Soowon Cho, Department of Plant Medicine, Chungbuk National University, Cheongju 361-763, Korea. E-mail: firstname.lastname@example.org.
Abstract The Heliothinae are a cosmopolitan subfamily of about 365 species that include some of the world’s most injurious crop pests. This study re-assesses evolutionary relationships within heliothines, providing an improved phylogeny and classification to support ongoing intensive research on heliothine genomics, systematics, and biology. Our phylogeny estimate is based on two nuclear gene regions, namely elongation factor-1α (EF-1α; 1240 bp) and dopa decarboxylase (DDC; 687 bp), and on the barcoding region of mitochondrial cytochrome oxidase I (COI; 708 bp), providing a total of 2635 bp. These were sequenced for 71 heliothines, representing all major genera and nearly all recognized subgenera and species groups, and for 16 outgroups representing all major lineages of trifine Noctuidae. Analysis of the combined data by maximum likelihood, unweighted parsimony and Bayesian methods gave nearly identical topologies, and the individual gene trees showed only one case of potentially strong conflict. Relationships among genera and subgenera are resolved with strong bootstrap support. The earliest-diverging lineages (c. 200 species in total) consist almost entirely of host specialists, reflecting the inferred ancestral heliothine host range under parsimony. The remaining species form a clade – the Heliothis group – that includes most of the polyphages (30% of heliothines) and all of the major pests. Many other species in the Heliothis group, however, are host specialists. Our results extend previous efforts to subdivide this large clade, and show the most notorious pest groups, the corn earworm complex (Helicoverpa) and the tobacco budworm (Heliothis virescens) group, to be closely related, joining with a small oligophagous genus in what we term the major-pest lineage. Thus, genomic/experimental results from one model pest may extrapolate well to other pest species. The frequency of evolutionary expansion and contraction in host range appears to increase dramatically at the base of the Heliothis group, in contrast to the case for earlier-diverging lineages. We ascribe this difference provisionally to differential evolutionary constraints arising from contrasting life-history syndromes. Host-specific behaviour and crypsis, coupled with low fecundity and vagility, may discourage host-range expansion in earlier-diverging lineages. By contrast, in the Heliothis group, the absence of host-specific traits, coupled with high vagility and fecundity, may more readily permit expansion or contraction of the host range in response to varying ecological pressures such as host species abundance or differential competition and predation.
The Heliothinae are a subfamily of about 365 species of noctuid moths (Matthews, 1999), cosmopolitan but most diverse in warm, dry regions. The larvae attack the flowers and fruits of herbaceous plants, and include a number of the world’s most injurious crop pests, such as the Old World bollworm [Helicoverpa armigera (Hübner)] and relatives, and the New World tobacco budworm [Heliothis virescens (Fabricius)], which cause damage amounting to billions of dollars annually (Fitt, 1989). A reliable classification and phylogeny are indispensable for the communication, organization, and prediction of facts about such pest-containing groups, and for understanding how the traits important to their management, particularly their characteristic broad host-plant ranges, evolve. Beginning with the landmark work of Hardwick (1965), a series of morphological studies has advanced the understanding of heliothine systematics substantially (Hardwick, 1970; Matthews, 1991; Poole et al., 1993; Matthews, 1999). Like advanced Noctuidae generally, however, Heliothinae have simplified morphology, making informative characters difficult to come by (Mitter et al., 1993). Thus, many aspects of heliothine phylogeny are not resolved satisfactorily by morphology.
For this reason, we have been re-examining heliothine relationships using sequences from two protein-encoding nuclear genes, namely longation factor-1α (EF-1α) and dopa decarboxylase (DDC). Previous papers demonstrated that both genes are highly informative about heliothine relationships, and that they yield phylogeny estimates largely in accord with each other and with groupings strongly supported by morphology (Cho et al., 1995; Fang et al., 1997). In this study we extended our sampling by 41 species to a total of 71 heliothines plus 16 outgroup species, attempting to resolve the major remaining uncertainties in heliothine phylogeny. We also added the barcoding region of the mitochondrial cytochrome oxidase I (COI) gene in the hope of obtaining better resolution within genera. Using separate and combined analyses of these three gene regions we now reassess relationships among heliothine species groups and genera. Based on this phylogeny, we offer an updated appraisal of the evolutionary history of host-plant specialization in heliothines, with emphasis on the origin of the polyphagous pest species.
Overview of heliothine classification and life histories
Heliothinae have had a complex taxonomic history since they were recognized formally by Boisduval in 1828 (Matthews, 1991, 1999), but were re-defined unequivocally in a landmark study by Hardwick (1970). Their monophyly is supported by the transverse arrangement of setae L1 and L2 on the larval prothorax and by the spiny larval skin. Two hypotheses about relationships within the subfamily have been proposed on the basis of adult morphology (Fig. 1). Hardwick (1970), in an arrangement of the North American genera (Fig. 1A), placed most species and all the major pests in what we will term the ‘Heliothis group’ (identified in Fig. 1B). The earliest-diverging member of this assemblage is Helicoverpa, the corn earworm and relatives. Hardwick regarded Schinia and its close relatives, a mostly North American group of about 150 host-specific flower feeders, as the most advanced members of this clade. The remainder is made up of Heliothis in Hardwick’s sense, defined only by the absence of characters unique to Helicoverpa and Schinia. The other main branch in Hardwick’s phylogeny contains a number of small genera that are mostly host-specific. Of these, only Pyrrhia, which contains several polyphages, reaches even minor pest status (Hardwick, 1965).
