Superfamily Membracoidea (Homoptera: Auchenorrhyncha). II. Cladistic analysis and conclusions

Abstract. Homologies among traditional morphological characters in the Membracoidea (sensu lato) are reassessed and the phylogenetic relationships among higher membracoid taxa are explored, incorporating new morphological evidence from nymphs and adults. Weighted and unweighted parsimony analyses of a matrix of sixty–three characters and thirty‐nine OTUs representing the families Aetalionidae, Cicadellidae, Melizoderidae and Membracidae, and an outgroup (superfamily Cercopoidea) yielded various topologies that were largely congruent but presented alternative hypotheses of relationships among the Membracidae. These analyses indicate that the superfamily consists of the following clades: Cicadellidae + (Melizoderidae + (Aetalionidae + Membracidae)). The family Membracidae, traditionally characterized by the presence of a posterior pronotal process, apparently gave rise to Nicomia Stål and other genera that lack this process.


Introduction
The phylogenetic relationships among the Membracoidea have been the subject of much discussion (Emel'yanov, 1987;Evans, 1946aEvans, . b, 1947Evans, , 1948Evans, , 1958Evans, , 1975bEvans, . 1977Hamilton, 1983;Haupt, 1929;Ross, 1957: Wagner, 1951, but have mainly focused on the family Cicadellidae. Few workers have attempted to estimate phylogenetic relationships among higher taxa of non-cicadellid Membracoidea (Haupt, 1929: Strumpel, 1972, and until now, no quantitative cladistic analyses of the membracoid family-group taxa have been published. Many aspects of Evans' (1046bEvans' ( , 1947Evans' ( , 1948Evans' ( , 1977 classification of the higher taxa of Membracoidea (as Jassoidea or Cicadelloidea), in which the Cicadellidae and the remaining Membracoidea are treated as sister groups, remain widely accepted (Nielson, 1985: Oman et ul., 1990. More recently, Hamilton (1983) suggcsted that the family Cicadellidae is paraphyletic, having given rise to the Membracidae and Aetalionidae. Both Evans and Hamilton underestimated the extent of morphological variation among the non-cicadellid Membracoidea. Our observations on the non-cicadellid Membracoidea (Deitz, 1975;Dietrich, 1989;and unpublished) complement the observations of other authors to form a substantial body of phylogenetically relevant information.
Here we present the results of our endeavours to determine the relationships among the non-cicadellid family-groups of Membracoidea.

