Genome‐wide SNPs of vegetable leafminer, Liriomyza sativae: Insights into the recent Australian invasion

Abstract Liriomyza sativae, the vegetable leafminer, is an important agricultural pest originally from the Americas, which has now colonized all continents except Antarctica. In 2015, L. sativae arrived on the Australian mainland and established on the Cape York Peninsula in the northeast of the country near the Torres Strait, which provides a possible pathway for pests to enter Australia and evade biosecurity efforts. Here, we assessed genetic variation in L. sativae based on genome‐wide single nucleotide polymorphisms (SNPs) generated by double digest restriction‐site‐associated DNA sequencing (ddRAD‐seq), aiming to uncover the potential origin(s) of this pest in Australia and contribute to reconstructing its global invasion history. Our fineRADstructure results and principal component analysis suggest Australian mainland populations were genetically close to populations from the Torres Strait, whereas populations from Asia, Africa, and Papua New Guinea (PNG) were more distantly related. Hawaiian populations were genetically distinct from all other populations of L. sativae included in our study. Admixture analyses further revealed that L. sativae from the Torres Strait may have genetic variation originating from multiple sources including Indonesia and PNG, and which has now spread to the Australian mainland. The L. sativae lineages from Asia and Africa appear closely related. Isolation‐by‐distance (IBD) was found at a broad global scale, but not within small regions, suggesting that human‐mediated factors likely contribute to the local spread of this pest. Overall, our findings suggest that an exotic Liriomyza pest invaded Australia through the Indo‐Papuan conduit, highlighting the importance of biosecurity programs aimed at restricting the movement of pests and diseases through this corridor.


| INTRODUC TI ON
Over the past 40 years, several polyphagous Liriomyza (Diptera: Agromyzidae) leaf-mining species have become recognized as pests capable of causing frequent and severe outbreaks in agricultural commodities (Murphy & LaSalle, 1999). Liriomyza sativae Blanchard is one of these leaf-mining pests, with damage resulting from tunneling activities of larvae that in turn decreases the vigor and the photosynthetic capacity of host plants (Johnson et al., 1983;Parrella, 1987). Leaf punctures caused by female flies of this species also facilitate the transfer of plant disease, with studies demonstrating that female L. sativae can transmit viruses of the potyvirus group (Zitter & Tsai, 1977). Liriomyza sativae is highly polyphagous across plant families, including Asteraceae, Cucurbitaceae, Fabaceae, Solanaceae, and Umbelliferae (Spencer, 1973), with its host range potentially expanding as it colonizes new areas, as in the case of an expansion to green onions in Hawaii (Carolina & Johnson, 1992).
Although L. sativae originated in the Americas, it has now expanded its geographic range and colonized many areas throughout the globe (Scheffer & Lewis, 2005). The movement of infested plant material through international trade and transportation has likely facilitated this process (Minkenberg, 1988). Eggs and larvae of L. sativae are embedded internally within plant leaves and can be easily moved from production areas to market without being noticed.
The pupae of L. sativae may also be transported through infested soil or plant debris, while strong winds facilitate long-distance adult dispersal (Fenoglio et al., 2019). Liriomyza sativae are typical secondary pests with outbreaks occurring after the indiscriminate use of broad-spectrum insecticides, which removes natural enemies as well as promoting the evolution of resistance (Mason et al., 1987;Oatman & Kennedy, 1976;Parrella & Keil, 1984;Reitz et al., 2013;Ridland et al., 2020). Once L. sativae invade new regions, it is difficult to eradicate (Scheffer & Lewis, 2005), highlighting the importance of industry preparedness and biosecurity measures.
Following invasion and establishment into new regions, L. sativae often has an immediate and detrimental impact on local horticultural industries (Murphy & LaSalle, 1999). For example, L. sativae led to substantial damage in vegetable crops in China after invading in 1993 and expanding into most agricultural areas, with over 2.7 million hectares affected by the end of 2005 at a cost of around 3 billion yuan annually in lost production (Chunlin et al., 2005). Similarly, L. sativae was the major pest attacking the foliage of commercial watermelon in Hawaii with an infestation rate of up to 70% of plants (Johnson, 1987) and caused substantial damage to cucumber crops in greenhouses in Iran (Alaei Verki et al., 2020).
Liriomyza sativae was first detected in Australia in 2008 when established populations were identified in the Torres Strait (Blacket et al., 2015). Subsequently, L. sativae was recorded at Seisia on the northern tip of the Cape York Peninsula, located in the northeastern part of the Australian mainland in 2015 (IPPC, 2017). It is possible this pest arrived through the "Indo-Papuan conduit," which covers lands and waterways, connecting southeastern Indonesia and New Guinea with Australia's Torres Strait and Cape York Peninsula (Horwood et al., 2018). This region is known to provide a pathway for the movement of exotic pests and diseases into Australia (Horwood et al., 2018;Thompson et al., 2003). Within Australia, modeling suggests that further spread of L. sativae is likely on the mainland unless this can be prevented by quarantine measures (Maino et al., 2019).
Pest activity is expected to overlap with the production cycle in Australia of several high-risk horticultural crops in the vegetable, production nursery, and melon industries (Maino et al., 2019).
Understanding the invasion history of pests like L. sativae can help inform quarantine and management strategies, including the likelihood of insecticide resistance genes entering colonizing populations (Ma et al., 2007). The invasion history of a pest can be determined not only from historical records but also through genetic approaches. To date, information on the genetic structure of L. sativae populations is limited. Previous genetic research on this species has mainly focused on mitochondrial DNA (mtDNA) for deciphering population structure (Blacket et al., 2015;Parish et al., 2017;Scheffer & Lewis, 2005;Tang et al., 2016;Xu, Coquilleau et al., 2021). While mtDNA markers provide some indication of the relatedness of populations, they have a relatively low resolution particularly in detecting fine-scale structure (Anderson et al., 2010). Patterns of population relatedness based on mtDNA can also be obfuscated in insects by processes like endosymbiont invasions and recombination across the mitochondrial genomes (Ballard & Whitlock, 2004).
With the increasing ease and speed of DNA sequencing, high-resolution genomic SNP-based methods like double digest restriction-site-associated DNA sequencing (ddRAD-seq) provide a new dimension to population-level studies (Peterson et al., 2012).
These markers have been successfully applied to understand patterns of movement in insects that act as pests or disease vectors (Ryan et al., 2019;Schmidt et al., 2019;Yan et al., 2021). They provide a much higher level of resolution than other nuclear marker systems like DNA microsatellites given the sheer number of marker loci that can be scored (Morin et al., 2004). An example of the resolution of population patterns based on SNPs versus microsatellites is provided by Rašić et al. (2014) who showed overlap between populations of the mosquito Aedes aegypti based on 8 microsatellite markers but clearly distinct populations when 2300 SNP markers were used. SNP markers scattered across the genome also provide the opportunity to use genetic approaches in tackling new questions such as estimating cross-generation movement rates of pests following SNP-based identification of related individuals (Cordeiro et al., 2019;Jasper et al., 2021) and the identification of genomic regions that may be under strong selection once candidate loci have been identified (Endersby-Harshman et al., 2020;Uchibori-Asano et al., 2019;Yang et al., 2020).
In this study, we used SNP-based methods to investigate in de- Laboratories, Gladesville, NSW, Australia) Xu, Coquilleau et al., 2021;. The criteria for assigning CO1 haplotypes were based on BLAST similarity in NCBI (Johnson et al., 2008) with 100% similarity being used to link to existing haplotypes, and assignment to L. sativae clades defined by a similarity threshold of <2.5% (Scheffer & Lewis, 2005). Once species identity was confirmed, whole fly bodies were used for ddRAD-seq libraries. and 2 μl of 10 μM Illumina PCR primers. PCR programs were run at 98°C for 30 s, followed by 15 cycles of 98°C for 10 s, 60°C for 30 s, 72°C for 90 s, and a final elongation step at 72°C for 5 min.

