Phylogenetic analysis and revision of subfamily classification of Belostomatidae genera (Insecta: Heteroptera: Nepomorpha)

Phylogenetic analysis and revision of subfamily classification of Belostomatidae genera (Insecta:... Abstract Recent investigations into relationships among the cosmopolitan Belostomatidae have led to recognition of the Belostomatinae, Horvathiniinae and Lethocerinae clades. Here we investigate the relationships among the genera of this family (with Appasus, Benacus and Kirkaldyia now resurrected, and three fossils: Cratonepa, Iberonepa and Lethocerus vetus) by including modifications or clarifications of both somatic and genitalic characters (including some spermatheca traits), as well as multiple genes (COI, 18S rDNA and 16S rDNA). A comparative study of these genera yielded putative homology hypotheses coded as 104 morphological characters and 1829 aligned characters. A majority-rule tree utilizing the Bayesian method [the ‘master’ tree with highlighted clades estimated by maximum parsimony and maximum likelihood (ML)] is as follows: (Cratonepa, Iberonepa, (Benacus, Kirkaldyia, Lethocerus), (((Horvathinia, Hydrocyrius), Limnogeton), (((Abedus, Weberiella), Belostoma), (Appasus, Diplonychus))))). Lethocerinae and Belostomatinae (now with Horvathinia included) form two monophyletic groups. Weberiella appears as the sister group of Abedus, and Belostomatini had to be reformulated. In all trees, the monophyly of Diplonychinitrib. nov. (Appasus+Diplonychus) was recovered with high support indices. This study highlights a global tree based on combining the molecular and morphological data representing topologies from separate analyses, irrespective of the existence of missing data and the type of dataset. INTRODUCTION Water bugs or electric-light bugs are common and well-known insects from aquatic habitats throughout the world’s subtropical and tropical areas (Lauck & Menke, 1961; Merritt & Cummins, 1996). The family Belostomatidae Leach, 1815 comprises 11 genera and approximately 150 species, most of which are reported from the New World (Polhemus, 1995; Schuh & Slater, 1995; Perez–Goodwyn, 2006; Estévez & Ribeiro, 2011; Moreira et al., 2011) and are found in limnic habitats such as rivers, streams, lakes and ponds. They are ambush predators, often hunting by night, and representatives of some genera (e.g. LethocerusMayr, 1853) are known to capture a wide variety of prey, including woodpeckers (Hungerford, 1919) and turtles (Ohba, 2011). Spooner (1938) intuitively proposed that Belostomatidae and Nepidae Latreille, 1802 are closely related, by indicating morphological similarity. Also, the conspicuous phallus of representatives of these two families, with the apex fully sclerotized, is a condition uniquely found among Nepoidea, so that this was highlighted as evidence of a relationship between them by Dupuis (1955). Using cladistic methods, Mahner (1993), Hebsgaard, Andersen & Damgaard (2004) and Hua et al. (2009) have suggested Belostomatidae and Nepidae as sister groups. This monophyletic group was defined by the non-retractable condition of the air straps and their origin from the eighth abdominal segment. Lauck & Menke (1961) intuitively subdivided Belostomatidae into three subfamilies based on aspects of male genitalia morphology and in a more critical analysis of other characteristics, such as the condition of the abdominal sternites divided or not by a suture and the degree of development of tarsal claws of the first pair of legs. Mahner (1993) undertook the most comprehensive phylogenetic analysis of the generic and suprageneric relationships within Belostomatidae based on morphology. Mahner also showed that the family is monophyletic, being supported by the presence of a metepistern quite developed in nymphs and a hairy conspicuous stripe visible on the connexivum of adults. Finally, based on combined analysis (with 16S+28S rDNA) by Hebsgaard et al. (2004), Belostomatidae is supported by the following two unambiguously optimized morphological changes: (1) conical metacoxae, firmly united with metapleuron; and (2) hind tibiae flattened, with swimming hairs. Lauck & Menke’s (1961) subfamily classification of Belostomatidae has been adopted by many authors (De Carlo, 1966; Nieser, 1975; Mahner, 1993; Schuh & Slater, 1995; Schnack & Estévez, 2005; Perez-Goodwyn, 2006; Estévez & Ribeiro, 2011; Moreira et al., 2011), and only one subfamily has been subjected to revision (Lethocerinae: Perez-Goodwyn, 2006). Belostomatinae is a widespread subfamily with the highest number of described genera among the family Belostomatidae. AbedusStål, 1862 is restricted to the Nearctic Region, Mexico, where there is the largest number of representatives, and to Central America. BelostomaLatreille, 1807, the most diverse genus, occurs in the New World, with the largest number of representatives in South America (Lauck & Menke, 1961; Moreira et al., 2011). AppasusAmyot & Serville, 1843 and Diplonychus Laporte, 1833 are distributed throughout Southeast Asia, Africa, India and Australia. HydrocyriusSpinola, 1850 is restricted to the African continent, and little is known about its biology. LimnogetonMayr, 1853, restricted to north-eastern Africa, comprises apparently strict predators of snails (Voelker apudSchuh & Slater, 1995). Its members bear smooth profemora, lacking sulci, and are thought to be the only representatives of Belostomatidae which do not have middle and posterior legs modified for swimming (Ribeiro et al. 2014). WeberiellaDe Carlo, 1966 is monobasic, constituted by Weberiella rhomboides (Menke, 1965), and it is recorded from the states of Amazonas, Mato Grosso and Rondônia in Brazil, as well as French Guyana and Venezuela (Menke, 1965; De Carlo, 1966; Estévez & Ribeiro, 2011). The record of this species from Roraima, Brazil, is incorrect (F. F. F. Moreira, pers. comm.), and its members exclusively inhabit the surface film of freshwaters (kinon by Fittkau, 1977). Horvathiniinae is a monotypic subfamily (with HorvathiniaMontandon, 1911), and its members are restricted to the southeast and central south of South America (Lauck & Menke, 1961). The genus comprises only two species, Horvathinia lentiDe Carlo, 1957 and H. pelocoroidesMontandon, 1911 (Schnack & Estévez, 2005). In addition, only a few representatives of these species have been collected or observed in their natural habitat, and almost nothing has been published about their biology (see Armúa-de-Reyes, Schnack & Estévez, 2005). Lethocerinae sensu stricto, now reformulated as a subfamily comprising the genera BenacusStål, 1861, KirkaldyiaMontandon, 1909 and LethocerusMayr, 1853 (according to Perez-Goodwyn, 2006), is a cosmopolitan subfamily, being better represented in the Neotropical Region. In this subfamily, Lethocerus maximusDe Carlo, 1938 is known to be the largest of all Heteroptera, and can attain at least 110.0 mm (Monte, 1945). Having undertaken his phylogenetic analysis of suprageneric relationships within Belostomatidae based on morphology, Mahner (1993) documented the generic recognition of several putatively monophyletic groups. In addition, Mahner’s (1993) phylogeny supports the monophyly of the subfamilies created by Lauck & Menke (1961), and many features of the external morphology and male genitalia erected by these authors were important synapomorphies in his analysis. The author also indicated that Horvathiniinae and Belostomatinae are sister groups, sharing the fusion of phallosoma and ventral diverticulum. Mahner created Belostomatini within Belostomatinae, a tribe consisting of Abedus and Belostoma, based on the presence of 2-segmented tarsi bearing only one visible claw. However, Mahner did not include Appasus, Benacus and Kirkaldyia in his analysis, invalid or unavailable genera at that time, and considered Belostoma and Weberiella as genera without putative synapomorphies. As Mahner’s phylogeny neither included such resurrected genera nor used molecular data as additional information, the relationship among Belostomatidae genera has been rendered incomplete and outdated. Moreover, Mahner did not present a matrix of characters nor consequently the distribution of their states he proposed, making it very difficult to reproduce his analysis. According to Andersen (1995), Mahner gave up the use of computer algorithms, because the characters he proposed were ‘reliable’ enough that these computational efforts would be superfluous. Here, for the first time, we present phylogenetic relationships among the genera of Belostomatidae based on characters of external morphology, spermatheca, and male and female genitalia, utilizing analysis of portions of the 18S and 16S rDNA genes and the mitochondrial gene COI. The goal was to conduct groups of different analyses, either with each gene separately (‘separated analysis’ of De Queiroz, Donoghue & Kim, 1995), with only morphological data, with the genes combined in the same matrix or with all available data (‘simultaneous analysis’ of De Queiroz et al., 1995). As the taxonomy of Belostomatidae is still in an incipient state regarding the study of females, we used some characters based on the morphology of the spermatheca. Since we found evidence to support Appasus and Diplonychus as a monophyletic group, the new tribe Diplonychini was established to accommodate that clade. As various data sources were used to obtain different phylogenetic relationships, we tested the hypothesis that the quality and the kind of information did not change the results. MATERIAL AND METHODS Taxon sampling and terminology Species exemplars of all 13 Belostomatidae genera were selected to estimate the genus-level phylogeny of belostomatids, including three fossil taxa and four monobasic genera. Type-species were selected when available, but otherwise preference was given to species recently collected in ethanol for the molecular analyses. Additional species were included for larger genera, especially those belonging to Belostoma, which are harder to characterize morphologically and some of which are apparently para- or polyphyletic. Outgroup taxa included were selected to test the monophyly of Belostomatidae. Eleven outgroup taxa representing four Nepidae genera were used: NepaLinnaeus, 1758, LaccotrephesStål, 1866, CurictaStål, 1861 and RanatraFabricius, 1790. These outgroups have been given as closely related to Belostomatidae in previous phylogenetic analyses (Mahner, 1993; Hua et al., 2009; Li et al., 2012; Brożek, 2014). Acronyms of the institutions that loaned or donated the specimens are listed in Table 1 together with a list of material studied. Table 1. Species studied for the phylogenetic analysis of Belostomatidae. The origin (country and region) and depository are provided for each species. Underlined localities refer to voucher specimens. An asterisk indicates no available information for that particular taxon in at least one data partition Taxa  Depositories†  Localities  BELOSTOMATIDAE  Abedus dilatatus (Say)  MNHN  MEXICO  Abedus indentatus (Haldeman)  LEBIP  MEXICO: Ciudad de México (Mpio. Temascaltepec)  Abedus signoreti Mayr*  LEBIP  MEXICO: Ciudad de México (Mpio. Luvianos), Baja California  Abedus ovatus (Stål)  LEBIP  MEXICO: Michoacán  Appasus ampliatus (Montandon)  MNHN  IVORY COAST: Foro-Foro; CENTRAL AFRICAN REPUBLIC: Lamaboké; SENEGAL: Kedougou  Appasus capensis (Mayr)  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga  Appasus grassei (Poisson)  MNHN  IVORY COAST: Lamto; SENEGAL: Dakar  Appasus japonicus Vuillefroy  LEBIP  CHINA: Gan Chouen Fou, Xiao Bei Lake; JAPAN (Sayo, Hyogo)  Appasus major (Esaki)*  MNHN/LEBIP  CHINA: Mandchourie; JAPAN: Sayo  Appasus nepoides (Fabricius)  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga; SENEGAL: Sangalkam; IVORY COAST: Adiopodoumé  Appasus procerus  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga; CAMEROON: Batouri; CENTRAL AFRICAN REPUBLIC: Lamaboké  Belostoma angustum Lauck*  LEBIP  BRAZIL: Rio Grande do Sul; URUGUAY: La Paloma  Belostoma cummingsi De Carlo  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma dilatatum (Dufour)  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma elongatum Montandon*  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma flumineum Say*  LEBIP  USA: Kansas  Belostoma harrisi Lauck  LEBIP  SURINAM: Paramaribo  Belostoma micantulum (Stål)  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma nicaeum Estévez & Polhemus  DZRJ  BRAZIL: Amazonas  Belostoma oxyurum (Dufour)  MNRJ  BRAZIL: Paraná  Belostoma plebejum (Stål)*  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma ribeiroi De Carlo*  LEBIP  BRAZIL: Mato Grosso do Sul  Belostoma testaceopallidum Latreille  DZRJ  BRAZIL: Rio de Janeiro, Santa Catarina  Benacus griseus (Say)*  MNHN  USA: New York; GUYANA  Diplonychus esakii Miyamoto & Lee*  LEBIP  REPUBLIC OF THE CONGO: Zanga-Di-Makodia; VIETNAM: Tainin  Diplonychus rusticus (Fabricius)*  LEBIP  REPUBLIC OF THE CONGO: Nduizi River; GenBank  Horvathinia pelocoroides Montandon  LEBIP/MLPA/MNRJ/ Nieser collection  BRAZIL: Minas Gerais, Mato Grosso, Rio Grande do Sul, Santa Catarina; PARAGUAY: Guaíra  Hydrocyrius colombiae Spinola  MNHN/MRAC  EGYPT; REPUBLIC OF THE CONGO: Myamirundi, Bas Ogoqué; CAMEROON: Yaoundé; SENEGAL: Saint Luis; MADAGASCAR: Farafangan; IVORY COAST: Adiopodoumé; SUDAN: Dogo  Hydrocyrius nanus Montandon  MNHN/MRAC  REPUBLIC OF THE CONGO; CAMEROON: Yaoundé; CENTRAL AFRICAN REPUBLIC: Lamaboké  Hydrocyrius punctatus Stål  MNHN  MADAGASCAR. Tananarive  Hydrocyrius rectus Mayr  MNHN/MRAC  CONGO: Brazzaville, Leopoldville  Hydrocyrius sp.  MNHN/LEBIP  MADAGASCAR: Tananarive; REPUBLIC OF THE CONGO: Zanga-Di-Makodia  Kirkaldyia deyrolli (Vuillefroy)  MNRJ/LEBIP  CHINA: Chekiang Province; JAPAN: Kumamoto, Tokyo, Alpes de Nikko; LAOS: Baudan  Lethocerus annulipes (Herrich-Schäffer)*  LEBIP  BRAZIL: Rio de Janeiro; Rio Grande do Sul  Lethocerus cordofanus Mayr  MNHN  SOMALIA: Glohar  Lethocerus delpontei De Carlo  LEBIP  BRAZIL: Rio Grande do Sul  Lethocerus maximus De Carlo  LEBIP  FRENCH GUYANA: Route de Pointe Combi  Limnogeton expansum Montandon  MACN/MNHN  REPUBLIC OF THE CONGO: Brazzaville; CAMEROON: Batouri, Nyaounderé; TANGANYIKA: Mlingano  Limnogeton fieberi Mayr  MNHN/MRAC  EGYPT; SUDAN; CAMEROON: Garoua  Limnogeton hedenborgi (Stål)  MNHN/MRAC  SENEGAL: Kolda; EGYPT; REPUBLIC OF THE CONGO; CAMEROON: Yaoundé; UGANDA: Victoria Nyanza  Limnogeton scutellatum Mayr  MNHN/MRAC  EGYPT; ETHIOPIA; CENTRAL AFRICAN REPUBLIC: Lamaboké; DEMOCRATIC REPUBLIC OF THE CONGO: Molindi River; REPUBLIC OF THE CONGO: Brazzaville; ZAMBIA: Muliba  Weberiella rhomboides (Menke)*  INPA/DZRJ/LEBIP /MNHN  BRAZIL: Amazonas, Mato Grosso; FRENCH GUIANA: Piste Coralie  NEPIDAE  Curicta borellii Montandon  LEBIP  BRAZIL: Rio Grande do Sul  Curicta cf. pelleranoi  LEBIP  BRAZIL: Mato Grosso do Sul  Curicta volxemi (Montandon)*  LEBIP  BRAZIL: Mato Grosso do Sul  Laccotrephes japonensis (Scott)  LEBIP  JAPAN: Kumamoto  Laccotrephes pfeiferiae (Ferrari)  LEBIP  CHINA: Tianjin (Wuqing Country)  Laccotrephes sp.*  LEBIP  CHINA: Tianjin (Wuqing Country)  Nepa cinerea Linnaeus  LEBIP  FRANCE: Vayral  Nepa hoffmanni Esaki*  LEBIP  JAPAN: Hyogo  Ranatra brevicauda Montandon  LEBIP  BRAZIL: Mato Grosso do Sul  Ranatra chinensis Mayr*  LEBIP  JAPAN: Kumamoto  Ranatra heydeni Montandon*  LEBIP  BRAZIL: Mato Grosso do Sul  Ranatra robusta Montandon  LEBIP  BRAZIL: Mato Grosso do Sul; Rio Grande do Sul  Ranatra sattleri De Carlo  LEBIP  BRAZIL: Mato Grosso do Sul  Taxa  Depositories†  Localities  BELOSTOMATIDAE  Abedus dilatatus (Say)  MNHN  MEXICO  Abedus indentatus (Haldeman)  LEBIP  MEXICO: Ciudad de México (Mpio. Temascaltepec)  Abedus signoreti Mayr*  LEBIP  MEXICO: Ciudad de México (Mpio. Luvianos), Baja California  Abedus ovatus (Stål)  LEBIP  MEXICO: Michoacán  Appasus ampliatus (Montandon)  MNHN  IVORY COAST: Foro-Foro; CENTRAL AFRICAN REPUBLIC: Lamaboké; SENEGAL: Kedougou  Appasus capensis (Mayr)  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga  Appasus grassei (Poisson)  MNHN  IVORY COAST: Lamto; SENEGAL: Dakar  Appasus japonicus Vuillefroy  LEBIP  CHINA: Gan Chouen Fou, Xiao Bei Lake; JAPAN (Sayo, Hyogo)  Appasus major (Esaki)*  MNHN/LEBIP  CHINA: Mandchourie; JAPAN: Sayo  Appasus nepoides (Fabricius)  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga; SENEGAL: Sangalkam; IVORY COAST: Adiopodoumé  Appasus procerus  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga; CAMEROON: Batouri; CENTRAL AFRICAN REPUBLIC: Lamaboké  Belostoma angustum Lauck*  LEBIP  BRAZIL: Rio Grande do Sul; URUGUAY: La Paloma  Belostoma cummingsi De Carlo  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma dilatatum (Dufour)  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma elongatum Montandon*  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma flumineum Say*  LEBIP  USA: Kansas  Belostoma harrisi Lauck  LEBIP  SURINAM: Paramaribo  Belostoma micantulum (Stål)  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma nicaeum Estévez & Polhemus  DZRJ  BRAZIL: Amazonas  Belostoma oxyurum (Dufour)  MNRJ  BRAZIL: Paraná  Belostoma plebejum (Stål)*  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma ribeiroi De Carlo*  LEBIP  BRAZIL: Mato Grosso do Sul  Belostoma testaceopallidum Latreille  DZRJ  BRAZIL: Rio de Janeiro, Santa Catarina  Benacus griseus (Say)*  MNHN  USA: New York; GUYANA  Diplonychus esakii Miyamoto & Lee*  LEBIP  REPUBLIC OF THE CONGO: Zanga-Di-Makodia; VIETNAM: Tainin  Diplonychus rusticus (Fabricius)*  LEBIP  REPUBLIC OF THE CONGO: Nduizi River; GenBank  Horvathinia pelocoroides Montandon  LEBIP/MLPA/MNRJ/ Nieser collection  BRAZIL: Minas Gerais, Mato Grosso, Rio Grande do Sul, Santa Catarina; PARAGUAY: Guaíra  Hydrocyrius colombiae Spinola  MNHN/MRAC  EGYPT; REPUBLIC OF THE CONGO: Myamirundi, Bas Ogoqué; CAMEROON: Yaoundé; SENEGAL: Saint Luis; MADAGASCAR: Farafangan; IVORY COAST: Adiopodoumé; SUDAN: Dogo  Hydrocyrius nanus Montandon  MNHN/MRAC  REPUBLIC OF THE CONGO; CAMEROON: Yaoundé; CENTRAL AFRICAN REPUBLIC: Lamaboké  Hydrocyrius punctatus Stål  MNHN  MADAGASCAR. Tananarive  Hydrocyrius rectus Mayr  MNHN/MRAC  CONGO: Brazzaville, Leopoldville  Hydrocyrius sp.  