Figures
Abstract
Background
The 1400 species of hawkmoths (Lepidoptera: Sphingidae) comprise one of most conspicuous and well-studied groups of insects, and provide model systems for diverse biological disciplines. However, a robust phylogenetic framework for the family is currently lacking. Morphology is unable to confidently determine relationships among most groups. As a major step toward understanding relationships of this model group, we have undertaken the first large-scale molecular phylogenetic analysis of hawkmoths representing all subfamilies, tribes and subtribes.
Methodology/Principal Findings
The data set consisted of 131 sphingid species and 6793 bp of sequence from five protein-coding nuclear genes. Maximum likelihood and parsimony analyses provided strong support for more than two-thirds of all nodes, including strong signal for or against nearly all of the fifteen current subfamily, tribal and sub-tribal groupings. Monophyly was strongly supported for some of these, including Macroglossinae, Sphinginae, Acherontiini, Ambulycini, Philampelini, Choerocampina, and Hemarina. Other groupings proved para- or polyphyletic, and will need significant redefinition; these include Smerinthinae, Smerinthini, Sphingini, Sphingulini, Dilophonotini, Dilophonotina, Macroglossini, and Macroglossina. The basal divergence, strongly supported, is between Macroglossinae and Smerinthinae+Sphinginae. All genes contribute significantly to the signal from the combined data set, and there is little conflict between genes. Ancestral state reconstruction reveals multiple separate origins of New World and Old World radiations.
Conclusions/Significance
Our study provides the first comprehensive phylogeny of one of the most conspicuous and well-studied insects. The molecular phylogeny challenges current concepts of Sphingidae based on morphology, and provides a foundation for a new classification. While there are multiple independent origins of New World and Old World radiations, we conclude that broad-scale geographic distribution in hawkmoths is more phylogenetically conserved than previously postulated.
Citation: Kawahara AY, Mignault AA, Regier JC, Kitching IJ, Mitter C (2009) Phylogeny and Biogeography of Hawkmoths (Lepidoptera: Sphingidae): Evidence from Five Nuclear Genes. PLoS ONE 4(5): e5719. https://doi.org/10.1371/journal.pone.0005719
Editor: Michael Knapp, Max Planck Institute for Evolutionary Anthropology, Germany
Received: February 20, 2009; Accepted: May 3, 2009; Published: May 28, 2009
Copyright: © 2009 Kawahara et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Financial support was provided by the U.S. National Science Foundation, award number DEB-0212910 and by NSF's Assembling the Tree of Life program, award number DEB-0604329. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
The hawkmoths (Lepidoptera: Sphingidae), comprising about 1400 species in some 200 genera and found on every continent except Antarctica [1], [2], are one of the most conspicuous and well-studied insects. Due in part to their large size, sphingids have long provided models for studies of genetics, physiology and development [3]–[8], functional morphology [9], [10], plant-herbivore interactions [e.g.], [ 11], [12]–[14], pollination biology [e.g.], [ 15], [16]–[21], and biogeography [22]. Sphingids are also a focal group for biodiversity and habitat quality assessment studies [23]. Some species are agricultural pests [24], [25], while others have been used as biological control agents [26]. Recently, sphingids have also become a model group for testing the reliability of DNA barcoding for species identification [27], [28].
Research of many kinds on hawkmoths would benefit from a more robust phylogenetic/comparative framework for this family than is currently available. Morphological and molecular analyses strongly support the monophyly of Sphingidae, and placement within the superfamily Bombycoidea [2], [29]–[31]. Within the family, however, there has yet to be a comprehensive study of relationships based on explicit phylogenetic methodology. A preliminary molecular analysis by Regier et al. [32], while spanning all three subfamilies, included just fourteen species. All other analyses have focused within a tribe or genus (e.g., Acherontiini [33], [34], Hyles [35]–[37], Proserpinus [38]).
The present study builds on past and ongoing efforts to reconstruct sphingid phylogeny using morphology (Figs. 1A–C). In their monumental revision, the starting point for subsequent classifications, Rothschild and Jordan [1] divided the hawkmoths into two “series,” “Sphingidae Semanophorae” and the “Sphingidae Asemanophorae,” distinguished by the presence or absence of a patch of short sensory hairs (microtrichia) on the inner surface of the first segment of the labial palp. These groups correspond approximately to the present-day Macroglossinae and (Smerinthinae+Sphinginae) respectively (Fig. 1A). Writing long before the formalization of modern cladistics [39]–[41], Rothschild and Jordan presented a “tree” on which all then-known sphingid genera were placed according to shared derived morphological structures, although they circumscribed some taxonomic groups on the basis of symplesiomorphic resemblance.
To facilitate comparison, all the trees have been reduced to show only the dispositions of the subfamilies, tribes and subtribes recognized by Kitching and Cadiou [2]. Nomenclature of previous authors, where different, is shown in quotes. A double line leading to a taxon name denotes paraphyly, a dotted line indicates uncertain relationships. Numbers above branches in Figure 1D are bootstrap values.
Within their two series, Rothschild and Jordan recognized five subfamilies, all of which correspond approximately to modern groups, although many names and ranks have changed (Fig. 1A). Janse [42] reduced Rothschild and Jordan's subfamilies to tribes, and treated Semanophorae and Asemanophorae as subfamilies (Semanophorinae and Asemanophorinae), names that Hodges [43] changed to Macroglossinae and Sphinginae, based on the type genus of each subfamily. Minet [29] separated Sphinginae sensu lato into Smerinthinae and Sphinginae sensu stricto, in part because he believed that Sphinginae sensu lato, defined by the absence of the microtrichial patch, might be paraphyletic.
The next major contribution to sphingid systematics came from Nakamura [44]–[46], who reclassified several sphingid tribes and proposed relationships among Japanese hawkmoth genera based on characters of the larva and pupa (Fig. 1B), although without explicit cladistic analysis. He considered Smerinthini and Sphingini to be sister groups. His remaining tribes formed a group that roughly corresponds to Macroglossinae as defined in the most recent comprehensive revision of sphingid classification, that of Kitching and Cadiou [2]. In addition to stabilizing sphingid taxonomy and revising the tribal and subtribal level classification, Kitching and Cadiou proposed provisional relationships among genera (Fig. 1C), based in part on unpublished morphological analyses. Unlike earlier authors, they tentatively placed the Smerinthinae at the base of the family, postulating that Sphinginae plus Macroglossinae constitute a monophyletic group. Within Smerinthinae, Smerinthini was hypothesized to be paraphyletic. While putative apomorphies were identified for each subfamily [29], Kitching and Cadiou expressed caution about the monophyly of most tribes and subtribes, particularly within Smerinthinae.