Matthews (1991, 1999) arrived at a similar arrangement in a morphological-phylogenetic survey of the world fauna (Fig. 1B). However, he removed Schinia from the Heliothis group, regarding its similarities to Heliothis sensu lato as superficial only, and instead grouped it with Adisura, an Old World genus not considered by Hardwick. He noted also the lack of clear synapomorphies for any groupings in Hardwick’s Pyrrhia clade, which he treated as an unresolved Pyrrhia group. Matthews erected the genus Australothis for several Heliothis species apparently allied to Helicoverpa, and restored generic status to Heliocheilus, synonymized with Heliothis by Hardwick (1970), a cosmopolitan set of about 50 species, which includes a major pest of millet in Africa (Gahukar et al., 1986). Matthews noted the probable monophyly of the Heliothis virescens group, documented later by Poole et al. (1993), and the possible monophyly of the Heliothis species sometimes segregated as the genera Masalia and Timora. Relationships among all these entities and the remaining ∼30 species of Heliothis were not resolved. Our discussion will use the group names of Matthews (1991; see Fig. 1B), except that we restrict the term ‘Schinia group’ to Schinia plus Heliolonche, as the molecular data do not support a relationship between these and Adisura.
Both Matthews (1991) and Hardwick (1970) postulated the nearest relatives of the Heliothinae to be the Stiriinae, a small subfamily (c. 50 spp.) of trifine noctuids that shares with heliothines the unusual trait of specializing on fruits, flowers and seeds of herbaceous host plants. However, an extensive survey of trifine noctuid phylogeny using the same two nuclear genes as employed in the current study (Mitchell et al., 2006) provided no support for this relationship. Instead, Heliothinae were allied consistently with the true cutworm clade of Lafontaine (1993; Lafontaine & Poole 1991) plus a few smaller groups in what was termed the pest clade. This name refers to the fact that the true cutworms (the Noctuinae sensu lato of Poole, 1995) and the heliothines together include a majority of the noctuids most harmful to agriculture. The pest clade, which numbers about 7000 species, is concentrated to an unusual degree in open, seasonal habitats, and on herbaceous food plants (Mitchell et al., 2006). Relationships within this clade are as yet little resolved, and thus the exact phylogenetic position of the heliothines remains unclear.
Food plants are known for about a quarter of the species of Heliothinae. Of these species, about 70% are monophagous or oligophagous, feeding on one or a few closely related species or genera of plants in the same family (Matthews, 1991; Mitter et al., 1993). Most host-specific heliothines (∼80%) feed on Asteraceae or related families such as Lamiaceae, Scrophulariaceae, Polemoniaceae or Solanaceae; a scattering of other families is also used, particularly Poaceae (grasses) and Fabaceae (legumes). By contrast, about 30% of known species, including most of the pests, are polyphagous, defined here as feeding on hosts in two or more plant orders. This variation has made Heliothinae a useful group for investigating the evolution of host range and other traits associated with pest status, and for addressing the much-debated question of why most phytophagous insects are host-specific (Mitter et al., 1993; Bernays & Chapman, 1994; Sheck & Gould, 1996; Winkler & Mitter, 2008). One goal of this study was to determine whether the polyphages represent the original feeding habit in Heliothinae, or instead were derived from a host-specialized ancestor.
Materials and methods
Taxon sampling and data generation
Our sample of 71 heliothines represents all of the major genera and nearly all of the major recognized subgenera or species groups within them. Of the 12 potentially generic-level entities not sampled, from the 26 possible recognized by Matthews (1991), most are monotypic and within the Pyrrhia group of Matthews (1991), and all are collected rarely. As the sister group to Heliothinae has not been identified definitively, we included data for 16 outgroups from Mitchell et al. (2006) that broadly sample the main lineages of the so-called trifine Noctuidae as inferred by that study. The exemplars sequenced and their source localities and GenBank numbers are listed in Table 1. Species names follow Poole (1989) except where superseded by Matthews (1999) or Hardwick (1996). For this study, 41 new sequences were generated for DDC, 57 for EF-1α, and 84 for COI. Methods of field collection, storage, and vouchering of exemplars followed Cho et al. (1995) and Cho (1997). Wing voucher images can be seen at www.Leptree.net (search for ‘voucher image gallery’). For EF-1α and DDC, methods of DNA extraction, polymerase chain reaction (PCR) and reverse transcriptase PCR (RT-PCR) amplification, DNA sequencing and data assembly, as well as primer sequences, followed Cho (1997) and Mitchell et al. (2006).
