergasser et al., 2012). Sequencing with Primer II produced a 575 bp sequencing read while a secondary primer (used in rare cases of primary primer failure; Primer III) produced a 352 bp read (Table   S3).
Because an individual's ace-1 alleles were sequenced concurrently (as in the vgsc genotyping assay in Major et al. (2018)), resulting sequences were manually examined and cleaned using IUPAC ambiguity codes. All sequences were cleaned, aligned, and trimmed using CLC Workbench v. 7.9.1 (https ://www.qiage nbioi nform atics. com/), and the resulting G119 genotype was scored by manual visualization of the sequence. Heterozygotes were indicated by a double peak at a single locus. When a secondary peak was 10% of the height of the primary peak or less at a locus, it was discarded as noise and the sequence was repeated with an alternative primer until a clean sequence was obtained. When secondary peaks were 10% of the height of the primary peak or greater, the individual was scored as a heterozygote at that locus. The ace-1 genotyping assay was verified by comparing genotyping assay results to the genotype profiles of the 16 individuals from which ace-1 amplicons were cloned.
Comparison showed that ace-1 assay genotypes were the same as those recorded from cloning for all individuals. In the present study, between 10 and 20 individuals from each of the survey populations of H. azteca were assayed for ace-1 genotype, including chlorpyrifos test survivors.

| Chlorpyrifos in sediment
Chlorpyrifos was detected at concentrations of ~2 ng/g in the sediments from four of the six Agricultural sites (MSH, CHL, CLG, and WHW). Chlorpyrifos was not detected at any of the LowOCU or Urban sites (Figure 1a; Table S1).

| Chlorpyrifos Sensitivity
The median chlorpyrifos 96-hr LC 50 for the UCB laboratory-cultured population was 154 ng/L (Table 2). Populations from all three LowOCU sites tested were comparably sensitive to chlorpyrifos with median 96-hr LC 50 s of 145-235 ng/L, and in most cases, with 95% confidence intervals overlapping those of the laboratory population (Table S2).
Of the three Urban sites, only one exhibited even a modest de-   Figure 1b). At all other Agricultural sites, the wild H. azteca were 47 to 1,000 times less sensitive to chlorpyrifos than the laboratory or sensitive wild populations. The LC 50 of the ULC population was 17,800 ng/L and that of the CLG population was 156,000 ng/L.
It was not possible to calculate a true LC 50 at CHL because the highest test concentration used of 20,500 ng/L was insufficient to produce 50% mortality.

| H. azteca species determination and pyrethroid resistance mutations
Analysis of COI sequences supported seven species groups previously reported by Major et al. (2018), designated as species B, C, D, E, F, Ps 17, and Ps 28 (Table 2). Pairwise COI sequence divergence between species groups ranged between 10% and 23% .
Overall, species identity did not explain the differences in chlorpyrifos sensitivity generally observed between LowOCU, Urban, and Agricultural populations. H. azteca from LowOCU sites were members of species B, D, E, Ps17, and Ps28, those from Urban sites were species B, C, and D, and those from Agricultural sites were species B, C, D, and F. The ace-1-genotyped survivors of chlorpyrifos toxicity tests were members of the same species group(s) as the overall population tested from their respective sites (Table 2). Only two sites yielded organisms with more than a single species group: RSN and MSH. In the initial MSH collection, ratios of the B and D species groups were relatively balanced (12 D and 8 B; Table 2), while a higher proportion of species D remained among chlorpyrifos survivors (11 D and 2 B).
In using the vgsc segment as a tool for species determination, we were also able to use these sequences to genotype for pyrethroid resistance alleles for the one wild population not previously included in Major et al. (2018). All ULC individuals were species D and had at least one pyrethroid resistance allele (either M918L, L925I, or L925V; Table S4). Frequencies of mutations were 0.15, 0.80, and 0.05 for M918L, L925I, and L925V, respectively.

