Microsatellite‐based analysis of genetic structure and gene flow of Mythimna separata (Walker) (Lepidoptera: Noctuidae) in China

Abstract The oriental armyworm, Mythimna separata, is a serious agricultural pest in China. Seasonal and roundtrip migration has recently led to sudden, localized outbreaks and crop losses. To evaluate genetic differentiation between populations in eastern and western China and elucidate gene flow, the genetic structure of 20 natural populations from nine provinces was examined using seven microsatellite markers. The results indicated high genetic diversity. However, little to moderate (0 < F ST < 0.15) genetic differentiation was detected, and there was no correlation between genetic distance and geographical distance. Bayesian clustering analysis identified three groups whereas discriminant analysis of principal components identified ten clusters that were considered as two clear‐cut clusters and one admixed group. Gene flow occurred frequently in most population pairs, and an asymmetrical migration rate was detected in several pairwise population comparisons. The bottleneck test showed that few populations had experienced recent bottlenecks. Correspondingly, large‐scale and long‐distance migration of M. separata has caused low genetic differentiation and frequent gene exchange. Our findings are important for studying genetic evolution and help to improve predictions of M. separata outbreaks in China.

The armyworm moth undertakes long-range and multigeneration roundtrip migration between southern and northern China from March to mid-September each year (Jiang, Luo, Zhang, Sappington, & Hu, 2011;Li, Wang, & Hu, 1964;Zhang, Zhang, Li, Jiang, & Zeng, 2013). This makes it difficulties to predict population dynamics and formulate proper management decisions. Its overwintering boundary indicates that M. separata is unable to overwinter in northern China and no diapause period occurs during development (Li, 1993).
Therefore, large-scale outbreaks of M. separata in northwestern China are probably attributed to immigration from the eastern regions (Cheng & Zhao, 2016;Zhou et al., 2017). However, seasonal dynamics of armyworm movement between eastern and western China remain unclear.
Based on isozyme analyses, population genetic studies of M. separata concluded that oriental armyworm populations formed a large panmictic population from five localities in China (Hao, Li, & Lin, 1992). Researchers further confirmed high genetic variation among individuals of studied populations, and little genetic differentiation was found among geographically distinct populations in China. This is likely due to increased gene flow as a result of their long-distance migration (Jiang, Luo, & Zhang, 2007;Li, Li, Wu, & Xu, 2018).
Furthermore, intersimple sequence repeat (ISSR) analysis implied that highly frequent gene flow among 23 geographically separated armyworm populations from various latitudes prevented genetic differentiation normally caused by genetic drift (Chen et al., 2017).
However, their conclusions were unable to reveal population genetic structure and evolutionary history due to limited sampling and the use of molecular markers of low reproducibility or diversity (Chen et al., 2017;Hao et al., 1992;Jiang et al., 2007).
Many studies have been conducted on the migratory flight, northern boundary of overwintering, and evolutionary history of armyworms (Chen et al., 1989(Chen et al., , 2017Hao et al., 1992;Jiang et al., 2007;Li, 1993;Li et al., 1964;Yin, Feng, Cheng, & Cao, 2003;Zhang et al., 2013). However, a comprehensive study of population genetic variation and population structure in China has not been accomplished. The present study analyzed the population structure of 20 F I G U R E 1 Sampling locations, repartition of microsatellite lineages based on Bayesian clustering analysis using STRUCTURE, and map of the migratory pathways of Mythimna separata in China. The inset in bottom left corner shows the life cycle of M. separata. Black triangles represent collection sites. Population codes refer to Table 1. HLJ, FJ, GX, and GD represent Heilongjiang, Fujian, Guangxi, and Guangdong provinces, respectively. Dimgray, Darkgray, and Lightgray of pie charts represented Cluster 1, 2, and 3, respectively. The emigration routes (solid arrows) depicted based on the mark-release-recapture study of Li et al. (1964), while immigration paths (dotted arrows) delineated in line with gene flow, recent migration rates, and also referring to Feng et al. (2008) and Jiang et al. (2011) geographically distinct populations of the oriental armyworm collected from nine provinces in China, using microsatellite markers.
Knowledge about genetic structure among subpopulations, population differentiation, and gene flow of M. separata in China is greatly important for determining migration patterns, and thus, the source of outbreaks; it also provides a theoretical framework for prediction and determines control strategies to prevent outbreaks.