The molecular analysis by Regier et al. [32], based on the nuclear genes elongation factor-1α (EF-1α) [47] and dopa-decarboxylase (DDC) [48], provided an initial test of Kitching and Cadiou's classification. Overall, their results (Fig. 1D) appeared to favor the conclusions of Rothschild and Jordan [1] and Nakamura [44], as Smerinthinae were grouped with Sphinginae to the exclusion of Macroglossinae. Further evidence was needed, however, as the number of genera was small and three key tribes, Acherontiini, Ambulycini, and Sphingulini, were not included.
In this report we expand gene sampling of Regier et al. [32] to five protein-coding nuclear genes, and taxon sampling to 131 exemplars representing 106 sphingid genera. We use these data to test the monophyly of, and to estimate phylogenetic relationships within and among, all of the subfamilies, tribes and subtribes in the classification of Kitching and Cadiou [2]. We then provide a preliminary assessment of the new phylogeny's utility for understanding sphingid biogeography. We investigate the scale, if any, on which distribution is phylogenetically conserved, allowing inference about the geographic setting of major events in hawkmoth evolution.
Materials and Methods
Taxon sampling
We sequenced all sphingid genera for which we were able to obtain alcohol-preserved material. We included one additional taxon (Kloneus babayaga) for which a nucleic acid extract was kindly provided by Mehrdad Hajibabaei of the Canadian Centre for DNA Barcoding at University of Guelph. Sequences for several taxa were obtained from previously published studies [31], [32], [49], [50]. In total, 131 sphingid species in 106 genera were included, and they are listed in Table S1. All generic names follow the classification of Kitching and Cadiou [2] except Lintneria, which was reinstated for 21 largely Neotropical species of Sphinx [51], and Arctonotus, which was recently synonymized with Proserpinus [52]. As outgroups we included ten exemplars sequenced by Regier et al. [53] that represent eight other families placed in the Bombycoidea sensu stricto of Minet [29], [54] and Lemaire and Minet [30].
Gene sampling
Characters for this study come from five nuclear gene regions, totaling 6793 bp, which have previously proven useful for lepidopteran phylogenetics [55]. These include 2929 bp of CAD [56], 1282 bp of DDC [48], 1228 bp of EF-1α [47], 951 bp of period [57], and 403 bp of wingless [58]. GenBank accession numbers are listed in Table S1 and the entire aligned data matrix is available as supporting information (Dataset S1).
Nucleic acid extraction, RT-PCR, and primer sequences
Our nucleic acid extractions, mostly from adult moths, were generally taken from the head or prothorax, though in a few cases a leg was used because the rest of the body was unavailable (e.g., Aleuron chloroptera, Deidamia inscriptum). Nucleic acid extractions were conducted with the Promega SV Total RNA Isolation System [59], with slight protocol modifications (exclusion of part IV. E. steps 4, 5) to permit extraction of both genomic DNA and RNA. All specimens and extractions are stored at −85°C in the AToLep Collection at the University of Maryland. Wing voucher images of specimens sequenced can be viewed at the LepTree website (www.leptree.net).
Selective amplification of gene coding regions was conducted using the reverse transcription polymerase chain reaction (RT-PCR) to avoid introns, and because RT-PCR yielded better results than DNA-PCR in previous experiments [55]. All primer sequences are bipartite [60], containing a specific, 18 nt, M13 sequence at the 5′ end, and a gene-specific, 17–32 nt, degenerate sequence at the 3′ end. Protocols and strategies used for RT-PCR amplification for each gene are available online [55].
Data matrix construction
Sequence chromatograms were checked for accuracy and contigs edited and assembled with the Staden GAP4 software package [61]. Sequence alignment for each locus was conducted with MAFFT 6.611 [62] and manually checked with the Genetic Data Environment (GDE) software [63]. Each matrix was converted from FASTA to NEXUS format [64] and sequentially combined to create a single concatenated matrix of five genes using the “New Matrix Merge” command in WinClada [65].
Phylogenetic analysis
An optimal model for nucleotide maximum likelihood (ML) inference was chosen under the Akaike Information Criterion (AIC) [66] from 56 different models as implemented in Modeltest 3.7 [67]. In all cases, the best model was determined to be the general-time-reversible substitution model [68], [69], with among-site-rate-heterogeneity modeled according to a gamma distribution (Γ) [70] while allowing for a proportion of invariable sites (I) [71]. ML analyses were conducted using the multi-threaded GARLI 0.96b8 [72]. Searches began with a random starting tree, and none of the default settings was changed except for the number of generations to termination, which was increased to improve the search for the optimal solution (genthreshfortopoterm = 20,000). To enhance further the search for best tree, the same search process was repeated 1000 times with a different starting seed, and the best tree was chosen. Optimal tree searches were executed in parallel using Grid computing [73] through The Lattice Project [74]. A GARLI Grid service was written using GSBL, a high-level Application Programming Interface (API) for writing Globus-based Grid services [75]. Non-parametric ML bootstrap analyses were also conducted on the Grid, each consisting of 2000 bootstrap replicates with default settings.
Maximum parsimony analyses (MP) were conducted with WinClada [65] and NONA [76]. Heuristic searches were computed with the following commands: hold 1000, hold/100, mult*100, max*. Congruence between multiple most-parsimonious cladograms (MPCs) was assessed with a strict consensus [77]. Branch support was assessed with Bremer support (BS) [78], [79] and the bootstrap [80]. Bremer support values were calculated in TreeRot 2c [81], and subsequent command files executed in PAUP*4b10 [82]. MP bootstrap values were computed in NONA [76] with 500 replications, 100 search replications (mult*100), and holding 10 starting trees (hold/10). Throughout this report, for the sake of consistency, we adopt the somewhat arbitrary conventions that “moderate” node support means bootstrap values of 70–79%, “strong” support means values of 80–89%, and “very strong” support means values of 90% and above.
To evaluate the relative contribution of each gene to the total phylogenetic signal, and to assess conflicts among genes, ML bootstraps were calculated separately for each of the one-gene data sets. As a complementary approach, we also carried out “partition addition bootstrap alteration”, or PABA [83], on a series of four-gene data sets obtained by removing each of the genes in turn. NEXUS format files of all combinations of the remaining four genes were constructed with the same methods as the combined five-gene data set.
Biogeography
To explore broad biogeographic patterns in hawkmoths, we first determined, from various sources [e.g.], [43], [84,85], the distribution of each exemplar species across the traditional global biogeographic regions [86], [87]. We then simplified these distributions into a two-state character contrasting Old and New World, to test Kitching and Cadiou's [2] postulate of an Old World ancestral origin for the Sphingidae. We used Mesquite 2.5 [88] to calculate ancestral state conditions for biogeography in both an ML and MP framework. The Mkv model of Lewis [89], using the default search strategy of Mesquite, was chosen for all ML ancestral state reconstructions. Branch lengths were estimated from the molecular phylogeny.