Table 1. Exemplars examined, source localities, and Genbank accession numbers for sequences
GenBank Accession Numbers
Helicoverpa armigera armigera
Helicoverpa armigera conferta
Argentina, lab colony
USA, Mississippi; Maryland
New Zealand, Central Otago
USA, Mississippi, lab colony
USA, Florida, lab colony
Israel, lab colony
Heliothis (Masalia) decorata
Heliothis (Masalia) terracottoides
Heliothis (Masalia) galathae
Heliothis (Masalia) nubila
Australia, northern Mataranka
Australia, Brunette Downs
Heliocheilus nr. pallida
Australia, Brunette Downs
Heliocheilus nr. venata
Australia, 80 Mile Beach, Mandora Stn
USA, Texas, Houston
USA, Georgia, lab colony
USA, Maryland, lab colony
The new COI sequences were obtained using the same specimens sequenced for the nuclear genes, using the previous extraction or a new one. New DNA extractions were performed using a GenElute™ Mammalian Genomic DNA Miniprep Kit (Sigma-Aldrich, Castle Hill, Australia). PCR amplifications were performed on a Perkin Elmer GeneAmp PCR System 2400 using either of the forward primers LepFm (5′-GTAAAACGACGGCCAGTCAATYTATCGCYTAAWMTTCAGCC-3′), which binds in the tRNA-Tyr gene, or BC1Fm (5′-GTAAAACGACGGCCAGTTCWACWAAYCAYAARGAYATYGG-3′), which binds in the COI gene, and the reverse primer BC3Rm (5′-CAGGAAACAGCTATGACGWARAATWARAATRTAWACYTCWGG-3′). These primers are degenerate versions of previously published primers K698D (Simon et al., 1994), LCO1490 and HCO2198 (Folmer et al., 1994), respectively, and amplify a much broader range of taxa than the original non-degenerate versions. The first 17 nt of each sequence (italicized) comprises an M13-vector sequence added to facilitate DNA sequencing. Each 30-μL PCR contained 1× PCR buffer (20 mm Tris-HCl, pH 8.4; 50 mm KCl), 2 mm MgCl2, 0.2 mm dNTPs, 10 pmol of each PCR primer, 1 unit of Platinum® Taq DNA polymerase (all supplied by Invitrogen, Mount Waverley, Australia), and 1 μL of genomic DNA. PCR was performed under the following conditions: 94°C for 2 min, 35 cycles of (94°C for 30 s, 50°C for 30 s, 72°C for 60 s), 72°C for 7 min, 4°C hold. PCR products were purified using a GenElute™ PCR Clean-Up Kit (Sigma-Aldrich, Castle Hill, Australia). DNA sequencing was performed using the ABI PRISM® BigDye™ Terminator ver. 3.1 Ready Reaction Cycle Sequencing Kit and an ABI 3730xl Genetic Analyzer (Applied Biosystems).
The fragment lengths amplified were 1240 bp, 689 bp and 717–732 bp for EF-1α, DDC and COI, respectively. However, the actual sizes we used for analyses were 1240 bp, 687 bp and 708 bp, respectively. In some species the COI fragment length was less than 708 bp after terminal cut-off owing to unconfirmed sequences and sequence ambiguities.
Phylogeny estimation from nucleotide sequences was attempted under both the maximum likelihood (ML) and unweighted parsimony (MP) criteria, for each gene separately, all three genes concatenated, and each possible pairwise combination of genes. The purpose of the one- and two-gene analyses was to assess the contribution of each gene to the total signal and to check for conflict between genes. As one gauge of the potential differences in evolutionary properties among genes, we calculated base composition separately for each codon position in each gene, using mega 3.1 (Kumar et al., 2004).
For the ML analysis of each partition we determined first the most appropriate substitution model, separately under the Akaike information criterion (AIC) and by hierarchical likelihood ratio tests (hLRTs), using modeltest 3.7 (Posada & Crandall, 1998). If the two criteria favoured different models, analysis was conducted under both. ML tree estimation employed garli 0.95 (Zwickl, 2006; Morrison, 2007). Following some experimentation with different starting trees, the garli default settings were used for all final analyses. For each data set, ten or more runs were made to estimate an optimal tree, after which 500 bootstrap replicates were performed to estimate node support.
Our MP analyses were conducted with paup 4.0b10 (Swofford, 2003). For the three genes together, the tree search consisted of 100 random addition sequences each followed by tree bisection−reconnection (TBR) branch swapping, with node support subsequently estimated by 500 bootstrap replicates, each employing ten random addition sequences plus TBR. Attempts to perform the same analyses on individual data partitions, however, proved difficult. Tree searches on the original datasets were slow and produced very large numbers of MP trees. For COI alone, for example, 169 509 MP trees were found in 82 random addition sequences before the search was stopped after 103 h, and for EF-1α and DDC combined, 37 795 MP trees were found in 100 random addition sequences. For no partition short of the entire dataset was it possible to complete more than a few bootstrap replicates without imposing severe restrictions on the number of MP trees to be saved in each replicate. For these reasons, MP analysis was discontinued for all but the three-gene dataset.
For purposes of discussion, bootstrap values are used to quantify the level of support at particular nodes, i.e. 70–79%: ‘moderate’; 80–89%: ‘strong’; 90–100%: ‘very strong’. Bootstrap values lower than 70% are considered ‘weak’ support. Gene-specific bootstrap values also are compared between incompatible resolutions, when present, and the lower of the two values is described as the level of conflict; for example, alternative groups with 75% and 90% bootstrap values would be considered to indicate a ‘moderate’ level of conflict. We adopt these largely arbitrary rules to ensure consistency, and hence, we hope, avoid confusion.
