The Spodoptera frugiperda Host Strains: What They Are and Why They Matter for Understanding and Controlling This Global Agricultural Pest

Abstract

Fall armyworm (Spodoptera frugiperda [J.E. Smith]) is a moth native to the Western Hemisphere where it is a major pest of corn (maize) and multiple other crops. It is now a global threat with its recent discovery and dissemination in the Eastern Hemisphere. Its broad host range is in part due to two subpopulations denoted as ‘host strains’ that differ in host plant specificity. Therefore, identifying the strains present in a location is critical to assessing what crops are at risk of infestation. However, much remains uncertain about how the strains differ and even on the fundamental issue of how they are identified. Complicating factors include the host strains are morphologically indistinguishable, the defining behavior of the strains (host plant specificity) is variable, and the existence of significant differences between geographical populations and laboratory colonies that are independent of strain identity. These factors contribute to substantial disagreements in the literature on presumptive strain differences. This paper presents a summary of strain characteristics and suggests the criteria that should be met before concluding a trait is ‘strain-specific’. The intent is to facilitate a common understanding of what the strains represent and to develop a more consistent experimental framework for studies on strain phenotypes. Evidence is summarized that supports a primary role for Z-linked genes in strain identity, which has potential implications for genetic approaches to define the strains, and we discuss the possibility that the strains arose from allopatric (rather than sympatric) speciation processes.

The noctuid moth Spodoptera frugiperda (JE Smith) (Lepidoptera: Noctuidae) is commonly known as fall armyworm and is native to the Americas where it is a significant agricultural pest. Spodoptera frugiperda was recently found in western Africa (2016), eastern and southern Africa (2017), India (2018), southeastern Asia (2018–2019), and most recently in Australia (2020), with the potential for devastating economic impact to that hemisphere estimated to be in the billions USD (Day et al. 2017, Cook et al. 2021, Wu et al. 2021). One frequently cited concern is that the species has an exceptionally broad host range having initially been reported on more than 80 plant species (Luginbill 1928), which has since been expanded to approximately 350 plant types (Montezano et al. 2018). However, infestations of economic importance in the Western Hemisphere are limited to a few crops that are differentially impacted by two populations historically described as host strains (Pashley 1986). The strains generally fulfill the criteria required for designation as ‘host races’, which includes the use of different hosts, sympatric distribution, genetic differences at multiple loci, and evidence of appreciable gene flow (Dres and Mallet 2002). As such they appear to be at an intermediate stage of speciation. The strains were initially named for the host plants they were first collected from, corn and rice, however, subsequent studies found that rice is an infrequent host with inconsistent strain specificity (Juárez et al. 2012, 2014). To avoid exaggerating the association of S. frugiperda with rice we have taken to using the terms C-strain and R-strain and will do so for the rest of this manuscript.

The differential association of the two strains with different crops means that assessments of potential economic damage by S. frugiperda depends on whether one or both strains are present in a region. This is made difficult by the inability to find consistent morphological differences between the strains, restricting identification to molecular methods. This limitation has greatly complicated efforts to describe or even define the two strains, particularly with respect to their behavior in the field, and has contributed to contradictory findings in the literature on various aspects of host strain biology. This review will summarize recent advances in our knowledge of the host strains, with a focus on how the strains are genetically defined (and the limits to that definition), what little we know about interstrain hybridization, and a discussion of what processes might be driving and maintaining strain divergence. Many of the same topics were covered in an excellent 2010 review (Groot et al. 2010), and we will try to not emphasize the information presented there but rather focus on relevant research since.

Defining the Host Strains

Dorothy (Prowell) Pashley and colleagues were the first to identify the host strains. In a survey of S. frugiperda populations in the Caribbean Basin, statistically significant differences in the distribution of electrophoretic allozymes were found between collections from different host plants (Pashley et al. 1985, Pashley 1986). These findings were confirmed and expanded by the development of DNA markers that displayed polymorphisms in both the mitochondrial and nuclear genomes that correlated with strain host preferences. To our knowledge, the only evidence of a morphological distinction between strains comes from studies of S. frugiperda in central Colombia where differences in both wing size and shape were reported (Cañas-Hoyos et al. 2014, 2016). However, a similar study using Florida populations found that variations in wing morphology are primarily influenced by diet, with the direct influence of strain genetics likely to be relatively minor (Nagoshi et al. 2020a). When S. frugiperda specimens from a common host plant were assayed, no significant differences in wing morphology were observed.

Without distinctive morphological features, the definition of the strains is limited to host plant association and molecular markers. However, the correspondence of these metrics is not absolute, which adds uncertainty to strain identification. For example, collections from a single host will frequently have some specimens with the opposing genetic markers and it is not uncommon to find collections where a majority of specimens carry markers opposite that predicted by the host plant (Prowell et al. 2004, Juárez et al. 2014, Murúa et al. 2015a). Such disagreements could be due to variability in the choice of host plant or that the genetic markers are polymorphic within each strain. Because of this uncertainty, the term ‘specificity’ in this context should be used loosely, with marker specificity being more of a bias than absolute, and strain-specific host usage appearing to be more of a preference than a physiological necessity.

Even with these variabilities, field studies using one or more molecular markers identified a consistent association of a majority C-strain in corn, sorghum, and cotton, while the R-strain predominates in alfalfa, pasture grasses, and millet (Pashley 1988, Nagoshi et al. 2007, Juárez et al. 2012, Murúa et al. 2015b). This pattern of strain specificity was consistent in surveys of both Americas, indicating the strains are present throughout the hemisphere. In contrast, S. frugiperda is a sporadic pest of peanuts where in collections from Florida a near equal number of the two strains were found (Nagoshi et al. 2020a). It is similarly sporadic in rice where the strain composition is highly variable but shows a tendency to the R-strain (Nagoshi et al. 2021b). These observations indicate variable host preferences, with relatively strong specificity for the most commonly used host plants and reduced or no specificity for host plants where infestations are more sporadic.