| Variation in ace-1
Cloning of the ace-1 alleles for sixteen H. azteca individuals yielded 20 different alleles (based on 876-bp alignments; Figure S2 through Figure S3), with most individuals showing evidence of more than two ace-1 alleles ( Table 3). The documented allele diversity was unlikely to be due to polymerase error, as base pair changes at a given site were screened and only considered true changes if other H. azteca clones varied similarly at the same site. If a base pair change was observed in a single allele but was never documented in any other allele from other populations, it was discarded as polymerase error (with the exception of Allele 3 in UCB-see below). Thus, our approach to delineate alleles was conservative. In addition, only five clones were sequenced from each individual, also likely leading to a conservative estimate of total alleles per individual.
All alleles had high homology to one another in both nucleotide (97.3% and greater) and amino acid sequences (98.3% and greater), providing evidence that alleles were all ace-1, as opposed to ace-2, which has not been found in the H. azteca genome , but is common in other arthropods. The single laboratory individual for which ace-1 amplicons were cloned had three ace-1 alleles, one of which included a premature stop codon (Allele 3, Figure S4). It is possible that the base pair substitution leading to the stop codon was a polymerase error given that it was not found in any other clones, but the allele also had distinct motifs TA B L E 2 Proportion of individuals with a given acetylcholinesterase (ace-1) genotype (wild-type wt) or resistant (res)) in a survey of wild H. azteca from sites in California and chlorpyrifos toxicity test survivors

Median population chlorpyrifos 96-hr LC 50 (ng/L) Species Sample size (n)
Proportion of individuals with a given genotype at G119

GG (wt) GS (res)
Laboratory This value is an average of two toxicity tests for which LC 50 s could not be obtained due to mortality <50% at the highest concentration, one with the highest concentration of 1,520 ng/L and the other at 20,500 ng/L. The actual LC 50 for these populations could not be determined. b Values following population name indicate the concentration of chlorpyrifos survived by these individuals that appeared unimpaired at the end of the 96-hr test. c Allele frequencies from populations with fewer than five individuals should be regarded with caution.
Only three amino acid residues associated with resistance in insects (as reviewed by Fournier (2005) Table 2). The proportion of individuals with a GS genotype was always higher among chlorpyrifos toxicity challenge survivors than in the control population for a given Urban or Agricultural site (Figure 3). In most instances, the difference in resistant genotype frequencies between the population overall and the survivors was dramatic (e.g., resistant genotype found in 8% of the species D from Mosher Slough overall, but in 100% of the Mosher species D survivors; resistant genotype found in 20% of the Medea

| Proportion of OP resistance genotypes in H. azteca
Creek individuals overall, but in 90% of the Medea survivors).   Table 2 for site abbreviations). For BCM, designated by an asterisk (*), fewer than five individuals were genotyped at ace-1, and proportions are not displayed. When two or more species were identified at a site and they differed from one another in G119S genotype composition, proportions for both species are shown