To infer the evolutionary history of larval host-plant ranges, each species for which food plant records exist was scored as oligophagous, defined as feeding on plants of a single family, or polyphagous, defined as feeding on two or more plant families. Food plant information was taken from compilations by Matthews (1991, 1999), Hardwick (1996) and Cho (1997). This two-state character was mapped onto the molecular phylogeny, to estimate the ancestral host range for each node. Ancestral host ranges were estimated under parsimony and under maximum likelihood. The ML analysis was conducted with mesquite 2.01 (Maddison and Maddison, 2006). Character change followed the Mkv model of Lewis (2001), and branch lengths were taken from the molecular phylogeny. The host range is not known for Erythroecia suavis, the only species of this genus included here, but is known for the closely related E. hebardi. To incorporate this information, thereby improving the estimate of the ancestral heliothine condition, we assigned the host range of E. hebardi to E. suavis for the purpose of this analysis. Similarly, although host ranges are not known for the three species of Adisura included here, host ranges are known for three of their congeners: all are oligophagous. To incorporate this information, a host range of ‘oligophagous’ was assigned to an arbitrarily chosen member of this genus in the analysis. The host range analysis was restricted to Heliothinae, as the phylogenetic position of heliothines is sufficiently unclear, and host-range evolution sufficiently rapid, that there is unlikely to be clear-cut homology in this trait between heliothines and our sample of outgroups.
Using MrModeltest, hLRTs preferred the GTR + I + G model for all genes combined as well as for each gene separately and for all pairs of genes, although SYM + I + G was an alternative choice for several partitions. For all genes combined and for COI only, GTR + I + G was also selected by the AIC. However, for the nuclear genes, the first choice under the AIC was SYM + I + G, and the second choice was GTR + I + G. This difference probably reflects the differences in base composition between the nuclear genes and the mitochondrial COI gene (see Table 2).
Table 2. Average base composition across taxa for each codon position in each gene.
% A + T
% (A + T)-50
EF-1α (1240 bp)
DDC (687 bp)
COI (708 bp)
The ML topology from the three genes combined, as estimated by garli under the GTR + I + G model, is shown in Fig. 2. Relationships among the outgroups, omitted here for the sake of readability, were consistent with those shown by Mitchell et al. (2006). Superimposed on the nodes of this tree are the bootstrap proportions (BP) for the nodes obtained in the all-genes ML analysis, the all-genes MP analysis, and the ML analyses on each gene separately and on the two nuclear genes combined. The strict consensus of the 63 most-parsimonious trees found for the three genes combined (not shown), although somewhat less resolved within genera, shows only minor differences from the ML tree of Fig. 2, with no alternative, preferred groupings supported by BP >57 in either analysis.
The ML trees found for individual genes (not shown) were very similar to each other in broad structure, the most notable exception being a slight shift of the root position for Heliothinae for COI alone (identified in Fig. 2). Alternative groupings among closely related species were more common, but supported by bootstraps of less than 50% except in the four instances shown in Fig. 2, all of which involve differences between COI and one or both nuclear genes, as follows.
(i) DDC very strongly supports the grouping of Rhodoecia aurantiago and Pyrrhia umbra to the exclusion of P. exprimens (BP = 100), rendering Pyrrhia paraphyletic. By contrast, COI moderately supports grouping of the two Pyrrhia species (BP = 74).
(ii) DDC very strongly supports sister group relationships between Schinia tertia and S. chrysella, and between S. nundina and S. pulchripennis (BP = 99 in each case), whereas COI strongly favours grouping of S. nundina with S. tertia (BP = 81).
(iii) Heliocheilus venata and H. cistella, placed as sister species by COI (BP = 72), are instead allied, respectively, with H. eodora by EF-1α (BP = 71) and with H. clathrata by DDC (BP = 60).
(iv) Heliolonche pictipennis and H. modicella are grouped together by EF-1α and DDC individually (BP < 50) and in combination (BP = 56). By contrast, these two species take separate, different placements within Schinia in analysis of COI alone. In the combined ML analysis, which estimates a single substitution model using all three genes, Heliolonche is paraphyletic with respect to Schinia jaguarina (BP = 56), whereas in MP analyses of the combined data Heliolonche is monophyletic.
Given the similarity of the topologies inferred from different data subsets, all three genes clearly contribute signal to the tree estimated from the combined data. The relative strength of these contributions, however, varies considerably. Table 3 shows the total numbers of nodes, in ML trees inferred (in separate analyses) from the combined data and from subsets thereof, that are supported by bootstrap proportions of 50 or higher, 70 or higher, and 90 or higher. Given that groups with such support rarely conflict among individual gene analyses, these bootstrap values can be interpreted as approximate measures of phylogenetic signal strength. On this criterion, DDC clearly provides the most signal, and COI the least. Both EF-1α and COI generally appear to provide signal complementary to that of DDC, as combined analysis of DDC with either yields equal or greater numbers of nodes under each BP support threshold than does DDC alone. However, EF-1α and COI contribute differently to the all-genes result. Subtracting EF-1α from the three-gene dataset (all genes vs. DDC + COI in Table 3) slightly but consistently decreases the number of nodes supported at successive bootstrap thresholds. By contrast, subtracting COI (all genes vs. EF-1α + DDC) actually increases bootstrap support within Heliothinae at moderately supported or higher BP thresholds, by eliminating the conflicts described above. The addition of COI did, however, result in strong support (BP = 83) for one relationship not previously identified, namely, the monophyletic group of three species in Fig. 2 that includes Heliocheilus melibaphes. COI data also provided very strong support, lacking in the nuclear gene datasets, even in combination, for the monophyly of Helicoverpa armigera.
Table 3. Number of nodes resolved at or above three thresholds of bootstrap support, for the various data sets and methods of analysis.
Data set, analysis
Number of nodes with
BP ≥ 50
BP ≥ 70
BP ≥ 90
The number on the left represents nodes within Heliothinae, and the number in parentheses represents nodes within outgroups.