Molecular Markers for Strain Identity

The number of molecular markers available to identify strains is small and generally shows similar levels of correspondence with the plant host association phenotype. Because genetic markers are currently the primary method to identify strains, it is important to recognize what is available and their general characteristics. For this paper, we’ve adopted the nomenclature convention of Haenniger et al. (2020) where the strain (C or R) is followed by a subscript indicating the marker upon which that identification is based.

Esterase-3 Allozyme

Pashley et al. (1986) compared S. frugiperda collections from multiple host plants and locations for the expression of 16 polymorphic enzymes. Five enzyme loci (adenylate kinase, esterase 3, hydroxybutyrate dehydrogenase, mannose-6-phosphate isomerase, and peptidase 1 and 2) were found to have different patterns of electrophoretic variants in collections from corn compared to those from Bermuda grass and rice. The most distinctive differences were observed with the esterase 3 locus (est3), which has been mapped to the Z chromosome (Lepidoptera has a ZW/ZZ sex determination mechanism where individuals with two Z chromosomes are male while ZW individuals are female). The est3 locus was deemed the most specific of the allozymes and was used in a survey of the Caribbean basin giving results comparable to DNA-based genetic markers (Prowell et al. 2004). In collections from Florida, Louisiana, and Puerto Rico, about 80% of the larvae collected from corn host plants showed the C_est3 pattern compared to only 5% in specimens from pasture grasses (Prowell et al. 2004).

Amplified Fragment Length Polymorphisms (AFLP)

Ten AFLP loci were identified as showing evidence of strain specificity (McMichael and Prowell 1999), with eight used to identify strains in collections from the U.S., South America, and the Caribbean (Prowell et al. 2004). There was a strong correlation between the AFLP-based strain designation with plant host. Out of 168 specimens, only 12 did not agree, with most of these arising from a single collection from corn in French Guiana where all nine specimens expressed a R_AFLP pattern. Other AFLP studies performed by different laboratories (Clark et al. 2007, Juárez et al. 2012, Lobo-Hernandez and Saldamando-Benjumea 2012), and a RAPD (Rapid Amplified Polymorphic DNA) study (Prowell 1998) failed to find strain-specific patterns, suggesting that such polymorphisms may be uncommon. However, an experiment using a related technique, Restriction Fragment Length Polymorphism analysis (RFLP), did identify probes that could distinguish an R-strain colony from other laboratory lines as well as one field collection from corn (Lu et al. 1992).

Fall Armyworm R-Strain (FR) Repetitive Elements

The FR element is 188 bp in length and organized in large tandem arrays reportedly found only in the R-strain where it is estimated to make up 0.05% of the nuclear genome (Lu et al. 1994). FR maps to the Z and W chromosomes and Southern blot analysis showed a 120-fold reduction in copy number in males compared to females, suggesting a particularly high abundance on the W chromosome of the R-strain. To simplify the detection of FR, a PCR-based method was developed that eliminated the need for Southern blotting (Nagoshi and Meagher 2003a, b; Nagoshi et al. 2007). This method found that FR elements were present in a subset of the C-strain specimens tested but these appeared to differ in both copy number and organization compared to those found in the R-strain (Nagoshi and Meagher 2003b). The PCR-base methodology could still distinguish C_FR from R_FR, but this identification was sensitive to PCR conditions. Comparisons with other strain markers indicate that the PCR-based FR method can substantially underestimate the R-strain population. In one study, only 40% of the specimens identified as R-strain by other methods showed the R_FR pattern (Nagoshi et al. 2008a).

Mitochondrial Genome (mtDNA) Haplotypes

Strain-specific restriction site polymorphisms in mtDNA were first identified from a segment that included the nd1 gene (Pashley 1989, Pashley and Ke 1992). An MspI strain-specific polymorphism in mtDNA was also reported (Lu and Adang 1996) that was subsequently mapped to the gene encoding the cytochrome oxidase subunit I (COI) gene (Levy et al. 2002). PCR-based detection of the nd1 marker has since been developed (Queiroz et al. 2016), as well as strain-specific polymorphisms identified in other mtDNA loci such as cytb and COIII (Otim et al. 2018). These indicate that strain-specific polymorphisms are dispersed across the mitochondrial chromosome. The advantage of using the COI gene as the mitochondrial marker for strain identification is the breadth of information that can be obtained. The 5’ portion of the gene is frequently used for DNA barcoding and can distinguish S. frugiperda from related Spodoptera species (Nagoshi et al. 2011). The 3’ segment of the gene carries polymorphisms that define haplotypes that can distinguish between North American S. frugiperda overwintering populations (Nagoshi et al. 2007, 2009). Single nucleotide polymorphisms (SNPs) in both segments can identify strains and include those that alter restriction sites convenient for RFLP analysis (Nagoshi et al. 2006). The COI gene is probably the most widely used S. frugiperda strain marker and so there is extensive evidence supporting its capacity to identify strains throughout the Americas.

The Triosephosphate Isomerase Gene (Tpi)

The Tpi gene has been extensively studied in multiple organisms to examine the evolution and function of introns (Gilbert et al. 1986, Logsdon et al. 1995). This gene was targeted for analysis because it maps to the Z-chromosome in Lepidoptera (Yasukochi et al. 2006), and Z-linked genes were proposed to play a disproportionately large role in S. frugiperda strain identity (Prowell 1998). This physical linkage to functions that define the strains makes Tpi a good candidate to carry strain-specific polymorphisms. Ten strain-specific SNPs were initially identified with one (gTpi183) contained within an MspI restriction site (Fig. 1). This led to a PCR-based method using MspI restriction enzyme digestion to rapidly identify strain without the need for DNA sequencing (Nagoshi 2012, Murúa et al. 2015a). Strain identification based on the single gTpi183 SNP was shown to be comparable to that achieved with multiple Tpi markers (Nagoshi 2012).