| ddRAD-seq library preparation
We pooled 10 such PCR products together and cleaned them with a 1.5× bead solution for making the final library. The ddRAD-seq libraries were sequenced on an Illumina NovaSeq 6000 platform using 2 × 150 bp chemistry by Novogene HK Company Limited, Hong Kong.

| Sequence data processing and genotyping
We processed the raw sequencing reads according to the STACKS v2.54 pipeline (Catchen et al., 2013). Quality filtering of raw reads was performed with the process_radtags program. This program first checks if the barcode and the RAD cut site are intact, and then demultiplexes the data, correcting errors within allowable limits. We first trimmed the reads to 115 bp in length (-t 115) and removed sequence reads with average Phred scores below 20 (-s 20), a cutoff used in similar studies (Chen et al., 2021;Schmidt et al., 2019). Given the absence of a reference genome for L. sativae, we used the Stacks workflow (ustacks, cstacks, sstacks, ts-v2bam, and gstacks) to build a catalog of loci de novo (Catchen et al., 2011). In brief, we first processed all samples with ustacks  ues between all pairs of populations using the "genepop" R package (Rousset et al., 2020). The geosphere R package v.1.5-5 was applied to calculate straight-line geographic distances between populations (Hijmans et al., 2017). p-Values smaller than 0.05 were considered significant. from Seisia, suggesting recent co-ancestry among these populations ( Figure S1). The phylogenetic tree showed that the individuals of Thursday Island and Seisia are mixed together and independent from groups of Masig Island and Boigu Island, and there is no structure between them. This analysis also revealed that the Australian F I G U R E 3 FineRADstructure plot with co-ancestry map and phylogenetic tree. The color scale bar indicates estimated co-ancestry, with light yellow suggesting low co-ancestry and darker yellow, orange, and red indicating progressively higher co-ancestry. The Australian populations formed a clade (labeled in red), which is shown in the solid black square. See Figure S1 for full-size figure mainland was likely invaded from the Torres Strait instead of other regions. Interestingly, the Australian populations were most closely related to Bali, followed by PNG, suggesting that Bali and PNG were important sources for the initial invasion of L. sativae into the Torres Strait Islands.