We also compared the levels of bootstrap support between ML and MP analyses of the combined data, which yielded nearly identical tree topologies. Table 3 shows slightly greater numbers (of the order of 10% more) of resolved nodes at all BP thresholds for ML analysis than for MP analyses of the combined data. Moreover, for the 28 nodes in Fig. 2 that are found in the best trees under both criteria (or found in one and unresolved in the other) but with differing BP support, BP under ML analysis was higher in 24 cases, whereas BP was higher under MP in just four cases. These observations, together with the difficulty of running MP analyses of data subsets to completion, suggest that ML analysis may be more effective than MP at extracting phylogenetic signal from these data.
We review first the bearing of our findings on previous proposals about phylogeny and classification in Heliothinae, and then examine briefly their implications for the evolution of pest status and host range. Except as otherwise specified, our discussion of relationships and bootstrap support will refer to the combined data ML analysis of Fig. 2.
Major lineages within Heliothinae: the Pyrrhia group sensu stricto
Our analysis yields strong evidence that the earliest divergence in Heliothinae falls within the Pyrrhia group as delimited by Matthews (1991), consistent with the absence of any definitive morphological synapomorphy for that group. In partial support of Hardwick’s hypothesis, our data provide strong evidence (BP = 99) for a monophyletic group comprising four of the five Pyrrhia group genera thus far sampled (nine genera not sampled), an assemblage we shall term the Pyrrhia group sensu stricto. Our remaining exemplar, Erythroecia, however, is grouped firmly with the remaining heliothines (BP = 96), rendering the Pyrrhia group sensu lato paraphyletic.
Host records exist for 14 species in nine genera of the Pyrrhia group sensu Matthews, with all genera but Pyrrhia being apparently mono- or oligophagous (Cho, 1997). In Pyrrhia, there is a single record for P. treitschkei, whereas P. purpurina and P. victorina have three and two recorded hosts, respectively, from two plant orders. Pyrrhia adela Hardwick (1996; formerly North American P. umbra), P. exprimens, and P. umbra each have many hosts from more than three plant orders (Matthews, 1991).
Our data also offer strong resolution of relationships within the Pyrrhia group sensu stricto, concordant with those proposed by Hardwick (Fig. 1) if the root of Hardwick’s subtree for this group is changed to lie on the branch leading to Erythroecia. Although a formal analysis is not attempted here, the molecular tree appears also to be consistent with the morphological characters mentioned by Hardwick (1970). For example, the pairing of Eutricopis and Heliothodes is consistent with shared possession of a well-sclerotized ovipositor, the absence of an ampulla on the male valve, and reduction of the eyes in association with diurnal adult activity.
A remaining point of uncertainty concerns the phylogenetic position of the monotypic genus Rhodoecia. The nuclear gene data (BP = 91), particularly DDC (BP = 100), group R. aurantiago with Pyrrhia adela, whereas COI alone groups the two Pyrrhia species together (BP = 74). Morphology is compatible with both alternatives because no synapomorphy is known for Pyrrhia. Rhodoecia aurantiago, restricted to eastern North America and feeding as a larva only on several genera of Scrophulariaceae, is narrower in both distribution and host plant range than the two Pyrrhia species, which are polyphagous. This observation lends plausibility to, but does not establish, the derivation of Rhodoecia from Pyrrhia in conjunction with a constriction of host range. If the gene tree conflict is real, both hypotheses for the species phylogeny require incomplete lineage sorting or introgression in one or more genes. Alternatively, the moderate branch support by COI (BP = 74) could be considered inconclusive.
The Schinia group sensu stricto
The remaining heliothines in our sample fall into three lineages, namely Erythroecia suavis, the Heliothis group plus its sister group Adisura, and Schinia + Heliolonche. The sequence shown in Fig. 2, with Erythroecia diverging first, is supported only weakly by the molecular evidence but is consistent with morphology (see the following section).
Our data strongly support the grouping of Heliolonche with Schinia (BP = 99), although the morphological synapomorphies proposed, namely the absence of cornuti on the vesica (Matthews, 1991) and of a scobinated plate near the base of the vesica (Hardwick, 1970), are homoplasious on our tree. Nearly all species in this clade, which we will term the Schinia group sensu stricto, are specialized feeders. The five species of Heliolonche for which host records are available (Hardwick, 1996) all appear to be monophagous, and restricted to the asteraceous tribe Lactuceae. Among the 79 species of Schinia with recorded hosts (Matthews, 1991; Hardwick, 1996), only S. chilensis, from Chile, is known to be polyphagous, although two undescribed species, from Mexico and Chile, share the unusual pad-like ovipositor and unspecialized maculation of S. chilensis, suggesting that these too might be polyphagous (Matthews, 1991).
Two functionally correlated synapomorphies have been proposed for Heliolonche, loss of the coiling in the male vesica and in the female appendix bursae (Hardwick, 1970; Matthews, 1991), which are characteristic of heliothines outside the Pyrrhia group. Heliolonche is also unusual in feeding only on the asteraceous tribe Lactuceae (see below), and in possessing several traits associated with diurnal adult flight, including reduced eyes. Our data are not decisive on the monophyly of this genus. The nuclear genes, individually and in combination, support monophyly, but with bootstraps always lower than 60%. COI favours polyphyly, but with even lower support. The combined data support monophyly (albeit weakly) under MP, but under ML place Schinia jaguarina within a paraphyletic Heliolonche, allied to H. pictipennis. Further evidence is required to determine whether molecular data support the monophyly of Heliolonche.