Real-Time PCR of SNPs

A new methodology was recently reported that uses real-time PCR-based TaqMan assays to identify strains (Tessnow et al. 2021). The assays differentiate four strain-specific SNPs, three that mapped to the Z-chromosome and one to chromosome 16, giving results comparable to the COI and Tpi markers. The method has so far been tested on North American populations and provides a new molecular tool for characterizing S. frugiperda field specimens.

Comparisons Between COI and Tpi

A total of 26 collections from multiple locations and times in the Western Hemisphere were analyzed with both the COI and Tpi markers (Table 1). Each collection had a sample size of at least 25 specimens with a total of 1,439 specimens tested (1,071, 300, and 68 from C-strain, R-stain, and peanut habitats, respectively). There was a strong positive correlation between the C_COI and C_TPI frequencies, confirming their capacity to identify strains (P < 0.0001, r = 0.9625). However, the Tpi marker detected a higher percentage of C-strain than COI in 24 of the 26 collections, with statistically significant differences observed in the data from C-strain preferred habitats but not in collections from R-strain hosts (Fig. 2). There were five collections from corn considered anomalous because the C-strain proportion based on C_COI was at or below 50%. In all but one of these collections the C_TPI group was in the majority, consistent with the expectations from the host plant. Comparable results were reported using a different set of strain markers where strain-specific polymorphisms in the mitochondrial nd1 gene identified fewer C-strain than the est3 or AFLP strain-specific alleles, both of which map to the nuclear genome (Prowell et al. 2004).

One consequence of this difference between COI and Tpi is that a substantial fraction of the S. frugiperda population carries the marker combination R_COIC_TPI that disagrees with respect to strain identity. We will refer to this discordant configuration as RC when describing mitochondrial and nuclear double marker combinations with the reciprocal CR describing the combination of C-strain mitochondria with R-strain nuclear markers and CC and RR denoting concordant C-strain and R-strain identity, respectively. Prowell et al. (2004) offered two possible causes for the discordant combinations. The first is that one or more of the markers did not achieve fixation at the time of strain divergence and as such may not be completely accurate in their strain identification. The second explanation is the intriguing possibility of hybridization between strains to be discussed below.

Cross-Hybridization Between Strains (Interstrain Mating)

The largely sympatric distribution of the two S. frugiperda strains would seem to necessitate the existence of biological limitations that restricts the generation, viability, or fertility of interstrain hybrids. Work in this area before 2010 was previously described (Groot et al. 2010). Here we will focus on more recent results and our interpretation of the current state of knowledge. This will be divided into laboratory studies that test for specific mechanisms of reproductive isolation and field studies that attempt to describe and quantify the degree of cross-hybridization in wild populations using genetic markers.

Laboratory Studies

Reproductive Incompatibility

A summary of the results from six mating studies is presented in Table 2. Two studies reported directionality in interstrain hybridization such that crosses between R-strain females and C-strain males (RxC cross) had approximately normal fertility while the reciprocal cross (CxR) showed reduced fertility (Table 2, af). However, these results were not replicated by other studies where similar levels of fertility resulted from both RxC and CxR crosses (Table 2, bcd). The studies differed in the sources for the S. frugiperda tested with respect to the age of the colonies and their geographical origins, which may explain some of this discrepancy. Quisenberry (1991) found evidence that colony age, particularly that of R-strain females, impacted the number of egg masses produced with fewer viable eggs found with increasing generations in colony.

It was previously noted that the studies are in general agreement that one of the products of cross-hybridization had reduced fertility (Groot et al. 2010). The hybrid females (RC) arising from RxC crosses were less likely to productively mate than the reciprocal CR hybrid females produced by CxR crosses (Table 2). In contrast, no reduction in fertility was typically observed with hybrid RC or CR males. The gene(s) associated with this reduced RC female fertility was mapped to the Z chromosome (Kost et al. 2016). The authors reasoned that since both C-strain and RC females have a single C-strain Z-chromosome and the former are fully fertile that the reduced RC fertility probably results from an incompatibility between the C-strain Z chromosome and one or more R-strain cytoplasmic or autosomal factors.

Strain Differences in Female Pheromones

Pheromones produced by female moths are used to attract males of the same species and so are an obvious candidate for imposing strain-specific mating behavior. Multiple studies have identified strain differences in the chemical composition of S. frugiperda female pheromones, though the exact compositions often differ between laboratories (reviewed in Groot et al. 2016). Field studies to date indicate that the observed differences in pheromone composition and the responses they elicit most likely reflect variations between geographical populations that are independent of strain (Unbehend et al. 2014). At present, there is no pheromone blend that can differentially attract the strains in field settings.

