ZSC_093.fm Page 3 Friday, January 18, 2002 1:55 PM PhylogenyBlackwell Science Ltd of the holometabolous insect orders: molecular evidence MICHAEL F. WHITING Accepted: 5 October 2001 Whiting, M. F. (2002). Phylogeny of the holometabolous insect orders: molecular evidence. — Zoologica Scripta, 31, 3–15. Phylogenetic relationships among the holometabolous insect orders were reconstructed using 18S ribosomal DNA data drawn from a sample of 182 taxa representing all holometabolous insect orders and multiple outgroups. Parsimony analysis supports the monophyly of all holo- metabolous insect orders except for Coleoptera and Mecoptera. Mecoptera is paraphyletic with respect to Siphonaptera, which is nested within Mecoptera. Coleoptera is scattered as a paraphyletic assemblage across the tree topology. These data support a monophyletic Halteria (Strepsiptera + Diptera), Amphiesmenoptera (Trichoptera + Lepidoptera), Neuropterida (Neuroptera + (Megaloptera + Raphidioptera)), but Antliophora (Halteria + Mecoptera + Siphonaptera) and Mecopterida (Antliophora + Amphiesmenoptera) are paraphyletic. The limitations of using 18S ribosomal DNA as the sole phylogenetic marker for reconstructing insect ordinal relationships are discussed. Michael F. Whiting, Department of Zoology, Brigham Young University, Provo, UT 84602, USA. E-mail: [email protected] Introduction on a single character associated with the female ovipositor Accounting for more than 80% of insect species and more (Mickoleit 1973; Achtelig 1975). The highly derived order than 50% of all animal species (Wilson 1988; Kristensen Siphonaptera has been associated with Diptera or Mecoptera 1999), Holometabola is the most diverse and successful group based on different character suites (Boudreaux 1979; Hennig of terrestrial organisms. Holometabola comprises 11 insect 1981; Kristensen 1991). The most perplexing question, and orders, four of which — Coleoptera, Hymenoptera, Diptera that which has received the most attention in recent years, and Lepidoptera — account for over 99% of the species has been the placement of Strepsiptera among the other insect diversity of this group. Mecoptera, Strepsiptera, Megaloptera orders. Strepsiptera has been associated with Coleoptera, and Raphidioptera each contain less than 1000 described either within Polyphaga (Crowson 1960) or as sister group species, and Trichoptera, Neuroptera and Siphonaptera to Coleoptera, based on wing morphology and function each contain less than 4000 species. The monophyly of each (Kristensen 1981, 1991; Kathirithamby 1989; Kukalova-Peck insect order is relatively well supported by morphological & Lawrence 1993). Detailed examination of these putative data (Kristensen 1995, 1999; Whiting et al. 1997), with the synapomorphies, however, suggests that they are based on exception of Mecoptera which appears to be paraphyletic mistaken descriptions of strepsipteran wing morphology with respect to Siphonaptera (see Whiting 2002). Apart and function (Kinzelbach 1990; Pix et al. 1993; Whiting 1998b; from Holometabola itself, Amphiesmenoptera (Lepidoptera + Beutel & Haas 2000). There have been a number of reviews Trichoptera) is the only well-established interordinal relation- of phylogenetically informative characters for Holometabola ship, being supported by over 15 synapomorphies (Hennig with their accompanying phylogenetic hypotheses (Kristensen 1981; Kristensen 1997; Whiting et al. 1997). Other postulated 1975, 1981, 1991, 1995; Boudreaux 1979; Hennig 1981). interordinal relationships are based on relatively few morph- Whiting et al. (1997) presented the first formal quantitative ological characters or characters of questionable phylogenetic analysis of holometabolan relationships based on a coded utility. For example, no characters support a firm placement character matrix and also generated molecular sequence data for Hymenoptera, which has been postulated as sister group to from 18S and 28S ribosomal DNA (rDNA) for Holometabola ‘Meronida’ (Mecopterida + Neuropterida) (Boudreaux 1979) and outgroups. The summary topology from the total evidence or to Mecopterida (Kristensen 1991; 1999; Whiting et al. analysis of Whiting et al. (1997) is given in Fig. 1. Kristensen 1997). While a sister group relationship between Coleoptera (1999) presented an excellent review of holometabolan morph- and Neuroptera appears to be widely accepted, it is based ology, and his phylogenetic conclusions largely agree with © The Norwegian Academy of Science and Letters • Zoologica Scripta, 31, 1, February 2002, pp3–15 3 ZSC_093.fm Page 4 Friday, January 18, 2002 1:55 PM Holometabolan phylogeny • M. F. Whiting A. (Hennig) B. (Boudreaux) C. (Whiting) D. (Kristensen) Fig. 1 Previous hypotheses for holometabolan phylogeny. Hennig (A.) (1981), Boudreaux (B.) (1979) and Kristensen (D.) (1999) are based primarily on morphological data; Whiting et al. (C.) (1997) is based on a combination of morphological and molecular data. Dotted lines refer to poorly supported relationships. those of Whiting et al. (1997), except for uncertainty as to Mecoptera and Siphonaptera. The Carmean et al. (1992) and whether Strepsiptera should be placed as sister group to Chalwatzis et al. (1996) analyses suggested a paraphyletic Diptera or Coleoptera. Holometabola, and the Pashley et al. (1993) analysis was Given that morphology has led to some ambiguous phylo- unable to test for Holometabola paraphyly. The analysis of genetic relationships within this diverse group, it is not Whiting et al. (1997) supported a monophyletic Holometabola surprising that in the past few years some effort has been and a sister group relationship between Megaloptera and placed on using DNA sequence data to decipher interordinal Raphidioptera. The most intriguing result of these molecular phylogenetic relationships within Holometabola. Carmean analyses, and certainly the most controversial, was the evidence et al. (1992) sequenced a portion of 18S rDNA from 19 taxa presented for a well-supported sister group relationship between representing six holometabolan orders, one hemipteran and Strepsiptera and Diptera ( Whiting & Wheeler 1994; Carmean & one spider outgroup (Fig. 2A). Pashley et al. (1993) used 17 Crespi 1995; Kristensen 1995; 1999; Whiting & Kathirithamby 18S rDNA sequences to represent nine holometabolan orders 1995; Huelsenbeck 1997, 1998; Whiting et al. 1997; Whiting and one hemipteran outgroup, and found a monophyletic 1998a,b). Beyond the question of phylogenetic affinity of a Amphiesmenoptera and Mecopterida, but other relationships remarkable group, this result has been centre stage in debates were unresolved (Fig. 2B). Chalwatzis et al. (1996) sequenced over competing methods of phylogenetic reconstruction and 22 exemplars for 18S rDNA to represent nine holometabolan the role of a homeotic mutation in giving rise to novel mor- orders and four outgroup taxa (Fig. 2C). Whiting et al. (1997) phology in an insect group. A recent summary and re-analysis used 87 exemplars for 18S rDNA and 54 exemplars for 28S of holometabolous relationships based on 18S ribosomal data rDNA to represent all 11 holometabolan orders and 15 out- can be found in Whiting (2001). The analysis in this paper pres- group orders (Fig. 2D). All of these molecular analyses concur ents a broader selection of taxa, particularly from Coleoptera, in supporting Amphiesmenoptera (except for Carmean et al. than in the Whiting (2001) review, and is thus better able to 1992 who omitted these taxa) and a close association between address coleopteran paraphyly based on molecular data. 4 Zoologica Scripta, 31, 1, February 2002, pp3–15 • © The Norwegian Academy of Science and Letters ZSC_093.fm Page 5 Friday, January 18, 2002 1:55 PM M. F. Whiting • Holometabolan phylogeny Fig. 2 Previous molecular hypotheses for Holometabola based on 18S rDNA including sampled outgroups. Numbers in parentheses refer to the number of exemplars used as terminals. Materials and methods Tipulidae), taxa under-represented in previous analyses (e.g. 18S rDNA sequences were acquired from GenBank and Neuroptera) or taxa whose phylogenetic position is still con- augmented with sequences generated in the laboratory. Only troversial (e.g. Strepsiptera). This sampling strategy resulted sequences with a length of 1 kb or greater were used in this in 147 ingroup sequences representing all holometabolous analysis, and an attempt was made to represent as many holo- orders and 111 families (Appendix 1). Outgroup taxa were metabolous families as possible. Multiple sequences were used selected from Paraneoptera, the hypothesized sister group to to represent diverse families (e.g. Carabidae, Scarabaeidae, Holometabola (Whiting et al. 1997; Kristensen 1999), and © The Norwegian Academy of Science and Letters • Zoologica Scripta, 31, 1, February 2002, pp3–15 5 ZSC_093.fm Page 6 Friday, January 18, 2002 1:55 PM Holometabolan phylogeny • M. F. Whiting Fig. 3 Scheme of alignment for blocked variable regions. Regions that were ambiguously aligned between orders, but unambiguously aligned within orders, were aligned as blocks for each holometabolous insect order (represented as light boxes). Taxa outside of the blocked regions were coded with missing data. These regions were spliced together to create a step-like formation in the total alignment, where the variable blocked regions are flanked by conserved regions. The blocks were combined for Mecoptera and Siphonaptera and excluded for the outgroups. Polyneoptera (sensu Boudreaux 1979), with a sequence from simultaneously