Investigating the Genetic Underpinnings of Ongoing Fall Armyworm (FAW) Range Expansion in Aotearoa New Zealand

TL;DR

This study examines the genetic makeup of fall armyworm populations in New Zealand and other regions to understand their spread and resistance to insecticides.

Contribution

The study provides new genomic data from the invasion front in Oceania and identifies insecticide resistance alleles in fall armyworm.

Findings

New invasive fall armyworm populations in New Zealand conform to the population structure seen in Benin.

Preliminary evidence suggests multiple introductions to New Zealand, increasing genetic diversity.

Putative insecticide resistance alleles were detected in new samples, highlighting the need for ongoing monitoring.

Abstract

Spodoptera frugiperda
(fall armyworm; FAW) is a major agricultural pest native to the Americas, with the first reported invasion of Africa in early 2016. Since then, FAW has spread rapidly across Africa and Asia before invading Australia (2020) and first being detected in Aotearoa New Zealand in February 2022. Here, we assessed the whole genomes of 34 novel FAW individuals along the invasion front (representing three new invasive populations from Cambodia, Australia, and New Zealand) with the largest publicly available global FAW genome dataset (n = 173), resulting in a dataset of 112 and 99 samples from the invasive and native range, respectively, to: (1) place the new invasive populations within the global invasion; (2) identify the potential geographic origin of the New Zealand invasion, including from a single or multiple incursion event; and (3) assess pre‐existing insecticide resistance potential at the invasion front. We confirm that these new invasions conform to the broad population structure of the initial invasive populations identified in Benin (West Africa), all of which belong to the invasive corn strain, as defined through previous triosephosphate isomerase (TPI) analysis and associated isolation from specific host plants. While we could not confidently assign the source population of the New Zealand invasion, we find preliminary support for a multiple introduction hypothesis in our data, which could contribute to increased genetic diversity within the New Zealand population. Further sampling is therefore required to fully characterise the origins of the New Zealand invasion. In novel samples, we detected putative insecticide resistance alleles previously reported in other invasive populations. These resistant loci should be tracked over time to understand the mechanisms enabling the invasion success of FAW in the Asia‐Pacific region. We emphasise that sharing of genomic resources between institutions and consortia is an essential first step in the control of this global invader.

We generated new genomic resources for the invasion front of fall armyworm in Oceania, confirming that the new invasions conform to the population structure initially identified in Benin. We also identified SNVs associated with insecticide resistance that should be a focus of future tracking of the invasion front and established populations.

Figures

Geographic location of native and invasive populations. Map indicates location of each of the samples used in this study, with populations specified to regional level and sample size. Novel sequencing data from recent invasion populations (2020–2023) is represented in the inset box, red indicates the native range; blue indicates invasive range (2016–2022).

Population structure analyses of FAW populations. (A) PCA derived from 1,080,182 SNVs of all 17 populations that cluster by corn, corn:mex and rice strains. Invasive strains separate on PC2 from the native corn strain. (B) Focused PCA derived from 434,359 SNVs to specifically establish the temporal relationships between the invasive populations of FAW. (C) Admixture plots of assignment probabilities from K = 2–5, where each bar represents an individual within a population. Population groups (invasive/native; country; SfR = rice strain) are noted in the x‐axis legend. White dashed line separates the invasive and native populations. (D) Pairwise weighted Weir and Cockeram F ST of populations. Colours correspond to the level of divergence between each population pair, where lower F ST (< 0.15) shows low‐minimal differentiation (light colours) and higher F ST (> 0.15) shows increased differentiation (dark colours).

Neighbour‐joining tree of fall armyworm samples constructed from SNV‐derived pairwise genetic distances. Branch lengths represent the genetic divergence, where sample names are highlighted by origin population and separation between invasive and native range is indicated. Highlighted nodes represent the most recent common ancestor for the corresponding colour‐coordinated population. Population status is highlighted in the key; invasive (i) and native (n).

The proportion of individuals of native and invasive populations either homozygous or heterozygous for the D327G missense variant associated with a putative cytochrome P450 gene on scaffold 14.

Tables

TABLE 1. Presence or absence of known insecticide resistance associated SNVs in the Australian, New Zealand, and Cambodian populations.