| Admixture analysis
To infer the genetic structure and the degree of relatedness among L. sativae populations with different genetic backgrounds, we ran an admixture analysis with cross-validation for values of K (subgroups) from 2 through 10 to examine patterns of ancestry based on the 193 individuals ( Figure S2). Note that a plot with K = 1 was not considered given the low likelihood of such a scenario based on existing literature (Yamashita et al., 2019). A 10 fold cross-validation (CV) procedure was performed to infer the optimal number of subgroups.
We found a minimum CV score was obtained when K = 5, and we When visualizing the population structure when K = 8 (the second-best K value), there are additional hierarchical structure levels, primarily involving Kenya, but also within Australian L. sativae populations, and within Hawaii ( Figure S3).

| Isolation-by-distance
To evaluate the possible association between genetic and geo-

| DISCUSS ION
With the increasing globalization of trade and an escalating volume of goods being moved, threats posed by invasive pest species are increasing (Hulme, 2009). Liriomyza sativae represents one such threat, having invaded many countries around the world in the past 40 years (Murphy & LaSalle, 1999;Scheffer & Lewis, 2005). We also found support for multiple invasions from different countries (especially Indonesia, PNG, and Timor-Leste) into the Torres Strait populations as reflected by results from the admixture analyses. Furthermore, we found that L. sativae populations in Hawaii are genetically distinct from other countries, whereas samples from Kenya, China, Vietnam, and Timor-Leste cluster as a group that is related to Bali and PNG populations.
Liriomyza sativae was first described from specimens collected from alfalfa (Medicago sativa) in Argentina in 1937 (Blanchard, 1938), although a South American origin for the species has not been validated with genomic markers. Liriomyza sativae (incorrectly named Agromyza pusilla) was reported to be damaging crops in mainland USA before World War 1 (Spencer, 1973). By 1980, L. sativae was common in southeast USA, Central America, and northern parts of South America (Hardy & Delfinado, 1980). The earliest recorded collection of L. sativae in Hawaii was in 1921 (Table S3), where it was described as Liriomyza canomarginis, but later synonymized with L. sativae (Spencer, 1973; Table S3). Given Hawaii imports most of its vegetables from mainland USA (Parcon et al., 2010), imported produce is the likely source of incursions. The CO1 haplotype of L. sativae in Hawaii belongs to the L. sativae-W clade , which is also widespread in North America, having been collected from crops in California, Florida, and Arizona (Scheffer & Lewis, 2005). In contrast, L. sativae-A clade is common in Florida and has not been recorded in Hawaii, suggesting that L. sativae in then to several South Pacific islands (Waterhouse & Norris, 1987) occurred at least 20 years before the pest moved through Asia (Martinez, 1994) and Africa (Martinez & Bordat, 1996). In turn, from the mid-1970s, Liriomyza trifolii (Burgess) has largely displaced L. sativae in southern parts of the USA (Parrella, 1982) and Hawaii (Johnson, 2005). The spread of L. sativae has speeded up since the 1990s when L. sativae moved eastward through tropical and subtropical areas like Africa, China, Indonesia, and Vietnam (Andersen et al., 2002;Chen & Zhao, 1999;Rauf et al., 2000;Ridland et al., 2020). Liriomyza sativae arrived in Indonesia in the 1990s and became a major pest in lowland vegetable crops (Rauf et al., 2000).   (Blacket et al., 2015;Xu, Coquilleau et al., 2021). Previous CO1 studies demonstrate that S.27 is the dominant haplotype in Indonesian and Australian populations whereas one population (Erub Island) in the Torres Strait possesses haplotype S.07, the dominant haplotype found in PNG (Blacket et al., 2015;Xu, Coquilleau et al., 2021). The SNP data generated in our study provide a much clearer signature of admixture in the Torres Strait populations, including Thursday Island and Seisia which are only separated by 34 km. In future studies, it will be interesting to track further incursions into this area and beyond.
Currently, the mainland population at Seisia is under quarantine to prevent the spread of L. sativae to other regions, but further spread seems inevitable given that many other regions on the mainland are suitable for this species to exist (Maino et al., 2019). Any further incursions into the Torres Strait may benefit L. sativae such as through the introduction of insecticide resistance genes. The geographical location of the Torres Strait is particularly suitable for multiple incursions as detected in species like the Asian tiger mosquito, Aedes albopictus (Beebe et al., 2013;Schmidt et al., 2021), island sugarcane planthopper Eumetopina flavipes (Anderson & Congdon, 2013), as well as Culicoides biting midges (Eagles et al., 2014). Pests can move into the Torres Strait by natural processes (wind currents) as well as human activities (vessel movements and fishing activity) (Kompas et al., 2015).

| CON CLUS IONS
In summary, our SNP-based analyses of L. sativae provide evidence of connections among populations that are mostly consistent with historical observations. Our study is the first to apply high-density SNP markers to determine the population structure of Liriomyza and provide a baseline and foundation to further track leafminer movements across the region and further into the Australian mainland.
The data provide a basis for quarantine detections to identify local versus international sources (c.f. Schmidt et al., 2019) and an ongoing assessment of new incursions can indicate risks associated with insecticide resistance genes entering local populations.

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