As no morphological synapomorphy is evident for Schinia, Matthews (1991) suggested that this genus might be paraphyletic with respect to Heliolonche. Consistent with this postulate, Schinia was not monophyletic in any of our analyses, but strong bootstrap support disfavouring monophyly is likewise lacking. Our study provides the first attempt to discern relationships within this large genus. None of the deeper splits are resolved strongly, but there are well-supported species pairs and triplets. Such relationships as are evident tend to be correlated with host taxon use. For example, the group consisting of S. felicitata, S. luxa, and S. gaurae (BP = 100) contains the only two Onagraceae feeders in our sample (the host of S. luxa being unknown). All Schinia apart from these and S. jaguarina, a legume feeder, form a clade (albeit weakly supported) that consists almost entirely of Asteraceae feeders (12 of 14 records). Within this lineage, the group consisting of S. oleagina, S. trifascia, and S. sanguinea (BP = 74) contains the only three species in our sample that specialize on the asteraceous tribe Eupatorieae, whereas the European sister pair S. cardui and S. cognata (BP = 100) are the only specialists on the tribe Lactuceae. Schinia chrysella and S. tertia (BP = 97) both feed on the tribe Astereae.
Adisura as sampled in this study is supported very strongly as monophyletic (BP = 100), and allied by all genes to the Heliothis group (BP = 100, combined data ML). Previously, Adisura had been thought to be allied to Schinia (Matthews, 1991), but the present analyses favour an alternative suggestion by Matthews (1999), that the unique form of coiling of the female appendix bursae and the male vesica that appears to unite these two genera is instead part of a broader synapomorphy for all heliothines apart from the Pyrrhia group sensu lato (including Erythroecia). So far as is known, all members of this tentative clade except Heliolonche retain some form of coiling in these functionally related genitalic structures. The coiling feature lends further support to the tentative coiled appendix/vesica clade shown in Fig. 2, for which molecular evidence is weak, and suggests that the non-sequenced genera of the Pyrrhia group sensu lato also will prove to lie outside this clade.
Judging from the three species for which hosts are recorded, Adisura can be characterized provisionally as feeding on legumes. Adisura atkinsoni and A. marginalis are legume crop pests in India (Thontadarya et al., 1982; Chakravarthy, 1983; Singh et al., 1991). There is one report of A. atkinsoni apparently feeding on Hibiscus mutabilis (Malvaceae), in a different plant order, leading Mitter et al. (1993) to characterize this species as possibly polyphagous, but this record is doubtful (Cho, 1997).
The Heliothis group
The Heliothis group sensu Matthews (1991), comparable in species diversity to the Schinia group s.s., was monophyletic in all of our analyses, and is supported strongly by the combined data (BP = 100). The taxonomic history of this group has been tumultuous, with nearly all species placed at one time or another in the genus Heliothis, but in recent decades several natural subgroups have been defined and named.
Our analyses distinguish three major elements among the Heliothis group species sampled, considered in turn below. The first is a clade containing both the corn earworm complex (the genus Helicoverpa) and the tobacco budworm complex (Heliothis virescens and relatives), which together include the heliothine pests of greatest economic impact on a global scale. All three genes support a close relationship of the two pest complexes. In the EF-1 α and CO-1 gene trees, the Heliothis virescens group and Helicoverpa, excluding Australothis, are sister groups, but with weak support. DDC favours insertion of Australothis into this cluster, as sister group to Helicoverpa. The combined data follow the DDC topology, providing moderate support (BP = 75) for alliance of the virescens group with Helicoverpa + Australothis, a grouping we shall term the major-pest lineage. This grouping has been a constant finding as our sampling of genes and taxa has grown (Cho et al., 1995; Fang et al., 1997).
The second major element of the Heliothis group is a species-rich, largely tropical clade, poorly known but possibly mostly host-specific, consisting of the genus Heliocheilus plus a set of Afrotropical species sometimes placed in the genus Masalia. The third element, an assemblage we will term typical Heliothis, comprises the residue of species not belonging to one of the aforementioned clades, and is probably paraphyletic with respect to the rest of the Heliothis group.
Major-pest lineage: Helicoverpa + Australothis
Monophyly of Helicoverpa (Hardwick, 1965), the 20 species of which include five major pests, is supported strongly by morphology (Matthews, 1991) and by all three genes in this study.
Despite our use of three genes, however, resolution within Helicoverpa remains limited, presumably owing to insufficient character information or taxon sampling (7/20 species). The few groupings that emerge are largely consistent with those postulated by Hardwick (1965) and those sketched on the basis of morphology and/or allozymes by Mitter et al. (1993) and by Matthews (1999). The only strong support, however, is for the pairing of Helicoverpa hawaiiensis and H. pallida, the two Hawaiian archipelago endemics, which are strongly to very strongly grouped by all three genes. For the pairing of the Old World bollworm, H. armigera, with the New World corn earworm, H. zea, both MP and ML analyses using all three genes provide moderate support. Given especially the powers of long-distance dispersal of some species, extensive sampling and more rapidly evolving markers will probably be necessary to resolve relationships within Helicoverpa.
All of the major pests in this genus are polyphagous except for H. assulta, which appears restricted mainly to Solanaceae. The Hawaiian endemics H. pallida and H. hawaiiensis have been associated consistently with Chenopodium (Chenopodiaceae) and Sida (Malvaceae), respectively, and may be specialized to these dominant plants of their respective habitats (Hardwick, 1965; S. Conant & A. Moore, personal communication).