Strain Differences in Diurnal Mating Patterns

Perhaps the best characterized potential reproductive barrier between strains is in the timing of nocturnal reproductive activity. In laboratory studies, the timing of first copulation occurred early in the dark cycle (scotophase) for C-strain moths from Louisiana (one hour) and Florida (3.5 hr) as compared to eight hours for R-strain moths from Louisiana (Pashley et al. 1992). Comparable time differences were obtained with colonies from Florida where first copulations occurred on average 4.9 hr into scotophase for the C-strain and 8.1 hr for the R-strain (Schöfl et al. 2009). Calling behavior showed a similar time shift where C-strain moths initiated the activity well before the R-strain (Pashley et al. 1992, Schöfl et al. 2009). A subsequent study confirmed this temporal pattern and demonstrated in choice tests (where an individual could choose to mate with partners of either strain) evidence for a preference for intrastrain mating (Schöfl et al. 2011). However, the overall differences between intrastrain and interstrain pairings were small and according to the authors ‘…raises the question of whether the temporal differences in mating activity alone suffice to explain the observed patterns of nonrandom mating’ (Schöfl et al. 2011). An important consequent finding was that this mating phenotype mapped to an autosomal location, where a candidate gene (vrille) contained DNA polymorphisms that were strain-specific in the colonies tested (Hänniger et al. 2017). Furthermore, the same study identified strain-specific SNPs that map to S. frugiperda homologs of several circadian rhythm genes (of which vrille is one) located on multiple chromosomes, though the comparisons were limited to two colonies derived from Florida (R-strain) and Puerto Rico (C-strain) populations. Confirmation that these markers are consistently strain-specific throughout the Western Hemisphere would be strong evidence that the linked circadian rhythm genes have an important role in strain divergence.

However, studies of S. frugiperda from Colombia failed to find statistically significant differences in the timing of mating activity between strains (Velásquez-Vélez et al. 2011, Saldamando-Benjumea et al. 2014). Furthermore, comparisons of different S. frugiperda lines generated from multiple corn sites in Mexico found copulation onset time ranging from less than an hour into scotophase to over four hours, with no apparent impact on mating success in crosses between lines (Rojas et al. 2018). These findings bring into question whether the differences in the timing of mating activity are generally strain-specific and, even if so, whether they significantly restrict mating between strains.

Summary

Overall, the laboratory studies describe an inconsistent picture of possible mating incompatibilities between strains. While there are intriguing indications of reproductive barriers, the universality and effectiveness of these phenotypes need to be confirmed as well as whether they are consistently observed in the field. Furthermore, the only observations of reproductive incompatibility that appear substantial enough to significantly limit the impact of interstrain hybrids are the reduced fertility of CxR mating combined with reduced fertility of RC hybrid females, though the former was only observed in a subset of studies (Table 2). In this scenario, most interstrain hybrids will come from RxC crosses but the resultant RC female progeny will have reduced fertility, thereby limiting the impact of such cross-hybridization (Groot et al. 2016).

Field Surveys

Combining Mitochondrial and Nuclear Markers

The identification of interstrain hybrids in the field is made difficult by the absence of distinguishing morphological characteristics, which limits the detection of hybrids to extrapolations from molecular markers. One method is to use a combination of mitochondrial and nuclear strain markers, with interstrain hybrids indicated by disagreement between the two. The first such study of this type was by Prowell et al. (2004) who applied a combination of the mitochondrial nd1 locus together with nuclear markers based on AFLP and est3. Discordance between markers was found in about 16% of the population. A statistically significant higher frequency (66%) of these putative hybrids were of the RC type (R-strain mitochondrial cytotype with C-strain nuclear marker) compared to CR, a bias consistent with what would be expected if RxC crosses were more frequent or productive than the CxR reciprocal.

Similar results were obtained in other field surveys using different marker combinations, including the COI with FR markers (Nagoshi and Meagher 2003a,b) and most recently COI with Tpi (Nagoshi 2010, 2012; Nagoshi et al. 2017a). The latter combination was used to analyze the Table 1 collections and the results are summarized in Fig. 3. The possible COI-Tpi configurations and a subset of the crosses from which they can be derived are depicted, with the maternally inherited mitochondrial cytotype indicated by shading and the Z-linked Tpi haplotype identified by font color (Fig. 3A). The W-chromosome does not carry Tpi and has no known role in strain identity. The frequencies of the COI-Tpi configurations found in our Western Hemisphere collections are shown immediately below the relevant crosses, with collections from C-strain and R-strain hosts treated separately to illustrate differences in host preferences (Fig. 3B). Despite large differences in the percentages of the two parental host strains (C_COIC_TPI and R_COIR_TPI) in the two host types, the total frequency of discordants (C_COIR_TPI + R_COIC_TPI) was similar in both habitats (17% in C-strain hosts, 14% in R-strain hosts), with the R-strain mitochondrial cytotype (R_COIC_TPI) predominating, consistent with the results of Prowell et al. (2004) (Fig. 4A).

Z-Chromosome Heterozygosity

A second method to identify hybrids is the use of co-dominant strain markers, with the simultaneous presence of both strain indicators in the same specimen diagnostic of a hybrid. For example, the expression of est3 allozymes for both strains in a single specimen indicates a male heterozygous for a Z-chromosome from each strain (Prowell et al. 2004). A modification of this strategy for the Tpi markers was developed based on the co-dominance of DNA sequence chromatography that allows detection of SNP heterozygotes through overlapping chromatographs (Fig. 1). Analysis of the strain diagnostic gTpi183 SNP on the Western Hemisphere collections identified an average of 9% of specimens from C-strain hosts and 10% from R-strain hosts as interstrain hybrids, with the majority carrying the R-strain cytotype produced from crosses between R-strain females and C-strain males (Fig. 4B).

Near the strain diagnostic gTpi183 site are two SNPs (gTpi192 and gTpi198) that are polymorphic within both the C_TPI and R_TPI populations (i.e., are nonstrain-specific; Fig. 1). This configuration provides a means to directly compare the relative frequencies of interstrain and intrastrain hybridization as heterozygosity at gTpi183 can only occur by a C_TPI X R_TPI cross (in either direction) while gTpi192 (or gTpi198) can become heterozygous by crosses both within and between strains. Because of this difference, reproductive barriers between strains will reduce heterozygosity at gTpi183 to a much greater degree than at gTpi192. Wright’s local inbreeding coefficient (F) provides a simple metric for quantifying the degree of observed heterozygosity (H_o) relative to what would be expected (H_e) based on allele frequencies. Since F = (H_e − H_o)/H_e, an F equal to 0 indicates a state of equilibrium where H_e = H_o, with values becoming increasingly positive if heterozygosity is suppressed (H_e > H_o). We previously demonstrated for collections from six locations in North America and the Caribbean that the mean F value for the strain-specific gTpi183 SNP was significantly higher (0.64) than for the nonspecific gTpi192 site (0.05, Fig. 4C). This indicates that hybridization between strains was much less than expected based on allele frequencies, consistent with a 4-fold reduction in the frequency of productive interstrain mating compared to matings within a strain (Nagoshi et al. 2017a). The data strongly suggest either that field populations are displaying assortative mating behavior that significantly favors within strain over between strain hybridization or that there is strong selection against hybrid progeny that has yet to be observed.

