Variable resistance to spinetoram in populations of Thrips palmi across a small area unconnected to genetic similarity

Abstract The melon thrips, Thrips palmi, is an increasingly important pest of vegetables in northern China. Some populations have developed resistance in the field to the insecticide spinetoram. Understanding the origin and dispersal of insecticide‐resistant populations can shed light on resistance management strategies. In this study, we tested susceptibility of seven greenhouse populations of T. palmi to spinetoram collected from a small area of about 300 km2 in Shandong Province and examined population genetic structure across the area based on a segment of mitochondrial cox1 gene and 22 microsatellite loci to infer the possible origin and dispersal of insecticide resistance. Levels of resistance to spinetoram differed among seven populations, which included one population with high resistance (LC50 = 759.34 mg/L), three populations with medium resistance (LC50 ranged from 28.69 to 34.79 mg/L), and three populations with low resistance (LC50 ranged from 7.61 to 8.97 mg/L). The populations were genetically differentiated into two groups unrelated to both levels of resistance and geographic distance. The molecular data indicated high levels of gene flow between populations with different levels of resistance to spinetoram and low gene flow among populations with the same level of resistance, pointing to a likely separate history of resistance evolution. Resistance levels of two tested populations to spinetoram decreased 23 and 4.6 times after five generations without any exposure to the pesticide. We therefore suspect that resistance of T. palmi most likely evolved in response to local applications of the insecticide. Our study suggests that the development of resistance could be avoided or resistance even reversed by reducing usage of spinetoram.


| INTRODUC TI ON
Applications of insecticide have led to the evolution of resistance in many insect pests (Nauen, Slater, Sparks, Elbert, & Mccaffery, 2019).
Insecticide resistance management (IRM) has become one of the components of pest control practices aimed at extending the useful life of a chemical against a pest (Brattsten, Holyoke, Leeper, & Raffa, 1986;Roush & Tab ashnik, 2012). Insecticide resistance can evolve and spread out from a single original population or independently evolve in multiple populations (Andreev, Kreitman, Phillips, & Beeman, 1999;Daborn et al., 2002;Shi et al., 2019).
Understanding the origin and dispersal of insecticide resistance especially in its early stage can provide information for identifying resistance mechanisms and managing further resistance evolution (Daborn & Le Goff, 2004;Hawkins, Bass, Dixon, & Neve, 2018).
Tracing the origin and spread of resistance can be challenging.
When resistance alleles spread out from a single origin via human activities, populations with pesticide resistance are often geographically and/or genetically connected through gene flow (Daborn et al., 2002;Raymond, Callaghan, Fort, & Pasteur, 1991). In cases where resistance has independent origins, developing in geographically distant populations, some populations may remain susceptible to an insecticide while others have varying levels of resistance, and these resistance patterns may be unconnected to geographic proximity (Shi et al., 2019). This pattern can be produced by population differences in local selection intensity. Population genetic approaches provide a useful approach to testing these scenarios, because they can trace the dispersal of individuals and possible spread and selection for insecticide resistance and also test how this spread coincides with resistance (Crossley, Chen, Groves, & Schoville, 2017;Fu, Epstein, et al., 2017;Pélissié, Crossley, Cohen, & Schoville, 2018;Shi et al., 2019;Yang et al., 2019).
Here, we compare molecular differentiation to resistance patterns in the melon thrips, Thrips palmi Karny (Insecta: Thysanoptera: Thripidae). This species is an economically important agricultural pest on vegetables. It causes severe injury to infested crops by ovulating, feeding directly, and transmitting plant virus from Orthotospoviruses (Rotenberg, Jacobson, Schneweis, & Whitfield, 2015;Stuart, Gao, & Lei, 2011). Originating from tropical countries of Asia, T. palmi was introduced and became established across South-East Asia, South America, the Caribbean, Florida, Australia, and West Africa (Cannon, Matthews, & Collins, 2007). In recent years, this species spread to northern China and became a serious pest in greenhouse vegetables . Unlike the invasive western flower thrips, Frankliniella occidentalis, which rapidly spread into most areas of China after initial reports, likely accelerated by human activities (Cao et al., 2017), T. palmi expanded its range of distribution in a pattern that fits a stepping stone model, forming genetic structure among geographically distinct populations .
Management of T. palmi has been heavily reliant on chemical control (Bao et al., 2014). However, the species has a high capability of developing resistance to numerous pesticides (Bao et al., 2014;Bao & Sonoda, 2012). Spinetoram is currently one of the remaining insecticides available to control thrips around the world (Cannon et al., 2007;Mouden, Sarmiento, Klinkhamer, & Leiss, 2017;Reitz et al., 2019). Resistance to spinetoram has developed in populations of T. palmi in Japan (Bao et al., 2014), but this insecticide continues to be widely used in controlling populations in China. However, a recent study showed that some populations of T. palmi have developed varying levels of resistance to spinetoram in northern China, although most populations remain susceptible . For example, the resistance of T. palmi to spinetoram is variable in the Beijing area of China, with LC 50 values among populations ranging from 1.69 to 19.69 mg/L .
It is still not clear whether the resistance of T. palmi to spinetoram has evolved in response to local selection pressures or whether it is dictated by gene flow. Understanding the development and spread of resistance in T. palmi to spinetoram will help to develop strategies for resistance management.
In this study, we investigated the development and spread of resistance of T. palmi to spinetoram in a small region of Shandong Province by comparing resistance levels with genetic structure among populations and changes in resistance after rearing thrips without exposure to pesticides. Our study was based on the following conjectures: When populations with similar resistance share a similar genetic background, we considered them as forming one cluster and having a single origin of resistance; on the other hand, when there are multiple populations with resistance and these are not connected genetically, we suspect that resistance has multiple origins. Based on our knowledge of the biology and resistance status of T. palmi, we hypothesized that resistance was more likely to develop independently in multiple populations rather than being solely a consequence of gene flow.