| An ace-1 mutation is associated with every case of chlorpyrifos resistance in H. azteca
We found a glycine-to-serine amino acid substitution (G119S) in Urban and Agricultural sites that was most prevalent in H. azteca populations with an elevated resistance to chlorpyrifos, and was in nearly all individuals that were able to survive chlorpyrifos exposures up to 10 times the recorded LC 50 for a population, or in organisms from Urban and Agricultural sites. The proportion of individuals with a GS genotype was always higher in survivors from the chlorpyrifos toxicity test than in the general wild population, providing strong evidence that this mutation is associated with the development of resistance to chlorpyrifos. While it is possible that sensitivity in While we measured chlorpyrifos sensitivity in populations containing both sexes, we preferentially genotyped males in order to eliminate the possibility of screening mate DNA through inclusion of offspring in the DNA preps. In other aquatic crustaceans, pesticide resistance is sex-linked (Carmona-Antonanzas et al., 2017). However, despite our preference for males, we genotyped females from many sites, including from populations in all categories (LowOCU, Urban, Agricultural, and Survivors) and in all species groups. The G119S mutation did not appear to be associated with sex; thus, the role that sex plays in OP resistance in H. azteca, if any, cannot be elucidated by the present study.
The same G119S mutation was originally documented in the ace-1 of mosquitoes (Culex pipiens and Anopheles gambiae). In vitro assays using recombinant wild-type and mutant enzymes showed that insensitivity to propoxur (a CM) in C. pipiens was explained by this single amino acid substitution (Weill et al., 2003). The glycine-to-serine substitution occurs in the oxyanion hole near the catalytic triad, reducing access of the insecticide to the target site (Weill, Malcolm, et al., 2004). Although the first example of resistance through G119Sconferred target site insensitivity was identified using a CM, it has been widely established that this mutation confers both OP and CM resistance across multiple species of mosquitoes all over the world (Essandoh et al., 2013;Liebman et al., 2015). The G119S mutation has also been associated with chlorpyrifos resistance in another insect, the brown planthopper (Nilaparvata lugens). Chlorpyrifos treatment of a field-collected population for nine generations produced a resistant strain (253-fold greater LC 50 ) with the G119S mutation (Zhang, Yang, Li, Liu, & Liu, 2017). Taken  In H. azteca from the present study, we only identified the G119S mutation in a heterozygous state. Among the 237 individuals genotyped at ace-1, we found no individuals harboring an SS genotype despite a large number with the GS genotype (78). Our finding suggests that the SS genotype has a strong fitness cost, potentially indicating the SS genotype in H. azteca is nonviable. In mosquitoes, the GS genotype shows overdominance (heterozygote advantage) in which individuals with the G119S allele are more fit in the presence of an OP or CM, but they are less fit in the absence of those contaminants. In the absence of an insecticide, the G119S allele reduces AChE activity by 60% in C. pipiens (Bourguet et al., 1997).
Other studies with mosquitoes have documented fitness costs associated with the G119S allele, including reproductive costs (Berticat et al., 2008) and developmental and physiological costs (Bourguet, Guillemnaud, Chevillon, & Raymond, 2004). The homozygous form of the G119S allele has been connected to a higher mortality rate for pupae in A. gambiae (Djogbenou, Noel, & Agnew, 2010). If, as these multiple studies with other arthropods have shown, the SS genotypes in H. azteca suffer a high fitness cost that prevents their development into adulthood, then the past OP and/or CM exposure F I G U R E 3 Proportion of individuals listed by site with acetylcholinesterase (ace-1) GS genotype before (initial wild collection) and after (survivors) a 96-hr chlorpyrifos water-only challenge. Proportions are not separated by species group (see Table 2 for site abbreviations), although only a single species (either C or D) was identified at ESC, CLG, and MED sites. For MSH, presented proportions represent the pooled sample of species B and D has imposed a significant cost on the reproductive capacity of the wild H. azteca populations.
The identification of the G119S mutation in both species C and D animals is evidence of a single emergence in one species group followed by introgressive hybridization or at least two independent origins of the mutation in H. azteca. In mosquitoes, the G119S mutation has also been independently and repeatedly selected multiple times as a result of OP and CM selective pressure Weill et al., 2003), although some evidence of introgression as a mechanism for the spread of G119S across species groups exists (Djogbenou et al., 2008).
However, introgression is only a plausible mechanism for transferring resistance alleles when two species can mate to produce viable offspring. Forced crosses in our laboratories do not produce fertile offspring (M. Lydy, personal communication), making adaptive introgression highly unlikely. Our results further suggest that interbreeding in wild populations containing two separate clades Weston et al., 2013) does not occur, supporting independent origins of resistance in H. azteca across species groups . The existence of sensitive, wild-type species C (laboratory) and species D (Mojave River) combined with the potential fitness cost associated with the G119S allele provides support that selection for G119S has occurred by OP and CM exposure rather than based on shared common ancestry given the high (14%) COI divergence between species C and D animals .

| Evidence for an ace-1 duplication in H. azteca
In the small subset of H. azteca cloned at ace-1 in the present study, we found evidence of between three and five different alleles for some individuals, with Agricultural group individuals (Calleguas Creek and Mosher Slough) sometimes harboring multiple versions of the G119S allele (two to three). The high homology (>97% nucleotide sequence identity) among all alleles (G119S or wild-type) suggests that these alleles come from a duplication of the ace-1 gene rather than an ace-2 gene, consistent with the failure to detect an ace-2 gene in the H. azteca genome .
A recent study of the genomic architecture of the ace-1 duplication showed that duplications are associated with all resistance alleles in A. gambiae, indicating that ace-1 duplication is an important mechanism for resistance in some mosquitoes (Assogba et al., 2016). In resistant mosquitoes, the ace D allele allows for fixed heterozygosis and has been shown to resorb many of the fitness cost associated with the G119S mutation (Assogba et al., 2015). In crustaceans, a duplication of the ace-1 gene was found in the salmon louse (Lepeophtheirus salmonis), which has developed resistance to OPs through a different point mutation in one of the ace-1 copies (Kaur, Helgesen, Helgesen, Bakke, & Horsberg, 2015). This provides the first example of a recent ace-1 duplication within this taxonomic group (Kaur, Bakke, Bakke, Nilsen, & Horsberg, 2015), although more work is needed to determine whether there is a relationship between the duplication and OP selective pressure.
It is possible, however, that the detection of multiple ace-1 alleles is instead explained by polyploidy rather than an ace-1 duplication event. Within the H. azteca species complex, genome size can vary by species group and evidence of polyploidy in some North American groups of H. azteca exists (Vergilino et al., 2012). However, the genome size variation and potential of polyploidy for groups in the present study remain uninvestigated, and future work will be needed to determine the genomic architecture associated with the multiple ace-1 alleles documented herein.
In the present study, we only cloned and sequenced a limited number of individuals. Given that the ace-1 genotyping assay did not Chlorpyrifos remains one of the more widely used insecticides in California agriculture and is by far the most heavily used of the OPs.
Therefore, as expected we found the greatest phenotypic resistance in the Agricultural sites (all four sites, up to 1000-fold resistance).
The mutant genotype was found in four of six Agricultural sites and was present in 90%-100% of the individuals at three of them. The data suggest a strong, ongoing selective pressure for the G119S mutation in many agriculture-influenced areas, driven by chlorpyrifos and/or other compounds within the OP and CM classes.
In a study of mutations conferring pyrethroid pesticide resistance in H. azteca, pyrethroid residues were consistently found in all sites where use of the compounds had been anticipated, and were often present at concentrations above acutely toxic thresholds for wild-type individuals, thus providing supportive chemical evidence of a selective pressure