| Sampling collection and DNA extraction
The oriental armyworm larvae samples were collected from maize (Z. mays) fields from nine provinces at 20 geographical sites in China ( Figure 1; Table 1). A total of 542 armyworms were sampled during the maize-growing season from May 2016 to August 2017. Each armyworm was collected at a distance of at least 50 m from any other sampled individuals, to avoid sampling siblings. Armyworm larvae were preserved in absolute ethanol and stored at −20°C. Total genomic DNA for genetic analysis was extracted from whole larvae using Biospin insect genomic DNA extraction Kit (Bioer Technology Co., Ltd).
Amplifications were initiated with a predegeneration 94°C for 2 min, followed by 30 amplification cycles consisting of 94°C for 10 s, 15 s at the primer-specific annealing temperature, 72°C for 15 s. This was followed by 10 cycles including 94°C for 15 s, 53°C for 20 s, and 72°C for 20 s, and ending with 72°C for 5 min for a final extension. The primerspecific annealing temperature of each primer was described in previous report (Li et al., 2018). The PCR products of the target band were genotyped by capillary electrophoresis with two-color fluorescent and were confirmed manually.

| Genetic structure and population differentiation
The population genetic variance was further analyzed by a global analysis of molecular variance (AMOVA) performed using Arlequin 3.5, and average F-statistics, pairwise differentiation (F ST ), and its significance levels (p-values) were estimated with 1,000 permutations (Excoffier & Lischer, 2010;Weir & Cockerham, 1984). We tested for isolation by distance by regressing pairwise genetic distance (F ST /(1 − F ST )) against the natural logarithm of geographical distance (km) across populations for microsatellite loci using the Mantel test (Mantel, 1967) with 1,000 permutations of ZT software package (Bonnet & Van der Peer, 2002). Population structure was analyzed based on Bayesian model-based clustering using STRUCTURE 2.3.4 with admixture ancestry and correlated allele frequency model (Pritchard, Stephens, & Falush, 2007). Twenty independent runs for each testing genetic clusters (K) values which ranged from 1 to 20 were performed with a burn-in period of 50,000 iterations and 1,000,000 Markov Chain Monte Carlo (MCMC) replicates. The most likely number of clusters (K) was determined by considering log-likelihood values of each K and the Delta K method described by Evanno, Regnaut, and Goudet (2005), implemented in Structure Harvester (Earl & vonholdt, 2012). CLUMPP 1.1 was used for model averaging of individual ancestry coefficients across the 20 independent runs (Jakobsson & Rosenberg, 2007). Then clusters were visualized using Distruct 1.1 (Rosenberg, 2004). Further, discriminant analysis of principal components (DAPC) was used to identify subpopulations of the species using the Adegenet package for the R 3.5.3 (Jombart, 2008 and a.optim.score functions were used to identify the optimal number of principal components to be retained. Twenty independent runs were executed and the average results were plotted.

| Analysis of gene flow and migration rate
Wright's method was employed to estimate gene flow (N e m) from genetic data based on the relationship F ST = 1/(4N e m + 1) (where N e is the effective size of each population and m is the migration rate between populations; Slatkin, 1987;Wright, 1931

| Bottleneck analysis
The infinite allele model (IAM), the strict stepwise mutation model (SMM) and two-phase model (TPM) were applied with 10,000 simulation iterations to identify whether populations have experienced bottleneck effects in history under the assumption of mutation-drift equilibrium using BOTTLENECK 1.2.02 (Piry, Luikart, & Cornuet, 2001). A qualitative descriptor of allele frequency distribution ("mode-shift" indicator), which discriminates between bottlenecked and stable populations, was also utilized (Luikart, Allendorf, Sherwin, Cornuet, & Sherwin, 1998