MNHN/LEBIP  MADAGASCAR: Tananarive; REPUBLIC OF THE CONGO: Zanga-Di-Makodia  Kirkaldyia deyrolli (Vuillefroy)  MNRJ/LEBIP  CHINA: Chekiang Province; JAPAN: Kumamoto, Tokyo, Alpes de Nikko; LAOS: Baudan  Lethocerus annulipes (Herrich-Schäffer)*  LEBIP  BRAZIL: Rio de Janeiro; Rio Grande do Sul  Lethocerus cordofanus Mayr  MNHN  SOMALIA: Glohar  Lethocerus delpontei De Carlo  LEBIP  BRAZIL: Rio Grande do Sul  Lethocerus maximus De Carlo  LEBIP  FRENCH GUYANA: Route de Pointe Combi  Limnogeton expansum Montandon  MACN/MNHN  REPUBLIC OF THE CONGO: Brazzaville; CAMEROON: Batouri, Nyaounderé; TANGANYIKA: Mlingano  Limnogeton fieberi Mayr  MNHN/MRAC  EGYPT; SUDAN; CAMEROON: Garoua  Limnogeton hedenborgi (Stål)  MNHN/MRAC  SENEGAL: Kolda; EGYPT; REPUBLIC OF THE CONGO; CAMEROON: Yaoundé; UGANDA: Victoria Nyanza  Limnogeton scutellatum Mayr  MNHN/MRAC  EGYPT; ETHIOPIA; CENTRAL AFRICAN REPUBLIC: Lamaboké; DEMOCRATIC REPUBLIC OF THE CONGO: Molindi River; REPUBLIC OF THE CONGO: Brazzaville; ZAMBIA: Muliba  Weberiella rhomboides (Menke)*  INPA/DZRJ/LEBIP /MNHN  BRAZIL: Amazonas, Mato Grosso; FRENCH GUIANA: Piste Coralie  NEPIDAE  Curicta borellii Montandon  LEBIP  BRAZIL: Rio Grande do Sul  Curicta cf. pelleranoi  LEBIP  BRAZIL: Mato Grosso do Sul  Curicta volxemi (Montandon)*  LEBIP  BRAZIL: Mato Grosso do Sul  Laccotrephes japonensis (Scott)  LEBIP  JAPAN: Kumamoto  Laccotrephes pfeiferiae (Ferrari)  LEBIP  CHINA: Tianjin (Wuqing Country)  Laccotrephes sp.*  LEBIP  CHINA: Tianjin (Wuqing Country)  Nepa cinerea Linnaeus  LEBIP  FRANCE: Vayral  Nepa hoffmanni Esaki*  LEBIP  JAPAN: Hyogo  Ranatra brevicauda Montandon  LEBIP  BRAZIL: Mato Grosso do Sul  Ranatra chinensis Mayr*  LEBIP  JAPAN: Kumamoto  Ranatra heydeni Montandon*  LEBIP  BRAZIL: Mato Grosso do Sul  Ranatra robusta Montandon  LEBIP  BRAZIL: Mato Grosso do Sul; Rio Grande do Sul  Ranatra sattleri De Carlo  LEBIP  BRAZIL: Mato Grosso do Sul  †List of specimen depositories is based on Arnett, Samuelson & Nishida (1993), except for DZRJ, LEBIP and MNRJ. DZRJ, Departamento de Zoologia, Instituto de Biologia, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil; LEBIP, Laboratório de Estudos da Biodiversidade do Pampa, São Gabriel, Universidade Federal do Pampa, Rio Grande do Sul, Brazil; MNRJ, Museu Nacional, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil. View Large Identification was based on the following authors: Menke (1960) for Abedus; Lauck (1962, 1963, 1964), Nieser (1975), Estévez & Polhemus (2001) and Ribeiro (2007) for Belostoma; Perez-Goodwyn (2006) for Benacus, Kirkaldyia and Lethocerus; Keffer (1996) for Curicta; Poisson (1949, 1954), Lee (1991) and Ribeiro et al. (2014) for Appasus, Hydrocyrius and Limnogeton; De Carlo (1964), Lansbury (1972), Nieser (1975) and Chen, Nieser & Ho (2004) for Ranatra; Lansbury (1972) and Keffer (2004) for Nepa; Polhemus & Polhemus (2013), Keffer (2004) and Nieser, Zettel & Chen (2009) for Lacotrephes. The terminology of wing parts follows Gorb & Perez-Goodwyn (2003), and of genitalia Dupuis (1955), Dupuis & Carvalho (1956), Scudder (1959), Lansbury (1972), Lalitha, Shyamasundari & Rao (1997), Ribeiro (2007) and Keffer (2004). Techniques for preparation of male and female genitalia were adapted from those of Ribeiro (2007). Dissected parts of the male and female genitalia are stored in microvials with glycerin. The spermatheca was studied by dissection of fresh or dried material according to the protocol proposed by Pluot-Sigwalt (1986, 2008). The spermatheca was examined after the abdomen had been removed from the specimen and placed into cold 10% KOH solution for 12–24 h, depending on the size of the insect. This treatment removes tissues and does not affect the cuticular intima of the ectodermal part of the internal genitalia. Following washing of the abdomen in distilled water and rinsing in 2% liquid detergent DECON90, the genital tract was transferred to glycerol. The organ was then dissected out and stained in noir chlorazol according to the method of Carayon (1969). Spermatheca samples were stored in vials with glycerol. The terminology used for spermatheca parts follows Pellerano & De Carlo (1985), Pluot-Sigwalt (2008) and Viscarret & Bachmann (1997). Many of the differences and peculiarities found therein are known to occur at genus level, which makes us believe that the morphology of the spermatheca is probably useful for understanding the evolution and systematics of these organisms (Pellerano & De Carlo, 1985; Viscarret & Bachmann, 1997). To give an accurate impression of structures such as antennae, brush-to-brush frictional surfaces of the hemelytra, the clavus–clavus clamp of hemelytra and spines on the ventral diverticulum of male genitalia, we obtained scanning electron micrographs of dorsal, lateral and ventral views of the aforementioned structures using a Hitachi scanning electron microscope. Preparations were critical-point-dried, mounted on holders and sputter-coated with gold–palladium (10 nm) when sufficient material was available. If not, such material was examined at 5 kV, without a coat of gold–palladium. DNA preparation Insect specimens collected in the field were placed directly in 95–100% ethanol and stored at −20 °C until processing. To amplify genes, genomic DNA was extracted from a single foreleg and associated muscles using a modified ethanol precipitation/resuspension protocol (Bender, Spierer & Hogness 1983) or the DNEasy tissue kit (Qiagen Inc.). In some cases, multiple individuals from the same species were extracted. All belostomatid vouchers were stored in 99.5% ethanol pro-analysis (PA) at −20 °C deposited at the LEBIP (Laboratório de Estudos da Biodiversidade do Pampa, Universidade Federal do Pampa, Rio Grande do Sul, Brazil), except for W. rhomboides, which was dried and is deposited at the MNHN (Museum National d’Histoire Naturelle, Paris, France). For amplification of genes from W. rhomboides, genomic DNA was extracted following protocols adapted from those of Gilbert et al. (2007), Thomsen et al. (2009) and Lis, Ziaja & Lis (2011), which allowed us to recover amplifiable DNA from dried museum specimens that were up to 10 years old. The specimens were pinned again and redeposited in the collection as vouchers. PCR, sequencing and alignment Modified primers based on Simon et al. (1994), von Dohlen & Moran (1995) and Simon et al. (2006) were used to amplify parts of the mitochondrial gene COI (~730 bp) and 16S rDNA (~480 bp). Nuclear 18S sequences were amplified using the primers Ns1 and Ns2a (Barker et al., 2002) (~600 bp) (see Table 2). Amplification was carried out in a 25-µL volume reaction, with 5 µL of Taq and Load Mastermix, 0.5 µL of each primer at 25 µM and 2 µL of extraction product. For 16S loci, all reactions were performed using HifiTaq DNA Polymerase – an enzyme mixture that greatly increases fidelity and amplification of genomic targets. For COI reactions, a ‘step-up procedure’ was used: 2 min at 95 °C, five amplification cycles to improve DNA stock of 1 min at 95 °C, 15 min at 45 °C and 1 min at 72 °C; followed by 35 cycles of 40 s at 95 °C, 1 min at 51 °C and 1 min at 72 °C, with a final extension at 72 °C for 5 min. The protocol for the 16S rDNA region was: 94 °C for 3 min and 80 °C for 20 min (hot start), followed by 35 cycles of 94 °C for 1 min, 50 °C for 1 min and 72 °C for 2 min, with a final extension at 72 °C for 7 min. The protocol for 18S rDNA was 94 °C for 2 min, 35 cycles of 94 °C for 30 s, 49 °C for 30 s and 72 °C for 30 s, with a final extension at 72 °C for 7 min. Negative controls (no template) were always run simultaneously with our PCR experiments. All reaction mixtures were discarded when any DNA appeared in the negative control. Table 2. Oligonucleotide primer sequences. Primers used in polymerase chain and sequencing reactions of COI, 18S and 16S Primer  Direction  Sequence 5′→ 3′  Alias  Reference  COI+  F  GGTCAACAAATCATAAAGATATTGG  LCO-1490  Folmer et al. (1994)   COI+a  F  GGAGGATTTGGAAATTGATTAGTTCC  C1-J-1718  Simon et al. (1994)   COI-  R  TAAACTTCAGGGTGACCAAAAAATCA  HCO-2198  Folmer et al. (1994)   COI+*  F  TGTAAAACGACGGCCAGTTTTCAACWAATCATAAAGATATTGG  LCO1490puc_t1  Cruaud et al. (2010)   COI+a*  F  TGTAAAACGACGGCCAGTTTTCAACTAAYCATAARGATATYGG  LCO1490Hem1_t1  Germain et al. (2013)   COI-*  R  CAGGAAACAGCTATGACTAAACTTCWGGRTGWCCAAARAATCA  HCO2198puc-t1  Cruaud et al. (2010)   COI-a*  R  CAGGAAACAGCTATGACTAAACYTCAGGATGACCAAAAAAYCA  HCO2198Hem2-t1  Germain et al. (2013)   COI-b*  R  CAGGAAACAGCTATGACTAAACYTCDGGATGBCCAAARAATCA  HCO2198Hem1-t1  Germain et al. (2013)   18S+  F  GTAGTCATATGCTTGTCTC  Ns1  Barker et al. (2002)   18S-  R  CGCGGCTGCTGGCACCAGACTTGC  Ns2a  Barker et al. (2002)   16S+  F  CCGGTYTGAACTCARATCA  LR-J-12887  Takiya et al. (2006)   16S-  R  CRMCTGTTTAWCAAAAACAT  LR-N-13398  Takiya et al. (2006)   16S-b  R  TAAGTGTGCAAAGGTAGC  16S-Lepto-F  Malm & Johanson (2008)   Primer  Direction  Sequence 5′→ 3′  Alias  Reference  COI+  F  GGTCAACAAATCATAAAGATATTGG  LCO-1490  Folmer et al. (1994)   COI+a  F  GGAGGATTTGGAAATTGATTAGTTCC  C1-J-1718  Simon et al. (1994)   COI-  R  TAAACTTCAGGGTGACCAAAAAATCA  HCO-2198  Folmer et al. (1994)   COI+*  F  TGTAAAACGACGGCCAGTTTTCAACWAATCATAAAGATATTGG  LCO1490puc_t1  Cruaud et al. (2010)   COI+a*  F  TGTAAAACGACGGCCAGTTTTCAACTAAYCATAARGATATYGG  LCO1490Hem1_t1  Germain et al. (2013)   COI-*  R  CAGGAAACAGCTATGACTAAACTTCWGGRTGWCCAAARAATCA  HCO2198puc-t1  Cruaud et al. (2010)   COI-a*  R  CAGGAAACAGCTATGACTAAACYTCAGGATGACCAAAAAAYCA  HCO2198Hem2-t1  Germain et al. (2013)   COI-b*  R  CAGGAAACAGCTATGACTAAACYTCDGGATGBCCAAARAATCA  HCO2198Hem1-t1  Germain et al. (2013)   18S+  F  GTAGTCATATGCTTGTCTC  Ns1  Barker et al. (2002)   18S-  R  CGCGGCTGCTGGCACCAGACTTGC  Ns2a  Barker et al. (2002)   16S+  F  CCGGTYTGAACTCARATCA  LR-J-12887  Takiya et al. (2006)   16S-  R  CRMCTGTTTAWCAAAAACAT  LR-N-13398  Takiya et al. (2006)   16S-b  R  TAAGTGTGCAAAGGTAGC  16S-Lepto-F  Malm & Johanson (2008)   *In some species of Ranatra and Laccotrephes COI sequences were amplified with these published primers. View Large Sequences were sent to the Genoscope (Centre National de Séquençage, Evry). The sequences obtained were read and cleaned with CodonCode (CodonCode Corporation, 2002–2015). All sequences were aligned using MUSCLE (Edgard, 2004), and MAFFT v.7 (Katoh & Standley, 2013) for non-coding sequences. In all regions, gaps were checked manually and treated as missing data. This procedure yielded approximately 1830 bp per taxon, although sequences for some taxa were not complete (Table 1). All Belostomatidae and Nepidae sequences are original accessions, except Abedus brevicepsStål, 1862, Curicta scorpioStål, 1862 and Nepa apiculata Uhler, 1862, which were taken from GenBank. NCBI accession numbers and references are summarized in Table 3. Table 3. Material examined with accession numbers for belostomatid genes. A dash (–) indicates that a region was not sequenced for that species because no PCR product could be obtained and hyphen (-) indicates sequences from different PCR products Taxon  GenBank acession(s): COI, 18S, 16S  BELOSTOMATIDAE  Abedus indentatus (Haldeman)  KY320459, KY389074, KY389116  Abedus signoreti Mayr  KY320460 - KY320461, KY389073 - KY389075, KY389110 - KY389122  Appasus japonicus Vuillefroy  KY320473, –, KY389104  Appasus major (Esaki)  KY320474, KY389067, KY389098 - KY389137  Belostoma angustum Lauck  KY320477, KY389084, –  Belostoma cummingsi De Carlo  KY320478, –, KY389115  Belostoma elongatum Montandon  –, KY389076, –  Belostoma flumineum Say  KY320475, KY389085, KY389121 - KY389127 - KY389132  Belostoma plebejum (Stål)  KY320476 - KY320479, KY389086 - KY389087, KY389097  Belostoma ribeiroi De Carlo  KY320462, KY389079, KY389103 - KY389109  Diplonychus esakii Miyamoto & Lee  KY320464 - KY320465, KY389068 - KY389069, KY389113 - KY389119 - KY389125 - KY389131  Diplonychus rusticus (Fabricius)  KY320466, KY389088, KY389136  Horvathinia pelocoroides Montandon  KY320467, KY389080, KY389100  Hydrocyrius colombiae Spinola  KY320481 - KY320482, KY389090, KY389139  Hydrocyrius sp.  KY320483, KY389089, –  Benacus griseus (Say)  –, –, –  Kirkaldyia deyrolli (Vuillefroy)  KY320484, KY389091, KY389106 - KY389112 - KY389118  Lethocerus annulipes (Herrich-Schäffer)  KY320486 - KY320487, KY389093, KY389133 - KY389138  Lethocerus delpontei De Carlo  –, KY389070 - KY389094, KY389128  Limnogeton expansum Montandon  KY320488, KY389095, KY389124 - KY389130 - KY389135  Weberiella rhomboides (Menke)  –, –, KY389102  NEPIDAE  Curicta borellii Montandon  KY320480, –, KY389141  Curicta cf. pelleranoi  –, KY389078, KY389123  Curicta volxemi (Montandon)  KY320463, KY389077, KY389129 - KY389134  Laccotrephes japonensis (Scott)  KY320485, KY389071, KY389108  Laccotrephes pfeiferiae (Ferrari)  KY320469, KY389092, KY389101  Laccotrephes sp.  KY320468, –, –  Nepa cinerea Linnaeus  KY320489, KY389096, KY389140  Nepa hoffmanni Esaki  KY320470, KY389072, KY389107  Ranatra brevicauda Montandon  KY320471, KY389083, KY389117  Ranatra chinensis Mayr  KY320490, –, KY389114 - KY389120 - KY389126  Ranatra heydeni Montandon  KY320491, KY389081, KY389099 - KY389105  Ranatra robusta Montandon  KY320472, KY389082, KY389111  Taxon  GenBank acession(s): COI, 18S, 16S  BELOSTOMATIDAE  Abedus indentatus (Haldeman)  KY320459, KY389074, KY389116  Abedus signoreti Mayr  KY320460 - KY320461, KY389073 - KY389075, KY389110 - KY389122  Appasus japonicus Vuillefroy  KY320473, –, KY389104  Appasus major (Esaki)  KY320474, KY389067, KY389098 - KY389137  Belostoma angustum Lauck  KY320477, KY389084, –  Belostoma cummingsi De Carlo  KY320478, –, KY389115  Belostoma elongatum Montandon  –, KY389076, –  Belostoma flumineum Say  KY320475, KY389085, KY389121 - KY389127 - KY389132  Belostoma plebejum (Stål)  KY320476 - KY320479, KY389086 - KY389087, KY389097  Belostoma ribeiroi De Carlo  KY320462, KY389079, KY389103 - KY389109  Diplonychus esakii Miyamoto & Lee  KY320464 - KY320465, KY389068 - KY389069, KY389113 - KY389119 - KY389125 - KY389131  Diplonychus rusticus (Fabricius)  KY320466, KY389088, KY389136  Horvathinia pelocoroides Montandon  KY320467, KY389080, KY389100  Hydrocyrius colombiae Spinola  KY320481 - KY320482, KY389090, KY389139  Hydrocyrius sp.  KY320483, KY389089, –  Benacus griseus (Say)  –, –, –  Kirkaldyia deyrolli (Vuillefroy)  KY320484, KY389091, KY389106 - KY389112 - KY389118  Lethocerus annulipes (Herrich-Schäffer)  KY320486 - KY320487, KY389093, KY389133 - KY389138  Lethocerus delpontei De Carlo  –, KY389070 - KY389094, KY389128  Limnogeton expansum Montandon  KY320488, KY389095, KY389124 - KY389130 - KY389135  Weberiella rhomboides (Menke)  –, –, KY389102  NEPIDAE  Curicta borellii Montandon  KY320480, –, KY389141  Curicta cf. pelleranoi  –, KY389078, KY389123  Curicta volxemi (Montandon)  KY320463, KY389077, KY389129 - KY389134  Laccotrephes japonensis (Scott)  KY320485, KY389071, KY389108  Laccotrephes pfeiferiae (Ferrari)  KY320469, KY389092, KY389101  Laccotrephes sp.  