Field Studies in Africa

Genetic analysis of S. frugiperda that recently became established in Africa identified additional complications. The dominant COI and Tpi marker combinations found in Africa are C_COIC_TPI and R_COIC_TPI that together make up over 98% of the S. frugiperda assayed to date from the continent (Nagoshi et al. 2019a). Infestations in Africa have for the most part been limited to corn and sorghum, indicating that both S. frugiperda populations are behaving like the C-strain with respect to host plant use (Nagoshi et al. 2021b). The females of both genotypes were found to be fertile, but express pheromones with altered composition and have diurnal mating times different from that described for either strain (Haenniger et al. 2020). These findings are consistent with the African R_COIC_TPI population having an interstrain hybrid identity but are difficult to reconcile with those laboratory studies indicating substantial reductions in the fertility of hybrid RC females (Table 2). Either the reproductive incompatibility responsible for the reduced fertility of RC females is not a universal trait or it can be overcome by environmental conditions.

Summary

The laboratory and field studies present a confusing picture of interstrain hybrids that underscores our incomplete understanding of their biology. The majority of laboratory data indicate that cross-hybridization between strains can occur in both directions, but field studies consistently find that marker combinations associated with RxC crosses are more abundant than those from the reciprocal CxR mating. There seems to be general agreement that hybridization between strains can occur with at least a subset of the hybrid progeny exhibiting near-normal fertility. This is consistent with field surveys showing that marker combinations indicative of such hybridization are routinely found at frequencies ranging from 10 to 20% of the total population, which would seem to represent a significant conduit for genetic introgression between sympatric populations with the potential to substantially compromise strain integrity.

Explanations for the Hybrid Marker Configurations and Their Distributions

The preponderance of the R_COIC_TPI configuration over C_COIR_TPI has been observed in both Americas and with different marker sets, indicating it is a universal characteristic of the species (Fig. 4A). The R-strain mitochondrial cytotype also predominates with C_TPI/R_TPI heterozygotes (H_TPI) where we consistently observe higher R_COIH_TPI than C_COIH_TPI frequencies (Fig. 4B). Here we briefly consider the plausibility of previously suggested explanations for the presence and distribution of discordant strain marker configurations that do not entail assortative mating.

Strain Polymorphism

Discordant configurations naturally arise if the strain markers are polymorphic within a strain, as would occur if they did not achieve complete fixation at the time of strain divergence (Prowell et al. 2004). A simple example of how this might work is described in Fig. 5A where the R-strain is modeled as being polymorphic for the Tpi marker. Matings within the R-strain can generate both R_COIC_TPI and R_COIH_TPI classes. In this scenario both these progeny types are biologically of the R-strain and so should be preferentially found in R-strain habitats like the R_COIR_TPI genotype. The current data does not support this hypothesis. The R_COIC_TPI and R_COIH_TP configurations are both observed in C-strain habitats at frequencies comparable to R-strain collections (Fig. 4A and B).

Habitat Differences in Allele Frequencies

A second explanation is based on allele frequencies associated with the strain-preferred habitats. Strain heterogeneity is typically higher in C-strain than R-strain locations, which should provide more frequent opportunities for the generation of hybrids (prediction 1, Fig. 5B) as was previously suggested (Prowell et al. 2004). In addition, if mating is random then females in C-strain habitats (including those of the R-strain) are more likely to mate with the majority C-strain male population. This means that RC hybrids should outnumber their CR counterparts, with the opposite expected in R-strain habitats where R-strain males predominate (Prediction 2, Fig. 5B). The results for prediction 1 are mixed. Prowell et al. (2004) found a significant majority of their discordant genotypes were associated with C-strain hosts. In contrast, we found that the discordant genotypes made up similar proportions of the collections from both habitat types (Fig. 4A). Our results also disagree with prediction 2 as RC hybrids always outnumbered CR hybrids regardless of habitat and plant host (Fig. 4B).

Summary

Explanations based on random mating, marker polymorphism, and allele frequencies are not consistent with the observed distribution of marker combinations in the field. Instead, we believe the observed RC > CR asymmetry is a consequence of strain behavior either in the form of assortative mating that favors RxC over CxR crosses or an unspecified fitness advantage or RC over CR progeny. At this point, we cannot distinguish between these possibilities.

What is Driving Strain-Specific Host Plant Preferences?

An important model for explaining insect host choice is the preference-performance hypothesis, which posits that females evolve to lay eggs on hosts most suitable for larval development (Jaenike 1978, reviewed in Gripenberg et al. 2010). Here we discuss how two important factors that determine the suitability of a host for larval development and viability, food quality and natural enemy pressure, might influence the host ranges of the two strains.

Strain Development on Different Plant Hosts

Whether the strains differ in the capacity of larvae to develop on different diets has been tested in multiple studies (reviewed in Groot et al. 2010, Meagher and Nagoshi 2012). The results are inconsistent and often conflicting as summarized in Table 3. Most studies showed no statistically significant strain difference in larval performance and the many disagreements between groups suggest that the phenotypic differences observed were likely independent of strain and may instead reflect the genetic variation within the species or the unpredictable consequences of prolonged laboratory culturing.