| Resistance in H. azteca has ecological and evolutionary implications
Although the potential fitness costs associated with OP-and CMresistant H. azteca have not been explored, the G119S allele has been associated with reduced AChE functionality (Bourguet et al., 1997) and reduced fitness in mosquitoes (Assogba et al., 2015).
However, the existence of ace D allele in mosquitoes is capable of absorbing much of the G119S cost of resistance measured as A.
gambiae larval mortality and development time, mating competition, and female fecundity and fertility (Assogba et al., 2015). This suggests that fitness costs associated with the G119S allele may also be ameliorated in fixed heterozygote H. azteca with duplicated ace-1 genes. However, other population-or community-level costs may be associated with the G119S allele, especially when considering that seven populations from the present study harbored both pyrethroid resistance alleles (M918L, L925I, and/or L925V; data shown in Major et al. (2018) and in Table S4) and the OP and CM resistance allele (G119S). For example, pyrethroid-resistant H.
azteca can act as a vector for pyrethroid bioaccumulation in fish (Muggelberg et al., 2017). Therefore, chlorpyrifos bioaccumulation of chlorpyrifos and its metabolites in fish predators may also be applicable for fish feeding on pesticide-resistant H. azteca, resulting in the potential for increased fish bioaccumulation not only of pyrethroids, but of OPs and/or CMs. Further, it is possible that selective sweeps for chlorpyrifos resistance have contributed to the reduced fitness already observed in some pyrethroid-resistant H. azteca (Heim et al., 2018). More research is warranted to determine the full extent of population and community-level effects of OP resistance.
The ecological implications of resistance are not limited to H.
azteca populations or the animals directly connected to this amphipod in the food web. Chlorpyrifos from agricultural runoff has been documented at levels that were acutely toxic to a variety of other ecologically important aquatic invertebrates including daphnids (Anderson et al., 2003), chironomids, and mayflies (Anderson et al., 2006). Given that such toxicity could act as the driver behind the development of pesticide resistance, it is possible that other species are undergoing similar OP and CM adaptive processes as those documented in H. azteca.

| CON CLUS ION
Agricultural and urban OP and CM use has been the driver behind the adaptive, genetically based chlorpyrifos resistance observed in numerous populations of the nontarget aquatic amphipod, H. azteca.
Resistant populations all share the same glycine-to-serine amino acid substitution at position 119 of ace-1 also found to be one of the primary mutations involved in mosquito OP and CM resistance.
Sensitive laboratory and wild populations with no history of pesticide exposure lacked the G119S mutation. This mutation has developed independently at least twice in two species groups within H. azteca species complex. Further, although our study design possessed limited ability to discriminate alleles at ace-1, we found that an ace-1 duplication (or polyploidy) may be common in some mem- have been shown to be resistant to pyrethroids Weston et al., 2013). Their sensitivity to the neonicotinoids has not yet been investigated, but they clearly have acquired resistance to many of the insecticides now in common use. This finding provides clear evidence that pesticide use throughout the state and the subsequent movement of those residues into aquatic systems have had a profound effect on evolution in H. azteca.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data used in these analyses and primer sequences used for sequencing are provided in the Supplemental Information.