Phylogenetic context
Outgroups Earlier attempts to estimate the phylogeny of the Auchenorrhyncha (reviewed by Garner, 1950, andEvans, 1963) produced no consensus. More recently, Hamilton (1981Hamilton ( , 1983 using cladistic criteria, treated either the superfamily Cercopoidea or the extinct family Jascopidae Hamilton. 1971, as the sister group of the Membracoidea. Hamilton's (1983) assertion that the Jascopidac represent a sister group of the Membracoidea requires that (1) the Jascopidae share at least one synapomorphy with Membracoidea, and (2) the taxa included in Jascopidae share at least one synapomorphy not present in Membracoidea. To date, only the first criterion is clearly documented (Hamilton, 1971a(Hamilton, , 1990(Hamilton, , 1992. The jascopid species Juscopus rtotuhi1i.s Hamilton and Hornopterulurn jelli Hamilton share putative synapomorphies with Cicadellidae: front and middle tibiae with setal rows, tarsal pectens narrow, hind tarsomere I with double row of plantar setae. These taxa also apparently differ from most Cicadellidae in having conical hind coxae and in lacking macrosetal rows on the hind tibiae, but these features are apparent plesiomorphies shared with Cercopoidea and other Auchenorrhyncha. Hamilton (1971aHamilton ( , 1992 listed no unique apomorphies shared by Jascopus Hamilton and Homopterulum Handlirsch, and evidence uniting these genera with other taxa placed in Jascopidae is weak. With the exception of Paracarsonus aphrodoides Hamilton, other taxa placed in the Jascopidae (Hamilton, 1990(Hamilton, ,1992 either have the hind legs poorly preserved or are represented by unassociated wings. Forms with preserved forewings have veins M and Cu fused for a short distance preapically, a putative synapomorphy: but the only species for which both forewings and hind legs are sufficiently preserved (P.aphrodoides) differs from Jascopus and Homopterulum in lacking macrosetae on the pro-and mesothoracic tibiae and plantar setae on hind tarsomere I. Thus, although taxa placed in the Jascopidae differ somewhat from modern Cicadellidae and Cercopidae, evidence supporting the monophyly of the Jascopidae and their status as a sister group of Membracoidea is tenuous (see Shcherbakov, 1992, for further discussion of the status of Jascopidae). Hamilton's (1981) treatment of the Cercopoidea and Membracoidea as sister groups seems reasonable. Apparent synapomorphies on the head (absence of a medial ocellus), thorax (pronotum extending to scutellum (except Cicadellidae)), wing (venation reduced) and abdomen (male with subgenital plate) unite these two superfamilies and distinguish them from the Cicadoidea and Fulgoroidea.
Unique apomorphies shared by most known Cercopoidea include: exposed areas of the integument (including wings) pubescent, forewing with subcostal ridge, meron of mesothoracic coxa large and acutely produced posterad, nymphs produce 'spittle'. Therefore the Cercopoidea apparently represent a monophyletic sister group of the Membracoidea and are an appropriate outgroup for our phylogenetic analysis.
Further support for the monophyly of the Membracoidea is found in the previously unstudied internal structure of the mesonotum. In the Membracoidea, the scutellar suture is associated with a pair of rod-shaped lateral apodemes, the ends of which are attached t o points along the suture (Figs 5-9, sa). The points of attachment of the scutellar apodemes are often indicated externally by the presence of pits. In those Membracidae having the scutellum com-pletely concealed by the pronotum, these pits are enlarged, shifted toward each other, and separated by an oblique external ridge (Fig. lo), and thus resemble those of Cicadoidea. The pits of cicadas, however, are not connected by an apodeme and are therefore not homologous. The presence of scutellar apodemes and their associated external pits is a feature unique to the Membracoidea.
In the Cercopoidea, the scutellar suture is associated with an internal ridge and a pair of elongate lateral fenestrae that are probably points of muscle attachment ( Fig. 4, f). The lateral fenestrae of the cercopoid scutellar suture occupy roughly the same position as the scutellar apodemes of Membracoidea and may be similar to the evolutionary precursors of the membracoid scutellar apodemes.
2. Beak apex: 0, reaching or surpassing metathoracic coxae: 3, not reaching metathoracic coxae. Pronotum 3. Posterior margin: 0, extending to or beyond scutellar suture: 3, not reaching scutellar suture. The scutum is entirely concealed by the pronotum in the Cercopoidea and most non-cicadellid Membracoidea. In the Cicadellidae, the posterior portion of the scutum and the scutellar suture are exposed.
41. Sternum IX and subgenital plate: 0, not fused; I , partially fused; 2, fused, but distinguishable; 3, fused, indistinguishable. The cercopoid subgenital plate is usually indistinguishably fused to sternum IX. Thus, based on the outgroup criterion, the articulated subgenital plates found in many membracoids may be derived from the fused subgenital plates similar to those found in the Cercopoidea. Yet Evans (1975a) provided evidence that the suhgenital plate of Membracoidea and Cercopoidea was derived from the gonocoxites and gonostyli of abdominal segment 1X of the primitive pterygote insect. Therefore, it remains unclear whether the fused condition was derived from the unfused condition or vice versa.