KY320468, –, –  Nepa cinerea Linnaeus  KY320489, KY389096, KY389140  Nepa hoffmanni Esaki  KY320470, KY389072, KY389107  Ranatra brevicauda Montandon  KY320471, KY389083, KY389117  Ranatra chinensis Mayr  KY320490, –, KY389114 - KY389120 - KY389126  Ranatra heydeni Montandon  KY320491, KY389081, KY389099 - KY389105  Ranatra robusta Montandon  KY320472, KY389082, KY389111  View Large Phylogenetic analyses Morphological characters were identified based on their topographical identity before proposing hypotheses of primary homology (de Pinna, 1991) by defining the states in the data matrix. All characters were assigned equal weights and treated as unordered. Multistate characters were treated as unordered under Fitch parsimony. The program Mesquite v.3.04 (Maddison & Maddison, 2015) was used for matrix preparation. Analyses of the morphological dataset were conducted using the cladistic method, considered as the current paradigm for systematic research based on morphological characters (de Pinna, 1991). Most-parsimonious trees were derived under the maximum-parsimony criterion using exact searches (branch-and-bound algorithm) with the addition sequence=furthest. The parsimony inference (PI) analyses utilizing morphological data were carried out under the equal weighting scheme (as for all the other analyses here). Because non-applicable characters are treated as missing data, we instructed the software to collapse zero-length branches (Brazeau, 2011). Multistate taxa were counted as uncertainty. All algorithms were implemented in PAUP* 4.0a152 (Swofford, 2002). Character states were optimized a posteriori, using the ACCTRAN criteria (de Pinna, 1991). Character states were scored as dashes (-) if inapplicable and as question marks (?) if ambiguous or missing (Appendix 1). Some of the characters we analysed were based on those of Lee (1991), Mahner (1993) and Keffer (2004), as well as Gorb & Perez-Goodwyn’s (2003) ideas. Character coding was contingent (Forey & Kitching, 2000) and the character statements were formulated following Sereno (2007). Character optimizations were reconstructed in MacClade (Maddison & Maddison, 2005) over trees based on morphological and total evidence analyses. Only unambiguously optimized characters are discussed in the text. Character optimizations were made both over morphological consensus and over combined morphological and molecular analyses, such that if any character optimization over their trees was different, both are treated in the Discussion. Branch support was assessed by 1000 non-parametric bootstrap replicates (Felsenstein, 1985) using heuristic searches, where each search was conducted using random additional sequences with branch-swapping by tree-bisection-reconnection (TBR) and ten replicates and hold=10. Apart from the parsimony analyses, we analysed our matrix of morphological characters using Bayesian inference (BI) of phylogeny using a likelihood model suitable for morphological data. We did so because this approach permits branch lengths to be estimated and helps in the resolution of clades in the presence of high levels of homoplasy, given that the likelihood approach takes the aforementioned branch lengths into consideration. Thus, by combining both approaches we could test whether our parsimony analysis could be under the influence of long-branch attraction (Magalhães & Santos, 2012). For that, two analyses were performed, one under the Mkv model and one under the MkvΓ model, which allows rate variation amongst characters to follow a gamma distribution. These are the variations of the ML models of Lewis (2001) for discrete morphological data (datatype=standard), implemented in MrBayes 3.2 (Ronquist et al., 2012). In the first model, we assumed equal rates of character change. The second one accounted for rate variation across characters (lset rates=gamma), i.e. Mkv model, for discrete morphological data allowing rate heterogeneity among characters. Flat priors were adopted here. For each of these analyses, four independent runs with four chains each were run for five million generations, sampling trees every 500 generations. Ten per cent of the trees were discarded as burn-in and the remaining were used to calculate the posterior probabilities. The best Bayesian topology was selected by comparing the harmonic mean of the log-likelihood of each of the two aforementioned analyses, as a way of estimating the marginal likelihood. The analysis with the harmonic mean closest to 0 is preferred. Also, we calculated twice the difference between log marginal likelihoods (Bayes factor) of these two BI analyses (Müller & Reisz, 2006). As a rule of thumb, a Bayes factor of greater than ten usually indicates strong support (Kass & Raftery, 1995). With this in mind, parsimony-uninformative autapomorphic characters were included in the data matrix as proposed by Yeates (1992), which made it possible to estimate all branch lengths (Lewis, 2001). All PI analyses utilizing molecular data were also run using PAUP* 4.0a152. Phylogenetic analyses were performed on each gene separately and on a combined molecular dataset. Combination in part of the molecular datasets was supported by 1000 replicate incongruence length difference (ILD) tests (Farris et al. 1994) (each with ten random addition replicates, with uninformative and constant characters excluded) of the partitions COI vs. 18S (P = 0.68). COI vs. 16S (P = 0.001) and mitochondrial vs. nuclear (p = 0.001) were not supported. All gaps were treated as missing data. Models of molecular evolution for use in ML and BI analyses were estimated by an Akaike Information Criterion (AIC) test (Akaike, 1974) using MrModeltest 2.3 (Nylander, 2004). As measures of phylogenetic content, the g1 (skewness) statistics of 1000 random trees and the percentage of clades supported by > 50% parsimony bootstrap among all resolved clades in the strict consensus trees were calculated (Table 4). Also, likelihood mapping was conducted by examining 10000 quartets using an approximate likelihood function based on the selected model parameters for each molecule. Models of molecular evolution are given in Table 4 for each molecule and codon position. Table 4. Descriptive statistics on sequence data. Models of molecular evolution were selected by AIC for each molecule and codon position in protein encoding genes. Per cent pairwise divergence (%PD) between Belostomatidae taxa is given. Skewness (g1) was calculated based on 1000 random trees. Percentage of clades with parsimony bootstrap support > 50% is based on strict consensus trees (%BS) Gene  Model selected  Length (bp)  Number of taxa  Number of variable sites (%)  Number of informative sites (%)  %A  %C  %G  %T  PD  g1  %BS  COI  GTR+I+Γ  735  38  569 (77)  408 (56)  40  19  9  33  0–59  −0.74  45  COI pos1  HKY+Γ  245                      COI pos2  GTR+Γ  245                      COI pos3  GTR+I+Γ  245                      18S  SYM+I+Γ  606  32  272 (45)  197 (33)  25  25  25  25  0–12  −0.71  81  16S  HKY+I+Γ  488  45  248 (51)  220 (45)  49  12  6  33  0–27  −0.58  64  Combined    1829  56  1829 (100)  834 (46)  34  21  17  28  0–30  −0.60  78  Gene  Model selected  Length (bp)  Number of taxa  Number of variable sites (%)  Number of informative sites (%)  %A  %C  %G  %T  PD  g1  %BS  COI  GTR+I+Γ  735  38  569 (77)  408 (56)  40  19  9  33  0–59  −0.74  45  COI pos1  HKY+Γ  245                      COI pos2  GTR+Γ  245                      COI pos3  GTR+I+Γ  245                      18S  SYM+I+Γ  606  32  272 (45)  197 (33)  25  25  25  25  0–12  −0.71  81  16S  HKY+I+Γ  488  45  248 (51)  220 (45)  49  12  6  33  0–27  −0.58  64  Combined    1829  56  1829 (100)  834 (46)  34  21  17  28  0–30  −0.60  78  View Large Despite a lack of support for the homogeneity of the signal from mitochondrial and nuclear partitions, a combination of the molecular and morphological datasets was supported (1000 replicate ILD tests, P = 0.97). In fact, we also performed phylogenetic analyses on that combined unsupported molecular dataset, as well as on the combined molecular and morphological datasets. Parsimony-based tree searches were performed using TBR branch swapping with 1000 random addition replicates, keeping ten trees at each step (hold=10). With regard to the combined molecular and morphological dataset, we conducted 10000 random addition sequence replicates and TBR branch swapping, keeping 100 trees at each step. All analyses used a random starting tree. BI analyses were performed with MrBayes, including that with molecular and morphological data. Mitochondrial protein encoding genes (COI) were classified by codon position and a mixed model approach (five partitions for the molecular analysis and six for the combined molecular and morphological data) was conducted. Models chosen for use in Bayesian analyses of combined molecular dataset were: HKY+Γ for COI pos1, GTR+Γ for COI pos2, GTR+I+Γ for COI pos3, SYM+I+Γ for 18S and HKY+I+Γ for 16S. Assuming flat priors, we performed each analysis consisting of four independent runs of 5000000 generations, each with four chains, sampling topologies from every 500 generations. For the molecular and morphological data combined, four Markov chain Monte Carlo chains were run for 10000000 generations sampling topologies from every 100 generations. Convergence was checked with the average standard deviation of split frequencies, which approach 0.01 when the runs converge. The data before the convergence were analysed with the software Tracer 1.6 (Rambaut et al., 2014) and burned-in. Ten per cent of the trees were discarded as burn-in after examination by Tracer 1.6 and the remaining trees were used to calculate the posterior probabilities. Calculations were performed on the parallel computing Linux cluster developed at the MNHN (80 CPUs in 23 nodes, 4 Go Ram per node). Descriptive information on the molecules studied is summarized in Table 4. Likelihood mapping diagrams for each molecule are given in Figure 1. The ML analyses were conducted with RAxML (Stamatakis, 2006). The parameters were estimated over 200 runs to ensure adequate searching, and the best scoring tree was based on the support for the likelihood-derived topologies estimated by bootstrap resampling (MLB) calculated using 1000 replicates. Figure 1. View largeDownload slide Likelihood mapping diagrams. Diagrams for each gene region based on 10000 quartets. Each point represents a four-taxon statement, with those quartets that are well resolved towards a certain topology falling in the three corners (high phylogenetic signal), while those in the central triangle are those that do not strongly support any of the three possible relationship statements (low phylogenetic signal). Figure 1. View largeDownload slide Likelihood mapping diagrams. Diagrams for each gene region based on 10000 quartets. Each point represents a four-taxon statement, with those quartets that are well resolved towards a certain topology falling in the three corners (high phylogenetic signal), while those in the central triangle are those that do not strongly support any of the three possible relationship statements (low phylogenetic signal). Support for nodes on the obtained trees were thus evaluated by non-parametric parsimony bootstrapping (PB) (Felsenstein, 1985), MLB and Bayesian clade probability (BCP) (as recovered from the majority-rule consensus of the 36000 or 360000 post-burn-in sampled trees). Bootstrap searches were conducted in general with 1000 pseudoreplicates each with ten random addition replicates and TBR branch swapping. For morphological and combined data, however, we used maxtrees=100000. For molecular and morphological data combined, partitioned Bremer decay (Bremer, 1994) (PBD) indices were also calculated based on one of the most parsimonious trees with the aid of TreeRot V.3 (Sorenson & Franzosa, 2007). The initial unconstrained and each constrained search were conducted with 10000 random addition replicates, keeping 100 trees at each step, and TBR branch swapping. As trees may differ in topology, we considered the MrBayes tree to be the ‘master’ tree, giving the figure its topology and branch lengths. The same splits were highlighted in the master tree. Support values from PI and ML bootstrap analyses and PBD indices (if applicable) were combined with the posteriors from the Bayesian tree. Topological agreement: support for classification changes We used parametric bootstrap analysis to test the monophyly of the Belostomatinae containing H. pelocoroides highlighted by some genes and morphological data. This technique is preferable to other methods of tree comparison in that it is less prone to type II statistical error (Goldman, Anderson & Rodrigo, 2000). Based on Smith, Page & Johnson’s (2004) procedure, model trees were estimated from the ML analyses with taxa constrained to be compatible with a hypothesis of monophyly for both the Belostomatinae and the Lethocerinae, and for the Belostomatinae taxa alone (loosest constraint possible). For each model tree (with branch length), we used the model of sequence evolution and parameters estimated to generate 500 simulated datasets of the same size as the original. Two heuristic searches were conducted on each replicate dataset using NNI branch swapping: once to find the overall optimal tree and again to find the best tree compatible with the constraint used to generate the model tree. Scores of these likelihood trees were then used to construct an expected distribution of likelihood differences under the null hypothesis being tested. Significance of the test statistic (the difference in log-likelihood values between the constrained and optimal trees) was assessed by direct comparison with the expected distribution. Data incongruence by evaluating the occurrence of unsupported and conflicting clades To indirectly assess resolution issues, some parts of the obtained phylogenies and the unsupported combination of COI and 16S partitions and mitochondrial and nuclear datasets, we used an alternative view of the correspondences between relationships among trees (by associating objective measures that quantify these correspondences) to examine the impact of using different partition schemes and methods of analysis, as well as the existence of missing data. We considered the nature of the data used for obtaining characters, as well as whether information for a particular taxon was available (e.g. a fossil specimen). We used the average number of conflicts between each of the proposed relationships (i.e. number of nodes that are contradicted by any other relationships involved by these nodes divided by the total number of nodes in the tree) and other different trees upon different treatments (not shown), by using the same logical approach when making the calculation of conflicts between super-trees created by Wilkinson et al. (2005). Two trees are understood to conflict if they assert logically contradictory relationships so that the tree clade cannot be present in any tree that includes the relationships in the input tree. Analyses were conducted among all different partitions, methods of analysis and taking into account the presence of missing data. A particular advantage of this approach is that it can be used to compare topologies with taxon sets that do not entirely match. In addition, not only did we adopt here the idea of lack of support, but also that tree clades conflicting with any input tree are more objectionable. Support values were obtained with stsupport (Wilkinson et al., 2005). Using such support for conflict among trees from different data sets as a means of assessing the independence between datasets can be seen as a test of heterogeneity (De Queiroz et al., 1995). Missing data were identified because of the degree of conservation of ancient exemplars of W. rhomboides, some problems in amplifying gene sequences of a small number of taxa, and the addition of fossil data, so that some gene sequences and morphological characters were missing from the analyses (see Table 1). To address the problem of the difference among the numbers of unsupported tree clades per tree in each treatment (i.e. using all available data or only those taxa with sequences of good quality) when different datasets could not be used, we performed a two-way ANOVA (Sokal & Rohlf, 1995). The number of conflicts per tree was the response variable, and datasets (Factor I) and quality of those data (Factor II) were used as factors, under the null hypothesis that the different datasets and the presence of missing data, with different methods of analysis (the replications) used to infer phylogeny (i.e. PI, ML and BI), did not affect the existence of phylogenetic signal and, in turn, the genus-level relationships in Belostomatidae. A failure to reject H0 suggests a strong congruence among such treatments. RESULTS Morphological analysis of Belostomatidae A morphological character matrix (Appendix 1) including 11 outgroup taxa (four genera) and 22 in-group species representing all 13 genera of Belostomatidae was coded based on 104 delimited morphological characters. Of these morphological characters, 16 were autapomorphies and 23 were multistate. The characters and their consistency and retention indices (CI and RI, respectively) are listed in Appendix 2. The PI analysis resulted in 42 most-parsimonious trees of 201 steps each and CI = 0.71, RI = 0.90 and RC = 0.64 (RC is the rescaled consistency index, which is obtained by multiplying the CI by the RI). A strict consensus among them was found by the branch-and-bound algorithm, as shown in Figure 2A, and is shown with unambiguous characters optimized. Conversely, Figure 2B–E are selected trees which have different topologies with regard to Belostomatinae (and position of Horvathinia). Figure 2. View largeDownload slide Morphological analysis. A, consensus of 42 most-parsimonious trees (length = 201, CI = 0.71, RI = 0.90, RC = 0.64) with non-ambiguous characters optimized; B, part of other most-parsimonious tree indicating Horvathinia as sister group of Diplonychinitrib. nov.; C, part of other most-parsimonious tree indicating Horvathinia as sister group of Belostomatini; D, part of other most-parsimonious tree indicating Hydrocyrius + Limnogeton as sister group of Horvathinia + Diplonychinitrib. nov. + Belostomatini; E, part of other most-parsimonious tree indicating Hydrocyrius + Limnogeton as sister group of Horvathinia + Diplonychinitrib. nov. Arrow indicates the node with main changes throughout the 42 most-parsimonious trees. (+), Fossils. Outgroups were omitted. Figure 2. View largeDownload slide Morphological analysis. A, consensus of 42 most-parsimonious trees (length = 201, CI = 0.71, RI = 0.90, RC = 0.64) with non-ambiguous characters optimized; B, part of other most-parsimonious tree indicating Horvathinia as sister group of Diplonychinitrib. nov.; C, part of other most-parsimonious tree indicating Horvathinia as sister group of Belostomatini; D, part of other most-parsimonious tree indicating Hydrocyrius + Limnogeton as sister group of Horvathinia + Diplonychinitrib. nov. + Belostomatini; E, part of other most-parsimonious tree indicating Hydrocyrius + Limnogeton as sister group of Horvathinia + Diplonychinitrib. nov. Arrow indicates the node with main changes throughout the 42 most-parsimonious trees. (+), Fossils. Outgroups were omitted. The monophyly of Belostomatidae s.l. (and as defined here: with Cratonepa enigmatica included) was not refuted in the PI analysis (PB = 96%) (Figs 2, 3). With low support, however, Belostomatinae is monophyletic (PB = 51%) with two well-defined clades, Belostomatini (Abedus+Belostoma and now W. rhomboides) (PB < 50%) and Diplonychini trib. nov. (Appasus+Diplonychus) (PB = 93%). Also, Lethocerinae was recovered as the sister group of Belostomatinae in the consensus tree, now with Benacus griseus and Kirkaldyia deyrolli resurrected and included, as well as the fossil taxa Lethocerus vetus and Iberonepa romerali. The main differences among the trees (Fig. 2B–E), however, are the sister group relationships of Horvathinia pelocoroides with Belostomatini and Diplonychini trib. nov., while Limnogeton and Hydrocyrius may be a monophyletic group and a sister group of Diplonychini trib. nov. + Horvathinia + Belostomatini. The support values are indicated in the preferred Bayesian topology (Fig. 3) referred below. Figure 3. View largeDownload slide Morphological analysis: Bayesian master tree (harmonic mean in log-likelihood = −787.68) selected by comparing the harmonic mean of the log-likelihood of trees obtained by the addition or not of the gamma parameter to the model. Support is based on Bayesian clade probability (BCP), maximum likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/MLB/PB. Clades are coloured according to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. ‘-’, Support values less than 50%. (+) with dashed lines, Fossils. Outgroups were omitted. Figure 3. View largeDownload slide Morphological analysis: Bayesian master tree (harmonic mean in log-likelihood = −787.68) selected by comparing the harmonic mean of the log-likelihood of trees obtained by the addition or not of the gamma parameter to the model. Support is based on Bayesian clade probability (BCP), maximum likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/MLB/PB. Clades are coloured according to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. ‘-’, Support values less than 50%. (+) with dashed lines, Fossils. Outgroups were omitted. The preferred Bayesian topology (Fig. 3) was selected by comparing the harmonic mean of the log-likelihood of each of the two aforementioned analyses presented in the Material and Methods. In this study, the addition of the gamma parameter to the model improved its fit to the data. The value of the harmonic mean of the Bayesian analysis without the gamma parameter was −799.71, whereas that of the analysis with the gamma parameter was −787.68. Because the latter value is closer to 0, the tree derived from the Bayesian analysis with gamma was preferred. The Bayes factor was approximately 24, which means that the difference between the harmonic means of the two analyses was significant. Some differences were found among the tree from BI analysis including gamma and those from parsimony analysis. The preferred BI analysis yielded trees less similar to those obtained through PI analysis with regard to the subfamily Lethocerinae (Fig. 3). Lethocerinae appear to be a paraphyletic taxon in relation to Belostomatinae, which was recovered as monophyletic only in the consensus tree based on parsimony (see above). Its strange configuration in the Bayesian trees, however, should be viewed cautiously, because it is probably an artefact of missing characters – the fossil specimen of L. vetus, which made it impossible to code characters from the wings and male genitalia for this species. In contrast, H. pelocoroides is recovered as the sister group to Diplonychini trib. nov. (BCP = 78%, MLB < 50%, PB < 50%), similar to some most-parsimonious trees with the almost same topology (e.g. Fig. 2B). One of those most-parsimonious trees (Fig. 2B) is slightly similar to that obtained by BI analysis. Another difference found here concerns that topology which deems Horvathinia as the sister group of Belostomatini, while Limnogeton is a sister group of Hydrocyrius + Diplonychini trib. nov. + Horvathinia + Belostomatini (Fig. 2C). Overall, the morphological data suggest the transfer of Horvathinia to Belostomatinae. The monophyly of Belostomatidae (BCP = 100%, MLB = 95%, PB = 96%), Belostomatinae (BCP = 88%, MLB < 50%, PB = 51%) and Belostomatini (BCP = 64%, MLB < 50%, PB < 50%) is not refuted in any of our analyses (i.e. PI and preferred BI and ML analyses) (Fig. 3), whereas Diplonychini trib. nov. was recovered as monophyletic across almost all analyses (BCP = 79%, MLB = 53%, PB = 93%), except ML. Molecular analyses of Belostomatidae The complete molecular dataset contained 1829 aligned characters, all of which were variable; 834 (about 46%) were parsimoniously informative. Within the COI gene fragment, 569 (77%) characters were variable, and 408 of these (56%) were parsimony informative, this being the partition with the largest fraction of variable sites and parsimony-informative sites (Table 4). As measures of phylogenetic content, COI and 18S contained higher values of g1 (skewness) of 1000 random trees (−0.74 and −0.71, respectively), while only 18S showed a higher number of the percentage of clades (81%) supported by > 50% parsimony bootstrap among all resolved clades in the strict consensus trees (Table 4). Likelihood mapping was conducted by examining 10000 quartets using an approximate likelihood function based on the model parameters selected for each molecule. The final distribution of points mapped into regions of the triangle, in each partition, reveals that COI and 18S fragments performed somewhat better than 16S, with a little over 83% of the quartets mapped into the strongly tree-like regions. Likelihood mapping diagrams for each molecule are given in Figure 1. Heuristic parsimony searches resulted in two most-parsimonious trees for COI (length = 1764, CI = 0.32, RI = 0.52, RC = 0.17), two for 16S (length = 687, CI = 0.48, RI = 0.78, RC = 0.37) and 144 for 18S (length = 327, CI = 0.64, RI = 0.78, RC = 0.50). Parsimony analyses still recovered 20655 most-parsimonious trees for the combined molecular dataset (length = 2,817, CI = 0.39, RI = 0.54, RC = 0.21) (Supporting Information: clades presented in the strict consensus compatible with the Bayesian consensus and ML method are highlighted in Figs S1–S3). The present phylogenetic analysis based on three gene regions provides further support for the classification changes proposed above for morphological characters. All molecular trees support the monophyly of Belostomatidae s.s. (i.e. Lethocerinae+Belostomatinae) (BCP = 69%, PB ≤ 50%, MLB = 100% for COI; BCP = 100%, PB = 65%, MLB = 100% for 18S; BCP = 100%, PB ≤ 50%, MLB = 100% for 16S), as well as of the Lethocerinae (BCP = 100%, PB = 100%, MLB = 100% for all gene partitions). However, only the 16S rDNA data support the monophyly of Belostomatinae (BCP = 100%, PB = 64%, MLB = 100%) with all methods (Supporting Information: Fig. S3). Conversely, COI and 16S rDNA support the monophyly of Belostomatini with all methods (BCP = 100%, PB = 100%, MLB = 100% for COI; BCP = 85%, PB = 60%, MLB = 100% for 16S). Interestingly, none of the molecular trees based on the separated gene regions supports the Diplonychini trib. nov., and COI gives low congruence. In this case Diplonychus is close to the base of the Belostomatinae, whereas Appasus is within the clade consisting of Appasus + Belostomatini. Finally, Horvathinia and Lethocerinae are sister groups (BCP = 95%, PB ≤ 50%, MLB = 100%) based on only the 18S fragment, and not all methods used here support such a clade (only BI and ML). With all other fragments, instead, Horvathinia comprises a clade, which includes all belostomatines, either close to the African genera Limnogeton and Hydrocyrius or close to the base of the Belostomatinae. Congruence amongst the molecular phylogeny with all datasets from PI, ML and BI is much higher (Fig. 4). BI analysis of the combined molecular dataset recovered the monophyly of the Diplonychini trib. nov. (BCP = 100%, PB = 58%, MLB = 100%), Belostoma (BCP = 99%, PB = 51%, MLB = 100%) and Belostomatinae (BCP = 100%, PB = 70%, MLB = 100%) (now with Horvathinia included), almost all with high levels of clade support. Lethocerinae were recovered as a monophyletic subfamily (BCP = 100%, PB = 99%, MLB = 100%) across all tree reconstruction methods, as well as Lethocerus (BCP = 100%, PB = 95%, MLB = 100%). Unfortunately, Benacus griseus was not included in those analyses because of problems in amplifying its gene sequences. Besides, all but PI trees support Belostomatidae as monophyletic (BCP = 98%, PB < 50%, MLB = 100%) as well, with low bootstrapping values instead. Additionally, PI, ML and BI supports the transfer of W. rhomboides to Belostomatini (BCP = 94%, PB = 80%, MLB = 100%). Conversely, neither Belostoma and Abedus nor Diplonychini trib. nov. + Belostomatini were recovered as a monophyletic group in the strict consensus of PI analysis. Figure 4. View largeDownload slide Molecular analysis: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the combined molecular dataset (five partitions: pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. Support is based on Bayesian clade probability (BCP), maximum-likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/PB/MLB. Clades are coloured according to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. (+), Fossils. Figure 4. View largeDownload slide Molecular analysis: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the combined molecular dataset (five partitions: pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. Support is based on Bayesian clade probability (BCP), maximum-likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/PB/MLB. Clades are coloured according to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. (+), Fossils. Morphological and molecular dataset analyses PI analysis yielded 25 most-parsimonious trees using the combined molecular and morphological dataset (length = 2592, CI = 0.45, RI = 0.56, RC = 0.25), and clades present in the strict consensus compatible with the Bayesian consensus and ML inference are highlighted in Figure 5. Partitioned Bremer analysis totalled 2592 characters and showed the majority of the signal coming from the first codon position of COI and 16S (COI pos1, 39.6%; 16S, 24.7%). Morphology and the third codon position of COI were minimal in their support (8.2 and 6.6%, respectively). Interestingly, for most well-supported clades, morphology is the partition agreeing with COI pos1 and nuclear partition 18S by showing positive values of decay indices. Figure 5. View largeDownload slide Combined analysis: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the combined morphological and molecular dataset (six partitions: morphology, pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. Support is based on Bayesian clade probability (BCP), maximum-likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/MLB/PB. Partitioned Bremer indices are shown as graphs near to respective clades, with the x-axis representing partitions: morphology, 16S, COI and 18S, and y-axis support ranging from −3.6 to 2.4. Clades are coloured accordingly to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. (+) with dashed lines, Fossils. Figure 5. View largeDownload slide Combined analysis: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the combined morphological and molecular dataset (six partitions: morphology, pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. Support is based on Bayesian clade probability (BCP), maximum-likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/MLB/PB. Partitioned Bremer indices are shown as graphs near to respective clades, with the x-axis representing partitions: morphology, 16S, COI and 18S, and y-axis support ranging from −3.6 to 2.4. Clades are coloured accordingly to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. (+) with dashed lines, Fossils. Belostomatidae s.l. was recovered as monophyletic across all analyses (BCP = 100%, MLB = 96%, PB = 100%, PBDmorphology = 2.3), and C. enigmatica is now thought to be included as a member of this family together with I. romerali. Although the phylogenetic analyses performed poorly with respect to clade posterior probabilities, parsimony bootstrap values and Bremer supports, Lethocerinae was recovered as a monophyletic subfamily (BCP = 58%, MLB < 50%, PB = 63%, PBDmorphology = 2.3) across all tree reconstruction methods. We recovered the position of H. pelocoroides as being the sister group of Hydrocyrius (BCP = 66%, MLB < 50%, PB = 91%, PBDmorphology = 2.3), and Limnogeton was recovered as the sister group of H. pelocoroides + Hydrocyrius (BCP = 51%, MLB < 50%, PB = 73%, PBDmorphology = 2.2) only in the BI and PI analyses. The monophyly of Belostomatinae was corroborated here (BCP = 61%, MLB = 60%, PB = 74%, PBDmorphology = 2.3) across all methods, despite performing poorly with respect to clade posterior probabilities and ML and parsimony bootstrap values. Instead, Belostomatini (BCP = 76%, MLB = 81%, PB = 85%, PBDmorphology = 2.3) and Diplonychini trib. nov. (BCP = 96%, MLB = 98%, PB = 89%, PBDmorphology = 2.2) were recovered as monophyletic and sister groups with slightly higher support values across all methods. Overall, congruence amongst the molecular and morphological datasets obtained from BI, PI and ML was much higher, despite the existence of the strange group comprising Horvathinia + Hydrocyrius + Limnogeton (Fig. 5). Topological agreement with morphological analysis: support for classification changes Parametric bootstrap analysis of the molecular data clearly rejects the null hypotheses of non-monophyly for both the Lethocerinae and the Belostomatinae (containing Horvathinia) clades predicted by some genes and morphology (P < 0.001, Figs 4–6), except for the 18S dataset (P = 0.99) (Supporting Information: Figs S4–S11). Under the constraint of Belostomatinae monophyly, the loosest constraint possible in this analysis, the log-likelihood difference between the simulated datasets with and without the constraint ranges from +100 to +300. However, the observed difference in log-likelihood for the actual dataset was +909. Thus, the probability of observing a difference of this magnitude (if the null hypothesis were true) is considerably less than 1%. BI, PI and ML trees share many common nodes, and the number of unsupported tree clades conflicting with all input trees seems to show strong congruence among different partitions when one considers the absence of some or all of the relevant data (F = 1.423, P = 0.944) (Tables 5–6). On average, 79.6% of the variation among the trees is due to differences among the matrices with or without some taxa, and only 5.4% of the variation is due to differences among the datasets. Table 5. Number of unsupported clades conflicting with all input trees per tree obtained by stsupport (Wilkinson et al., 2005), taking into account maximum-parsimony (PA), maximum-likelihood (ML) and Bayesian analysis (BI) approaches and the type of data partition. Figures in parentheses are U = number of unsupported clades conflicting with all relevant input trees and TC = number of tree clades, respectively   Quality of the data    With all data  Incomplete data and fossils absent  Partitions  PA vs. ML + BI  ML vs. BI + PA  BI vs. ML + PA  PA vs. ML + BI  ML vs. BI + PA  BI vs. ML + PA  COI  0.30 (10/33)  0.11 (4/35)  0.07 (2/30)  0.39 (7/18)  0.11 (2/18)  0.06 (1/17)  18S rDNA  0.05 (1/20)  0.08 (2/26)  0.08 (2/26)  0 (0/20)  0.13 (3/24)  0 (0/21)  16S rDNA  0.17 (6/35)  0 (0/35)  0 (0/34)  0.14 (5/35)  0 (0/35)  0 (0/33)  Morphology  0.22 (4/18)  0.14 (3/21)  0 (0/20)  0 (0/15)  0.21 (4/19)  0 (0/18)  Molecular dataset  0.11 (4/36)  0.04 (2/46)  0.04 (2/46)  0.22 (5/23)  0.14 (5/35)  0 (0/32)  Molecular + morphology  0.11 (4/38)  0.31 (16/51)  0.12 (5/43)  0.29 (6/21)  0.06 (2/36)  0.03 (1/34)    Quality of the data    With all data  Incomplete data and fossils absent  Partitions  PA vs. ML + BI  ML vs. BI + PA  BI vs. ML + PA  PA vs. ML + BI  ML vs. BI + PA  BI vs. ML + PA  COI  0.30 (10/33)  0.11 (4/35)  0.07 (2/30)  0.39 (7/18)  0.11 (2/18)  0.06 (1/17)  18S rDNA  0.05 (1/20)  0.08 (2/26)  0.08 (2/26)  0 (0/20)  0.13 (3/24)  0 (0/21)  16S rDNA  0.17 (6/35)  0 (0/35)  0 (0/34)  0.14 (5/35)  0 (0/35)  0 (0/33)  Morphology  0.22 (4/18)  0.14 (3/21)  0 (0/20)  0 (0/15)  0.21 (4/19)  0 (0/18)  Molecular dataset  0.11 (4/36)  0.04 (2/46)  0.04 (2/46)  0.22 (5/23)  0.14 (5/35)  0 (0/32)  Molecular + morphology  0.11 (4/38)  0.31 (16/51)  0.12 (5/43)  0.29 (6/21)  0.06 (2/36)  0.03 (1/34)  View Large Table 6. Two-way ANOVA. The number of unsupported tree clades conflicting with all input trees was the response variable and datasets and quality of those data were used as factors, under the null hypothesis that the different dataset and the presence of missing data, with different methods of analysis used to infer phylogeny, did not affect the existence of phylogenetic signal and, in turn, the genus-level relationships in Belostomatidae. A failure to reject H0 suggests strong congruence among such treatments. Abbreviations: d.f. = degree of freedom; SS = sum of squares; MS = mean of squares Source  d.f.  