Results
Phylogenetic analyses of five concatenated genes
The ML tree for the five genes combined, with associated bootstrap (BP) values, is shown in Fig. 2 and 3, while the corresponding MP tree is shown in Figure S1. The two trees were similar in topology, fully resolved or nearly so, and showed strong bootstrap support (≥80%) for more than two-thirds of the possible nodes. However, MP was susceptible to long-branch attraction (see discussion), and for this reason, the exposition below focuses on the ML tree except as otherwise noted. Relationships among outgroups, not shown here, were identical to those reported by Regier et al. [53].
The classification used here follows Kitching and Cadiou [2]. Larger-font numbers below branches are bootstrap values (>50%), smaller-font numbers to right of nodes are node numbers. Pie diagrams on nodes show proportions of total ancestral biogeographic state likelihood contributed by the states Old World (black) versus New World (gray), under ML (MK1).
The ML tree showed strong signal either for or against nearly all of the fifteen subfamily, tribal and sub-tribal entities recognized by Kitching and Cadiou [2], as well as many other groupings within and among these postulated by those authors (see also Table 1). Thus, our data set provides a strong test of those authors' hypothesis of classification and phylogeny. A detailed comparison of the present results to the hypothesis of Kitching and Cadiou [2] is provided in the Discussion.
Contributions of, and conflict among, individual genes
To characterize the individual gene contributions (or lack thereof) to the overall phylogenetic signal, we focused on a set of twenty relatively deep divergences (Table 1) that were strongly supported by the combined data (all BP≥88%; 16/20 with BP≥95%). When each gene was analyzed independently, all made strong contributions to at least some of these nodes (Table 1). The gene with the longest sequence, CAD, provided strong support (BP≥80%) for 13 of those nodes when analyzed independently. DDC alone strongly supported 14 nodes, despite a sequence length less than half that of CAD. EF-1α strongly supported five of the 20 nodes, period five, and wingless four. Most nodes were strongly supported by one or more individual genes, but four (nodes 2, 3, 27 and 29) had high combined-gene BP despite lack of strong support from any single gene, suggesting that additivity of weak but compatible signal from multiple genes accounts for about a fifth of these twenty instances of strong overall resolution. No gene strongly supported relationships that conflicted with those found in the all-gene tree, suggesting a general lack of marked conflict among genes.
The four-gene analyses provided a complementary assessment of individual gene contributions and conflict (Table 1). A decrease in BP for a node when a given gene is excluded, as compared to the all-genes analysis, was evidence that the gene provided support for that node. An increase in BP, conversely, implied that the gene conflicted with the node. We adopted the somewhat arbitrary heuristic criterion that a ≥10% change in BP represents a “significant” difference. Results from the four-gene analyses mirrored those from the single-gene analyses. Exclusion of CAD resulted in decreased BP support for thirteen nodes, four of which showed highly “significant” differences (≥25% BP). For the remaining genes, the corresponding numbers were: DDC, eight decreases, two “significant”; EF-1α, five decreases, one “significant;” period, four decreases, three “significant”; and wingless, six decreases, two “significant.” There was at least one apparent conflict (increase in BP upon exclusion) for each gene except wingless (CAD: 3; DDC: 1; EF-1α: 2; period: 1). However, these differences were mostly less than 5%, and all less than 10%, providing further evidence against strong conflict among genes.
Biogeography
On a broad, intercontinental scale, geographic distribution appeared to be strongly conserved on the ML phylogeny (Figs. 2, 3). Ancestral distribution can be confidently inferred for most clades at all levels. Thus, as we have seen, Smerinthinae+Sphinginae comprise an ancestrally and still predominantly Old World lineage, giving rise to substantial secondary New World radiations within Ambulycini, Smerinthini and Sphingini. Within these larger clades restricted mainly to one hemisphere are occasional inferred dispersals to the other. Examples include apparent dispersals to the Nearctic by ancestral Amorpha, to the Palearctic within Sphinx, and to the Neotropics within the Cocytius group and Agrius. Within Macroglossinae, there is a strong dichotomy between the clades comprising the predominantly Old World Macroglossini sensu stricto and New World Dilophonotina sensu lato, but there also appear to have been recent dispersals back to the New World in at least four nested lineages within Choerocampina and Macroglossina. Parsimony and likelihood also assigned an Old World ancestor to Macroglossinae (Fig. 3), and to the family as a whole, but this inference is somewhat less secure given our limited taxon sampling (see Discussion).
Discussion
Monophyly of Sphingidae, basal divergences, and subfamily definitions/relationships
Our analysis strongly corroborates morphological evidence for the monophyly of the Sphingidae [1], [2], [29], and supports a basal divergence within the family between Smerinthinae+Sphinginae (BP = 92%) and Macroglossinae (BP = 91%). Overall, the Smerinthinae+Sphinginae lineage most closely matches the morphological trees of Rothschild and Jordan [1], and Nakamura [44], but as yet no definitive morphological synapomorphy is known. While not yet objectively tested using morphology, monophyly of Macroglossinae was previously hypothesized on the basis of labial palp and pupal characters [2], [44]. Support for the monophyly of Macroglossinae came predominantly from CAD, as it provided the highest bootstrap support when analyzed alone, and bootstrap dropped from 91% to 49% when it was excluded (Table 1). CAD constitutes nearly 45% of the data set, so its major contribution to the node in question is expected. Monophyly of Smerinthinae+Sphinginae, and also for the Macroglossinae, however, was substantially lower in the MP analysis. Examination of the individual MP bootstrap trees revealed several long-branched taxa that were frequently grouping with outgroups and lowering bootstrap support (see below for further discussion on long-branch attraction). Basal divergences within Sphingidae are congruent with those of the much smaller molecular study by Regier et al. [32].
Within the Smerinthinae+Sphinginae lineage, our tree renders both Smerinthinae and Smerinthini paraphyletic with respect to Sphinginae by favoring very strongly a basal split between Langia zenzeroides and all remaining taxa (node 4, BP = 98%, Fig. 2). The position of Langia is unique in that it may illustrate the plesiomorphic condition of the Smerinthinae+Sphinginae. While an explicit analysis of optimization is necessary, Langia, Sphingulini, and most Smerinthinae share a larva that has a pointed triangular head, a full set of oblique lateral stripes, a rough “shagreened” skin, and an adult with a non-functional tongue.
Monophyly of Smerinthinae sensu lato, Smerinthini, and Sphingulini had previously been doubted on multiple morphological grounds (e.g., possession by various subgroups of Smerinthini of features found in other hawkmoths [1], [2], [44]). Smerinthinae sensu lato seem clearly paraphyletic, but our data nonetheless very strongly support monophyly for the majority of this subfamily, consisting of Ambulycini and Smerinthini minus Langia. Sphingulini is also paraphyletic, with Dolbina+Kentrochrysalis, to the exclusion of an Australian group represented here by Hopliocnema, comprising the sister group to Sphinginae (node 11, BP = 100%). We make no formal nomenclatural proposals here, but it would be reasonable to treat Sphinginae+Sphingulini as Sphinginae sensu stricto (node 7, BP = 100%), and its sister group, Ambulycini and Smerinthini minus Langia, as Smerinthinae sensu stricto (node 6, BP = 97%).
