Complex multiple introductions drive fall armyworm invasions into Asia and Australia

Strain assessment

We did not detect any C-strain individual following analysis of 138 fully assembled mitochondrial DNA genomes (mitogenomes) from Australian samples. Our results, particularly that from Northern Territory, are not dissimilar to the finding of Piggott et al.⁵⁶ who detected only two (i.e., 4.2%) C-strain mtCOI haplotype individuals from a much larger (i.e., n = 48) Northern Territory sample size. Proportions of C-strain to R-strain also varied significantly across the different SEA populations (Table S1) in contrast to the patterns observed in China, India, and African nations (e.g.,^{22,33,34,39,57}). All Australian populations analysed for their corn or rice mitochondrial haplotypes via mitogenome assemblies of whole genome sequencing data therefore contrasted with the invasive populations from SEA where in some countries (e.g., Myanmar, Vietnam) FAW with the C-strain mtCOI haplotypes made up approximately 50% of the populations examined (see Table S1 for C- and R-strains mitogenome proportions, see also Fig. 1 ‘C-strain’ and ‘R-strain’ Maximum Likelihood cladograms).

Mutations in the Triosephosphate isomerase (Tpi) gene used to differentiate the C- and R-strains suggested that virtually all individuals from the SEA, PA, and South Korean (i.e., EA) carried the C-strain genotype at the Tpi gene. Of all SEA and PA individuals assessed, none was homozygous for the rice allele at this locus, while a number of samples (n = 10) were heterozygous C-/R-strains, primarily from the Philippines as well as one sample each from Malaysia and South Korea. Similar to the conclusions of Zhang et al.³⁴, at least in the invasive range, our finding suggested that the Tpi is not a useful marker: not only are the majority of samples of ‘corn’ preference, with limited evidence of hybrids and therefore contrasting greatly with whole genome analysis, but both the Tpi and the mitochondrial markers frequently contradicted each other, and do not provide accurate genealogy in, e.g., hybrid females or their offspring, especially if the female also mated with a hybrid male³¹.

Mitochondrial genome analysis

In the invasive populations, we identified for the C-strain at least 12 unique maternal lineages forming a total of five clades. For the R-strain, conservatively we identified 19 maternal lineages from nine clades (i.e., I–IX; Fig. 1 ‘C-strain cladogram’). Combining results from this study and the study of Tay et al.³³, we therefore identified at least 27 unique maternal lineages in the invasive FAW populations from across 14 countries from Africa, South Asia, East Asia, SEA/Pacific, and Oceania.

The C-strain mitochondrial genomes detected in SEA (Clade II) differed from the one present in the Yunnan CY and Indian populations, although the four haplotypes clustered with high (87–100%) bootstrap support. One other C-strain mitogenome was detected in one individual from China³³, and this remained a singleton haplotype amongst the other eight unique C-strain mitogenomes detected from African and SE Asian populations. For the R-strain mitogenomes (Fig. 1 ‘R-strain cladogram’), Clade I consisted of invasive individuals from all countries sampled to-date, although it did not contain individuals from all populations (e.g., no individuals from Australia Sf20-4, Sf20-5, or Mackay populations). R-strain Clade I also contained unique mitogenomes of individuals from the three China Yunnan populations (i.e., CY, XP, YJ), although other Chinese R-strain mitogenome haplotypes were also detected in clades II and VI. The second largest clade (Clade V) contained R-strain mitogenomes shared between individuals from six Australia populations, East Africa (Tanzania, Uganda), India, South Korea, Laos, PNG, and Malaysia (i.e., blue branches), although excluded other populations from SEA (i.e., Myanmar, Vietnam, Philippines), China, and Africa (i.e., Malawi, Benin). Clades II, III, IV, VIII, IX contained mitogenome haplotypes of individuals from either SEA and/or Asia (China, India) but not individuals from any of the African countries. Clades VII, VIII, and IX also have unique mitogenome haplotypes of invasive individuals that were more similar to French Guiana (FG code) or the Caribbean (i.e., Puerto Rico (PR), FG, or Guadeloupe) populations. The overall low bootstrap confidence at various nodes for both C-strain and R-strain phylogenies reflect the high number of maternal lineages with low mitogenome nucleotide diversity (i.e., mitogenome haplotypes differed by only very few nucleotide substitutions) across both invasive and native range populations. These limitations therefore lower their powers for use to confidently infer invasive population origins.