SS  MS  F-value  P-value  Quality of data  5  0.074  0.015  1.235  0.324  Partitions  1  0.001  0.001  0.083  0.776  Interaction  5  0.014  0.003  0.233  0.944  Residuals  24  0.288  0.012  –  –  Source  d.f.  SS  MS  F-value  P-value  Quality of data  5  0.074  0.015  1.235  0.324  Partitions  1  0.001  0.001  0.083  0.776  Interaction  5  0.014  0.003  0.233  0.944  Residuals  24  0.288  0.012  –  –  View Large TAXONOMY Since we found here evidence to support Appasus and Diplonychus as a monophyletic group, the new tribe Diplonychini is herein described to accommodate that clade. Family Belostomatidae Leach, 1815 Subfamily Belostomatinae Leach, 1815 Tribe Diplonychini trib. nov. Type genus: Diplonychus Laporte, 1833 Diagnosis: Diplonychini differs from other belostomatine suprageneric groups by the frons rounded or curved in dorsal view (Fig. 9D, G, H), as well as the following male genitalia features: (1) the transverse bridge of basal plate of male genitalia clearly jointed and entire (Fig. 16B); (2) plate of phallotheca as long as ventral diverticulum. Likewise, females of Diplonychini differ from the others by the presence of an ampulla located at the basal part of spermatheca (Fig. 21A–C). Distribution: Africa, Australia, East Indies, southern Asia (Estévez & Ribeiro, 2011). Description: Measurements. – Total length (from apex of head to apex of abdomen at rest): from 8.2 to 27.7 mm. General coloration. – Almost uniformly brown. External morphology. – Body ovate with wings usually covering abdomen. Frons rounded or curved (Fig. 9D, G, H); vertex without median longitudinal carina (Fig. 9D); antennae with segments 2 and 3 not flattened ventrally, with fourth segment similar to or slightly more bulbous than prolongations of segments 2 and 3 (Fig. 11B); frontogenal suture slightly convergent and opened distally (Fig. 9D). Pronotum without longitudinal median carina (e.g. Fig. 9D); prosternal keel usually poorly elevated, except for some Appasus species; hemelytra with rounded pruinose area; clamp of clavus with its outer projection overlapping inner part, always far from the margin of hemelytra (Fig. 14A); its outer carina with three rows of microtrichiae along its external margin, covering small portion in dorsal view (Fig. 13A, B); tile-like microtrichiae rounded at apex, never toothed along the margin of its apex (Fig. 12B); foretarsi with two segments, externally usually appearing one-segmented, with segment 1 conspicuous, with two symmetrical grooves; claws vestigial; hind trochanters carinated, with short hairs or bristles along outer margins. Pilosity developed, covering half of connexivum, slightly constricted between spiracles, extending posteriorly along about half of or almost entire genital operculum; pubescence of ventral laterotergites 3 and 4 not attaining entire external margin; air straps lanceolate, with somewhat uniform width along its extension. Male genitalia: Phallosoma fused to ventral diverticulum; arms of phallosoma well developed, extending nearly to apex of ventral diverticulum, enclosing ventral diverticulum in some Appasus (Fig. 16E), somewhat laterally directed; orifice strongly developed, dorsally located on apex of phallosoma; ventral diverticulum contiguous, never bilobed, with its apex without ventroapical protuberance, not showing spines or tubercles in ventral view; transverse bridge of basal plate of male genitalia clearly jointed and entire (Fig. 16B); plate of phallotheca somewhat developed, fused to or close to ventral diverticulum. Female genitalia: Operculum of females with two tufts of setae on apex; apex of second valvulae with an inconspicuous spine; basal part of spermatheca without distinct apodemes, clearly with the presence of an ampulla; median vagina area below spermatheca without pouch (Fig. 21A–C). Taxonomic notes: In general, members of the new tribe Diplonychini share with W. rhomboides the prosternal carina poorly elevated (‘prosternal keel rounded’ according to Estévez & Ribeiro, 2011: 51) (character 27: 0 > 1) and with H. pelocoroides the surface of apex of ventral diverticulum (in ventral view) without spines or tubercles (character 89: 0 > 1). Our findings support Diplonychini trib. nov. as the sister group of the tribe Belostomatini by the following unambiguous homoplastic synapomorphies: (1) clamp of clavus with its outer projection not flattened, overlapping the inner part (similar to Fig. 14A) (character 31: 2 > 3); and (2) outer carina of the clamp with three rows of microtrichiae along its external margin (character 32: 2 > 1) (Fig. 13A, B). The new tribe Diplonychini shares with other members of the subfamily Belostomatinae the foretarsi bearing both claws vestigial (character 44: 0 > 3), phallosoma fused with ventral diverticulum (character 71: 0 > 1), ventral diverticulum contiguous, never bilobed (character 85: 0 > 1) (Fig. 16E), and apex of second valvulae with a spine and/or protuberance (character 94: 0 > 1). All morphological character states cited above are exclusive to this subfamily (Fig. 6). Figure 6. View largeDownload slide Combined analysis: Bayesian consensus phylogram of Figure 5. Mixed-model Bayesian analysis of the combined morphological and molecular dataset (six partitions: morphology, pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Optimized tree with non-ambiguous characters, except for the clades including Hydrocyrius + Horvathinia + Limnogeton and Belostoma ribeiroi + B. plebejum. (+), Fossils. Outgroups were omitted. Figure 6. View largeDownload slide Combined analysis: Bayesian consensus phylogram of Figure 5. Mixed-model Bayesian analysis of the combined morphological and molecular dataset (six partitions: morphology, pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Optimized tree with non-ambiguous characters, except for the clades including Hydrocyrius + Horvathinia + Limnogeton and Belostoma ribeiroi + B. plebejum. (+), Fossils. Outgroups were omitted. DISCUSSION Support for classification changes in the composition of groups The present study is the first higher-level phylogenetic study to include all Belostomatidae genera. Lee (1991) concurred with the decision made by Lauck & Menke (1961) of separating Belostomatidae into the Lethocerinae, Horvathiniinae and Belostomatinae, highlighting the existence of further characters shared by Lethocerus (Lethocerinae) and Limnogeton (Belostomatinae), and also by Horvathinia (Horvathiniinae) and Hydrocyrius (Belostomatinae). The results of all the datasets do not corroborate the restriction of Belostomatidae to the subfamilies Lethocerinae, Horvathiniinae and Belostomatinae as previously proposed by Mahner and earlier by Lee, and by the intuitive phylogeny produced by Lauck & Menke (1961), but rather suggest the inclusion of H. pelocoroides within the subfamily Belostomatinae instead of comprising the monobasic subfamily Horvathiniinae related to the subfamily Belostomatinae (Figs 2–6). All analyses support the inclusion of C. enigmatica in Belostomatidae s.l. and the inclusion of I. romerali in Belostomatidae s.s., based at least on moderately high clade support (combined dataset, BCP = 100%, MLB = 96%, PB = 100%, PBDmorphology = 2.3 and BCP = 81%, MLB = 63%, PB = 70%, PBDmorphology = 2.3, respectively) with the use of the combined morphological and molecular datasets (Fig. 6). The present study provided evidence and assigned C. enigmatica to Belostomatidae based on the following putative synapomorphy of Mahner: the siphon retracted into the abdomen (character 104: 0 > 1) (Figs 2, 3A, 4, 6). As indicated by Jattiot et al. (2012), the abdominal sternites divided into parasternites and median sternites (cited herein as character 53: 1 > 0) (Fig. 7A) is a condition also considered by Mahner (1993) as a putative synapomorphy. Our results, however, considered such a character as a plesiomorphy for Belostomatidae. The present study also provides further evidence for the monophyly of Belostomatidae s.l., based on the following unambiguous non-homoplastic synapomorphies: (1) eyes somewhat projected laterally with angled borders (character 8: 0 > 1) (Figs 8, 9); and (2) tarsi 2- or 3-segmented, rarely 1-segmented (character 37: 0 > 1). Figure 7. View largeDownload slide General aspect of abdomen. A, ventral laterotergites (= connexivum), parasternites and sternites, ventral view. B, subcylindric body, transversal view. C, connexivum and sternites, ventral view. Parasternites are not visible. D, flattened body, transversal view. Adapted from Mahner (1993). con, connexivum; gop, genital operculum; lfs, like-fold suture; par, parasternites; spi, spiracles; ste, sternites. Figure 7. View largeDownload slide General aspect of abdomen. A, ventral laterotergites (= connexivum), parasternites and sternites, ventral view. B, subcylindric body, transversal view. C, connexivum and sternites, ventral view. Parasternites are not visible. D, flattened body, transversal view. Adapted from Mahner (1993). con, connexivum; gop, genital operculum; lfs, like-fold suture; par, parasternites; spi, spiracles; ste, sternites. Figure 8. View largeDownload slide Head and pronotum, dorsal view. A, Ranatra sp. [female]. B, Lethocerus collosicus [male]. C, Abedus ovatus [male]. D, Belostoma bifoveolatum [male]. E, Belostoma dentatum [female]. F, Hydrocyrius sp. [male]. cfg, distally closed frontogenal suture; ofg, opened frontogenal suture. Figure 8. View largeDownload slide Head and pronotum, dorsal view. A, Ranatra sp. [female]. B, Lethocerus collosicus [male]. C, Abedus ovatus [male]. D, Belostoma bifoveolatum [male]. E, Belostoma dentatum [female]. F, Hydrocyrius sp. [male]. cfg, distally closed frontogenal suture; ofg, opened frontogenal suture. Figure 9. View largeDownload slide Head and pronotum, dorsal view. A–E, general aspects: A, Horvathinia pelocoroides [male]. B, Nepa hoffmani [male]. C, Weberiella rhomboides [female]. D, Diplonychus sp. [female]. E, Hydrocyrius colombiae [male]. F–I, comparison among the species with streamlined and rounded eyes: F, Limnogeton fieberi [female] (streamlined eyes). G, Appasus sp. [female] (streamlined eyes). H, Diplonychus esakii [female] (streamlined eyes). I, Weberiella rhomboides [male] (rounded eyes). car, developed longitudinal carina; cfg, distally closed frontogenal suture; fov, fovea; ofg, opened frontogenal suture; ver, vertex. Figure 9. View largeDownload slide Head and pronotum, dorsal view. A–E, general aspects: A, Horvathinia pelocoroides [male]. B, Nepa hoffmani [male]. C, Weberiella rhomboides [female]. D, Diplonychus sp. [female]. E, Hydrocyrius colombiae [male]. F–I, comparison among the species with streamlined and rounded eyes: F, Limnogeton fieberi [female] (streamlined eyes). G, Appasus sp. [female] (streamlined eyes). H, Diplonychus esakii [female] (streamlined eyes). I, Weberiella rhomboides [male] (rounded eyes). car, developed longitudinal carina; cfg, distally closed frontogenal suture; fov, fovea; ofg, opened frontogenal suture; ver, vertex. On the other hand, Hebsgaard et al. (2004) considered the presence of a metacoxa conical, firmly united with metapleuron (character 36: 0 > 1), as well as the hind tibiae flattened, with swimming hairs (considered here as characters 45 and 46), synapomorphies for the clade Belostomatidae. These characters were therefore thought to be characteristic of Belostomatidae s.l. according to the aforementioned authors; however, this result is in conflict with the present study. The condition of the metacoxae and hind tibiae was herein not recovered in C. enigmatica (Jattiot et al., 2012), but rather in Belostomatidae s.s. as follows (Figs 2A, 6): (1) metacoxae firmly united with metapleuron as an unambiguous non-homoplastic synapomorphy; (2) middle and hind tibia and tarsus completely different from each other, with metathoracic legs wide in relation to its length (character 45: 0 > 1) as an unambiguous non-homoplastic apomorphy of I. romerali, either with the use of morphological or combined morphological and molecular datasets. In fact, for those authors C. enigmatica remained a taxon of enigmatic position within the Nepoidea, with several plesiomorphic characters that imply that it could be a basal sister group of Nepidae, or even the sister group of all the Nepoidea. Curiuosly, the simple, slender and cursorial middle and hind legs, probably not modified for swimming, are unambiguously a homoplastic apomorphy in the consensus tree based on the morphological dataset because it supported Limnogeton as a reversal (character 46: 2 > 0) (Fig. 2A). The biology of these aquatic predators was probably different from those of Cenozoic and Recent Nepidae, based on the absence of an elongate siphon that stopped these insects breathing under water deeper than its length. According to Jattiot et al. (2012), the non-flattened hind tibiae of Cratonepa suggest that this species had a means of locomotion similar to that of the recent Nepidae, walking on the bottom of a pond or on emergent vegetation. In fact, the elongation of the tibiae and the tarsi was not considered in the present study but the specialized condition of the middle and hind tibiae and tarsi broadly flattened with swimming setae (character 46: 0 > 1) shared by lethocerines is reported here as an ambiguous synapomorphic trait with either the morphological or the combined morphological and molecular datasets. Likewise, Mahner (1993), in his phylogeny, proposed the presence of pubescence along ventral laterotergites (character 54: 0 > 1) as a synapomorphy shared by all belostomatids. The present analysis gives evidence for this, having reported it instead as an ambiguous non-homoplastic synapormorphy of Belostomatidae s.l. with either the morphological or the combined morphological and molecular datasets. Popov (1971) considered the Stygeonepinae to be a subfamily of Belostomatidae. Schlüter (1981), however, placed Stygeonepa among the Nepidae although he gave no reason for doing so and Nel (1991) restated that no precise reason exists for such an allocation. As stated by Martínez-Delclòs, Nel & Popov (1995), the present study shows that I. romerali (a Stygeonepinae) is a Belostomatidae (considered herein Belostomatidae s.s.) (BCP = 81%, MLB = 63%, PB = 70%, PBDmorphology = 2.3, in combined morphological and molecular datasets) and not proximal to the Nepidae based on the metacoxae being firmly united with metapleuron, although ML analysis did not give more than 70% support for such a clade using combined gene partitions. Only with the morphological dataset (Fig. 2A), the posterior portion of its hind tibiae, divided in two parts, both separated by a sulcus (character 49: 0 > 1), was Iberonepa grouped together with members of the Lethocerinae, as reported in the consensus tree. Also, this trait appears to arise independently in the Belostomatinae (Hydrocyrius). Likewise, in analysis of the combined morphological and molecular datasets (Fig. 6) the aforementioned condition is also homoplastic (convergent) and either ambiguously supports Belostomatidae s.l. or it is an unambiguous synapomorphy of Hydrocyrius. Finally, there is confusion promoted by Mahner (1993) regarding the existence of a third pair of valvulae in Mahner’s Belostomatidae. According to the author, Belostomatidae are defined by the absence of these valvulae, but we are inclined to accept the opinion of Lalitha et al. (1997) that the third pair of valvifers and valvulae both are fused so that only one plate can be observed. Thus, Mahner possibly incorrectly reported this trait and it is not considered in this study. Lee (1991) continued to characterize the subfamily Belostomatinae based on the symplesiomorphic character state of the absence of divisions in the abdominal sternites. Mahner (1993) probably assigned neither W. rhomboides nor H. pelocoroides to Belostomatinae because of the supposed lack of synapomorphies and misunderstandings concerning some apomorphic states of the abdomen; however, the present analyses based either on morphological or on combined morphological and molecular datasets gives evidence for the monophyly of such a subfamily (with H. pelocoroides and W. rhomboides included), based on 20 morphological synapomorphies (when optimized only on trees based on the morphological dataset), of which six are unambiguous (Fig. 2A), and 20 morphological synapomorphies (when optimized on the tree based on the combined morphological and molecular datasets), of which six are unambiguous (Fig. 6). Excluding the homoplasies, the non-homoplastic synapomorphies are as follows: (1) foretarsus bearing both claws vestigial (character 44: 0 > 3); (2) phallosoma fused with ventral diverticulum (character 71: 0 > 1); (3) ventral