Gene	Mutations	Presence/Absence	Origin/Known distribution
ABCC2	P799K/R	Not present	Brazil (Boaventura, Ulrich, et al. 2020)
	GY deletion	Not present	Brazil (Boaventura, Ulrich, et al. 2020)
	G1088D	Not present	Brazil (Boaventura, Ulrich, et al. 2020)
	12 bp insertion	Not present	Brazil (Guan et al. 2021)
	GC insert/FS	Not present	Puerto Rico (Banerjee et al. 2017)
AChE	F290V	Present	Uganda, Malawi, Brazil, China, India, Benin, Puerto Rico, Mexico, Mississippi, Florida, Australia (Boaventura, Martin, et al. 2020; Yainna et al. 2021; Chen et al. 2024)
	A201S	Present	Uganda, Malawi, Brazil, Benin, Mississippi, Florida, Australia (Guan et al. 2021; Yainna et al. 2021; Chen et al. 2024)
	G227A	Not present	Brazil; Puerto Rico; Florida (Boaventura, Ulrich, et al. 2020)
VGSC	T929I	Not present	Brazil (Carvalho et al. 2013)
	L932F	Not present	Brazil (Carvalho et al. 2013)
	L1014F	Not present	Brazil, Indonesia (Carvalho et al. 2013)

Full text

Introduction

1

Biological invasions are a major cause of economic loss to agriculture and economic sectors, resulting in US$1.288 trillion dollars of global costs between 1970 and 2017 (Diagne et al. 2021). The globally expanding invasive fall armyworm, Spodoptera frugiperda (FAW; Lepidoptera: Noctuidae), provides a spectacular example of the destruction to agroeconomies and livelihoods that invasive species can cause. As a multivoltine, polyphagous insect that is also a strong flier, FAW has severely impacted the yield of important crops such as corn, rice, and soybean (Fiteni et al. 2022) across Africa, Asia, and the Americas, and is now threatening the same crops in Australia and the South Pacific.

In early 2022 FAW was first detected in the North Island of New Zealand, with its arrival likely assisted by El Niño weather events, featuring strong and persistent Westerly winds in the South Pacific (MPI 2025). Confirming model predictions (Benjamin 2021), FAW has now been identified across at least nine regions of New Zealand and is thought to have persistent overwintering populations in the northern region of the North Island (MPI 2025; Benjamin 2021), where average winter temperatures are known to exceed its survival threshold (winter averages in Kaitaia 12°C–13°C, and Whangarei 11°C–12°C; NIWA 2025). Potentially facilitated by local seasonal migration patterns from November to February, FAW is now found as far south as Canterbury and the West Coast of New Zealand's South Island, and with warming climates and milder winters, the risk of populations overwintering in more regions increases.

Concern is also rising on the potential for FAW to carry over or locally evolve resistance alleles for commonly applied insecticides. For example, the F209V acetylcholinesterase (AChE) resistance mutation is present in higher numbers of individuals of susceptible invasive populations than native (Yainna et al. 2021). Whether these resistance alleles are persisting along the invasion front in recently invaded countries remains crucial to management. Existing large genomic datasets have revealed salient features, such as host specificity, geographic structure of the native and invasive lineages, and classification of the global FAW populations into corn (native), corn:mex (native), and rice (native) strains, and established the invasive lineage to be derived from the corn strain (Yainna et al. 2022).

In this study, we aimed to place FAWs range expansion in New Zealand within the context of its global invasion, with the aid of a publicly available genomic dataset representing a large proportion of the native and invaded range (Figure 1). We aimed to resolve the potential origin of the New Zealand lineage against Australian and Southeast Asian lineages. We also sought to identify loci associated with insecticidal resistance that could have downstream effects on management of established populations in the wider Pacific region, as well as highlight regions under selection that may reflect the evolutionary trajectory of the FAW as it continues expanding globally.

Geographic location of native and invasive populations. Map indicates location of each of the samples used in this study, with populations specified to regional level and sample size. Novel sequencing data from recent invasion populations (2020–2023) is represented in the inset box, red indicates the native range; blue indicates invasive range (2016–2022).

Materials and Methods

2

Sampling and Data Generation

2.1

Samples from New Zealand were collected in the summers of 2022/23 from maize fields (n = 13), either manually or using FAW pheromone traps, by collaborators at AgResearch and the Ministry for Primary Industries (MPI) following confirmation of FAW field reports. Australian samples (n = 4) were intercepted in Queensland in March 2021 and provided by MPI. Cambodian samples (n = 17) from the provinces of Siem Reap, Kandal, Kampong Thom and Preah Vihear were collected from corn in June 2022 by collaborators at Plant and Food Research (Table S1). This dataset was combined with the global FAW published dataset (Yainna et al. 2021; Figure 1), which included field collected samples from the native range (Mississippi = 17, Florida = 24, Puerto Rico = 15, Guadeloupe = 4, French Guiana = 3, Brazil = 10, Mexico = 26) and invaded range (China = 2, India = 14, Malawi = 16, Uganda = 7, Benin = 39) samples. The combined dataset represents the scope of the invasion from 2016 in Africa up to the most recent invasion in New Zealand (2022–2023). The total number of geographic regions includes 14 countries, with 211 individuals (112 from the invasive range; 99 from the native range).