| Population genetic structure
According to the Bayesian analysis performed with STRUCTURE The Bayesian information criteria (BIC) curve in the DAPC analysis supported 10 clusters ( Figure S2). Clusters 2, 4 and 10 were clearly differentiated, which indicated that LNF, SDW, and HNL belonged to disparate subgroups. Individuals in the other seven clusters had no distinct division (Figure 3). NME and SXB had similar genetic structure, and SDJ and HBB also had similar genetic structure but were differed from the structure of NME and SXB (Table S3). Inspection of the DAPC plot also revealed the presence of a genetic structure within these populations and was more likely to divide them into three groups (Figure 3). Per Bayesian clustering, no clear associations between the groups were identified using DAPC.
Combining results from Bayesian analysis in STRUCTURE and DAPC, individuals from NME and SXB to be genetically differentiated from LNF, as well as HBC (Figures 2 and 3). Although the four populations from 2016 were in various subgroups, they were weakly differentiated. Individuals from the JLC and YNZ populations belonged to different clusters and were partial overlapped in DAPC analysis despite they divided into multiple genetic clusters, while they had a similar genetic partitioning in Bayesian analysis. Genetic clustering of individuals in each population presented mixed ancestry (Figure 2 and 3), which led to multiple genetic clusters found in armyworm populations. Consequently, these armyworm populations were considered to originate from three differentiated subpopulations.

| Population genetic differentiation
The global analysis of molecular variance (AMOVA) test results attributed 4.54%, 19.72%, and 75.74% of the variation to differences among populations, among individuals within populations, and within individuals, respectively, based on the pairwise differences distance method (  (Wright, 1978). The pairwise F ST values (Table   S4) were low to moderate, with a maximum between SXB and HBB  (Tables S4 and S5), demonstrating that the population genetic structure of M. separata in China did not conform to the isolation by distance model (Figure 4). That is, no correlation between pairwise genetic distance and geographical distance of armyworm populations was evidenced. Furthermore, geographical distance did not reflect a barrier to gene flow.

| Gene flow and migration rate
Observations of the effective number of migrants per generation (N e m) of several population pairs were greater than 4 (Table S5), indicating that active dispersal ability among those populations was not limited. It is generally assumed that N e m equal to 1 is sufficient to bring the two populations genetically closer to each other, and if N e m > 4, then local populations belong to one panmictic population (Wright, 1938). The value of N e m between SXH and SYC populations   Figure 1).

| Population demography
An attempt to test heterozygote excess was inconsistent under the IAM, TPM, and SMM models ( Table 5) (Table 5).

| Genetic diversity
In this study, 193 alleles were found and M. separata had a higher genetic diversity based on seven polymorphic microsatellites with low null allele frequency (Tables 2 and 3). The Fis values for all populations were greater than zero, indicating a heterozygote deficiency, which caused a deviation from HWE (Table 3). Overall, Ho values for all populations were lower than He, which might be attributed to null alleles, Wahlund effect, and the population not being in HWE (Joanna et al., 2011). There was no complete linkage of any locus, indicating that the selected loci were evenly distributed in the genome of armyworm and were relatively independent in the process of generational transmission (Torriani, Mazzi, Hein, & Dorn, 2010).
The genetic indices showed that the level of genetic diversity in most populations was similar. Higher genetic diversity was observed in the second and third generations, except the HBB population.
Furthermore, the YNZ population from the overwintering region exhibited a higher genetic diversity than others from nonoverwintering areas. Abundant genetic diversity was shown to be the genetic F I G U R E 4 Scatter plots of pairwise genetic distance (F ST / (1 − F ST )) against the natural logarithm of geographical distance (km) for pairwise population comparisons basis for its strong adaptive potential (Chen et al., 2017;Wang, Yang, Lu, Zhou, & Wu, 2017), which aids in understanding why M. separata localized, large-scale outbreaks occurred in China in recent decades (Cheng & Zhao, 2016;Jiang et al., 2011;Jiang, Zhang, et al., 2014a;Zeng et al., 2013).