Nuclear SNP phylogenetic analysis

Maximum Likelihood (ML) phylogenetic analysis of the global S. frugiperda populations based on 870 neutral single nucleotide polymorphic (SNP) markers provided evidence of complex genetic relationship especially between individuals representing the invasive populations (Fig. 2). While the New World native populations (USA, Puerto Rico, Guadeloupe, Mexico, French Guiana, Peru, Brazil) clustered together and shared a basal (ancestral) lineage, the invasive populations did not identify Africa as the founding invasive population nor support the axiom of rapid west-to-east spread across the Old World. Selected individuals from Uganda and Benin appeared to cluster close to China/India populations, but there was no strong bootstrap support for a shared common ancestor (i.e., bootstrap value < 74%). China populations (i.e., CY, YJ, XP from Yunnan; see Fig. 1B in⁵⁸) formed two closely related sister clades with long branch lengths, suggesting these likely represented separate introduction events. Based on analyses of nuclear SNP data, the main Australian FAW population clade consisted of multiple sister clades and were clustered with varying degree of bootstrap support values. Some of these Australian populations appeared to share close relationship with the FE (i.e., South Korean) population (> 87–100% bootstrap support; e.g., selected Burdekin individuals), while others had largely unknown invasive origin(s).

Southeast Asian countries (i.e., Malaysia, Philippines, Laos PDR, Vietnam, Myanmar) and Pacific/Australia (i.e., PNG) have populations that were mixed, and interestingly were ancestral to East African populations (i.e., Malawi, Uganda), while some Australian populations (i.e., NSW; NT) were shown to be closely related to these SE Asian populations with strong bootstrap support (e.g., > 87–100% support for NT and NSW populations). Myanmar and China appeared to not share FAW populations with similar SNP profiles despite being neighbours. Multiple introductions of FAW with diverse genetic profiles and potentially similarly diverse origins were therefore likely the main reason for their wide spread detection across the Old World.

Clear evidence of multiple introductions can therefore be seen from the phylogenetic analysis, with unique introduction events likely underpinning the establishment of FAW populations in EA. (e.g., China) and SEA (e.g., Malaysia). In Australia, while the NT/NSW populations shared closer evolutionary relationships with SE Asian populations, the population origin for most of Qld and WA populations was unclear, and could be due to a sampling effect (e.g., this study could not source FAW populations from e.g., Indonesia, Cambodia, Thailand, Sri Lanka, Bangladesh, etc.). Some Burdekin and South Korean individuals appeared to share closer evolutionary lineage, although the factor leading to this detected pattern remained as yet unknown.