Relationships within “Smerinthinae sensu stricto” (Ambulycini+Smerinthini [minus Langia])
Monophyly of Ambulycini is strongly supported by molecules (node 53, Fig. 2) and by morphology (I. J. Kitching, unpublished data, 1993). All pupae examined so far in this tribe have a unique notch at the base of the cremaster [45] and many species also typically share larval host plants in Anacardiaceae or Lauraceae [90]. The tribe as sampled here divides into Neotropical (Adhemarius+Protambulyx) and Old World (Ambulyx+Amplypterus) sister groups. Kitching and Cadiou [2] hypothesized, however, that some Old World Ambulycini not sampled here (e.g., Akbesia, Batocnema, Compsulyx) may be more closely allied to the Neotropical species than to Ambulyx+Amplypterus, based on similarities of the hindwing eyespot pattern and a spinose gnathos in the male genitalia. These relationships would reinforce the inference of an Old World ancestor for Ambulycini.
Smerinthini, even excluding Langia and Sphingulini, are a morphologically heterogeneous group for which morphological synapomorphies are lacking. Basal divergences in this assemblage are weakly supported by molecular data, and monophyly can be neither confirmed nor confidently rejected. However, the molecular analysis provides strong support for a number of subgroups previously identified, at least in part, on morphological grounds [2]. Thus, monophyly is very strongly supported for both the Marumba group (node 72) and the Polyptychus group (with the addition of Chloroclanis, node 43) of Kitching and Cadiou [2], as well as a sister group relationship between these. Together these form a very strongly supported predominantly Afrotropical clade (node 32) with an apparent behavioral synapomorphy: in all species for which the immature stages are known, the first instar larva consumes only the eggshell before molting into the second instar [84], [91], [92]. Strong support is also found for another predominantly African clade, identified in part by Kitching and Cadiou [2], that is centered on Clanis (node 54).
Laothoe, Pachysphinx, Paonias, and Smerinthus form a very strongly supported Holarctic clade (node 31), identified in part by Kitching and Cadiou [2], within which generic relationships are strongly resolved. Laothoe, distributed from Ireland to Japan, feeds on Populus and Salix [84], which are also the sole larval hosts of the Nearctic Pachysphinx [90]. In contrast, one of the two most-recently diverging genera, Paonias, is polyphagous, feeding on a variety of host plants that differ in chemical composition [90]. Relationships among these genera suggest a single invasion from Old World to New World, followed by development of the full adult hindwing eyespot in the Nearctic Paonias and Smerinthus, a shift from monophagy to polyphagy in the former genus and reinvasion of the Palearctic by one or more lineages of the latter. Finally, the molecular data strongly corroborate the predicted sister group relationship between the eastern Nearctic Amorpha and the eastern Palearctic Phyllosphingia, both restricted to Juglandaceae [2], as well as a close relationship of these to the Palearctic Mimas.
Relationships within Sphinginae: Acherontiini and Sphingini
Basal divergences within Sphinginae are strongly resolved in our analyses (BP≥89%, Fig. 2). The first lineage to branch off comprises the Palaeotropical genus Xanthopan plus the Neotropical Cocytius/Neococytius (node 74, BP = 100%). This grouping, previously suggested by the shared possession of extremely long tongues [93], had been questioned on other morphological grounds [33], [34]. The remaining sphingines are split, with 100% BP support, into a larger, mostly New World clade containing the great majority of Sphingini (node 18), and a smaller, Old World clade consisting of the Acherontiini and their nearest relatives among the Sphingini (node 46). Monophyly of Acherontiini, strongly corroborated here, is supported by a labial palp feature and three characters of the genitalia [33], [34]. Relationships among the acherontiine genera included here are congruent with Kitching's expanded morphological parsimony analyses of this tribe [33], [34]. The sister clade to the Acherontiini is a set of genera centered on Psilogramma (node 78, BP = 100%), phylogenetic proximity among which had been previously predicted [2].
The phylogenetic placement of Acherontiini within Sphingini renders the latter paraphyletic. One reasonable way to achieve monophyly for tribes of Sphinginae would be to confer tribal status separately on the Psilogramma group, the Xanthopan/Cocytius group, and the remaining Sphingini, in addition to Acherontiini. Relationships among these four potential tribes according to the molecular analysis are largely congruent with the morphological phylogeny of Kitching [33] if the trees in the latter are re-rooted at the Xanthopan/Cocytius group.
Basal divergences within “Sphingini sensu stricto” (node 18, Fig. 2) are resolved with very strong support. As currently sampled, this clade appears to have an Old World tropical origin, as the Afrotropical Dovania is strongly placed as sister group to the remaining genera. We predict that the other, currently unsampled Old World sphingine genera placed near Dovania by Kitching and Cadiou [2] (e.g., the Afrotropical Hoplistopus, Litosphingia, Praedora, and the Oriental Apocalypsis and Pseudodolbina), will also prove to fall near the base of this clade.
The remaining Sphingini sensu stricto (node 24, Fig. 2) appear to represent an ancestrally and predominantly New World radiation, with repeated invasion of the north temperate regions from the tropics in both main lineages (centered on Manduca and Sphinx). We predict that the latter pattern will become clearer with further sampling, as our current exemplar selection is biased towards the Nearctic. Our sampling of the large genera Lintneria and Sphinx provides a probable instance of this postulated latitudinal trend. The two Neotropical species of Lintneria, L. istar and L. merops, are strongly placed (node 102, BP = 100%) as the sister group of Sphinx (plus two smaller genera; node 38, BP = 100%). The latter are all north temperate. Until recently, Lintneria was treated as part of a larger concept of Sphinx, despite Forbes [94] separation of it on the basis of a unique mesothoracic dorsal hump in the larva. Recent larval rearings have now corroborated these earlier observations by showing that the first four instars of all members of Lintneria for which they are known have this large, fleshy, anteriorly angled, dorsal protuberance, and the adults also have unique wing markings that are not found in Sphinx sensu stricto [51]. Our results are consistent thus with the reinstatement of Lintneria [51]. Relationships within the north temperate clade of Sphinx are not strongly resolved, possibly reflecting a very recent radiation, but the conifer feeding genera Isoparce and Lapara are nested within Sphinx and their generic status may need to be reconsidered.
Relationships within Macroglossinae
Basal divergences within Macroglossinae are less strongly resolved than those within its sister group Smerinthinae+Sphinginae, and correspond poorly to current classification, paralleling the lack of clear morphological divisions within this subfamily [2]. Three main lineages can be provisionally recognized. While only three samples could be included, there is very strong support for monophyly of Dilophonotini: Hemarina (node 80, BP = 100%, Fig. 3), as predicted by morphology [2]. The remaining macroglossines are grouped into two clades, each moderately well supported; these are placed as sister groups in our tree, to the exclusion of Hemarina, but with very weak support. These two clades correspond strikingly to geographic distribution, but less so to current tribal/subtribal boundaries. One (node 19, BP = 68%), restricted entirely to the New World, which we tentatively call, “Dilophonotina sensu lato”, consists of Dilophonotina, Philampelini, and the Proserpinus group and a set of genera, typified by Amphion and Sphecodina, both currently placed in Macroglossini: Macroglossina. Its ostensible sister group (node 8, BP = 72%), consisting of all remaining Macroglossina as well as Choerocampina, is ancestrally and predominantly confined to the Old World.