Whole Genome Assembly and Variant Calling

2.2

Genomic DNA was extracted from individuals using the Nucleospin Tissue kit (MN), with libraries for Illumina sequencing generated from 100 to 299 ng DNA input using the Illumina DNA PCR‐free Library prep tagmentation kit (Illumina), as per manufacturer's instructions. Resequencing was performed on samples collected in Cambodia (n = 17), Australia (n = 4), and New Zealand (n = 13). Paired‐end genome sequencing was performed using the NovaSeq 6000 system (150 bp PE; LIC, New Zealand) resulting in an average of 21× coverage. We followed the assembly and variant calling approach outlined in Yainna et al. (2022) to ensure comparability. Reads were mapped to the FAW reference genome v7 (Fiteni et al. 2022) using BWA‐mem (v07.1; Li 2013) to ensure interoperability with the global dataset. Assemblies with genome coverage greater than 5× (to include additional lower coverage Australian samples and remain in the threshold to accurately assess genomic variation; Benjelloun et al. 2019) and mapping at 80% were retained, and haplotypes were locally reassembled using GATK v4.1.2.0 (van der Auwera and O'Connor 2020). Mapping rates for novel invasive samples did not differ significantly between populations, indicating no evidence of population‐specific reference bias. Per sample identification of duplicate regions was undertaken using the MarkDuplicates function of GATK before we performed base quality score recalibration (BQSR) using a set of previously validated SNVs as known variant sites to recalibrate base quality scores and correct errors in sequencing data. Variants were then called using GATK Haplotypecaller (Poplin et al. 2018) following GATK best practices, with all GVCFs merged and a final variant set of n = 94,772,931 single nucleotide variants (SNVs) called using GATK GenotypeGVCF function. Filtering using GATK VariantFiltration retained 33,138,472 SNVs that satisfied the following conditions: QD > 2.0, FS < 60.0, MQ > 40.0, MQRankSum < −12.5 or ReadPosRankSum > −8.0.

Novel FAW samples (Cambodia, Australia, and New Zealand) were positively identified as FAW using a phylogenetic approach reconstructed from the mitochondrial COI gene (Figure S1). The mitochondrial genome was identified using NCBI (Accession number KM362176) and extracted using SAMtools (Danecek et al. 2021). The COI gene was identified using BLAST functions against other Spodoptera sequences (Camacho et al. 2009). Mitochondrial COI genes for each sample were then aligned using MAFFT (Katoh et al. 2002), and a maximum likelihood tree was reconstructed using IQTREE (Danecek et al. 2021). We used the GTR + G model that accounts for different evolutionary rates between nucleotides, as well as performing 1000 bootstrap replicates to calculate statistical support at each node. The resulting tree was visualised using the ggtree v4.0.1 package (Yu et al. 2017) in R (R Core Team 2022). As there is evidence in invasive populations of FAW suggesting hybridisation between SfC and SfR, the triosephosphate isomerase (TPI) gene has been identified as a reliable marker to differentiate between host plants (Fiteni et al. 2022; Nam et al. 2024). We extracted the TPI locus from the complete dataset (n = 211) using Tabix v0.2.6 (Li 2011) before performing a principal component analysis (PCA) on this locus using Plink2. Strain designation was then assigned based on host plant, as SfR (rice) or SfC (corn) according to the clustering identified in the PCA relative to known corn and rice strain identifications from previous studies (Figure S2), as well as the host plant the individual was isolated from (corn = 34).