Our analysis also strongly supports monophyly for the genus Australothis (BP = 100), erected by Matthews (1991). The genus is unique in bearing a long spiral male vesica with a band of minute cornuti dispersed along its length. This condition appears intermediate between that in Helicoverpa, in which the cornuti are larger and fewer in number, and that in the remainder of the Heliothis group, which lack cornuti altogether (Matthews, 1991), and has been taken as evidence of a sister group relationship between Australothis and Helicoverpa. That pairing is strongly corroborated by our data, although the evidence comes almost exclusively from one gene, DDC.
Of the four Australothis species, the two with known hosts appear mostly restricted to Asteraceae, although there are occasional records of A. rubrescens from other host families (Matthews, 1999).
Major-pest lineage: the Heliothis virescens group
All three genes uphold the pairing of Heliothis virescens and H. subflexa, consistent with the strong morphological evidence for monophyly of the H. virescens group (13 species; Poole et al, 1993). Relationships and life histories of these species are of especial interest, because several are pests, and because the contrast between H. virescens and H. subflexa, which can be crossed in one direction, has become a model system for investigating the genetic basis of differences in host-plant range (Sheck & Gould, 1996), pheromone communication systems (Groot et al., 2006; Sheck et al., 2006), and other traits. A phylogeny estimate based on adult morphology (Poole et al., 1993) shows H. virescens and H. subflexa not to be sister groups, a conclusion that needs testing through increased taxon sampling with molecular data. Life histories in this group, although not well studied, appear quite variable. Heliothis molochitina, sister to the remaining species according to morphology (Poole et al., 1993), appears to be polyphagous, as is H. virescens. Heliothis subflexa and H. tergemina, closely related according to morphology, and both occasional pests, are oligophagous and restricted to Solanaceae. The single record for H. distincta is from the legume Cajanus indicus (chick pea).
Heliocheilus + Heliothis (Masalia)
Monophyly of Heliocheilus, suggested initially by the male forewing modified for sound production (Matthews, 1987), is supported very strongly by our data (BP = 98), which include 18 of the 43 known species. Within Heliocheilus, there are very strong resolutions of deeper divergences (Fig. 2), including robust support for monophyly of the large Australian radiation (BP = 93; 29 species), and against monophyly of the African species (BP = 99 for African H. confertissima + Australian clade), which are interspersed with those from North America. Within the Australian clade, however, divergences are very shallow and resolution is mostly weak. Of the seven species of Heliocheilus for which host records exist, all appear to be restricted to grasses (Poaceae).
The c. 60 palaeotropical species sometimes segregated as the genera Masalia and Timora were placed within Heliothis by Matthews (1991), but with a note that they could prove to be monophyletic. Seymour (1972) had resurrected Masalia, formerly synonymized with Timora, defined by a distinctive scale-like cornutus on the vesica. Matthews (1991), however, found similar traits in Heliothis peltigera and other Heliothis species. Furthermore, he found no strong synapomorphy for Masalia plus Timora, although these species tend to share a number of characters: light spining on the mid- and hind-tibia, stout forelegs, two heavy claws on the fore-tibia, and a longitudinally streaked pattern on the forewings (similar to many Heliocheilus). Our data provide modest support for uniting the four Masalia species sampled (BP = 67), and strong evidence for the grouping of these with Heliocheilus (BP = 96), a relationship not suggested previously. The Masalia/Timora assemblage, for which no host records are available, clearly needs further characterization.
There are c. 30 species of ‘typical’Heliothis, i.e. those not belonging to one of the monophyletic entities described. Of the six exemplars in our sample, the four primarily holarctic species form a very strongly supported monophyletic group (BP = 92) consisting of two very strongly supported species pairs. Were this previously unrecognized clade to require a name, it could appropriately be called Heliothis sensu stricto, as it includes the type species of Heliothis, viriplaca. We predict that most or all of the other north temperate typical Heliothis, including H. acesia, H. australis, H. belladonna, H. borealis, H. maritima, and H. ononis, will prove also to belong to this group. The remaining species in our sample, the Australian H. punctifera and the paleotropical/subtropical H. peltigera (which regularly migrates into northern Europe), are of uncertain position, not obviously related to each other or to any other group. Although strong support is lacking, the typical Heliothis collectively appear to be paraphyletic with respect to the remainder of the Heliothis group. As is suggested by Fig. 2, we predict that the root of the Heliothis group will prove to lie among the typical Heliothis of the tropics and/or southern hemisphere, exemplified by H. punctifera and H. peltigera.
Of the 12 species of ‘typical’Heliothis for which host records exist, three have been reported from just a single host plant, although their apparent specificity could be an artifact of limited sampling. Heliothis prorupta and H. belladonna (a single larva) have been reported only from Castilleja (Scrophulariaceae; Hardwick, 1996), and H. scutiligera only from Helichrysum (Asteraceae; Pinhey, 1975). Most species of ‘typical’Heliothis, however, appear to be polyphagous.
Evolutionary history of pest status and host range
The ancestral state reconstructions shown in Fig. 2 clearly imply that heliothines ancestrally were host-specific, whereas polyphagy has had multiple independent origins. Most of the present-day polyphagous species and nearly all the major pests belong to a single clade, the Heliothis group, in which polyphagy is prevalent and might be ancestral (but see below). Within that clade, the entities containing the most notorious and polyphagous pests, the corn earworm complex (Helicoverpa) and the tobacco budworm (Heliothis virescens) group, appear to be closely related. This result suggests that findings from ongoing intensive genetic studies of several pest species (e.g. Heckel, 1998; Krieger et al., 2004; Gahan et al., 2005) may extrapolate well to other members of these complexes.