Genera within the New World clade fall into two main groups. One is a very strongly supported subgroup of dilophonotines (node 20, BP = 100%) that we term “Dilophonotina sensu stricto”, for which there are several synapomorphies in the male genitalia [2]. These sphingids, like Hemarina and many Old World Macroglossina, feed most commonly on euasterids, particularly Rubiaceae and/or Apocynaceae, and/or on other plants sharing with Apocynaceae the possession of latex or resin canals [95]. Cautethia spuria, whose placement within the New World clade is not strongly resolved, also feeds on Rubiaceae [90].
The other main New World lineage, moderately supported (node 28, BP = 72%), consists of additional Dilophonotina, Philampelini (represented here only by Eumorpha, as the Hawaiian endemic Tinostoma was not available for study), and most New World genera of Macroglossina. This grouping is a novel hypothesis, and no morphological synapomorphies are yet apparent, although a close relationship among some of the included Macroglossina had been tentatively postulated on the basis of their shared spinose tibiae [1], [2]. Its constituent species, however, differ strikingly from “Dilophonotina sensu stricto” and most Old World Macroglossina in host plant use, being almost entirely restricted to a recognized syndrome of sphingid host plant families [95] that includes Actinidiaceae, Dilleniaceae, Onagraceae and Vitaceae, all which are thought to share the presence of raphide crystals.
Within Macroglossini, there is strong support for inclusion of Choerocampina in a derived subset of Macroglossina; the two closest relatives in our sample are the Southeast Asian Gnathothlibus and Eupanacra (nodes 21, 27, BP≥95%). Monophyly of Choerocampina, strongly corroborated here (node 29, BP = 87%), is supported by a unique bat-detecting hearing organ formed from the swollen, scale-less, air-filled second labial palp segment and differentiated pilifer [2], [96]. Rothschild and Jordan [1] recognized three groups within Choerocampina, based largely on variations in the labial palp-pilifer hearing organ: (1) Chaerocina and Euchloron; (2) Deilephila, Hyles, Rhodafra, Xylophanes; (3) Basiothia, Cechenena, Centroctena, Hippotion, Pergesa, Rhagastis, and Theretra. Relationships among genera in our analysis, on the whole are strongly resolved, agree partially with this postulate. Three of the members of group 2, Hyles, Rhodafra, and Xylophanes, are indeed closely related, but the smallest clade including them also contains Chaerocina, and Deilephila is placed with group 3 rather than here. Euchloron is the sister taxon to group 2 thus re-defined (node 42, BP = 69%), rather than to Chaerocina alone, rendering group 1 polyphyletic. Members of group 3 are indeed phylogenetically contiguous, but the group is basal and paraphyletic with respect to group 2. Rothschild and Jordan [1] also postulated a sister group relationship between Cechenena and Rhagastis, which is very strongly supported in the present analysis (node 87, BP = 100%).
If we provisionally accept the monophyly of the three main lineages of Macroglossinae identified here, substantial changes to the current classification will be required to achieve a set of formally named monophyletic tribes and subtribes. As currently constituted, Dilophonotini and Macroglossini are both polyphyletic. If we further accept the weak to modest support for nesting of both Philampelini and part of Macroglossina within Dilophonotina (the strongest being node 28, BP = 72%), then Dilophonotina is also polyphyletic. One implication of this finding is that phylogenetic evidence for retaining Philampelini at tribal rank with its current composition is weak. Macroglossini can be restored to monophyly (“Macroglossini sensu stricto,” node 8) if the New World genera grouping with Dilophonotina are removed. Subtribe definition therein will be problematic, however, as Choerocampina are nested deep within Macroglossina.
Basal relationships within Macroglossinae were the main point of difference between our ML and MP results. In the MP strict consensus (Fig. S1), Hemarina do not lie at the base of Macroglossinae; rather, they fall at the base of Old World Macroglossina, as sister group to Neogurelca+Sphingonaepiopsis. The conflict cannot be called strong, as basal divergences are resolved even less strongly under MP than under ML. Inspection of the ML phylogram, however, suggested that the discrepancy might reflect long-branch attraction under MP [97], as both Hemarina and Neogurelca+Sphingonaepiopsis appear to be long-branched (Fig. S2). To explore this possibility further, the MP analysis was repeated with each group excluded in turn. When Neogurelca and Sphingonaepiopsis were excluded, Hemarina moved to base of the Macroglossinae as in the ML tree, whereas when Hemarina were excluded, Neogurelca+Sphingonaepiopsis moved to base of Sphingidae, allied with the longest branches in the analysis, namely those leading to the outgroups. These results are at least consistent with long-branch attraction.
Biogeography
A notable finding of this study is that broad-scale geographic distribution in sphingids is more phylogenetically conserved than previous classifications had implied. The ancestral distribution for Smerinthinae+Sphinginae is very clearly Old World, while the great majority of Macroglossinae are divided into two probable sister groups with sharply contrasting Old World (Macroglossini sensu stricto; node 8, Fig. 3) versus New World (Dilophonotina sensu lato; node 19, Fig. 3) distributions. The most notable uncertainty is the ancestral distribution for the basal macroglossine lineage, Hemarina, which contains only Hemaris and Cephonodes. This ambiguity is probably the result of our limited sampling of Hemaris. The single species included here is atypical in being North American; most of the genus, like Cephonodes, is restricted to the Old World. We predict that further phylogenetic study of Hemaris will show this genus to be ancestrally Old World, thereby more securely establishing an Old World origin for Hemarina, Macroglossinae and Sphingidae as a whole. Appeal to related families, in contrast is unlikely to further clarify the issue. Molecular studies now place Sphingidae in a clade with Saturniidae and Bombycidae sensu stricto [31], [98]. Saturniidae appear to have a New World origin, and the Bombycidae divide basally into New World versus Old World sister groups [53].
Supporting Information
Table S1.
The sampled 131 ingroup and 10 outgroup taxa with specimen localities, LepTree voucher identification numbers, and GenBank accession numbers.
https://doi.org/10.1371/journal.pone.0005719.s001
(0.29 MB DOC)
Figure S1.
Strict consensus of the 12 MPCs (length = 42618 steps, CI = 0.15, RI = 0.53) resulting from five-gene simultaneous MP analysis. Nodes are labeled to the right of each internal branch. Bootstrap values below branches, Bremer supports above.
https://doi.org/10.1371/journal.pone.0005719.s002
(2.03 MB TIF)
Figure S2.