Genomic Analysis

2.3

To minimise bias introduced when using uneven sample sizes among regions, population structure was assessed using complementary approaches to emphasise genetic relationships at the individual level. Results from sparsely sampled regions should, however, be interpreted with caution. Principal component analysis (PCA) on the full dataset was performed to validate the grouping of new invasive populations of fall armyworm relative to Yainna et al. (2022). Further filtering of the variant call file first excluded SNVs with a missing genotyping rate higher than 5% (‐geno 0.05) and a minor allele frequency (MAF) below 0.05 (‐MAF > 0.05). Highly correlated SNVs were also removed using LD pruning with a sliding window of 50 SNVs and an r ^2^ threshold of 0.2 (LD using ‐indep‐pairwise 50 5 0.2) using Plink2 (Chang et al. 2015). The resulting hard filtered SNV file used for population structure consisted of 1,080,182 SNVs. The resulting VCF was then used to perform PCA in Plink2 and visualised in R, where 10 principal components were extracted, with the first two explaining 50% of the variation between groups. We followed the same filtering method with a reduced dataset of invasive populations (n = 112) for population analysis, resulting in a hard filtered SNV file of 434,359 SNVs. We tested genome coverage sensitivity for variation among samples by performing population analyses with and without lower coverage genomes and found no change in structure. As groups align with the clusters observed in Yainna et al. (2022), we used the previous designations of corn, corn:mex and rice strains to fit additional samples to those previously defined groups within this study. FastStructure (Raj et al. 2014) was used to determine population structure at a range of K‐values (K = 1–10) before determining the appropriate model complexity range using the integrated tool chooseK.py, where the optimal K‐value was determined as 2, with up to 5 components contributing to population structure. We then visualised the population structure within this range (K = 2–5) to elucidate hierarchical structure between invasive and native populations. F ST was calculated from Weir and Cockerham's (1984) test from non‐overlapping 500 kb windows using VCFtools (Danecek et al. 2011) to determine significant variation between populations and groups (invasive vs. native) while accounting for unequal sample sizes between populations. We also assessed sample relatedness using the KING‐robust kinship estimator of Plink2, an integrated function of the KING algorithm (Manichaikul et al. 2010). A genome‐wide SNV phylogeny was constructed using pairwise genetic distances calculated using the ‐distance function of Plink v1.9 (Purcell et al. 2007). A neighbour‐joining tree was then inferred from the distance matrix before being visualised in R using ggtree.

For tracking spread of insecticide resistance alleles to invasive populations in Australasia and Cambodia, analysis of relevant genes was undertaken. The ATP‐binding cassette subfamily C member 2 gene (ABCC2; Boaventura, Martin, et al. 2020; Boaventura, Ulrich, et al. 2020)—associated with Bacillus thuringiensis (Bt) insecticidal toxin resistance in native populations—and acetylcholinesterase (AChE; organophosphate and carbamates resistance; Chen et al. 2024) and VGSC (voltage gated sodium channel; synthetic pyrethroid resistance; Carvalho et al. 2013) genes were annotated by aligning the protein sequences for each gene respectively from NCBI and using the protein2genome model from Exonerate v2.4.0 (Slater and Birney 2005) to locate chromosomal scaffold coordinates. Genotype status (homozygous reference, heterozygous, or homozygous) of known resistance‐associated SNVs was subsequently verified from the novel data generated in this study by manual inspection in IGVtools (Robinson et al. 2011). Positive identification of alleles of populations from Yainna et al. (2021) was used to validate the novel findings from new populations. As there is evidence for gene duplication in the AChE gene in FAW (Chen et al. 2024), we assessed allele copy number and depth support in this region relative to the whole genome. To further investigate this region, we isolated the homologous region in scaffold 14 corresponding to the primers reported by Chen et al. (2024). BAM files from newly sequenced individuals were assessed for read coverage of AChE (HiC_scaffold_14:11,311,618–11,312,093). We assessed relative copy number by normalising the mean read depth of the AChE gene to both genome‐wide mean depth and the single copy gene region encoding glutamate‐gated chloride channels (GluCl). We then undertook copy number variant (CNV) discovery using the software CNVnator v0.4.1 which utilises a statistical approach against a reference sequence (Abyzov et al. 2011) to assign read depths to bins, using 100 bins for sequences with coverage between 20 and 40× and ~500 bins for those between 4 and 10× coverage. The output was then screened for CNVs where size > 1 Kb and q0 < 0.5. The resulting VCF file and BAM files were manually examined using IGVtools to inspect variants.

Following from Yainna et al. (2022), who identified markers of selective sweeps on chromosomes 8 (134,913 sites), 14 and Z, we selected these scaffolds for initial analysis with BayPass v3.0 (Gautier 2015) using the recently developed C_2_ statistics (Olazcuaga et al. 2020) to identify candidate loci under selection within invasive populations versus the native populations, specifically as hypothesis testing for the identification of alleles associated with insecticide resistance. We selected these chromosomes to impose a biological prior that would minimise the burden of uneven or small sampling sizes from some populations, as well as the impact of false positives that could be driven by demographic history. We ran BayPass core model with default parameters, where histograms show a ‘flat’; distribution of p‐values associated with the C_2_ statistics between 0 and 1. A control for multiple testing (resulting in q‐value calculation) was implemented in RStudio using the package qvalue, and SNVs with a q‐value < 0.01 were used for the invade range SNVs.