diverticulum contiguous, never bilobed (character 85: 0 > 1) (Fig. 10A, B); (4) genitalic protuberances spiny-shaped (character 88: 0 > 1) (Fig. 18); (5) phallothecal plate developed related to the ventral diverticulum, in lateral view (character 90: 0 > 1); and (6) apex of second valvulae with spine and/or protuberance (character 94: 0 > 1). All the aforementioned characters supported that clade in both the morphological and the combined morphological and molecular datasets. Analyses of the morphological, combined molecular dataset and the morphological + molecular dataset presented here recover Horvathinia + Belostomatinae with moderate to high support with all methods (i.e. PI, ML and BI) (Figs 3–6). It therefore seems plausible to transfer H. pelocoroides to Belostomatinae, although nothing is known about brooding behaviour in this genus. Horvathinia is supported here by the abdominal sternites 3–7 divided laterally by weak, suture-like folds into median- and parasternites (character 57: 1 > 0), always a homoplasy (reversal) according to our results based on the morphological and the combined morphological and molecular datasets. This study hence is in conflict with previous hypotheses (Lauck & Menke, 1961; Lee, 1991; Mahner, 1993) that define Lauck & Menke’s Belostomatinae as having abdominal sternites 3–7 undivided laterally by weak folds (character 57: 0 > 1). The present study highlights that this trait ambiguously supports Belostomatidae s.s. as a homoplastic synapomorphy with either morphological or combined morphological and molecular datasets. However, note that this result should be viewed cautiously, because it is probably an artefact of missing characters (see below) in I. romerali. Figure 10. View largeDownload slide Scanning electron micrographs of ventral diverticulum. Bilobed ventral diverticulum, aspect of apex in ventral view. A, Benacus griseus. B, Kirkaldyia deyrolli. Figure 10. View largeDownload slide Scanning electron micrographs of ventral diverticulum. Bilobed ventral diverticulum, aspect of apex in ventral view. A, Benacus griseus. B, Kirkaldyia deyrolli. Appasus and Diplonychus may be formally recognized as a separate, higher taxon based on the present study. This clade is herein elevated to tribal status, Diplonychini trib. nov. (type-genus: Diplonychus) based on the present analyses (Figs 2–6). When combining morphological and molecular datasets, our results give strong evidence for the monophyly of such a tribe (BCP = 96%, MLB = 98%, PB = 89%, PBDmorphology = 2.2) based on ten morphological synapomorphies, of which six are unambiguous (Fig. 6). The non-homoplastic synapomorphies are as follows: (1) transverse bridge of basal plate clearly jointed, entire (character 69: 1 > 2) (Fig. 16D); and (2) phallothecal plate developed related to ventral diverticulum (in lateral view) as long as ventral diverticulum (character 91: 0 > 1). Only the morphological dataset (Fig. 2A), the aforementioned characters and the dorsal arms completely laterally located (character 75: 0 > 2), grouped Appasus together with members of the Diplonychus. According to Lee (1991), the genera Appasus (= MuljarusLee, 1991) and Belostoma converged to share the elongate ellipsoidal body shape and developed a hemelytral membrane of two successive grades, whereas the genera Diplonychus and Abedus have in common a broad oval body and reduced membrane. The present study highlights that differentiation in the hemelytral membrane appears independently in some instances in Belostomatidae. A hemelytrum not differentiated from the elytra (character 28: 0 > 1) is an unambiguous non-homoplastic autapomorphy of D. rusticusFabricius, 1781, while a hemelytrum with the width of membrane and translucent margin combined in relation to the greatest width of clavus smaller than the greatest width of clavus (character 29: 0 > 2) is an unambiguous non-homoplastic synapomorphy supporting members of the Neartic Abedus and W. rhomboides as sister groups in the analysis using the combined morphological and molecular datasets (Fig. 6). Aside from members of Belostomatini, a hemelytrum with its membrane and translucent margin combined is approximately equal to the greatest width of clavus (character 29: 0 > 1) in Limnogeton, which is, again, an unambiguous non-homoplastic synapomorphy in the analysis using either the morphological or the combined morphological and molecular datasets. According to Polhemus (1995), the shapes of the eyes, antennae and male genitalia of Appasus and Diplonychus might be used in characterizations of those genera, which is why he resurrected Appasus. Despite the necessity of modification or clarification of some characteristics erected by Lee (1991), his separation of Diplonychus into two genera is also corroborated in the present study. Accordingly, considering the optimization done over combined morphological and molecular analysis, Appasus is supported herein by four unambiguous synapomorphies, of which the ventral diverticulum fused with phallosoma laterally flattened (character 81: 5 > 6) (Fig. 16D, E) is non-homoplastic. A pygophore with abrupt sculptured shoulder between basal portion and apical semitubular portion (character 93: 0 > 1) (as mentioned by Lee, 1991;,Polhemus, 1995) is homoplastic (parallelism) instead, also supporting B. griseus (Fig. 6) (see below). In the same way, Diplonychus is supported herein by segment 1 of antennae with four segments as long as the lateral prolongations of segments 2 and 3, as well as longer than segment 4 (character 19: 0 > 2) (Fig. 11), an unambiguous homoplastic synapomorphy, and the margins of dorsal arms (in lateral view) each with low angular medial projection (character 74: 0 > 1) (reported by Polhemus, 1995), an unambiguous non-homoplastic synapomorphy. As reported by Mahner (1993), BI, PI and ML inferences of either the morphological or the combined morphological and molecular datasets corroborate Belostomatini as a monophyletic clade. The analyses furthermore group W. rhomboides in Belostomatini together with other members of the Neartic Abedus (Figs 2–6) and thus indicate it should be transferred to this tribe. Considering the optimization done over combined morphological and molecular analysis, this tribe is supported by three unambiguous synapomorphies, of which two are non-homoplastic: (1) foretarsi bearing one anterior claw and one vestigial to absent posterior claw (character 44: 3 > 1); and the ventral diverticulum fused with phallosoma slightly convex (character 83: 3 > 2) (Fig. 6). In fact, Mahner (1993) also reported the former synapomorphy; however, according to him, character 44 (state 1) is a specialized condition (reversal) shared by Belostoma and Abedus. The genus Belostoma was recovered as invalid by Mahner (1993) who considered it very conservative and mostly distinguished by its plesiomorphic set of characteristics, such as membrane of hemelytrum distinctly larger than the greatest width of clavus, thus being considered equivalent to an ancestral Belostomatini. Herein Belostoma is recovered as a monophyletic group in all BI, PI and ML trees based on the combined morphological and molecular dataset (Fig. 5) (BCP = 78%, MLB = 64%, PB = 83%, PBDmorphology = 2.2), despite having grouped it with low support in the analysis using the morphololgical dataset (Fig. 3). The present status of Belostoma is based on the following unambiguous homoplastic synapomorphy: antennae (with four segments) with the form of fourth segment with only one finger-like projection slightly more bulbous than prolongations of segments 2 and 3 (character 16: 1 > 2) (similar to Fig. 11B). Abedus is supported herein by the apex of the ventral diverticulum (fused with phallosoma), which is straight (character 83: 2 > 1) (Figs 16C, 17C) – a unique unambiguous non-homoplastic synapomorphy (Fig. 6). As with Mahner (1993), the width of hemelytral membrane and translucent margin combined smaller than greatest width of clavus (character 29: 0 > 2), tile-like microtrichiae on outer carina of clamp strongly acute at apex (character 33: 0 > 1) (Fig. 12A), pubescence of ventral laterotergite 4 not attaining external margin (character 56: 0 > 1) and ventral diverticulum quadrangular-shaped along its median portion (character 81: 5 > 3) (Fig. 16C), all unambiguous non-homoplastic synapomorphies, here supported Abedus + W. rhomboides as a monophyletic group (Fig. 6). Finally, also reported by Mahner (1993), Weberiella is supported herein by the lateral margins of abdomen in adults not smooth but interrupted at the borders between the segments (scale-like abdomen) (character 52: 0 > 1) – a unique unambiguous non-homoplastic apomorphy (see Estévez & Ribeiro, 2011). Figure 11. View largeDownload slide A–F, scanning electron micrographs of antennae. General aspects of antennae (number of segments, aspect of second and third segments, aspect of fourth segment, form of fourth segment with only one finger-like projection, form of segment 2 ventrally, form of segment 3 ventrally, length of segment 1 in relation to lateral prolongations of segments 2 and 3, and segment 4). A, Nepa sp. B, Appasus sp. C, Limnogeton sp. D, Hydrocyrius sp. E, Lethocerus sp. F, Horvathinia pelocoroides. G, hind trochanter of Limnogeton scutellatum, aspect of its outer margin (with a conspicuous spine directed laterally). H–J, general aspect of posterior portion of hind tibiae without sulcus. H, Lethocerus sp. I, Laccotrephes sp. J, Horvathinia pelocoroides. fsa, first segment of antenna; ssa, second segment of antenna; tsa, third segment of antenna; fos, fourth segment of antenna; spi, spine. Figure 11. View largeDownload slide A–F, scanning electron micrographs of antennae. General aspects of antennae (number of segments, aspect of second and third segments, aspect of fourth segment, form of fourth segment with only one finger-like projection, form of segment 2 ventrally, form of segment 3 ventrally, length of segment 1 in relation to lateral prolongations of segments 2 and 3, and segment 4). A, Nepa sp. B, Appasus sp. C, Limnogeton sp. D, Hydrocyrius sp. E, Lethocerus sp. F, Horvathinia pelocoroides. G, hind trochanter of Limnogeton scutellatum, aspect of its outer margin (with a conspicuous spine directed laterally). H–J, general aspect of posterior portion of hind tibiae without sulcus. H, Lethocerus sp. I, Laccotrephes sp. J, Horvathinia pelocoroides. fsa, first segment of antenna; ssa, second segment of antenna; tsa, third segment of antenna; fos, fourth segment of antenna; spi, spine. Figure 12. View largeDownload slide Scanning electron micrographs of outer carina of the clavus–clavus clamp of hemelytra. General aspect of tile-like microtrichiae. A, Weberiella rhomboides. B, Appasus japonicus. C, Belostoma cummingsi. D, Belostoma flumineum. E, Hydrocyrius colombiae. F, Horvathinia pelocoroides. G, Limnogeton hedenborgi. Figure 12. View largeDownload slide Scanning electron micrographs of outer carina of the clavus–clavus clamp of hemelytra. General aspect of tile-like microtrichiae. A, Weberiella rhomboides. B, Appasus japonicus. C, Belostoma cummingsi. D, Belostoma flumineum. E, Hydrocyrius colombiae. F, Horvathinia pelocoroides. G, Limnogeton hedenborgi. Figure 13. View largeDownload slide Scanning electron micrographs of clavus–clavus clamp of hemelytra. Microtrichiae on outer carina. A–B, Appasus japonicus, margin with three rows of microtrichiae along it. C–D, Benacus griseus, margin with microtrichiae clearly distributed throughout it, comprising more than three rows. E–F, Curicta borelli, margin with just one or two rows of microtrichiae along it. Arrows indicate rows of microtrichiae. I, first row; II, second row; III, third row. Figure 13. View largeDownload slide Scanning electron micrographs of clavus–clavus clamp of hemelytra. Microtrichiae on outer carina. A–B, Appasus japonicus, margin with three rows of microtrichiae along it. C–D, Benacus griseus, margin with microtrichiae clearly distributed throughout it, comprising more than three rows. E–F, Curicta borelli, margin with just one or two rows of microtrichiae along it. Arrows indicate rows of microtrichiae. I, first row; II, second row; III, third row. Figure 14. View largeDownload slide Scanning electron micrographs of clavus–clavus clamp of hemelytra. General aspect. A, Laccotrephes japonensis, outer projection not flattened, extending onto inner part, far from the margin of hemelytrum. B, Limnogeton expansum, both inner and outer parts contiguous with each other, not overlapping. C, Nepa cinerea, hole system. D, Ranatra chinensis, outer projection flattened extending onto inner part, near the margin of hemelytrum. inn, inner part; mhe, margin of hemelytrum; out, outer projection. Figure 14. View largeDownload slide Scanning electron micrographs of clavus–clavus clamp of hemelytra. General aspect. A, Laccotrephes japonensis, outer projection not flattened, extending onto inner part, far from the margin of hemelytrum. B, Limnogeton expansum, both inner and outer parts contiguous with each other, not overlapping. C, Nepa cinerea, hole system. D, Ranatra chinensis, outer projection flattened extending onto inner part, near the margin of hemelytrum. inn, inner part; mhe, margin of hemelytrum; out, outer projection. Lauck & Menke (1961) considered Benacus as a subgenus and Kirkaldyia as a synonym for Lethocerus. Lee’s (1991) opinion that Lethocerus has uniquely differentiated with respect to the functional structures of the male genitalia and adaptations for swimming seems to be corroborated herein by the morphological and molecular dataset. On the other hand, Lauck & Menke’s (1961) view of the retention of plesiomorphic features in the Lethocerinae, greatly resembling Nepidae, is also reported in this study. Mahner (1993) suggested the validity of Lethocerinae to be enhanced by the presence of metathoracic scent glands in adults, a condition found only in this subfamily. In fact, this condition has been reported in six species of Lethocerus, in B. griseus and in K. deyrolli (Carayon 1971; Staddon & Thorne, 1979), but we failed to find glands in females of K. deyrolli (J. R. I. Ribeiro & D. Pluot-Sigwalt, unpubl. data); this trait was therefore not treated here. The present phylogenetic analysis based on the combined morphological and molecular dataset provides support for Lethocerinae as a monophyletic group, now with K. deyrolli and B. griseus included, despite having low support in that tree (Fig. 5). Also, BI and ML analyses using only the morphological dataset indicate the Lethocerinae to be paraphyletic with respect to Belostomatinae (Fig. 3). In contrast, the subfamily Lethocerinae now comprise lineages including Lethocerus, Benacus and Kirkaldyia, corroborating Perez-Goodwyn (2006) and Estévez & Ribeiro (2011), and this was supported in the BI, PI and ML trees using the combined dataset (Fig. 5). In analyses based on the morphological dataset, only the PI analysis reported Lethocerinae as monophyletic (Fig. 2A), and it seems that the Lethocerinae clade in our parsimony analysis could be under the influence of long-branch attraction (Magalhães & Santos, 2012). In their characterization of this subfamily, Estévez & Ribeiro (2011) used several plesiomorphic traits, such as: antennal segment 3 with long, sometimes angular, finger-like projection (character 14, state 0); segment 4 with two projections (character 15: state 0) (Fig. 11E); foretarsi 3-segmented (often appearing 2-segmented externally) (character 42: state 0) and bearing one long claw (character 44: state 0); mesial margins of ventral laterotergites meeting genital plate near its apex; tibia and tarsus of hind leg thinly compressed, much more dilated than middle tibia and tarsus; aedeagus and ventral diverticulum separate (character 71: state 0); and genital operculum of females with spines and acutely rounded at apex (characters 62 and 64: state 0). Aside from this, Lethocerinae is supported here using the combined dataset by the following unambiguous synapomorphies: (1) frontogenal suture distally closed (character 21: 0 > 1) (Fig. 8B), which is homoplastic; (2) abdominal spiracles located on mesial margins of lateral plates close to lateral sulci (character 58: 1 > 0), also homoplastic; (3) lateral lobes of segment 7 of abdomen with proximal portions subdivided into sublateral and lateral plates (character 60: 1 > 0), which is non-homoplastic and also supports that clade using the morphological dataset; and (4) median vaginal area below spermatheca with a conspicuous pouch (character 99: 0 > 1) (Fig. 19C), also non-homoplastic. Lee (1991) stated that the phallus in Lethocerus, originally composed of two simple tubules corresponding to the exophallotheca and endophallus, is strong evidence of its extreme specialization. According to him, the male genitalia in Lethocerus are characterized by the separation of aedeagus from the ventral diverticulum (also reported by Dupuis, 1955), and the construction of the phallus with various parts (i.e. the free aedeagus, ventral diverticulum, levers, knob and ligament), which are all functionally connected with the basal plate, parameres and pygophore as well as the muscles. However, the condition of aedeagus separated from the ventral diverticulum is recovered as plesiomorphic