| Genetic structure
Bayesian clustering and DAPC analyses (Figures 2 and 3) suggested that the populations were mostly clustered into three groups. All populations were admixtures of these clusters, which could be only explained by the highly migratory ability of the armyworm gave rise to extensive gene flow. Although genetic structure appeared irrelevant to both geographical divisions and occurrence generations of M. separata populations throughout China, four populations from 2016 had a similar genetic partitioning which grouped in cluster 3 ( Figure 2). As in Coulon et al. (2008), individuals with maximum inferred ancestry <0.6 were not assigned to any group and were considered to exhibit admixed assignation. That was, nine of the populations could not be assigned to any cluster (Table S1). Furthermore, DAPC confirmed that there was little genetic distinction among clusters 1, 3, and 5-9, which mainly contained 15 populations excluding: HNL, LNF, SDW, SXF, and SXW ( Figure 3). Based on these results, we inferred that individuals from 20 geographic populations in the same cluster had the same immigration source; they likely shared a common gene pool in the past. AMOVA analysis found low to moderate genetic differentiation (global F ST = 0.045, p < .001) based on F ST among populations. Our results indicated that migratory dispersal led to differences in population structure but they still belonged to a panmictic population owing to their high capacity for long-distance migration.

| Genetic differentiation and migration
There was no significant correlation between geographical distance and genetic distance among any of the 20 populations ( Figure 4). Our estimates (Table S5) (Table S5) had a homogenizing effect on the genetic variation over geographic populations, counteracting random drift, selection, and mutation (Sun, Li, Yang, & Hong, 2012).
Notably, the analyses of 20 geographically separated M. separata populations showed little to moderate genetic differentiation and a high gene flow (Tables S4 and S5). Insect species might be divided into genetically diverse populations that cover large and heterogeneous geographical regions with limited gene flow (Chen, Tan, Liu, Wang, & Li, 2000). In contrast, geographically distinct armyworm populations did not exhibit genetic differentiation, and isolation by distance was absent due to extensive gene flow, which further validated our findings. Previous research elucidated that frequent of Helicoverpa zea and that these populations did not show any signs of differentiation based on location (Perera & Blanco, 2011).
The lack of genetic differentiation more likely reflected genetic mixing between eastern and western North American monarchs, Danaus plexippus (Lyons et al., 2012). Therefore, migratory insects form a large panmictic population experience enhanced gene flow because of genetic exchange, migratory flight, and long-range dispersal. Additionally, the present study articulated that there was no population differentiation caused by genetic drift and no clear evidence for isolation by distance, which is consistent with other studies of this insect (Chen et al., 2017;Li et al., 2018).
Furthermore, the Wilcoxon signed-rank test of heterozygote excess (  Figure 1). This migratory pathway agrees with the seasonal north-south migration pattern, based on mark-release-recapture data and trajectory simulation analysis of northeast populations from Shanxi, southern HB, northern HN, and most of the SD regions (Jiang et al., 2011;Li et al., 1964). The next generation migrated through the central and western regions of HB and IM, and some flew directly from the SD Peninsula across the Bohai Sea to the north of LN and center of JL by radar observation (Chen et al., 1989;Zhang et al., 2013).
It was further observed that autumn migration of M. separata that originated in northeastern China (i.e., LN, JL, HLJ, and part of IM regions) could immigrate into east-central China (SD) and subsequently to southern China (i.e., FJ, GD, and GX; Table 1; Figure 1) within a week for overwintering (Feng et al., 2008). Considering the lack of data on the overwintering generation, more research is needed to elucidate the migratory patterns in the western areas, in order to better predict and prevent outbreaks.

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