| Samples
To compare spinetoram resistance and genetic structure in populations of T. palmi, we sampled seven populations from greenhouses in six villages, involving two collections from eggplant, three from pepper, and two from cucumber (Table 1). These populations were collected from Shouguang in Shandong Province, where is a large area for vegetable production and where the control of T. palmi has been heavily reliant on spinetoram in the past few years. Due to the reduced control efficacy of spinetoram, the frequency of application of this pesticide has been reduced to 1-3 times per growing season in recent years. All samples were collected on November 20, 2018, except for SGL1, which was collected on July 11, 2018.
Samples were collected from a core planting area covering about 300 km 2 ( Figure 1, Table S1). A set of around 3,000 thrips collected from one host plant crop in a greenhouse was considered to represent a population. These thrips were collected from at least 24 sites scattered across the greenhouse crop. The populations were kept in net bags and taken to the laboratory for bioassays and immediate genotyping. For each population, both randomly selected male and female adults were used for bioassays to estimate their susceptibility to spinetoram. Due to haplodiploidy in thrips (haploids develop into males and diploids into females) (Moritz, 1997), we used 24 diploid adult females per population for genetic analyses so that heterozygosity could be computed. One individual was randomly selected from each of the 24 collection sites in a crop for genotyping to decrease the likelihood of close relatives being included in the population sample.
Additionally, we used another population on cucumber in July 2018 from the Beijing area (BJFS)  , about 400 km away from Shouguang, where the level of resistance was expected to be low compared to samples collected from Shouguang due to the low usage of spinetoram in the Beijing area.
This population and one population collected from Shouguang (SGL1) as described above were reared in the laboratory for five generations without pesticide exposure on cucumber to examine changes in spinetoram susceptibility across generations (Table 1).

| Bioassay
Spinetoram 6% SC (Dow AgroSciences Company) was used in a leaf-dipping bioassay for testing susceptibility of T. palmi (Wang et al., 2016). The original concentrations of spinetoram were determined based on pretests and then were serially diluted into eight concentrations using distilled water containing 0.1% Triton X-100 (Beijing Solar BioScience and Technology Limited Company). Three duplicates were set for each concentration of spinetoram. All leaves used for bioassays were grown in a greenhouse without exposure to insecticides, and leaves came from the same plant type as where the TA B L E 1 Host plant, pesticide usage status, and resistance level to spinetoram for populations of Thrips palmi species was collected. The leaves were cut to fit plastic containers and dipped in spinetoram solutions for ten seconds before air drying at room temperature. We used 0.2% agar in the bottom of the containers to avoid the leaves drying out. In total, 20-25 adults T.
Mortality was recorded after 48 hr. Individuals unable to move were considered as dead. Control leaves were treated with 0.1% Triton X-100 solution. All control moralities were below 10% in the bioassays. The lethal concentrations of 50% (LC 50 ) and 95% (LC 95 ) and other parameters were estimated through DPS software (Tang & Zhang, 2013).
Conditions for PCR amplification were described in Gao, Gong, Ma, et al. (2019). The size of amplified PCR products was deter- Polymerase chain reaction (PCR) was conducted with the following program: an initial denaturation for 3 min at 94°C, followed by 35 cycles of 30 s at 94°C, 15 s at 52°C and 1 min at 68°C, and a subsequent final extension for 10 min at 68°C. Amplified products were purified and sequenced directly from both strands using an ABI 3730xl DNA Analyzer by Tsingke Biotechnology Co. Ltd.