Population statistics and estimates of substructure

The basic population diversity statistics for each population are listed in Table 1. Nucleotide diversity (π) varied across a narrow range with the lowest (0.237) being from the Malaysian Kedah State population (i.e., reflected its lab-colony background), and the highest from the Malawian population (0.324), similar to that reported (³³; π = 0.279–0.329) in native and invasive populations based on the same set of 870 SNP loci. The high nucleotide diversity estimates from the SEA, Australia and South Korean populations likely reflected effects of limited (i.e., 870) highly polymorphic SNPs from non-coding genomic regions being used, and are comparable to the findings from Tay et al.³³ that used the same sets of SNP loci. We did not detect significant overall differences between the various invasive populations, and for the SE Asian and Australian populations the nucleotide diversity estimates were generally between 0.258 and 0.291. All invasive populations from Africa (Benin, Malawi, Uganda), SA/EA (India, China (CY, XP, XJ), South Korea), SEA (Malaysia (Johor, Kedah, Penang), Laos, Vietnam, Myanmar, Philippines), and Pacific (PNG)/Australia (WA (Kununurra), NT, QLD (Strathmore, Walkamin, Burdekin), NSW (Wee Waa)) showed higher average observed heterozygosity (Het_obs) than the average expected heterozygosity (Het_exp), with the highest Het_obs seen in the Malawian population as also reported previously³³. Within the Australian populations, highest average observed heterozygosity was seen in the Strathmore population, while in populations in SEA (i.e., Myanmar, Philippines) and the EA (i.e., South Korea) all showed similar average Het_obs. Negative average F_IS values for all populations were consistent with Het_obs being higher than Het_exp, and suggested populations were out-breeding (i.e., avoidance of matings between related individuals;⁵⁹). Similar to the previous findings³³, the lower Het_exp (i.e., Het_obs > Het_exp; see⁶⁰) could likewise indicate recent mixing of distinct populations from SEA that suggest multiple introductions (e.g.,^33,39 cf.^46,47,61,62; i.e., due to a recent bottleneck from a recent western Africa founder event).

The observed heterozygosity excess detected in all invasive range populations could be further explained as due to population sub-structure and isolation breaking through periodic migration. Significant numbers of loci (ca. 30%) were also shown to not be in Hardy–Weinberg equilibrium (HWE) especially for the Malaysian (i.e., Kedah), but also Australian (i.e., Wee Waa, NT, Kununurra), Chinese (e.g., XP), South Korean, and Malawian populations. Taken as a whole, genetic diversity results from this study therefore suggested that the invasive Asian (i.e., SA, SEA, EA) FAW populations exhibited signatures of recent mixing of previously separated populations. Simulated patterns of moth migration of various invasive FAW populations such as between Myanmar and China (e.g.,^41,42,55) and to Australia⁵⁴ are incompatible with the population genomic data, which suggests these were likely discrete and non-panmictic FAW populations with the most probable explanation being due to multiple origins of founding populations.

Genetic differentiation analysis

Estimates of pairwise genetic differentiation (F_ST) between populations varied significantly (Table 1) and extended to between populations within a country (e.g., Mackay vs. rest of Australia; Kedah vs. rest of Malaysia). Of interest are the pairwise estimates between different Australian FAW populations from Kununurra (Western Australia), Northern Territory, Queensland (Strathmore, Walkamin, Burdekin, Mackay) and New South Wales (Wee Waa) that represented the most recently reported invasive populations in this study, and predominantly showed significant differentiation amongst themselves (with the exception of the two Queensland populations of Mackay and partially for Walkamin) and with other SEA/SA/EA countries. The majority of non-significant population genetic differentiation estimates were in SEA where the presence of FAW was reported earlier, i.e., since 2018 (e.g.,^63,64 or as early as 2008^65,66; see also³³), while across Asia (e.g., China) since 2016 but also potentially pre-2014 (^16,67; see also³³).

Interestingly, significant genetic differentiation was observed between populations from Yunnan province in China and populations from Myanmar, Laos, and Vietnam. Penang and Johor (Malaysia) populations were not significantly differentiated from other SE Asian populations, nor with Ugandan and Malawian populations from east Africa. Individuals from Benin and Mackay (Queensland, Australia) showed non-significant genetic differentiation with all populations except with Kedah, and for Mackay also surprisingly with the Wee Waa population from New South Wales. The South Korean population exhibited significant genetic differentiation with SE Asian population except with Mackay, India and the Yuanjiang (YJ) population in Yunnan Province. Finally, the Kedah population, being one of the earliest collected samples from Malaysia and having been maintained as a laboratory population, showed strong differentiation with all populations (and lowest nucleotide diversity, π = 0.237; Table 2) further supporting unique, non-African, introduction events in SEA. Strong genetic differentiation suggested there was limited gene flow to breakdown sub-structure between populations, and the F_ST estimates from these invasive populations therefore failed to support a west-to-east spread pathway for the FAW. This observation instead suggested the widespread presence of genetically distinct FAW populations, likely due to independent introductions and therefore also highlighting likely biosecurity weaknesses especially in East Asia (e.g., China, South Korea) and SEA (e.g., Malaysia).