here (see above). Likewise, the abdominal sternites 3–7 divided laterally by weak, suture-like folds into median and parasternites (character 57: 1 > 0) has been used in the characterization of this subfamily by other authors (e.g. Nieser, 1975; Nieser & Melo, 1997; Perez-Goodwyn, 2006). Such a condition is an ambiguous homoplastic synapomorphy here. According to the optimization done over combined morphological and molecular analysis (Fig. 6), Benacus was herein supported by the following three unambiguous homoplastic synapomorphies: (1) microtrichiae of lateral border of hemelytrum absent (character 35: 1 > 0) (similar to Fig. 15B, C); (2) forefemora apparently smooth, without grooves (character 38: 1 > 2); and (3) pygophore with abrupt sculptured shoulder between basal portion and apical semitubular portion (character 93: 0 > 1). Likewise, Kirkaldyia is supported by four unambiguous synapomorphies, of which two are non-homoplastic: (1) pronotum with strongly developed lateral expansions (character 22: 0 > 1) and also supports that clade using the morphological dataset; and (2) furrows of forefemur with two asymmetrical grooves (character 40: 0 > 1). Finally, if we consider Lethocerinae with I. romerali included, as recovered by the morphological dataset (Fig. 2A), the following unambiguous homoplastic synapomorphies can be reported: (1) angled eyes oblong anteriorly (character 10: 1 > 0) (Fig. 8B); and (2) posterior portion of hind tibiae divided into two parts, both separated by a sulcus (character 49: 0 > 1). Figure 15. View largeDownload slide A–C, scanning electron micrographs of part of hemelytra. Microtrichiae of lateral border of hemelytrum, degree of development. A, Belostoma elegans (developed). B, Weberiella rhomboides (absent). C, Curicta borelli (absent). Arrows indicate rows of microtrichiae. Figure 15. View largeDownload slide A–C, scanning electron micrographs of part of hemelytra. Microtrichiae of lateral border of hemelytrum, degree of development. A, Belostoma elegans (developed). B, Weberiella rhomboides (absent). C, Curicta borelli (absent). Arrows indicate rows of microtrichiae. Major diagreements between morphology and molecules The Neotropical monobasic genus Horvathinia had been considered to be very different from other genera, so that it has been assigned to the subfamily Horvathiniinae by several authors (Lauck & Menke, 1961; Mahner, 1993; Schnack & Estévez, 2005; Estévez & Ribeiro, 2011), based mainly on the morphology of antennal segments, which are large, expanded and flattened ventrally, and the presence of foretarsi that are 2-segmented (often appearing 1-segmented), each bearing two very short, vestigial claws. The characters were thought to be characteristic of Horvathiniinae, and this study highlights that only the former (character 14: 0 > 1) (Fig. 11F) is an unambiguous non-homoplastic synapomorphy of Horvathinia either in some trees using the morphological dataset (Fig. 2A) or in the tree based on the combined morphological and molecular dataset (Fig. 6). In fact, this genus was supported by three other unambiguous non-homoplastic synapomorphies: (1) parameres large basally, tapering abruptly near the apex (character 66: 0 > 1) (Fig. 16B); (2) ventral diverticulum slightly concave (character 83: 3 > 0); and (3) short air lanceolate straps large at apex (character 103: 0 > 1). In the present paper, considering only morphology, the foretarsi bearing very short, inconspicuous claws (character 44: 3 > 2), the genital operculum of female fringed with hairs (character 64: 3 > 1) and the phallotheca without a pair of elongate dorsal arms almost extending to apex of phallosoma (character 76: 0 > 1) are three additional unambiguous non-homoplastic synapomorphies supporting Horvathinia (Fig. 2A). Conversely, a frons rounded (in dorsal view) (character 3: 0 > 1) (Figs 8, 9), and the lateral eye margins flush with lateral margin of pronotum and frons (character 7: 0 > 2) are unambiguous non-homoplastic synapomorphies in other trees, in which Horvathinia and Diplonychini trib. nov. comprise a monophyletic group (Figs 2B, D, E). Also, the foretarsi 2-segmented but externally usually appearing 1-segmented is an unambiguous non-homoplastic synapomorphy in other trees, in which Horvathinia and Belostomatini comprise a monophyletic group (Fig. 2C). In addition, the surface of apex of ventral diverticulum without any spines or tubercles (in ventral view) (character 89: 0 > 1) (Fig. 17A) appears to arise independently in many instances in the Belostomatinae, Lethocerinae and Nepidae. Considering the optimization done over morphological analysis, such homoplastic synapomorphy supported Horvathinia in that tree, in which Horvathinia and Diplonychini trib. nov. comprise a monophyletic group (Fig. 2B). Likewise, this homoplastic synapomorphy unambiguously supported Horvathinia in the analysis using the combined morphological and molecular dataset (Fig. 6). Figure 16. View largeDownload slide A–B, general aspect of parameres. A, Belostoma elegans (tapering slowly at apex). B, Horvathinia pelocoroides (large basally, tapering abruptly near apex). C–D, general aspect of the transverse bridge of basal plate. C, Abedus ovatus (not entire). D, Appasus sp. (clearly jointed and entire). E, Appasus major, male genitalia (dorsal view). bri, transverse bridge; dar, dorsal arms. Figure 16. View largeDownload slide A–B, general aspect of parameres. A, Belostoma elegans (tapering slowly at apex). B, Horvathinia pelocoroides (large basally, tapering abruptly near apex). C–D, general aspect of the transverse bridge of basal plate. C, Abedus ovatus (not entire). D, Appasus sp. (clearly jointed and entire). E, Appasus major, male genitalia (dorsal view). bri, transverse bridge; dar, dorsal arms. Figure 17. View largeDownload slide Scanning electron micrographs of part of ventral diverticulum. General aspect of the apex of ventral diverticulum fused to aedeagus. A, Horvathinia pelocoroides (without apicoventral protuberance). B, Limnogeton scutellatum (without apicoventral protuberance). C, Abedus signoreti (with apicoventral protuberance highlighted by dashed line). D, Weberiella rhomboides (with apicoventral protuberance highlighted with dashed line). apr, apicoventral protuberance. Figure 17. View largeDownload slide Scanning electron micrographs of part of ventral diverticulum. General aspect of the apex of ventral diverticulum fused to aedeagus. A, Horvathinia pelocoroides (without apicoventral protuberance). B, Limnogeton scutellatum (without apicoventral protuberance). C, Abedus signoreti (with apicoventral protuberance highlighted by dashed line). D, Weberiella rhomboides (with apicoventral protuberance highlighted with dashed line). apr, apicoventral protuberance. Figure 18. View largeDownload slide Scanning electron micrographs of part of the apicoventral protuberance of ventral diverticulum. Surface of apex of ventral diverticulum showing spines and tubercles. A–B, Belostoma elegans (spine-shaped). C–D, Belostoma plebejum (tuberculous-shaped). Figure 18. View largeDownload slide Scanning electron micrographs of part of the apicoventral protuberance of ventral diverticulum. Surface of apex of ventral diverticulum showing spines and tubercles. A–B, Belostoma elegans (spine-shaped). C–D, Belostoma plebejum (tuberculous-shaped). Figure 19. View largeDownload slide Female genital tract after KOH treatment with details of the spermatheca (aspect of the basal portion bearing muscular apodemes). A, Kirkaldyia deyrolli. B, Lethocerus maximum. C, Benacus griseus, genital tract and detail. D, Horvathinia pelocoroides, genital tract and detail. E, Limnogeton scutellatum, genital tract and detail. apo, apodemes; ovi, common oviduct; pou, pouch; spe, spermatheca; vag, vagina. Figure 19. View largeDownload slide Female genital tract after KOH treatment with details of the spermatheca (aspect of the basal portion bearing muscular apodemes). A, Kirkaldyia deyrolli. B, Lethocerus maximum. C, Benacus griseus, genital tract and detail. D, Horvathinia pelocoroides, genital tract and detail. E, Limnogeton scutellatum, genital tract and detail. apo, apodemes; ovi, common oviduct; pou, pouch; spe, spermatheca; vag, vagina. Figure 20. View largeDownload slide Female genital tract after KOH treatment showing details of the spermatheca: aspect of the basal part greatly inflated or not. A, Belostoma cummingsi, genital tract and basal part of spermatheca detailed. B, B. flumineum, genital tract and detail. C, B. elegans, genital tract and part of spermatheca that enters inside vagina detailed. D, B. plebejum, genital tract and part of spermatheca that enters inside vagina detailed. inf, inflated part; ovi, common oviduct; spe, spermatheca; vag, vagina. Figure 20. View largeDownload slide Female genital tract after KOH treatment showing details of the spermatheca: aspect of the basal part greatly inflated or not. A, Belostoma cummingsi, genital tract and basal part of spermatheca detailed. B, B. flumineum, genital tract and detail. C, B. elegans, genital tract and part of spermatheca that enters inside vagina detailed. D, B. plebejum, genital tract and part of spermatheca that enters inside vagina detailed. inf, inflated part; ovi, common oviduct; spe, spermatheca; vag, vagina. Figure 21. View largeDownload slide Female genital tract after KOH treatment showing details of the spermatheca: opening of the spermatheca into the vagina simple or through a vaginal ampulla. A, Appasus major, genital tract and detail. B, A. procerus. C, Diplonychus esakii. D, Weberiella rhomboides, genital tract and detail. amp, ampulla; apo, apodemes; ovi, common oviduct; spe, spermatheca; vag, vagina. Figure 21. View largeDownload slide Female genital tract after KOH treatment showing details of the spermatheca: opening of the spermatheca into the vagina simple or through a vaginal ampulla. A, Appasus major, genital tract and detail. B, A. procerus. C, Diplonychus esakii. D, Weberiella rhomboides, genital tract and detail. amp, ampulla; apo, apodemes; ovi, common oviduct; spe, spermatheca; vag, vagina. BI, PI and ML inferences of the combined molecular dataset presented here recovered (Limnogeton + (Horvathinia + Hydrocyrius)) with high support (BCP = 97%, PB = 80%, MLB = 100%) (Fig. 4). As in the morphological study, 16S and 18S partitions alone herein did not provide any further insight into the relationships of this group, thus being one of the major disagreements between morphology and molecules found (Supporting Information: Figs S1–S3). In the present paper, such a relationship could also be recovered by analysis of the combined morphological and molecular dataset (Fig. 6) based only on ambiguous non-homoplastic synapomorphies: frons (rounded) strongly curved (character 4: 0 > 1) (Fig. 9A, D, H), posterior portion of hind tibiae (without sulcus) with a conspicuous projection (character 51: 0 > 1) (Fig. 11H–J), and phallotheca without a pair of elongate dorsal arms almost extending to apex of the phallosoma (character 76: 0 > 1). In contrast, taking into account only the 18S partition dataset, the Limnogeton–Hydrocyrius relationship was recovered (Supporting Information: Fig. S2). However, according to the trees based on morphology (Fig. 2D, E), that clade was supported by only two unambiguously non-homoplastic sinapomorphies: (1) orifice of phallosoma postero-ventrally located (character 77: 1 > 0); and (2) small orifice on phallosoma, not strongly developed (character 78: 1 > 0). Additionally, when that clade is the sister group of Horvathinia + Diplonychini trib. nov. (Fig. 2E), only one unambiguous homoplastic synapomorphy supports this monophyletic group: lateral margin of eyes flush with at least the lateral margin of pronotum or frons (character 6: 0 > 1) (Figs 8, 9). Between ten and 15 unambiguous synapomorphies supported Limnogeton depending on whether trees were recovered by only the morphological or the combined morphological and molecular dataset (Figs 2A, 6). Nevertheless, the following five non-homoplastic synapomorphies supported Limnogeton in the analyses using both the morphological and the combined morphological and molecular datasets: (1) segment 3 of the antennae flattened, with short and broad projection (character 18: 0 > 1) (Fig. 11C); (2) width of the membrane and translucent margin combined in relation to greatest width of clavus approximately equal to greatest width of clavus (character 29: 0 > 1); (3) tile-like microtrichiae rounded at apex, usually toothed along the margin of its apex (character 34: 0 > 1) (Fig. 12G); (4) outer margins of the hind trochanters with a conspicuous spine directed laterally (character 47: 0 > 1) (Fig. 11G); and (5) ventral diverticulum fused with phallosoma acute at apex (character 83: 3 > 4). As already mentioned for other genera, the general body shape and colour make the members of Limnogeton easy to recognize in the field, but such traits are quite superficial and variable. In fact, they have little value in a phylogenetic sense (Ribeiro et al., 2014). There seems to be agreement with the characterization of these genera given by Poisson (1947), Estévez & Ribeiro (2011) and Ribeiro et al. (2014) and the phylogeny of the present study. According to Estévez & Ribeiro’s (2011) key to Belostomatidae genera, it is unlikely that the forefemora with one groove for reception of the tibia (a wide and flat furrow) (character 38: 1 > 0) in Limnogeton is an unambiguous non-homoplastic synapomorphy using the combined morphological and molecular dataset. Using only the morphological dataset, this condition is instead a parallelism, because such a condition is also herein reported in W. rhomboides (Fig. 2A). For Hydrocyrius, two unambiguous non-homoplastic synapomorphies supported it using the morphological dataset (Fig. 2A): (1) foretarsi bearing two large, equal or subequal claws (character 44: 3 > 4); and (2) setae on genital operculum of female with only one apical tuft (character 64: 3 > 2). However, no non-homoplastic unambiguous synapomorphies supported this genus using the combined morphological and molecular dataset (Fig. 6). The major disagreement found between morphology and molecules, regarding the position of Horvathinia, might be a problem that stems from combining noisy datasets. Parametric bootstrap analysis of the 18S dataset does not reject the null hypotheses of non-monophyly for both the Lethocerinae and the Belostomatinae (containing Horvathinia) clades. The ILD test rejects combining the COI and 16S datasets, as well as the mitochondrial and nuclear datasets (see above). Likewise, likelihood mapping diagrams revealed here that the COI fragment performed somewhat better than other partitions (Fig. 1); this could indicate that rates of nucleotide substitution were quite different (Barker & Lutzoni, 2002). On the other hand, the present study highlights that a global tree based on combining the molecular and morphological data might represent a compromise between the topologies from separate analyses, irrespective of the existence of taxa with missing data and type of data. Only relationships between Horvathinia, Limnogeton and Hydrocyrius have not been found with some methods of analysis or any combination of datasets. Topological congruence among different analyses of the molecular datasets was relatively high. Because of the apparent independence among datasets, they should in general not mislead in the same way. Thus, areas of agreement are likely to represent real groups (De Queiroz et al., 1995). Areas of disagreement occur, however, mainly between PI and other methods (Table 6), and missing data, present for some gene fragments and for fossils, appear to interfere with the parsimony reconstructions. In fact, it is likely that the ML and BI methods are less sensitive to this (Smith et al., 2004). SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article at the publisher’s web-site: Figure S1. COI: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the COI dataset (3 codon partitions: pos1=HKY+Γ, pos2=GTR+Γ, pos3=GTR+I+Γ). Thicker clades refer to those also found by maximum likelihood and maximum parsimony analyses. Support based on Bayesian clade probabilities (BCP), maximum likelihood bootstrap (MLB), and parsimony boostrap (PB). Percentages are given near to the branches as BCP/PB/MLB. Clades are colored accordingly to proposed classification. The colored names refer to the taxa affected by the present classification proposal. (+), Fossils. Figure S2. 18S rDNA fragment: Bayesian consensus phylogram. Bayesian analysis of the 18S rDNA dataset (SYM+I+Γ). Thicker clades refer to those also found by maximum likelihood and maximum parsimony analyses. Support based on Bayesian clade probabilities (BCP), maximum likelihood bootstrap (MLB), and parsimony boostrap (PB). Percentages are given near to the branches as BCP/PB/MLB. Clades are colored accordingly to proposed classification. The colored names refer to the taxa affected by the present classification proposal. (+), Fossils. Figure S3. 