| Population genetic analyses
To examine genetic structure across the populations, phylogenetic relationships among the populations were inferred with POPTREE2 (Takezaki, Nei, & Tamura, 2009)

| Genetic diversity of T. palmi in seven greenhouse populations
For the mitochondrial cox1 gene, four haplotypes were found, and nucleotides diversity was low, ranging from 0 to 0.0004. Hap1 was the most common haplotypes followed by Hap2, and these two haplotypes were found in all populations except for SGZJ that only had Hap1. Additionally, there was one Hap3 individual and one Hap4 individual in the populations SGNC and SGL2, respectively ( Figure 1, Table 2). Proportions of the four cox1 haplotypes did not differ significantly across the seven populations ( Figure 1, Table 2).
For the microsatellite loci, we found 696 alleles among the 168 females of T. palmi characterized for the 22 microsatellite loci. The average allelic richness (A R ) varied from 4.14 to 4.86. The observed heterozygosity (Ho) tended to be lower-than-expected heterozygosity (He) and the inbreeding coefficients (F IS ) was low (−0.01-0.10) ( Table 2).
Nine out of 154 population-locus pairs showed deviation from HWE (p < .05); however, none of the loci showed deviation in all populations, and no population showed HWE deviation at all loci (Table S2).

| Population genetic structure of seven greenhouse populations
For microsatellite loci, pairwise F ST values among seven populations ranged from 0.0005 to 0.1258 (Table 3)

| Correlation between genetic differentiation and geographic distance or resistance difference
Mantel tests indicated that there was no correlation between genetic distance and geographic distance (r = .089, p = .308, Figure 3a) or susceptibility to spinetoram (r = .456, p = .077, Figure 3b). The highest genetic differences were found between SGL1 with low susceptibility and the other populations, followed

| Gene flow among seven greenhouse populations
A relatively high level of contemporary gene flow was found between four populations with varied levels of resistance to population SGDY (m ranged from 0.2575 to 0.2671) and between SGZJ and SGL1 which had a large difference in resistance to spinetoram (m = 0.1291 from SGZJ to SGL1, m = 0.0523 in the reverse direction) (Figure 4). Low levels of gene flow were found among the three populations with a medium level of resistance to spinetoram.

| Variation of susceptibility and genetic structure in the absence of spinetoram
After rearing the two populations in the laboratory for five generations without exposure to the pesticide, the resistance level of T. Genetic diversity estimated from microsatellites did not change significantly after five generations. The F IS increased from −0.0079 to 0.0627 in SGL1 and decreased from 0.0247 to 0.0052 in BJFS (Table 2).

DNA
For the mitochondrial cox1 gene, the type of haplotype (Hap 1 and Hap 2) did not change in both populations after five generations, while the nucleotide diversity increased in SGL1 and decreased in BJFS (Table 2). Pairwise

| Development of spinetoram resistance in populations of T. palmi
We palmi to spinetoram will be an increasing problem in China  with an ongoing selection expected in populations where the LC 50 remains below field rates.
We also found that resistance levels of T. palmi declined after five generations without insecticide exposure. Due to the need for large numbers of individuals when undertaking bioassays, we were unable to examine the susceptibility of each generation reared successively.
This made it impossible to clearly separate environment-induced resistance from genetically controlled resistance. However, with the source and laboratory populations coming from the same hosts, we suspect that genetic changes are involved in the reduction of resistance, reflecting fitness costs associated with resistance alleles. We also found significant genetic differentiation between the F5 generation and the source generation, which was unrelated to resistance.
These may occur as a consequence of genetic drift in the lines or rapid genetic changes as a consequence of adaptation to laboratory conditions (Hoffmann & Ross, 2018), assuming that some microsatellite loci are linked to loci under laboratory selection. A decrease in spinetoram resistance in the absence of ongoing selection for resistance was previously described in Thrips hawaiiensis (Fu et al., 2018).
Fitness costs of spinetoram resistance have also previously been suggested for F. occidentalis and T. hawaiiensis (Fu, Li, et al., 2017;Li et al., 2017). Current variation in resistant levels to spinetoram in greenhouse populations of T. palmi might partly reflect these costs balanced against ongoing selection for resistance.

| Contribution of gene flow to resistance in greenhouse populations of T. palmi
In a genetic analysis of T. palmi across a larger area than in the current study but including the area from which the current populations were sourced, populations from northern China formed a genetic cluster . This is consistent with our analysis of the mitochondrial cox1 gene, which showed that all individuals were dominated by one haplotype. Nevertheless, we have found substantial substructuring within this northern cluster based on resistance alleles. As recombination occurs across generations, any association between microsatellite alleles and resistance is expected to break down rapidly, although linkage disequilibrium may be maintained for many generations in polymorphisms near selected resistance alleles (Daborn et al., 2002). Tight linkage is unlikely for the few microsatellite loci scored here, whereas markers in linkage disequilibrium are much more likely to be discovered when thousands of SNP markers are scored across the genome (Endersby-Harshman et al., 2019).
In the present case, the lack of a correlation between resistance and genetic distance may reflect a difference in the intensity of selection for resistance in a population on resistance mutations found throughout the geographic range of the species and/or a separate geographic origin for resistance alleles. The latter seems less likely since our study showed that all tested populations exposed to spinetoram had developed at least some resistance to spinetoram compared to previously reported levels of susceptibility (Bao et al., 2014;. Therefore, there may have been a single origin of resistance alleles, with local selection pressures driving them to different frequencies in populations within the area we examined. Further studies on spinetoram resistance mechanisms in these T. palmi populations and patterns of polymorphism around the selected alleles involved will help to resolve the role of local selection versus independent origin.