The genetic diversity of Australian populations identified surprisingly complex sub-structure patterns given the short time frame of population detections across different northern Australian regions. Significant genetic differentiation between, e.g., Kununurra (WA), Northern Territory (NT), Queensland (e.g., Strathmore, Burdekin), and Wee Waa (NSW) populations suggests these populations likely derived from separate establishment events. The WA Kununurra population was not significantly differentiated from the Johor State (Malaysia), India and the Cangyuan (CY) China populations, suggesting a potential south-eastern route from SA/SEA into north-western Australia. Contrasting this, Walkamin and Mackay populations showed non-significant genetic differentiation with the Madang (PNG) population, suggesting a potential second pathway for SEA individuals to arrive at the north-eastern region of Australia. Significant genetic differentiation between WA, NT, and Qld populations suggested that at least during the early stage of pest establishment in northern Australia, there was limited gene flow to homogenise the unique genetic background carried by these distinct individuals, some of which exhibited also distinct insecticide resistance profiles^48,49.

PCA

We selected specific populations to compare using Principal Component Analysis (PCA) as examples to support evidence of independent introductions, as seen from Fig. 3a between China (CY, YJ, XP) populations vs. Myanmar, in Fig. 3b (within Malaysian populations between those collected from Penang and Johor States vs. Kedah State), in Fig. 3c for between China and East Africa (e.g., Uganda, Malawi), and where Benin and India individuals that grouped with either China or east Africa; and in Fig. 3d between China, Malaysia (Kedah State), and Australia (NT, NSW)). Genetic variability between Australian populations (e.g., Strathmore (QLD) vs. NT and NSW) was also evident (Fig. 3d).

PCA also showed that differences existed between FAW populations from the Madang Province in PNG and with the Strathmore population from Qld (Fig. 3e). The SEA FAW populations from Lao PDR/Vietnam also exhibited diversity from the South Korean population (Fig. 3f), with the South Korean and Strathmore populations largely exhibiting similar diversity patterns, while the Madang population shared similarity with Laos and Vietnam populations. Plotting all SEA populations against China clearly showed that populations from SEA were distinct from the Chinese FAW populations (Fig. 3g), while in Australia, individuals from various populations shared similarity with both Chinese and SEA FAW. Despite the connectedness of the landscape between SEA and China, SEA largely appeared to have their own FAW populations, with FAW in SEA and in China differing in their genome compositions overall as shown via PCA.

PCA further enabled visualisation of genetic diversity amongst Australia FAW populations, suggesting that arrival and establishment of FAW likely involved separate introduction events that followed closely after each other and over a short timeframe. While it had been anticipated that the southward spread of FAW from SEA would necessarily lead to Australia FAW and PNG FAW to share similar genetic backgrounds, the Madang Province FAW population appeared to be different from the Strathmore (Qld) population, with the Madang population being more similar to Lao PDR/Vietnam populations, and the Strathmore population more similar to FAW from South Korea.