The choice of R-strain host material may also be a confounding factor. There is growing evidence that the plant hosts categorized as R-strain and frequently used in feeding studies can vary substantially in their capacity to support S. frugiperda larval development. For example, an R-strain colony derived from Louisiana populations outperformed C-strain larvae (from GA) when raised on bermudagrass (Cynodon dactylon (L.) Persoon) varieties Tifton 292 and Coastal, but not when fed the Grazer or OSU71x6–7 lines (Table 3, a). Similar differences were observed in studies with the perennial pasture grasses Miscanthus × giganteus Greef and Deuter ex Hodkinson and Renvoize and switchgrass, Panicum virgatum L. In comparisons of colonies derived from Florida populations, R-strain larvae were larger and developed faster than their C-strain counterparts on Miscanthus but showed no significant differences when raised on switchgrass (Prasifka et al. 2009). A possible physiological basis for these diverse plant responses comes from observations of strain dependent variations in the larval salivary proteins after herbivory, which can elicit different defense responses depending on plant variety and so potentially influence host plant associations under field conditions (Acevedo et al. 2017). In additions, switchgrass varieties showed substantial variations in their biochemical responses to S. frugiperda herbivory that corresponded to differences in larval growth rates (Palmer et al. 2019). These results suggest that plant hosts associated with the R-strain can differ significantly in their susceptibility to one or both strains, which could explain at least some of the inconsistencies in the feeding studies comparing strain performance.

Summary

While it appears that host-dependent differences in larval development can be detected between some colonies, whether these represent strain differences remains uncertain. Pashley et al. (1995) concluded that the C-strain was less affected by host than the R-strain, though they also noted that this ran counter to field observations that showed a strong association of the C-strain in corn while the R-strain could be found in substantial numbers in both corn and pasture settings. We believe that if strain differences in host utilization exist, they appear to be relatively modest or limited, perhaps explaining their inconsistent detection under laboratory conditions. This is in general agreement with the summation by Groot et al. (2010): ‘It thus seems unlikely that the consistent host differentiation found in field samples from a wide range of populations can be explained by differential host-plant adaptations of the larvae’.

Natural Enemy Pressure

Negative selection pressure from natural enemies has long been considered a potential driver of host range specialization in phytophagous insects (reviewed in Bernays 1991), and was specifically implicated as contributing to S. frugiperda strain divergence (Pashley et al. 1995). However, we know of few studies directly addressing the role of natural enemies in the evolution of the strains. In field collected specimens from Louisiana, larval mortality caused by parasitoids and pathogens was modestly greater in pastures (53%) than in corn (36%), with unpublished observations of much higher predator populations in pastures (Pashley et al. 1995). In another study, laboratory-reared C_COI and R_COI larvae were released in corn and pasture habitats in northern Florida then recovered and examined in the laboratory for parasitism (Hay-Roe et al. 2016). While there was no significant difference in parasitism rates between strains, larval recovery rates in the pasture were much poorer than in corn fields, an observation consistent with higher predation in pasture habitats.

Clearly our current knowledge on how natural enemy pressure impacts the distribution of the two S. frugiperda strains is very limited despite its potential influence on strain divergence. To facilitate further research in this area a summary of natural enemy species reported to be associated with S. frugiperda is listed in Table 4 with a more detailed description provided in Supp Table 1 (online only). The higher numbers of natural enemy species listed for C-strain habitats most likely reflects more intense surveying of corn agroecosystems because of their economic importance and indicates a need for additional studies in turf and pasture habitats.

Female Oviposition or Larval Behavior?

While the preference-performance hypothesis assumes host specificity is dictated primarily by oviposition choice, there is growing evidence in S. frugiperda that larval behavior must also have a significant role. Spodoptera frugiperda females frequently lay eggs on nonhost plants and even nonplant structures (Luginbill 1928, Thomson and All 1982, Prowell et al. 2004), with the C-strain appearing to be more indiscriminate than the R-strain (Meagher et al. 2011). Furthermore, there is currently little evidence in S. frugiperda for a correlation between oviposition preference and larval performance. No difference in ovipositional preference was observed for modern corn [Zea mays ssp. mays L. (Poaceae)] versus its ancestor Balsas teosinte (Zea mays ssp. parviglumis Iltis & Doebley), despite four-fold higher parasitism and three-fold higher predation mortalities on the latter (Bernal et al. 2015). Similarly, in comparisons of corn varieties that differed significantly in their resistance to S. frugiperda, females displayed no ovipositional preference for the cultivar most supportive of larval development and, contrary to expectations, seemed to prefer the most resistant host (Rojas et al. 2018).

However, given the limitations in larval mobility and the distances that can separate strain-preferred habitats, it seems necessary for female oviposition behavior to have a significant role in determining the field distribution of the two strains. It may be the case that females first localize to habitats acceptable for larval development but optimize egg survival in the final determination of where to lay eggs, with subsequent location to specific hosts dependent on larval dispersal (Prowell et al. 2004, Bernal et al. 2015, Rojas et al. 2018). Further refinement of larval distribution within a plant could occur through cannibalism, which can mitigate predator mortality by reducing larval density and thereby the attraction and aggregation of predators (Chapman et al. 2000). In addition, intraguild competition studies demonstrate that when cannibalistic behavior is directed to competing species (i.e., predatory behavior), S. frugiperda has a competitive advantage against at least some other corn pests (Bentivenha et al. 2016). This same aggressive behavior could presumably provide some protection against natural enemies. Based on these observations, if natural enemies are a major factor driving strain divergence, then we should expect to see strain differences in larval dispersal and cannibalism, with more aggressive behaviors found in the strain associated with habitats with greater parasitoid and predator pressure.