| Potential factors influencing spinetoram resistance in T. palmi
The development of insecticide resistance can be influenced by many factors and resistance itself can have a complex or simple genetic basis (Crossley et al., 2017). The resistance of spinetoram in insect pests may be associated with target-site mutation, enhanced detoxification and changes in gene transcription (Bao et al., 2014;Baxter et al., 2010;Berger et al., 2016;Somers, Nguyen, Lumb, Batterham, & Perry, 2015;Wan et al., 2018;Wang et al., 2016). For T.
palmi, resistance to spinetoram can be conferred by the G275E mutation in the target nicotinic acetylcholine receptor α6 subunit and cytochrome P450-mediated detoxification as identified in Japanese populations (Bao et al., 2014). However, the resistance mechanism  Insecticide resistance can be influenced by environmental factors, such as temperature and host plant (Dermauw, Pym, Bass, Van Leeuwen, & Feyereisen, 2018) and transgenerational effects of insecticides (Brevik, Lindstrom, McKay, & Chen, 2018). T. palmi is one of the main pests on greenhouse vegetables in the Shouguang area.
The greenhouses provide suitable conditions for the continuous presence of thrips throughout the year. To control this pest, farmers need to spray pesticides intensively, which poses a high selective pressure and leads to insecticide resistance of this species.
In our study, we used plant species from which the thrips were collected for bioassays to avoid any change of susceptibility due to the host shift. When we consider the resistance of T. palmi to spinetoram in the tested population from the perspective of the host plant, we note that the resistance of populations collected on pepper and eggplant is higher than that from cucumber (  (Bielza, Quinto, Fernandez, Gravalos, & Contreras, 2007). The evolutionary dynamics of genes in haplodiploids share many features with X-linked genes and are different from diploid (autosomal) genes in many respects (Hedrick & Parker, 1997). A simulation study showed that resistance develops at a faster rate under haplodiploid reproduction than under diploid reproduction when a resistance allele is recessive, and at a similar rate when a resistance allele is dominant or semi-dominant (Denholm, Cahill, Dennehy, & Horowitz, 1998). Many studies have found that spinosyn resistance is recessive in insects (Bielza et al., 2007;Wang et al., 2019). Recessive inheritance and haplodiploidy may contribute to the development of resistance in T. palmi to spinetoram (Bielza et al., 2007). However, males are generally less tolerant to pesticides than females in both haplodiploid and diploid arthropods. Simulations considering between-sex differences in insecticide tolerance show that resistance evolution can then be slower in haplodiploids than in diploids (Carrière, 2003). In our study, we randomly selected adults from each population irrespective of sex to represent average resistance levels of natural populations, and further studies on sex-specific differences in resistance are needed to understand the influence of haplodiploidy on development of insecticide resistance in T. palmi.

| Implications for pest management
Field control of thrips is heavily reliant on spinetoram globally (Cannon et al., 2007;Mouden et al., 2017;Reitz et al., 2019 (Bielza et al., 2008;Brødsgaard, 1994), we found that resistance to spinetoram rapidly declined in T. palmi in the absence of the pesticide. This suggests that it may be possible to avoid further development of resistance or even restore susceptibility by reducing pesticide applications. We also found high levels

| CON CLUS ION
We found varying levels of spinetoram resistance among populations collected from a small area. Spinetoram resistance was unrelated to genetic distance, indicating that resistance of T. palmi most likely evolved in response to local applications of the insecticide, as further highlighted by independent changes in susceptibility to spinetoram and genetic differentiation after thrips were reared in the laboratory without insecticide. The results provide information on the regional management of insecticide resistance and the possibility of recovering susceptibility in this pest before resistance alleles become fixed across China . Our study indicates how the incorporation of population genetic approaches into insecticide resistance research can help to elucidate patterns of resistance development in the field and inform insecticide resistance management.

DATA A R C H I V I N G S TAT E M E N T
Microsatellite data and mitochondrial cox1 gene sequences used in the study were deposited in Dryad repository: https://doi. org/10.5061/dryad.bnzs7 h476.