Data matrix
Our data (Table 2) consisted of the coded character states of the above characters based on adult and nymphal morphology of thirty-nine Operational Taxonomic Units (OTUs) ( Table 2). The OTUs (Table 1) represented all genera of the Aetalionidae and Melizoderidae, two unplaced membracoid genera, and an outgroup, the Cercopoidea. We also included representatives of four undescribed genera, including a fossil genus from Oligo-Miocene Dominican amber, that have character state combinations absent in other Membracoidea; formal descriptions will be presented elsewhere.
Because we were most interested in the relationships among the non-cicadellid Membracoidea, we included only a few representatives of the family Cicadellidae. We   also included only a few representatives of the 'higher Membracidae' -those having a n apically oblique clavus.  To give binary and ordered multistate characters equal wcight , we assigned states to each character ranging from 0 to 3; binary characters had states 0 a n d 3, and the intermediate states of multistate characters were assigned values of 1 and 2 for four-state characters and 2 for threestate characters. Thus, a change from the least t o most derived state in an ordered multistate character added the same length as a change in a binary character. Transformation series seemed dubious in characters 1 and 29, so we treated then as unordered initially and, for some analyses, gave them weights = 3 (thus, any transformation in o n e of these characters would add length equivalent t o a transformation in a binary character), W e searched for maximally parsimonious trees using Hennig86 (ver. 1.5, Farris, 1988) and used PAUP (ver. 3.0, Swofford, 1990)

t o check results and output apomorphy lists. T h e implicit enumeration algorithm of
Hennig86 proved too time consuming, so we used the heuristic branch breaking routine (command sequence 'mh; bb";'), which searches for as many equally parsimonious tress as will fit in memory, but is not guaranteed to find maximally parsimonious tree(s). To increase o u r chances of finding as many maximally parsimonious cladograms a s possible, we repeated each analysis several times using different orderings of the OTUs. W e also rearranged topologies manually (using the 'xx' facility of HennigS6).