ML phylogram. lnL = −187418.656372. The scale bar indicates the estimated substitutions per site.
https://doi.org/10.1371/journal.pone.0005719.s003
(1.73 MB TIF)
Dataset S1.
Data matrix. The aligned sequence data are presented in sequential Nexus format.
https://doi.org/10.1371/journal.pone.0005719.s004
(0.96 MB DOC)
Acknowledgments
This manuscript has profited from stimulating discussions with Michael P. Cummings, Soowon Cho, Jae-Cheon Sohn, and Isaac S. Winkler. Adam L. Bazinet and Greg Hess assisted in conducting many computationally intensive analyses. We thank Nathan Jud, Michael Knapp, Rodolphe Rougerie, and Andreas Zwick for their comments on the manuscript. We are very grateful to the many collectors who provided specimens used in this study, including: James K. Adams, Vitor O. Becker, Charles W. Bordelon, Tom Burbidge, Jeff Crolla, Robert F. Denno, John De Benedictis, Jurate and Willy De Prins, Keitaro Eda, Janet Farr, Timothy P. Friedlander, Michael Fibiger, Wayne W. Hsu, Roger W. Hutchings, John Ismay, Daniel Janzen and Winifred Hallwachs, William J. Kelly, Ed C. Knudson, Marcus J. Matthews, Charyn J. Micheli, Jacqueline Y. Miller, Mogens C. Nielsen, Richard S. Peigler, A. R. Pittaway, Daniel Rubinoff, D. Craig Rudolph, J. Bolling Sullivan, Pierre Tripotin, James P. Tuttle, Bruce Walsh, Kirby L. Wolfe, and Andreas Zwick.
Author Contributions
Conceived and designed the experiments: AYK AAM JR IJK CM. Performed the experiments: AYK AAM JR. Analyzed the data: AYK AAM. Contributed reagents/materials/analysis tools: JR. Wrote the paper: AYK JR IJK CM.
References
- 1. Rothschild LW, Jordan K (1903) A revision of the lepidopterous family Sphingidae. Novitates Zoologicae 9: 1–972.
- 2.
Kitching IJ, Cadiou J-M (2000) Hawkmoths of the world: annotated and illustrated revisionary checklist. Ithaca: Cornell University Press.
- 3.
Willis JH, Wilkins AS, Goldsmith MR (1995) A brief history of Lepidoptera as model systems. In: Goldsmith MR, Wilkins AS, editors. Molecular model systems in the Lepidoptera. Cambridge: Cambridge University Press. pp. 1–20.
- 4. Jindra M, Huang JY, Malone F, Asahina M, Riddiford LM (1997) Identification and mRNA developmental profiles of two ultraspiracle isoforms in the epidermis and wings of Manduca sexta. Insect Molecular Biology 6: 41–53.
- 5. Jochova J, Quaglino D, Zakeri Z, Woo K, Sikorska M, et al. (1997) Protein synthesis, DNA degradation, and morphological changes during programmed cell death in labial glands of Manduca sexta. Developmental Genetics 21: 249–257.
- 6. Kelber A, Balkenius A, Warrant EJ (2003) Colour vision in diurnal and nocturnal hawkmoths. Integrative and Comparative Biology 43: 571–579.
- 7. Wannenmacher G, Wasserthal LT (2003) Contribution of the maxillary muscles to proboscis movement in hawkmoths (Lepidoptera : Sphingidae) - an electrophysiological study. Journal of Insect Physiology 49: 765–776.
- 8.
Roe AD, Weller SJ, Baixeras J, Brown J, Cummings MP, et al. (2009) Evolutionary framework for Lepidoptera model systems. In: Goldsmith M, Marec F, editors. Genetics and molecular biology of Lepidoptera. Gainesville: CRC Press.
- 9. Eaton JL (1971) Morphology of the head and thorax of the adult tobacco hornworm, Manduca sexta (Lepidoptera: Sphingidae). International Journal of Insect Morphology and Embryology 3: 47–66.
- 10.
Eaton JL (1988) Lepidopteran Anatomy. New York: John Wiley and Sons.
- 11. Jackson DM (1990) Plant-insect behavioral studies: examples with Heliothis and Manduca species. Florida Entomologist 73: 378–391.
- 12. Osier TL, Traugott MS, Stamp NE (1996) Allelochemicals in tomato leaves affect a specialist herbivore Manduca sexta negatively but with no ill effects on a generalist insect predator, Podisus maculiventris. Oikos 77: 481–488.
- 13. Kessler A, Baldwin IT (2002) Manduca quinquemaculata's optimization of intra-plant oviposition to predation, food quality, and thermal constraints. Ecology 83: 2346–2354.
- 14. Agosta SJ, Janzen DH (2005) Body size distributions of large Costa Rican dry forest moths and the underlying relationship between plant and pollinator morphology. Oikos 108: 183–193.
- 15. Haber WA, Frankie GW (1989) A tropical hawkmoth community: Costa Rican dry forest Sphingidae. Biotropica 21: 155–172.
- 16. Gregory DP (1963–1964) Hawkmoth pollination in the genus Oenothera. Aliso 5: 357–419.
- 17. Nilsson LA (1998) Deep flowers for long tongues. Trends in Ecology and Evolution 13: 259–260.
- 18. Nilsson LA, Jonsson L, Rason L, Randrianjohany E (1985) Monophily and pollination mechanisms in Angraecum arachnites Schltr (Orchidaceae) in a guild of long-tongued hawk-moths (Sphingidae) in Madagascar. Biological Journal of the Linnean Society 26: 1–19.
- 19. Nilsson LA, Rabakonandrianina E, Pettersson B (1992) Exact tracking of pollen transfer and mating in plants. Nature 360: 666–668.
- 20. Raguso RA, Willis MA (2002) Synergy between visual and olfactory cues in nectar feeding by naive hawkmoths, Manduca sexta. Animal Behaviour 64: 685–695.
- 21. Wasserthal LT (1997) The pollinators of the Malagasy star orchids Angraecum sesquipedale, A. sororium and A. compactum and the evolution of extremely long spurs by pollinator shift. Botanica Acta 110: 343–359.
- 22. Beck J, Kitching IJ, Linsenmair KE (2006) Wallace's line revisited: has vicariance or dispersal shaped the distribution of Malesian hawkmoths (Lepidoptera: Sphingidae)? Biological Journal of the Linnean Society 89: 455–468.
- 23. Beck J, Kitching IJ, Linsenmair KE (2006) Effects of habitat disturbance can be subtle yet significant: biodiversity of hawkmoth-assemblages (Lepidoptera: Sphingidae) in Southeast-Asia. Biodiversity and Conservation 15: 465–486.
- 24. Bellotti AC, Arias VB, Guzman OL (1992) Biological control of the cassava hornworm Erinnyis ello (Lepidoptera: Sphingidae). Florida Entomologist 75: 506–515.