The most striking trend evident in Fig. 2 is an apparent dramatic shift in the evolutionary pattern of host range at the base of the Heliothis group, in contrast to the case for all the other major lineages of Heliothinae. In the latter (e.g. Adisura and the Schinia and Pyrrhia groups sensu stricto), which include the earlier-diverging heliothine lineages, host specificity is the rule, and evolutionary change in host range is quite rare. More sampling is needed, but from present evidence it appears possible that, in the entire history leading to the c. 200 total species in these lineages (94 with known habits, of which 23 are included here; see Fig. 1), there have been only two origins of polyphagy, one each in Pyrrhia and in Schinia chilensis and relatives (not included here; see Matthews, 1991), and a single reversion to host specialization, in Rhodoecia. By contrast, among the c. 175 species of the Heliothis group (35 with known habits; Fig. 1), shift in host range has been sufficiently frequent that for many internal nodes we cannot confidently assign ancestral conditions. Parsimony reconstruction requires at least six changes in host range within the Heliothis group given just the 23 species with known habits included in Fig. 2. Presumably, many more changes would be required with additional sampling, as the species with known habits but not included here are divided about equally between specialists and generalists. Moreover, nearly all of this evolution is concentrated in the c. 65 species in Australothis, Helicoverpa and Heliothis sensu lato (28 with known habits, 17 included here), as Heliocheilus appears to be restricted to Poaceae. Dramatic shifts of host range have taken place even between close relatives within the major pest complexes, giving rise to contrasts such as Heliothis virescens vs. H. subflexa, and H. assulta vs. Helicoverpa armigera and H. zea. There are clear instances of both broadening and narrowing of host range, but the relative frequencies of these are not clear.
Why should there be such marked variation in the propensity for change in host range? Hardwick (1965, 1970), who compiled a lifetime of detailed observations on heliothine natural history, pointed to a suite of life-history features that vary between groups with differing typical host ranges [although Matthews (1991) noted significant exceptions]. Thus, species of Schinia, which typically attack one or a few closely related plant species, often stay on or near the host, on which they appear frequently to be specifically modified for crypsis, as both larvae and adults. Schinia typically lay relatively few (tens to hundreds) but relatively large eggs, often inserting them deep into host flowers with their elongate, sclerotized ovipositors. Hardwick contrasted these traits with their opposites in Helicoverpa, species of which are typically highly polyphagous, lay hundreds to thousands of eggs scattered across the plant or even on the ground with their pad-like ovipositors, are highly vagile to migratory, and lack specific crypsis to particular host species. They also show larger adult body sizes on average than do Schinia.
Following Hardwick’s lead, we postulate that these contrasting life-history syndromes differentially constrain the evolutionary lability of host range. For species with the life history typical of Schinia, low fecundity, low vagility and host-specific crypsis and oviposition behaviour should lower the frequency of the oviposition mistakes that probably underlie expansions of host range (Janz et al., 2001), and lower the chance that a mistakenly placed individual will survive. These constraints should keep host shifts and evolutionary expansion of host range rare, even if such changes would potentially be favoured by ecological factors such as reduced competition or predation on alternative hosts. By contrast, high fecundity and vagility, and lack of host-specific morphology or behaviours, prevalent traits in the Heliothis group, not only are conducive to host-range expansion, but also are permissive of, that is do not strongly disfavour, host-range contraction. Thus, host range should more readily evolve in response to varying ecological selection pressures (Winkler & Mitter, 2008). A somewhat analogous argument was made by Morse & Farrell (2005), who found that expansions of host range in legume-feeding Stator seed beetles are more frequent in lineages that oviposit on seeds after the pods have been opened than in ones ovipositing on intact seed pods; the latter are harder to get into. Rigorous evaluation of the constraint postulate will require tests for phylogenetic correlation between heliothine life-history features and host range and its evolutionary rate.
This study could not have been undertaken without the guidance, initial collaboration and expert advice of Robert W. Poole. Michael Pogue kindly provided us with updated species counts from his unpublished compilation. We are also greatly indebted to many generous colleagues for supplying fresh specimens for this study, including James Adams, Matti Ahola, Miriam Altstein Joaquin Baixeras, Richard Brown, Sheila Conant, Chris Conlan, Bob Denno, Dave Dussourd, Ted Edwards, Charles Ely, Randy Furr, Matt Greenstone, David Hardwick, Milton Huettel, Nils Hyden, Ed Knudson, Don Lafontaine, Harry Lonka, Eric Metzler, Audrey Moore, Brian Patrick, Ric Peigler, Laszlo Peregovits, Bob Poole, Rick Roche, Amy Sheck, Eva Silverfine, Neal Spencer, and Andy Warren. Suwei Zhao and Kongyi Jiang provided technical assistance and Andreas Zwick kindly performed the tedious task of Genbank submission. Financial support was provided by the USDA-NRICGP, an NSF Dissertation Improvement Grant, and the University of Maryland Graduate School. This paper is dedicated to the memory of Dr. David Hardwick, unrivalled master of heliothine systematics and biology, whose life work and generous initial help made our own effort possible.