Results and Discussion
Our preferred estimate of cladistic relationships among the Membracoidea (Fig. 31; apomorphy list, Table 3) was: (1) one of ten equally parsimonious tress found when characters 1 and 29 were unordered with weights of 1 or 3; and (2) the most parsimonious tree found when all characters were ordered with weights = 1. This tree had length = 683, consistency index (ci) = 77, and retention index (ri) = 63 when all characters had weight = 1 and characters 1 and 29 were unordered. Other topologies found were less parsimonious than Fig. 31 for one or more of the alternative weighting/ordering schemes. Nonetheless, these alternative trees varied only within component 73. Four iterations of successive weighting using the original most parsimonious tree, with all characters weighted equally and 1 and 29 unordered, as the initial estimate, produced a stable solution of one tree. This tree differed from the preferred topology only within component 72, but was twenty-five steps longer based on the original character weights and orderings. We therefore rejected the tree produced by successive weighting because it was substantially less parsimonious than the original tree (Fig. 31) for the overall data.
When wc treated all characters as unordered and of  Cicadellidae are united in having the mesonotum exposed posteriorly (character 3 ) , a feature that was apparently derived independently within the Biturritiinae. This result differs from that of Hamilton (1983), who suggested that the Aetalionidae and Membracidae were derived from within the Cicadellidae. Additional data are needed to elucidate relationships among the Cicadeliidae and to test our hypothesis of monophyly for this family.
Those Membracidae having the clavus of the forewing oblique and the pronotum at least partially concealing the scutellum also consistently formed a monophyletic group (component 58). Results of analyses of these 'higher' membracid taxa will be presented elsewhere.
Nymphs of the family Melizoderidae, component 44 (Fig. 31), have the anal opening completely concealed by tergum IX dorsally, a unique synapomorphy. This family consistently formed a sister group of the remaining non-cicadellid Membracoidea. Both genera of Melizoderidae have tergum IX of the nymph ensheathing segment X (Figs 23, 24). an apparently unique derivation shared with all Membracidae for which nymphs are known. Yet the Melizoderidae have parapsidal clefts on the mesonotum, a feature also present in Cicadellidae and Cercopoidea (Figs 4,5 ) , but absent in other Membracoidea (Figs 6-8, 10). The Melizoderidae resemble the Aetalionidae (component 51) in having a keeled scutellum, but lack other derived aetalionid features (such as fused prothoracic femora and trochanters). Thus, the Melizoderidae represent a distinct lineage, intermediate between Cicadellidae and Membracidae, but independent of Aetalionidae.
Component 51, the family Aetalionidae, was invariant among the maximally parsimonious topologies obtained. The monophyly of this component is supported by the fusion of the prothoracic trochanter and femur and by the presence of posterior processes on the female pygofer (Figs 19, 20), both unique derivations. The biturritiine genera of component 50 are united in having the head strongly produced dorsally and the scutellum enlarged. Aetafion and Darthula, component 45, are united with the Oligo-Miocene fossil in having reticulate forewing venation and two or fewer rows of cucullate setae on the hind tibia.
Component 73, the family Membracidae, is supported by the following synapomorphies: forewing veins M and Cu fused basally; male pygofer with lateral plate; nymphal abdominal tergum 1 with pair of tubercles (Fig. 30) Procyrtini. However, the relationships among these three clades and other members of component 72 varied considerably among the cladograms obtained. According to Fig. 31, the membracid subfamily Stegaspidinae is monophyletic, but on some topologies this group was paraphyletic, giving rise to various other membracid groups. Much of this instability may be attributable to the lack of data on immatures or males for several taxa.
Despite their occurrence as distinct lineages on the cladogram (Fig. 31), Antillotolania and Deiroderes share an apomorphy with the Stegaspidinae: crossvein m-cuz connects to vein M basad of its fork (character 27). Genera arising between nodes 69 and 58 (not inclusive) may also share this feature, but because they either have reticulate venation or have vein M 3-branched, we were unable to interpret their venational homologies.
The topologies obtained from unweighted data imply that the posterior process of the membracid pronotum arose twice, once in the Stegaspidinae (component 71) and once in component 63. However, morphological evidence suggests that the posterior pronotal process arose once and was lost in various lineages. In the Centrotinae and Stegaspidinae, this process usually rests in a groove or emargination in the scutellum. The genera Tolaniu, Deiroderes, Antillotolania, Euwalkeria, and new genera A and C, all of which lack posterior pronotal processes, nonetheless have the scutellum similarly grooved or emarginate. This suggests that the ancestors of these genera had posteriorly produced pronota. The posterior pronotal process was apparently also lost in two centrotine genera (Deitz, 1985). Among the Membracidae, only the subfamily Endoiastinae arose prior to the initial derivation of the posterior pronotal process.

Biogeography
The predominantly Neotropical distribution of the noncicadellid Membracoidea (Fig. 31, component 75) suggests that this lineage originated in the New World. The most plesiomorphic group, family Melizoderidae, is known only from Chile and Argentina. Among the Aetalionidae, the subfamily Biturritiinae is exclusively Neotropical; but the disjunct Neotropical (Aetalionini) and Oriental (Darthulini) distribution of the subfamily Aetalioninae, and the presence of fossils in Oligo-Miocene Dominican amber, imply that the family was once much more widespread and may have originated prior to the Tertiary.
The cosmopolitan distribution of the subfamily Centrotinae also suggests a pre-Tertiary origin for the family Membracidae (Wood, 1993). However, because the most plesiomorphic membracid subfamilies (Endoiastinae, Stegaspidinae, Nicomiinae and Centronodinae) are restricted to the New World, either these subfamilies had a restricted distribution in tropical Gondwanaland (Wood, 1993), or the Membracidae did not arise until after the Tertiary isolation of South America. Under the latter scenario, the Centrotinae would have arisen in the New World and reached the Old World subsequently through dispersal. This hypothesis might be tested by assessing the cladistic relationships among New and Old World Centrotinae. Clearly, the Membracidae of the Eastern and Western Hemispheres diversified in isolation (Wood, 1993), for the two regions have one subfamily, one tribe, and, at most, one endemic genus in common (Deitz, 1975).