We then used SnpEff v.5.2f (Cingolani et al. 2012) to manually construct a database from the annotated S. frugiperda reference genome to and then annotated using SnpSift investigate downstream impacts of candidate outlier SNVs and identify those impacting target genes. From this, we could determine whether candidates had a low, medium, or high effect on associated genes. From those candidate outliers, we extracted CDS regions and putative annotations from OGS7.0 (Yainna et al. 2020) before validating any P450s against OGS2.2 (Gouin et al. 2017) using NCBI BLAST+ v.2.16 (Camacho et al. 2009), where annotations include manually curated P450 genes. Sequences were secondarily validated in UniProt BLAST (The Uniprot Consortium 2024), before manual inspection of allele presence in individuals and at population level using IGVtools (Robinson et al. 2011).

Results

3

The three recently invaded countries, Cambodia, New Zealand, and Australia represent the continuing spread of the globally prevalent highly invasive corn strain (Figure 2A,B). Through TPI‐based PCA analysis, these were identified as belonging to the corn (SfC) strain, as with most of the invasive population (Figure S2) correlating with the known plant host of individual samples. The rice strain (SfR) included the known populations of Florida, French Guiana, Guadeloupe and a singular individual from Mississippi, but also three invasive individuals: two from the Benin population (ACH20 and ACH6) and one from a novel Cambodian individual (03‐B). Phylogenetic analysis of the mito‐COI gene showed two clades, with representative invasive samples present in both. As all novel FAW samples (except 4 from Nelson, New Zealand) were collected from corn, this supports prior observations and evidence that the invasive populations represent a hybrid strain between SfR and SfC (Wang et al. 2025). When SNV‐based PCA analysis of the complete dataset (n = 211) was undertaken, the first principal component (PC1) axis explained 33.81% of the variance in allele frequency and broadly separated the native and invasive groups from the native corn: mexico group, whereas PC2 separated the native and invasive groups (Figure 2A). Two invasive individuals from Malawi showed less variation to native strains than other invasive populations (Benin, India, China, Uganda, Cambodia, Australia, New Zealand). PCA groupings by strain were supported by admixture plots (Figure 2C) and pairwise F ST values (Figure 2D), which showed higher differentiation among population pairs in line with invasion status (native/invasive) and strain classification (rice/corn). Once the invasive dataset was subset (n = 112), populations demonstrated a temporal and spatial trend in variance across PC1 (14.04%) (Figure 2B), where the initial invasion into Benin showed stronger population structure compared to the consecutive invasions from 2017 to 2019, likely reflecting the introduction of a small founding population from the initial invasion source. The Australian individuals appear to be distinct from other invasive populations across PC1 (Figure 2B). New Zealand and Australian populations were separated along PC2 (Figure 2B), where New Zealand shared a strong signal instead with IndoPacific countries, suggesting that the investigated Australian samples intercepted in Queensland were not the origin of the New Zealand invasion, though broader sampling to capture the full population diversity is needed to confirm this hypothesis. The separation of samples along PC3 and PC4 (which explain 9.5% and 11% of the variance, respectively; Figure S3) was primarily driven by modest variation within the Indian population, and some individuals from Malawi, Cambodia and an individual from New Zealand. This may reflect localised differences arising from multiple introduction sources and demographic processes. To minimise batch effects, novel samples (n = 34) were randomised in DNA extraction and library preparation. As we did not see clustering in any principal component by batch, or between novel and contemporary sequences, patterns likely reflect biological differences. Kinship coefficients between samples calculated in Plink2 showed that none of the populations or samples were closely related to the New Zealand invasion to be defined as third degree relations or closer (< 0.0442), suggesting an absence of close familial relationships among samples.

Population structure analyses of FAW populations. (A) PCA derived from 1,080,182 SNVs of all 17 populations that cluster by corn, corn:mex and rice strains. Invasive strains separate on PC2 from the native corn strain. (B) Focused PCA derived from 434,359 SNVs to specifically establish the temporal relationships between the invasive populations of FAW. (C) Admixture plots of assignment probabilities from K = 2–5, where each bar represents an individual within a population. Population groups (invasive/native; country; SfR = rice strain) are noted in the x‐axis legend. White dashed line separates the invasive and native populations. (D) Pairwise weighted Weir and Cockeram F ST of populations. Colours correspond to the level of divergence between each population pair, where lower F ST (< 0.15) shows low‐minimal differentiation (light colours) and higher F ST (> 0.15) shows increased differentiation (dark colours).