- 25. Winder JA (1976) Ecology and control of Erinnyis ello and E. alope, important insect pests in the New World. Proceedings of the National Academy of Sciences of the United States of America 22: 449–466.
- 26. Batra SWT (1984) Establishment of Hyles euphorbiae (L.) (Lepidoptera; Sphingidae) in the United States for the control of weedy spurges Euphorbia esula L. and E. cyparissias L. Journal of the New York Entomological Society 91: 304–311.
- 27. Janzen DH, Hajibabaei M, Burns JM, Hallwachs W, Remigio E, et al. (2005) Wedding biodiversity inventory of a large and complex Lepidoptera fauna with DNA barcoding. Philosophical Transactions of the Royal Society B-Biological Sciences 360: 1835–1845.
- 28. Hajibabaei M, Janzen DH, Burns JM, Hallwachs W, Hebert PDN (2006) DNA barcodes distinguish species of tropical Lepidoptera. Proceedings of the National Academy of Sciences of the United States of America 103: 968–971.
- 29. Minet J (1994) The Bombycoidea: phylogeny and higher classification (Lepidoptera, Glossata). Entomologica Scandinavica 25: 63–88.
- 30.
Lemaire C, Minet J (1998) The Bombycoidea and their relatives. In: Kristensen NP, editor. Lepidoptera, Moths, and Butterflies, Volume 1: Evolution, Systematics, and Biogeography. Berlin, New York: Walter de Gruyter. pp. 321–353.
- 31. Regier JC, Grant MC, Mitter C, Cook CP, Peigler RS, et al. (2008) Phylogenetic relationships of wild silkmoths (Lepidoptera: Saturniidae) inferred from four protein-coding nuclear genes. Systematic Entomology 33: 219–228.
- 32. Regier JC, Mitter C, Friedlander TP, Peigler RS (2001) Phylogenetic relationships and evolution of hostplant use in Sphingidae (Lepidoptera): initial evidence from two nuclear genes. Molecular Phylogenetics and Evolution 20: 311–316.
- 33. Kitching IJ (2002) The phylogenetic relationships of Morgan's Sphinx, Xanthopan morganii (Walker), the tribe Acherontiini, and allied long-tongued hawkmoths (Lepidoptera: Sphingidae, Sphinginae). Zoological Journal of the Linnean Society 135: 471–527.
- 34. Kitching IJ (2003) Phylogeny of the death's head hawkmoths, Acherontia [Laspeyres], and related genera (Lepidoptera: Sphingidae: Sphinginae: Acherontiini). Systematic Entomology 28: 71–88.
- 35. Derzhavets YA (1993) Phylogenetic interrelations of the sphinx moths of the genus Hyles Hbn. (Lepidoptera, Sphingidae). Entomologicheskoe Obozrenie 73: 648–663.
- 36. Hundsdoerfer AK, Kitching IJ, Wink M (2005) A molecular phylogeny of the hawkmoth genus Hyles (Lepidoptera : Sphingidae, Macroglossinae). Molecular Phylogenetics and Evolution 35: 442–458.
- 37. Hundsdoerfer AK, Kitching IJ, Wink M (2005) The phylogeny of the Hyles euphorbiae complex (Lepidoptera: Sphingidae): molecular evidence from sequence data and ISSR-PCR fingerprints. Organisms Diversity & Evolution 5: 173–198.
- 38. Rubinoff D, Le Roux JJ (2008) Evidence of repeated and independent saltational evolution in a peculiar genus of sphinx moths (Proserpinus: Sphingidae). PLoS ONE 3: e4035.
- 39.
Hennig W (1950) Grundzüge einer Theorie der phylogenetischen Systematik. Berlin: Deutscher Zentralverlag.
- 40. Hennig W (1965) Phylogenetic systematics. Annual Review of Entomology 10: 97–116.
- 41.
Hennig W (1966) Phylogenetic systematics. Urbana: University of Illinois Press.
- 42.
Janse AJT (1932) The moths of South Africa. Durban: E. P. & Commercial Printing.
- 43.
Hodges RW (1971) The moths of North America North of Mexico. Fascicle 21: Sphingoidea. London: E.W. Classey Limited & R.B.D. Publications.
- 44. Nakamura M (1976) An inference on the phylogeny of Sphingidae in relation to habits and the structures of their immature stages. Yugatô 63: 19–28.
- 45. Nakamura M (1977) Supplement to the pupae of Japanese Sphingidae (Lepidoptera). New Entomologist 26: 1–13.
- 46. Nakamura M (1978) The “cell length ratio” of the wing in Sphingidae, and its application to the classification. Yugatô 74: 111–116.
- 47. Cho S, Mitchell A, Regier JC, Mitter C, Poole RW, et al. (1995) A highly conserved nuclear gene for low-level phylogenetics: elongation factor-1α recovers morphology-based tree for heliothine moth. Molecular Biology and Evolution 12: 650–656.
- 48. Fang QQ, Cho S, Regier JC, Mitter C, Matthews M, et al. (1997) A new nuclear gene for insect phylogenetics: dopa decarboxylase is informative of relationships within Heliothinae (Lepidoptera: Noctuidae). Systematic Biology 46: 269–283.
- 49. Regier JC, Mitter C, Peigler RS, Friedlander TP (2000) Phylogenetic relationships in Lasiocampidae (Lepidoptera): initial evidence from elongation factor-1α sequences. Insect Systematics and Evolution 31: 179–186.
- 50. Caterino MS, Reed RD, Kuo MM, Sperling FAH (2001) A partitioned likelihood analysis of swallowtail butterfly phylogeny (Lepidoptera : Papilionidae). Systematic Biology 50: 106–127.
- 51.
Tuttle J (2007) The hawk moths of North America: a natural history study of the Sphingidae of the United States and Canada. The Wedge Foundation.
- 52. Rubinoff D, Osborne KH, Kawahara AY (2009) Synonomization of the euphonious Arctonotus lucidus (Sphingidae) based on recent DNA evidence. Journal of the Lepidopterists' Society.
- 53. Regier JC, Cook CP, Mitter C, Hussey A (2008) A phylogenetic study of the ‘bombycoid complex’ (Lepidoptera) using five protein-coding nuclear genes, with comments on the problem of macrolepidopteran phylogeny. Systematic Entomology 33: 175–189.
- 54. Minet J (1991) Tentative reconstruction of the ditrysian phylogeny (Lepidoptera, Glossata). Entomologica Scandinavica 22: 69–95.
- 55.
Regier JC (2008) Protocols, Concepts, and Reagents for preparing DNA sequencing templates. Version 12/4/08. www.umbi.umd.edu/users/jcrlab/PCR_primers.pdf.
- 56. Moulton JK, Wiegmann B (2003) Evolution and phylogenetic utility of CAD (rudimentary) among Mesozoic-aged eremoneuran Diptera (Insecta). Molecular Phylogenetics and Evolution 31: 363–378.