Defining Strain-Specificity

The above survey of S. frugiperda is striking for the number of reported strain differences that cannot be consistently replicated in the laboratory. It appears that the genetic diversity of the species and the apparent sensitivity of many S. frugiperda behaviors to laboratory rearing conditions are confounding factors that must be accounted for when comparing strains. One example of this comes from studies using an R-strain colony kept in culture for greater than 25 generations (Nagoshi 2011). Lines from this colony were subjected to selection based on developmental time for two generations, and this resulted in significant differences in larval weight and duration. Larvae selected for fast development were 50% heavier after 14 d and initiated pupation approximately two days earlier than their ‘slow’ selected counterparts. This indicates that even in highly inbred colonies of a single strain there remains substantial phenotypic variation in an important developmental metric. We, therefore, believe that adequate quantification of the variation of a trait or marker within each strain (with sampling that is representative of the general population) is needed before making conclusions about observed differences between strains. ‘Strain differences’ based on comparisons of a few colonies, or in the case of genome studies a few individuals, may be suggestive but should not be assumed to be generalizable. At this time, we believe that the standard for being strain-specific has only been met for genetic markers mapping to the mitochondria or the Z-chromosome, and to the phenotype of host plant preference.

Speculations on Strain Evolution

The Z Chromosome is Critical for Strain Divergence

Sperling (1994) documented that a disproportionate number of traits distinguishing closely related Lepidopteran species are sex-linked, consistent with the proposition that early speciation is predominantly driven by genes on the sex chromosome (see also Haldane 1922, Coyne and Orr 1989, Oppenheim and Hopper 2010). Prowell (1998) proposed that the S. frugiperda strains are an example of this phenomenon based on a bias for strain-specific traits and markers to map to the Z chromosome. Subsequent descriptions of the Z-linked FR element and Tpi as being strain-specific were consistent with this proposal.

With evidence for significant levels of interstrain hybridization in field populations, the localization of strain-determining factors to the Z chromosome becomes more appealing both to explain the continued linkage of Tpi, est3, and FR to strain-specific phenotypes as well as to limit the destabilizing impact of interstrain hybrids. If strain identity were dependent on autosomal genes, then an F1 interstrain hybrid would be heterozygous for these loci in both sexes, with substantial mixing occurring in subsequent generations. In contrast, if strain identity is solely determined by Z-linked genes, then only males (who are ZZ) can be genetically hybrid, with heritable mixing of the strain genes limited to relatively infrequent genetic crossing over events. This means that most of the Z chromosomes transmitted by hybrids will be of a single strain. In short, the more important the Z chromosome is for determining strains, the less consequential are the effects of interstrain hybridization on strain integrity.

We believe that the current data is most consistent with one or more genes on the Z-chromosome being the primary determinants of strain identity, with autosomal loci at best contributing to secondary and (perhaps) less consistent strain phenotypes. The experimental consequences of this perspective are potentially significant, particularly with respect to the use of whole-genome sequencing methodologies to identify strain differences. If the strains are defined by a small number of genes on a single chromosome, then the number of strain-specific polymorphisms will be small and difficult to identify by this method. Furthermore, such polymorphisms will be mostly limited to the Z-chromosome and so give results similar to existing Z-linked markers like Tpi. The strongest evidence against a Z-chromosome based model of strain evolution are indications that autosomal circadian rhythm genes are associated with strain-specific polymorphisms (Hänniger et al. 2017) and the identification of an autosomal strain-specific SNP (Tessnow et al. 2021). However, in the latter study it was noted that ‘Virtually all diagnostic SNPs identified were located on the Z-chromosome’ (Tessnow et al. 2021), consistent with the primacy of Z-linked genes and the exceptional nature of the autosomal SNP. Confirmation that these autosomal markers are consistently strain-specific in other S. frugiperda populations, particularly those from South America, is needed to clarify this issue.

Strain Divergence by Geographical Isolation

The possibility that the S. frugiperda strains are undergoing sympatric speciation is frequently suggested in the literature (Prowell 1998, Groot et al. 2010, Dumas et al. 2015, Nagoshi and Meagher 2016). This is based primarily on modern observations of that the strains display widely overlapping distributions. However, genetic studies based on molecular clock extrapolations have placed the time of strain divergence from 290,000 (Lewter et al. 2006) to over 2 million years (Kergoat et al. 2012, 2021) ago, well before most of the plant types (corn, sorghum, cotton) currently determining strain habitat distributions were present in the Western Hemisphere (Dillon et al. 2007, Gross and Strasburg 2010, Kistler et al. 2018). We have no information as to how the ancestral strain populations were distributed at the time of their divergence, nor do we know what was driving that divergence.

An alternative scenario is suggested by recent evidence that the R-strain is very rare and possibly absent in Ecuador, both on the mainland and the Galapagos Islands (Nagoshi et al. 2020b, 2021a). This observation of a continental region that we believe is predominated by a single strain, demonstrates the plausibility of strain divergence due to the more traditional mechanism of geographical isolation. This is not to imply that Ecuador is the original source of the C-strain as other incidences of geographical isolation are certainly possible before strain divergence. But Ecuador has features that make it a plausible candidate, most notably its location between the Pacific Ocean and Andes Mountain Range, two imposing physical barriers to natural migration. In this scenario, a geographically isolated S. frugiperda population diverged into the C-strain and then, perhaps due to human activity, reemerged relatively recently back into the areas populated by the more ancestral R-strain. Such an allopatric mechanism for the initial differentiation of the two strains could explain why evidence of reproductive isolation between strains is so variable. With geographical isolation there is no need for behavioral mechanisms to keep the populations reproductively separated, as seems necessary for the process of sympatric speciation (Groot et al. 2016).