To further explore the genetic similarity among novel samples collected at the invasion front, we constructed a neighbour‐joining tree based on the pairwise genetic distances from the SNV dataset (Figure 3). Native populations formed distinct and well‐supported clusters associated with geographic origin, especially the Mexican corn strain, which showed stronger divergence than other native populations. Phylogenetic patterns among invasive populations suggested a single common ancestor, with some evidence of genetic admixture between SfC and SfR strains (Figure 2C), supported by the formation of a monophyletic clade composed of all invasive individuals (Figure 3). Samples from Benin also formed a monophyletic clade, consistent with an early divergent lineage responsible for the invasive populations. The clustering of other invasive populations (including the novel populations described here) supports the hypothesis that the original invasive population was the dominant source of the globally invasive lineage, and this was further supported by the reduced genetic distance between these groups in PCA and admixture plots (Figure 2B,C). In the SNV phylogeny, Australian samples were primarily clustered together, with a single New Zealand sample from Waikato and two individuals from India (Figure 3). Exploration of PC3 and PC4 (Figure S3) showed that this New Zealand sample (Wai_01) was more differentiated from the core population clustering, demonstrating tentative evidence in support of a hypothesised multiple invasion scenario in New Zealand. The lack of clustering of this individual with other Waikato samples could suggest it had a differential origin, while other New Zealand individuals more closely aligned with the IndoPacific countries in our dataset (Figure 2B). Further sampling would be necessary to define this link.

Neighbour‐joining tree of fall armyworm samples constructed from SNV‐derived pairwise genetic distances. Branch lengths represent the genetic divergence, where sample names are highlighted by origin population and separation between invasive and native range is indicated. Highlighted nodes represent the most recent common ancestor for the corresponding colour‐coordinated population. Population status is highlighted in the key; invasive (i) and native (n).

Signatures of Adaptive Traits for Informing Management Strategies

3.1

We evaluated the presence and spread of known mutations on the ABCC2 (Bt resistance), AChE (organophosphate and carbamate resistance), and VGSC (pyrethroid resistance) genes in recent invasive populations (Table 1). Two targeted mutations were identified in the populations assessed in this study. Resistant alleles were identified across all populations for F290V in either heterozygous or homozygous form (88%). A201S was present but with low frequency, where 26% of samples carried one copy of the mutation. AChE is known from previous studies to be duplicated, so we calculated the sequence coverage relative to the whole genome and a single copy gene (GluCl). Read coverage of the AChE gene was consistently increased relative to the whole genome (0.8–3×) and GluCl (1.4×–2.5×), supporting the presence of a gene duplication event and further supporting work by Chen et al. (2024). The exception was 06‐A and 08‐A samples (Cambodia), for which AChE had a similar coverage to that of GluCl and the whole genome (Table S2). Read depth‐based inference with CNVnator supported a duplication overlapping the acetylcholinesterase locus in the three novel populations (Cambodia, New Zealand and Australia). While CNVnator failed to produce reliable calls in a subset of low coverage samples (< 10×), consistent signals across independent individuals and agreement between statistical based and manual read depth analysis support the presence of this duplication.

As some populations are known to have resistance to synthetic pyrethroids (Carvalho et al. 2013), we used BayPass to find candidate targets of selective sweeps in invasive populations to assess adaptation to insecticides. Previous work had identified chromosomes 8, 14, and Z as important candidates under adaptive evolution (Yainna et al. 2022), so we specifically looked in these regions, focusing on glutathione S‐transferases (GSTs), cytochrome P450s (P450s), and esterases (Chen et al. 2023). Chromosomal scaffolds 8 and Z returned no outliers associated with known genes related to insecticide resistance. However, we identified five SNVs on chromosomal scaffold 14 in three cytochrome P450 genes (CYP321B3, CYP332A1, CYP321B4) and a single carboxylesterase gene that differentiated invasive and native populations, also found in Yainna et al. (2024). Of these, four were synonymous variants resulting in no change to amino acid translation, while one (HiC_scaffold_14: 12470120) resulted in a missense variant at a stop codon (D327G) on cytochrome P450 CYP321B1 (Table S3). Native populations from the USA had the lowest proportion of individuals that were positive for this mutation with only 39% carrying this variant. While this was not statistically significant from invasive populations (two‐tailed Fishers Exact; p > 0.05), it contrasted significantly with the high frequencies (> 75%) observed across invasive populations (Figure 4 and Table S4). Significant differences were detected between invasive populations (Benin, Cambodia, Uganda, New Zealand, and India) and native range populations of Mexico and Puerto Rico, where the mutation was entirely absent.

The proportion of individuals of native and invasive populations either homozygous or heterozygous for the D327G missense variant associated with a putative cytochrome P450 gene on scaffold 14.

Discussion

4

Rapid insecticidal resistance and long‐distance dispersal events have cemented FAW's position as a notorious unwanted invader globally. Its ability to acclimate (Benjamin 2021) and adapt to local conditions is underpinned by high existing genetic diversity and ongoing population admixture occurring in tandem with the global invasion (Yainna et al. 2022). Thus, understanding the genetic mechanisms that underpin FAW's invasion success is crucial to successful management of new range expansions and the mitigation of impacts to local economies.