- 57. Regier JC, Fang QQ, Mitter C, Peigler RS, Friedlander TP, et al. (1998) Evolution and phylogenetic utility of the period gene in Lepidoptera. Molecular Biology and Evolution 15: 1172–1182.
- 58. Brower AVZ, DeSalle R (1998) Patterns of mitochondrial versus nuclear DNA sequence divergence among nymphalid butterflies: the utility of wingless as a source of characters for phylogenetic inference. Insect Molecular Biology 7: 73–82.
- 59.
Promega (2004) SV Total RNA Isolation System. Catalog no. Z3100, TM048. Madison: Promega Corp.
- 60. Regier JC, Shi D (2005) Increased yield of CR product from degenerate primers with nondegenerate, nonhomologous 5′ tails. BioTechniques 38: 34–38.
- 61.
Staden R, Beal KF, Bonfield JK (1999) The Staden package, 1998. In: Misener S, Krawetz SA, editors. Computer Methods in Molecular Biology, Bioinformatics Methods and Protocols. Totowa: The Humana Press. pp. 115–130.
- 62.
Katoh K (2008) MAFFT ver. 6.611, available online at: http://align.bmr.kyushu-u.ac.jp/mafft/software/.
- 63. Smith SW, Overbeek R, Woese CR, Gilbert W, Gillevet P (1994) The Genetic Data Environment. An expandable GUI for multiple sequence analysis. Computer Applications in the Biosciences 10: 671–675.
- 64. Maddison DR, Swofford DL, Maddison WP (1997) NEXUS: an extensible file format for systematic information. Systematic Biology 46: 590–621.
- 65.
Nixon KC (2002) WinClada ver. 1.00.08. New York: Published by the author.
- 66.
Akaike H (1973) Information theory as an extension of the maximum likelihood principle. In: Petrov BN, Csaki F, editors. Second International Symposium on Information Theory. Budapest: Akademiai Kiado. pp. 267–281.
- 67. Posada D, Crandall KA (1998) Modeltest: Testing the model of DNA substitution. Bioinformatics 14: 817–818.
- 68. Lanave C, Preparata G, Saccone C, Serio G (1984) A new method for calculating evolutionary substitution rates. Journal of Molecular Evolution 20: 86–93.
- 69. Tavaré S (1986) Some probablistic and statistical problems on the analysis of DNA sequences. Lectures on Mathematics in the Life Sciences 17: 57–86.
- 70. Yang Z (1994) Maximum-likelihood phylogenetic estimation from DNA sequences with variable rates over sites: approximate methods. Journal of Molecular Evolution 39: 306–314.
- 71. Gu X, Fu Y-X, Li W-H (1995) Maximum likelihood estimation of the heterogeneity of substitution rate among nucleotide sites. Molecular Biology and Evolution 12: 546–557.
- 72.
Zwickl DJ (2006) Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion. Ph.D. dissertation: The University of Texas at Austin. http://www.bio.utexas.edu/faculty/antisense/GARLI/GARLI.html.
- 73. Cummings MP, Huskamp JC (2005) Grid computing. EDUCAUSE Review 40: 116–117.
- 74.
Bazinet AL, Cummings MP (to appear) The Lattice Project: a Grid research and production environment combining multiple Grid computing models. In: Weber WHW, editor. Distributed & Grid Computing - Science Made Transparent for Everyone. Principles, Applications and Supporting Communities. Marburg: Tectum Publishing.
- 75. Bazinet AL, Myers DS, Fuetsch J, Cummings MP (2007) Grid Services Base Library: a high-level, procedural application program interface for writing Globus-based Grid services. Future Generation Computer Systems 22: 517–522.
- 76.
Goloboff P (1999) NONA (NO NAME), Version 2. Tucumán, Argentina: Published by the author.
- 77. Sokal RR, Rohlf FJ (1981) Taxonomic congruence in the Leptodomorpha re-examined. Systematic Zoology 30: 309–325.
- 78. Bremer K (1988) The limits of amino-acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 795–803.
- 79. Bremer K (1994) Branch support and tree stability. Cladistics 10: 295–304.
- 80. Felsenstein J (1985) Confidence-limits on phylogenies - an approach using the bootstrap. Evolution 39: 783–791.
- 81.
Sorenson MD (1999) TreeRot, Version 2. Boston: Boston University.
- 82.
Swofford DL (2002) PAUP*: Phylogenetic analysis using parsimony (*and other methods). Version 4.0b10. Sunderland: Sinauer Associates.
- 83. Struck TH, Purschke G, Halanych KM (2006) Phylogeny of Eunicida (Annelida) and exporing data congruence using a partition addition bootstrap alteration (PABA) approach. Systematic Biology 55: 1–20.
- 84.
Pittaway AR (1997–2006) Sphingidae of the Western Palaearctic. http://tpittaway.tripod.com/sphinx/list.htm.
- 85.
Pittaway AR, Kitching IJ (2006) Sphingidae of the Eastern Palaearctic. http://tpittaway.tripod.com/china/china.htm.
- 86.
Udvardy MDF (1975) A classification of the biogeographical provinces of the world. Morges: IUCN.
- 87.
Pielou EC (1979) Biogeography. New York: Wiley.
- 88.
Maddison WP, Maddison DR (2008) Mesquite: a modular system for evolutionary analysis. Version 2.5 http://mesquiteproject.org.
- 89. Lewis PO (2001) A likelihood approach to estimating phylogeny from discrete morphological character data. Systematic Biology 50: 913–925.
- 90.
Robinson GS, Ackery PR, Kitching IJ, Beccaloni GW, Hernández LM (2009) HOSTS - a database of the World's lepidopteran hostplants. http://www.nhm.ac.uk/research-curation/projects/hostplants/.
- 91.
Bell TRD, Scott FB (1937) Sphingidae. Fauna of British India, Vol. 5. London: Taylor and Francis.
- 92.
Sevastopulo DG (1949–1975) The early stages of east African Lepidoptera. Sphingidae I & II. London: Manuscript notes and photographs in the Entomology Library of The Natural History Museum.
- 93. Wasserthal LT (1998) Deep flowers for long tongues. Trends in Ecology and Evolution 13: 459–460.
- 94. Forbes WTM (1911) A structural study of the caterpillars. II. The Sphingidae. Annals of the Entomological Society of America 4: 261–279.
- 95. Harris P (1972) Food-plant groups of Semanophorinae (Lepidoptera-Sphingidae): A possible taxonomic tool. Canadian Entomologist 104: 71–80.
- 96. Roeder KD, Treat AE (1970) An acoustic sense in some hawkmoths (Choerocampinae). Journal of Insect Physiology 16: 1069–1086.
- 97. Felsenstein J (1978) Cases in which parsimony or compatibility methods will be positively misleading. Systematic Zoology 27: 401–410.
- 98. Zwick A (2008) Molecular phylogeny of Anthelidae and other bombycoid taxa (Lepidoptera: Bombycoidea). Systematic Entomology 2008: 190–209.