This hypothesis has at least a superficial resemblance to what has been happening with the sibling species Helicoverpa armigera (Hübner) (Lepidoptera: Noctuidae) and H. zea (Boddie) (Lepidoptera: Noctuidae). Helicoverpa armigera is estimated to have entered the Western Hemisphere about 1.5 million years ago where it is proposed to have diverged into H. zea, a species that has become increasingly specialized to corn as its principal host (Behere et al. 2007). This parallels the supposition that the corn-adapted C-strain arose from the more generalist R-strain (Juárez et al. 2014, Groot et al. 2016), which may have occurred at about that same time period (Kergoat et al. 2012, 2021). Helicoverpa armigera reentered South America in 2013 or earlier, creating the current situation where it has an overlapping distribution with H. zea (Kriticos et al. 2015). Like the S. frugiperda strains, the two Helicoverpa species are capable of cross-hybridization in the laboratory (Rios et al. 2022), although there is (so far) no evidence for the formation of a significant wild hybrid population (Valencia-Montoya et al. 2020). If the S. frugiperda strains are undergoing the same sequence of allopatric speciation followed by sympatry due to migration, then the two strains may be in the process of converging rather than diverging as expected from the assumption of sympatric speciation. Hybridization between strains to produce nonparental genotypes, the frequencies of which might vary by region, could explain the observed variabilities in reproductive compatibility and host plant preferences.

Why Do Strains Matter for Pest Mitigation?

Besides its relevance to understanding speciation, the study of strain biology is important to mitigation efforts, with the most immediate need being risk assessment. Yield losses to corn and sorghum are primarily due to the C-strain while the R-strain is predominantly responsible for infestations in millet, turf, and forage grasses. Therefore, the identification of the strains present in a location defines the crops most at risk of economic damage. This is particularly relevant in the Eastern Hemisphere where there are concerns that such staples as millet and rice are at risk in addition to the observed infestations in corn and sorghum (Early et al. 2018). This possibility appears to be supported by COI marker analysis suggesting that the R-strain is present in substantial numbers in Africa, India, and southeastern Asia (Cock et al. 2017, Kalleshwaraswamy et al. 2019, Wu et al. 2021). However, other studies using the Tpi marker indicates that the R-strain is rare or absent in these locations, suggesting that (at least for the moment) consistent economic damage will be limited to hosts preferred by the C-strain, thereby simplifying mitigation strategies (Nagoshi et al. 2019a,b, 2020c, 2021b). These observations underline the importance of accurate methods of strain identification as well as a better understanding of how the markers differ and how disagreements between markers should be interpreted. Furthermore, we note that the original concerns remain relevant as outbreaks of the R-strain are still possible and perhaps even likely given the globalization of trade, which if not discovered and controlled would substantially increase the range of crops at economic risk. Therefore, continued surveillance in the Eastern Hemisphere for the R-strain is recommended.

Over the longer term, a better understanding of how the strains differ could lead to the development of nonchemical, environmentally benign means of mitigation. The defining phenotype of the two strains is their differential distribution in the wild, which occurs despite few, if any, consistent differences in the capacity of the strains to feed and develop on different plant hosts. This suggests that environmental factors may have a critical role in defining strain-specific habitat choice, which if identified could become the basis for the development of landscape strategies to reduce infestations in cornfields. We, therefore, believe that additional research directed to why the C-strain is not found in pasture and turf grass habitats could have significant economic benefits.

Finally, we note that a likely confounding factor in many studies on S. frugiperda is the variability in the behaviors and physiology of the laboratory colonies used that lead to disagreements between findings and limit the degree to which the results can be generalized. To facilitate comparisons between laboratory findings we believe that laboratory colonies need to be molecularly characterized for strained identity, preferably with the Tpi marker, and ideally with more accurate methods as they become available.

Conclusion

Despite research efforts extending over 30 yr our understanding of the two S. frugiperda strains remains frustratingly inadequate. The only strain-specific phenotype that we are confident is broadly expressed throughout the Western Hemisphere remains the preference for different host plants, the physiological basis of which remains uncertain. While there is intriguing evidence for reproductive barriers between strains, particularly with respect to the timing of mating activity and reproductive incompatibility in a subset of interstrain hybrids, the example of previously reported strain differences in female pheromone composition (reviewed in Groot et al. 2016) provides an important reminder that the generality of a phenotypic difference observed in the laboratory cannot be assumed. Furthermore, it remains to be demonstrated whether the observed reproductive limitations are sufficient to explain the haplotype patterns observed in field studies.

The continued difficulty in identifying the selection pressures and reproductive isolation mechanisms driving strain divergence brings into doubt whether the strains are undergoing sympatric speciation. Instead, the S. frugiperda strains may be undergoing convergence through cross-hybridization made possible from a breakdown in geographical isolation, with the original divergence of the strains driven by factors that no longer exist and are therefore difficult to identify. The S. frugiperda strains are probably best considered as being at an intermediate stage of speciation with the direction of the process uncertain and dependent on habitat. Regions with high plant diversity and abundant plant hosts available for both strains would tend to promote habitat isolation and the divergence of the strains. More limited habitats increase sympatry as plant hosts are chosen more out of necessity than preference. In these locations genetic introgression between strains through hybridization is more likely, leading to increasing strain similarity. These considerations have important implications for studies attempting to find phenotypic or genetic differences between the strains as the type and extent of such differences are likely to vary significantly depending on location.

Acknowledgments

Support came from the Agricultural Research Service of the United States Department of Agriculture (6036-2200-30-00D) and USAID PASA (908-0210-012) and the Agency for International Development of the United States Department of Agriculture. The use of trade, firm, or corporation names in this publication is for the information and convenience of the reader. Such use does not constitute an official endorsement or approval by the United States Department of Agriculture or the Agricultural Research Service of any product or service to the exclusion of others that may be suitable.

References Cited