We assessed the invasion trajectory of FAW in the Asia‐Pacific region, providing new genomic resources from the invasion front. We demonstrated that recently established populations do not show genomic predispositions to known insecticide resistance alleles aside from the characterised AChE gene, with our data providing a baseline for monitoring the process of gene flow and local adaptation over time within New Zealand. We also confirmed that the New Zealand invasion is derived from the prevalent globally invasive strain. Genomic signals within New Zealand populations were strongly correlated with IndoPacific countries, and less so with the single Australian population assessed here. Further, within the New Zealand populations studied, we found a geographically unstructured signal that could be an artefact of the recency of the invasion and a lack of local adaptation since arrival that could be resolved in future with further analysis on an expanded dataset. In the absence of a strong signal of similarity between the Queensland and New Zealand populations, we cannot rule out that the New Zealand populations derive from a single mixed‐population event or via multiple invasion or migratory events from disparate Australian or other IndoPacific populations, though further sampling in both regions will be required to confirm this multiple‐invasion hypothesis. Finally, as the Australian samples were sourced from a single interception in Queensland where the origin was unknown and where previously multiple disjunct populations have been described in addition to the basal Queensland clade, our determination of origin remains incomplete (Plant Health Australia Ltd. 2023). Australian analysis was also limited by sample size, where additional samples, notably from Queensland, New South Wales, Victoria, and Northern Territory, are thus essential to include in future analysis to determine with certainty the origin of the New Zealand invasion of FAW, and relationship to overwintering in Northland.

Pesticide resistance presents a major challenge in the development of management strategies. In FAW, pre‐existing resistance to some pyrethroids and other synthetic pesticides, such as carbamates, carried by the founding invasive populations and identified across the populations in Australia (Bird et al. 2022) highlight the threat to New Zealand of rapid resistance. Screening in Australian FAW between 2022 and 2024 showed that resistance to some oxadiazines and anthranilic diamide class insecticides were present, but rare (Bird et al. 2022). However, Bird et al. (2022) also found that, relative to Helicoverpa armigera, FAW had decreased sensitivity to synthetic pyrethroids. We showed that, of previously described resistance alleles (AChE; Bt, ABCC2; organophosphate and VGSC; pyrethroid resistance), both F290V and A201S were present within Australian, New Zealand, and Cambodian samples studied here. This is consistent with work by Nguyen et al. (2021) and Chen et al. (2024) who each found that ~80% of the Queensland population in Australia was positive for the F290V AChE mutation. However, we confirmed prior observations of candidate outlier SNVs in three P450 loci in the invasive populations, two of which (CYP321B3 and CYP321B4) were previously found to be under selection in populations on maize hosts (Yainna et al. 2024). We also identified a single candidate SNV on a stop codon within CYP321B1 that had high prevalence (> 75%) in invasive populations but was only present at low frequency in the native USA population. This suggests a putative origin of this variant in either Florida or Mississippi. CYP321B1 has undergone functional characterisation by Wang et al. (2017), where RNAi silencing significantly increased mortality in Spodoptera litura larvae exposed to synthetic the pyrethroid β‐cypermethrin and the organophosphate Chlorpyrifos. In silico simulations to assess the binding capacity of insecticides and their target proteins in fall armyworm have found that known mutations can reduce their efficacy by minimising space within binding sites (Cai et al. 2024). Together, these observations raise the hypothesis that, should this variant be placed in a crucial area of the docking domain for β‐cypermethrin Chlorpyrifos, it could reduce the efficacy of the insecticide. However, these inferences would need to be characterised explicitly, and functional assays are required to confirm any putative resistance mechanisms. While phenotypic resistance data for candidate outlier CYP321B1 is not available for the populations studied here, Khan and Ali (2025) demonstrated that populations of invasive fall armyworm in Punjab, Pakistan were 61.9–540.6× more resistant to cypermethrin than susceptible lab strains, the highest of any of the insecticides assessed. Our work independently confirmed that this SNV is proliferating at the invasion front and could be involved in P450‐mediated insecticide resistance. Thus, careful monitoring for resistance in invasive populations should be undertaken, especially across recent range expansion extents (Australia and New Zealand), where localised adaptation and selection may yet occur. However, further work would need to be undertaken to define susceptibility at the invasion front and to assign functionality to the variant. Further sampling within these regions would also improve the detection of signatures relevant to adaptive traits.