DivMigrate analysis

Directionality of gene flow between African, South Asia (Indian), East Asia (China) and SE Asian populations were predominantly from China to east African and SE Asian populations (e.g., Figs. 4a, b, S-1; see also Table 3), while movements of FAW in Laos and Vietnam (i.e., the Indochina region) were predominantly with other SEA countries (e.g., with Myanmar and East Africa; Figs. 4c, d, S-2; see also Table 3) but with no directional movements to the three Yunnan populations (CY, XP, YJ). Migration directionality with other SE Asian populations (e.g., Johor (JB; Fig. S-3) and Penang (PN, Fig. S-4)) showed that these two populations (but especially the Johor population) were predominantly source populations for Uganda, Malawi, Philippines, Vietnam, and PNG (Fig. S-3). Bidirectional migration between Myanmar and Laos PDR populations were also detected with the Johor population from Malaysia (Fig. S-3). When India was selected as the source population, bidirectional migration events were detected with Myanmar and with the Cangyuan (CY) populations (Fig. S-5) while unidirectional migration events from India to Uganda and Malawi and to Laos were detected, and the China Yuanjian (YJ) population showed unidirectional migration to India. Unidirectional migration events from CY and YJ populations to the PNG Madang population were detected, while bidirectional migration events between PNG and Myanmar, Laos PDR, Philippines, Vietnam, and with Uganda and Malawi were also detected (Fig. S-6). No migration events were detected between the West African Benin population and with the South Korean population.

Admixture analysis

Admixture analyses involving all Australian, Southeast Asian and South Korean populations from this study; and native populations from the Americas and Caribbean Islands, and invasive populations from Africa (Benin, Uganda, Malawi), India, and China³³, provided an overall complex picture of population structure that reflected the species’ likely introduction histories across its invasive ranges.

Admixture analysis that excluded New World, African and Indian populations identified four genetic clusters (i.e., K = 4) to best describe these invasive populations from SEA, and EA (i.e., China, South Korea), and Pacific/Australia (Fig. 5a). At K = 4, Australian populations from NT and NSW, YJ population from China, South Korean, and Malaysia’s Kedah population, each showed unique admixture patterns (i.e., some individuals from NT and NSW populations lacked cluster 3; most of YJ (but also some CY and XP) individuals lacked clusters 1 and 2; South Korean (e.g., MF individuals) lacked cluster 2; Malaysia’s Kedah population lacked evidence of admixture (i.e., reflecting its laboratory culture history) and was made up predominantly by individuals that belonged to cluster 4. Populations from China also differed from most populations from SEA due to the overall absence of genetic cluster 4. Taken as a whole, establishment of the FAW populations in China, Malaysia, vs. other SE Asian populations, and between Australian populations (e.g., NT/NSW cf. WA/Qld), likely involved individuals from diverse genetic background (i.e., multiple introductions). At K = 4, the majority of Australian populations appeared to contain genetic clusters similar to China (i.e., cluster 3) and to SEA (i.e., cluster 2).

Overall admixture patterns at best K = 4 in China and SEA remained unchanged when analysed together with African and Indian individuals (Fig. 5b; excluded Australia). Benin individuals were either similar to China or to SEA, while eastern African populations (e.g., Uganda, Malawi) were similar to Southeast Asian populations from e.g., Vietnam, Laos, and is in agreement with the phylogenetic inference (Fig. 3) that identified these African individuals as having loci that were derived from Southeast Asian populations.

Genome-wide SNP loci demonstrated that invasive FAW populations from SEA and Australia exhibited admixed genomic signatures similar to that observed in other invasive populations^33,34. While the current invasive populations in Africa and Asia likely arrived already as hybrids as suggested by Yainna et al.⁶⁸, the Malaysia Kedah State population was potentially established by offspring of a non-admixed female. Distinct admixture patterns in Malaysian FAW populations between Kedah and Johor/Penang states therefore suggested that establishment of these populations was likely as separate introduction events. As reported also in Tay et al.³³, the Chinese YJ population appeared to have admixed signature that differed from XP and CY populations, and suggested that the YJ population could have a different introduction history than the XP and CY populations. Similar multiple genetic signatures based on lesser nuclear markers by Jiang et al.³⁹ also supported likely multiple introductions of China Yunnan populations.

Source: Ecology - nature.com