Currently, the WGS data for Australia and New Zealand is limited, with small sample sizes (Australia n = 4; New Zealand n = 13) reducing statistical power and the ability to reliably detect adaptive genetic variation. Small sample sizes can increase the likelihood of sampling bias, whereby observed allele frequencies may not accurately represent population‐level variation (Nazareno and Jump 2012). This is supported by Li et al. (2020) who found the minimum sample size for populations of the invasive harlequin lady beetle to estimate diversity within populations was six individuals. This can obscure subtle signals of selection or lead to spurious associations being interpreted as adaptive (Subramanian 2016), limiting inference of signals associated with selective sweeps that may be difficult to distinguish from demographic effects. Therefore, loci identified here—such as cytochrome P450 CYP321B1—warrant further investigation. Expanding sample sizes across geographic and environmental gradients for FAW would strengthen the ability to detect adaptive signatures and provide more robust insights for evidence‐based management interventions. Subsequently, ongoing tracking of key alleles at an early stage of future invasion trajectories will enable the study of subsequent adaptation, gene flow, and evolution of novel resistance capacities. Management of this global invader will require development of effective management strategies that rely on multiple modes of management, such as selective use of pesticides and biocontrol organisms. We call for greater global collaboration to enable exchange of knowledge and resources, in particular open access publication of genetic datasets, to ensure that future analyses can inform affected parts of the world—especially countries and territories with limited resources, such as those in the Pacific, where this pest is now also advancing and remains a major concern (International Plant Protection Convention 2023).

In sum, our results demonstrate that the source of the ongoing range expansion of FAW into the Pacific aligns genetically with the globally invasive population that originated with the invasion of west Africa in 2016. While the New Zealand population shares similarities to the original invasive strain, initial evidence putatively suggests multiple introduction events, which comprehensive sampling across the New Zealand and the Australian range will be essential to fully resolve. Some evidence of insecticide resistance loci is present within the novel populations in our study, warranting further monitoring and exploration as this pest continues to spread across the Pacific.

Author Contributions

Amy L. Vaughan: conceptualization (equal), data curation (equal), formal analysis (equal), funding acquisition (equal), investigation (equal), methodology (equal), visualization (equal), writing – original draft (equal), writing – review and editing (equal). Angela McGaughran: conceptualization (equal), funding acquisition (equal), methodology (equal), project administration (equal), supervision (equal), writing – original draft (equal), writing – review and editing (equal). Kiwoong Nam: data curation (equal), methodology (equal), resources (equal), writing – original draft (equal), writing – review and editing (equal). Manpreet K. Dhami: conceptualization (equal), funding acquisition (equal), methodology (equal), project administration (equal), resources (equal), supervision (equal), writing – original draft (equal), writing – review and editing (equal).

Funding

This work was funded by Manaaki Whenua Landcare Research, Bioeconomy Science Institute: Matawhānui Visionary Science grant to A.L.V. and by a Catalyst Dumont d'Urville NZ‐France Science & Technology Support Programme grant by Royal Society Te Apārangi Cataylst Dumont D'urville France‐New Zealand Science & Technology Exchange Programme (New Zealand) and French Embassy in New Zealand to A.M. and M.K.D.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Phylogenetic tree of cytochrome oxidase (COI) gene for fall armyworm samples. Phylogeny constructed using IQTREE. The branch containing the novel Australia and New Zealand samples is indicated in green.

Figure S2: Principal component analysis (PCA) of the triosephosphate isomerase (TPI) gene to show the clustering of SfC and SfR strains. Designation of host plant strains as SfC or SfR was assigned relative to the predetermined strains of Puerto Rico and Mississippi (Gimenez et al. 2020). Labels are shown for the populations previously identified as rice strain by Yainna et al. (2022), and for the invasive samples that clustered with them in the PCA.

Figure S3: Principal component analysis (PCA) of PC3 and PC4, derived from 434,359 SNVs to specifically establish the temporal relationships between the invasive populations of FAW. Sample labels are shown for individuals where variation is different from the basal cluster of invasive populations.

Table S1: Sample collection and sequencing metadata for 34 samples for which novel data was generated in this study.

Table S2: Sequencing coverage of the AChE gene relative to the single copy GluCl gene and the whole genome of fall armyworm for novel samples, as well as the output of copy number variant discovery using CNVnator.

Table S3: Similarity and BLAST results (UniProt and NCBI+ against OGS2.2; Gouin et al. 2017 with curated P450s) for closest related gene, and the associated SNV with functional impact.

Table S4: Presence of the D327G mutation in individuals of invasive and native populations. Heterozygous and homozygous variant forms shown. Individuals that didn't alter from the reference are shown by *. Missing data are shown with.