16S: Bayesian consensus phylogram. Bayesian analysis of the 16S dataset (HKY+I+Γ). Thicker clades refer to those also found by maximum likelihood and maximum parsimony analyses. Support based on Bayesian clade probabilities (BCP), maximum likelihood bootstrap (MLB), and parsimony boostrap (PB). Percentages are given near to the branches as BCP/PB/MLB. Clades are colored accordingly to proposed classification. The colored names refer to the taxa affected by the present classification proposal. (+), Fossils. Figure S4. Results of the parametric bootstrap analysis using COI gene. Test of monophyly of Belostomatinae with Horvathinia pelocoroides included. Model trees were constructed for each hypothesis by conducting maximum likelihood searches with taxa forced to the constraint tree (shown inset). The distributions of the difference in likelihood scores between the optimal trees and the best trees that fit the constraint are shown for 500 simulations for each hypothesis. In each case, the difference in score between the model and observed trees for the original data (arrow) was considerably greater than expected if the corresponding hypothesis were true. Thus, such hypothesis is rejected at P < 0.001. F, fossil. Figure S5. Results of the parametric bootstrap analysis using COI gene. Test of monophyly of Belostomatinae with Horvathinia pelocoroides included and Lethocerinae. Model trees were constructed for each hypothesis by conducting maximum likelihood searches with taxa forced to the constraint tree (shown inset). The distributions of the difference in likelihood scores between the optimal trees and the best trees that fit the constraint are shown for 500 simulations for each hypothesis. In each case, the difference in score between the model and observed trees for the original data (arrow) was considerably greater than expected if the corresponding hypothesis were true. Thus, such hypothesis is rejected at P < 0.001. F, fossil. Figure S6. Results of the parametric bootstrap analysis using 18S rDNA gene. Test of monophyly of Belostomatinae with Horvathinia pelocoroides included. Model trees were constructed for each hypothesis by conducting maximum likelihood searches with taxa forced to the constraint tree (shown inset). The distributions of the difference in likelihood scores between the optimal trees and the best trees that fit the constraint are shown for 500 simulations for each hypothesis. In each case, the difference in score between the model and observed trees for the original data (arrow) was not considerably greater than expected if the corresponding hypothesis were true. Thus, such hypothesis is not rejected at P < 0.001. F, fossil. Figure S7. Results of the parametric bootstrap analysis using 18S rDNA gene. Test of monophyly of Belostomatinae with Horvathinia pelocoroides included and Lethocerinae. Model trees were constructed for each hypothesis by conducting maximum likelihood searches with taxa forced to the constraint tree (shown inset). The distributions of the difference in likelihood scores between the optimal trees and the best trees that fit the constraint are shown for 500 simulations for each hypothesis. In each case, the difference in score between the model and observed trees for the original data (arrow) was not considerably greater than expected if the corresponding hypothesis were true. Thus, such hypothesis is not rejected at P < 0.001. F, fossil. Figure S8. Results of the parametric bootstrap analysis using 16S gene. Test of monophyly of Belostomatinae with Horvathinia pelocoroides included. Model trees were constructed for each hypothesis by conducting maximum likelihood searches with taxa forced to the constraint tree (shown inset). The distributions of the difference in likelihood scores between the optimal trees and the best trees that fit the constraint are shown for 500 simulations for each hypothesis. In each case, the difference in score between the model and observed trees for the original data (arrow) was considerably greater than expected if the corresponding hypothesis were true. Thus, such hypothesis is rejected at P < 0.001. F, fossil. Figure S9. Results of the parametric bootstrap analysis using 16S gene. Test of monophyly of Belostomatinae with Horvathinia pelocoroides included and Lethocerinae. Model trees were constructed for each hypothesis by conducting maximum likelihood searches with taxa forced to the constraint tree (shown inset). The distributions of the difference in likelihood scores between the optimal trees and the best trees that fit the constraint are shown for 500 simulations for each hypothesis. In each case, the difference in score between the model and observed trees for the original data (arrow) was considerably greater than expected if the corresponding hypothesis were true. Thus, such hypothesis is rejected at P < 0.001. F, fossil. Figure S10. Results of the parametric bootstrap analysis using combined gene dataset. Test of monophyly of Belostomatinae with Horvathinia pelocoroides included. Model trees were constructed for each hypothesis by conducting maximum likelihood searches with taxa forced to the constraint tree (shown inset). The distributions of the difference in likelihood scores between the optimal trees and the best trees that fit the constraint are shown for 500 simulations for each hypothesis. In each case, the difference in score between the model and observed trees for the original data (arrow) was considerably greater than expected if the corresponding hypothesis were true. Thus, such hypothesis is rejected at P < 0.001. F, fossil. Figure S11. Results of the parametric bootstrap analysis using combined gene dataset. Test of monophyly of Belostomatinae with Horvathinia pelocoroides included and Lethocerinae. Model trees were constructed for each hypothesis by conducting maximum likelihood searches with taxa forced to the constraint tree (shown inset). The distributions of the difference in likelihood scores between the optimal trees and the best trees that fit the constraint are shown for 500 simulations for each hypothesis. In each case, the difference in score between the model and observed trees for the original data (arrow) was considerably greater than expected if the corresponding hypothesis were true. Thus, such hypothesis is rejected at P < 0.001. F, fossil. [Version of Record, published online 23 August 2017; http://zoobank.org/urn:lsid:zoobank.org:pub:7EB66AA1-09CD-428F-A70A-4A6E718287CC] ACKNOWLEDGEMENTS Thanks to Nico Nieser and Ping-Ping Cheng (The Netherlands), and Ana L. Estévez (Argentina) for their support, suggestions and discussion of the manuscript. Thanks to Neusa Hamada (INPA) for providing photos and a specimen of W. rhomboides for our study. Thanks to Axel O. Bachmann (MACN), C. Magalhães (INPA), Eliane De Connick (MRAC), Gabriel Mejdalani (MNRJ), Jorge L. Nessimian (DZRJ) and colleagues from MLPA for the loan of specimens and access to collections. Thanks to Luana F. da Silva (Agência de Desenvolvimento de Águas da Prata, Meio Ambiente, São Paulo, Brazil) for collecting specimens of W. rhomboides. We are indebted to Robert W. Sites (Enns Entomology Museum, University of Missouri, USA) and Nico Nieser for donating Abedus, Diplonychus, Hydrocyrius and Nepa specimens. J.R. is indebted to colleagues Teng Li, Qiang Xie and David Redei for help in China. The first author would also like to thank his colleagues and friends Camille Desjonquères, Diego Llusia and Maram Caeser (MNHN). This research was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) - Edital Universal (2011) (n. 471034/2011-8). Additional support came through funding by UMR 7205 CNRS OSEB-MNHN. We have no conflicts of interest to declare. REFERENCES Akaike H. 1974. A new look at the statistical model identification. IEEE Transactions on Automatic Control  19: 716– 723. Google Scholar CrossRef Search ADS   Amyot CJB, Serville JGA. 1843. Histoire naturelle des insectes. Hémiptères. Librairie Encyclopédique de Roret  86: 427– 430. Andersen NM. 1995. Phylogeny and classification of aquatic bugs (Heteroptera, Nepomorpha). An essay review of Mahner’s ‘Systema Cryptoceratum Phylogeneticum’. Entomologica Scandinavica  26: 159– 166. Google Scholar CrossRef Search ADS   Armúa-de-Reyes C, Schnack JA, Estévez AL. 2005. Primer descubrimiento de Horvathinia pelocoroides Montandon como um habitante acuático (Hemíptera: Belostomatidae): em una laguna de la província de Corrientes, Argentina. Comunicaciones Científicas y Tecnológicas, Universidad Nacional Del Nordeste, Resumen B-007 . Arnett RH Jr, Samuelson GA, Nishida GM. 1993. The insect and spider collections of the world , 2nd edn. Gainesville: Sandhill Crane Press, Inc. Barker FK, Lutzoni FM. 2002. The utility of the incongruence length difference test. Systematic Biology  51: 625– 637. Google Scholar CrossRef Search ADS PubMed  Barker SC, Whiting M, Johnson KP, Murrell A. 2002. Phylogeny of the lice (Insecta, Phthiraptera) inferred from small subunit rRNA. Zoologica Scripta  32: 407– 414. Google Scholar CrossRef Search ADS   Bender W, Spierer P, Hogness DS. 1983. Chromosomal walking and jumping to isolate DNA from the ace and rosy loci and the bi thorax complex in Drosophila melanogaster. Journal of Molecular Biology  168: 17– 34. Google Scholar CrossRef Search ADS PubMed  Brazeau MD. 2011. Problematic character coding methods in morphology and their effects. Biological Journal of the Linnean Society  104: 489– 498. Google Scholar CrossRef Search ADS   Bremer K. 1994. Branch support and tree stability. Cladistics  10: 295– 304. Google Scholar CrossRef Search ADS   Brożek J. 2014. Phylogenetic signals from Nepomorpha (Insecta: Hemiptera: Heteroptera) mouthparts: stylets bundle, sense organs, and labial segments. The Scientific World Journal  2014: ID 237854. Carayon J. 1969. Emploi du noir chlorazol en anatomie microscopique des Insectes. Annales de la Société Entomologique de France (Nouvelle série)  5: 179– 193. Carayon J. 1971. Notes et documents sur l’appareil odorant métathoracique des Hémiptères. Annales de la Société Entomologique de France (Nouvelle série)  7: 737– 770. Chen P, Nieser N, Ho JZ. 2004. Review of Chinese Ranatrinae (Hemiptera: Nepidae), with descriptions of four new species of Ranatra Fabricius. Tijdschrift voor Entomologie  147: 81– 102. Google Scholar CrossRef Search ADS   Cruaud A, Jabbour-Zahab R, Genson G, Cruaud C, Couloux A, Kjellberg F, van Noort S, Rasplus J-Y. 2010. Laying the foundations for a new classification of Agaonidae (Hymenoptera: Chalcidoidea), a multilocus phylogenetic approach. Cladistics  26: 359– 387. Davis CM, Roth VL. 2008. The evolution of sexual size dimorphism in cottontail rabbits (Sylvilagus, Leporidae). Biological Journal of the Linnean Society  95: 141– 156. Google Scholar CrossRef Search ADS   De Carlo JA. 1938. Los Belostómidos americanos. Anales del Museo Argentino de Ciencias Naturales ‘Bernardino Rivadavia’ Buenos Aires  39: 189– 260. De Carlo JA. 1957. Identificación de las especies del género Horvathinia Montandon. Descripción de tres especies nuevas (Hemiptera - Belostomatidae). Revista de la Sociedad Entomológica Argentina  20: 45– 52. De Carlo JA. 1964. Los Ranatridae de America (Hemiptera). Revista del Museo Argentino de Ciencias Naturales Bernardino Rivadavia e Instituto Nacional de Investigación de las Ciencias Naturales. Entomologia  1: 133– 227. De Carlo JA. 1966. Un nuevo género, nuevas especies y referencias de otras poco conocidas de la familia Belostomatidae (Hemiptera). Revista de la Sociedad Entomologica Argentina  28: 97– 109. de Pinna MCC. 1991. Concepts and tests of homology in the cladistic paradigm. Cladistics  7: 367– 394. Google Scholar CrossRef Search ADS   De Queiroz A, Donoghue MJ, Kim J. 1995. Separate versus combined analysis of phylogenetic evidence. Annual Review of Ecology and Systematics  26: 657– 681. Google Scholar CrossRef Search ADS   Dupuis C, Carvalho JCM. 1956. Heteroptera. In: Tuxen SL, ed. Taxonomist’s glossary of genitalia in insects . Copenhagen: Munksgaard. Dupuis C. 1955. Les génitalia des hémiptères hétéroptères (génitalia externes des deux sexes; voies ectodermiques femelles). Revue de la morphologie. Lexique de la nomenclature. Index bibliographique analytique. Mémoires du Muséum National d’Histoire Naturelle (série A, Zoologie)  6: 183– 278. Edgar RC. 2004. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Research  32: 1792– 1797. Google Scholar CrossRef Search ADS PubMed  Estévez AL, Polhemus JT. 2001. The small species Belostoma (Hemiptera: Belostomatidae). I. Introduction, key to species groups and a revision of the denticolle group. Iheringia, Série Zoologia  91: 151– 158. Google Scholar CrossRef Search ADS   Estévez AL, Ribeiro JRI. 2011. Weberiella De Carlo, 1966 (Insecta: Heteroptera: Belostomatidae) revisited: redescription with a key to the genera of Belostomatidae and considerations on back-brooding behaviour. Zoologischer Anzeiger  250: 46– 54. Google Scholar CrossRef Search ADS   Fabricius JC. 1781. Species Insectorum Exhibitens Eorum Differentias Specificas, Synonyma Auctorum, Loca Natalia, Metamorphosin Adjectis Observationibus, Descriptionibus, Vol. 2 . Hamburgi et Kilonii: Impensis Carol, Ernest, Bohnii. Fabricius JC. 1790. Nova insectorum genera. Naturhistorie Selskabet  1: 213– 228. Farris JS, Källersjö M, Kluge AG, Bult C. 1994. Testing significance of congruence. Cladistics  10: 315– 320. Google Scholar CrossRef Search ADS   Felsenstein J. 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution  39: 783– 791. Google Scholar CrossRef Search ADS PubMed  Fittkau EJ. 1977. Kinal and kinon, habitat and coenosis of the surface drift as seen in Amazonian running waters. Geo-Eco-Trop  1: 9– 20. Folmer O, Black M, Hoeh W, Lutz R, Vrijenhoek R. 1994. DNA primers for amplification of mitochondrial cytochrome c oxidase subunit I from diverse metazoan invertebrates. Molecular Marine Biology and Biotechnology  3: 294– 299. Google Scholar PubMed  Forey PL, Kitching IJ. 2000. Experiments in coding multistate characters. In: Scotland R, Pennington RT, eds. Homology and systematics: coding characters for phylogenetic analysis . London: Taylor and Francis, 54– 80. Germain J, Chatot C, Meusnier I, Artige E, Rasplus J-Y, Cruaud A. 2013. Molecular identification of Epitrix potato flea beetles (Coleoptera: Chrysomelidae) in Europe and North America. Bulletin of Entomological Research  103: 354– 362. Google Scholar CrossRef Search ADS PubMed  Germar EF. 1839. Die versteinerten Insecten Solenhofens. Nova Acta Physico-Medica Academiae Caesareae Leopoldino-Carolinae Naturae Curiosorum  19: 187– 222. Gilbert MTP, Moore W, Melchior L, Worobey M. 2007. DNA extraction from dry museum beetles without conferring external morphological damage. Plos One  2: e272. Google Scholar CrossRef Search ADS PubMed  Goldman N, Anderson JP, Rodrigo AG. 2000. Likelihood-based tests of topologies in phylogenetics. Systematic Biology  49: 652– 670. Google Scholar CrossRef Search ADS PubMed  Gorb SN, Goodwyn PJP. 2003. Wing-locking mechanisms in aquatic Heteroptera. Journal of Morphology  257: 127–1 46. Google Scholar CrossRef Search ADS PubMed  Hebsgaard MB, Andersen NM, Damgaard J. 2004. Phylogeny of the true water bugs (Nepomorpha: Hemiptera-Heteroptera) based on 16S and 28S rDNA and morphology. Systematic Entomology  29: 488– 508. Google Scholar CrossRef Search ADS   Hennig W. 1968. Elementos de una Sistemática Filogenética . Buenos Aires: EUDEBA. Hua J, Li M, Dong P, Cui Y, Xie Q, Bu W. 2009. Phylogenetic analysis of the true water bugs (Insecta: Hemiptera: Heteroptera: Nepomorpha): evidence from mitochondrial genomes. BMC Evolutionary Biology  9: 134– 145. Google Scholar CrossRef Search ADS PubMed  Hungerford HB. 1919. Notes on the aquatic Hemiptera. The Kansas University Science Bulletin  11: 141– 151. Jattiot R, Bechly G, Garrouste R, Nel A. 2012. An enigmatic Nepoidea from the Lower Cretaceous of Brazil (Hemiptera: Heteroptera). Cretaceous Research  34: 344– 347. Google Scholar CrossRef Search ADS   Kass RE, Raftery AE. 1995. Bayes factor. Journal of the American Statistical Association  90: 773– 795. Google Scholar CrossRef Search ADS   Katoh K, Standley DM. 2013. MAFFT: Multiple sequence alignment software version 7: Improvements in performance and usability. Molecular Biology and Evolution  30: 772– 780. Google Scholar CrossRef Search ADS PubMed  Keffer SL. 1996. Systematics of the New World waterscorpion genus Curicta Stål (Heteroptera: Nepidae). Journal of the New York Entomological Society  104: 117– 215. Keffer SL. 2004. Morphology and evolution of waterscorpion male genitalia (Heteroptera: Nepidae). Systematic Entomology  29: 142– 172. Google Scholar CrossRef Search ADS   Lalitha TG, Shyamasundari K, Hanumantha-Rao K. 1997. Morphology and histology of the female reproductive system of Abedus ovatus Stål (Belostomatidae: Hemiptera: Insecta). Memórias do Instituto Oswaldo Cruz  92: 129– 135. Google Scholar CrossRef Search ADS   Lansbury I. 1972. A review of the Oriental species of Ranatra Fabricius. Transactions of the Royal Entomological Society of London  124: 287– 341. Google Scholar CrossRef Search ADS   Laporte FLD. 1832. Essai d’une classification systématique de l’ordre des hémiptéres (hémiptères hétéroptères, Latr.). Magasin de Zoologie  2: 1– 88. Latreille PA. 1807. Genera Crustaceorum et Insectorum Secundum Ordinem Naturalem in Familias Disposita, Iconibus Exemplisque Plurimis Explicata, Vol. 3 . Parisiis et Argentorati: Amand Koenig. Lauck DR, Menke AS. 1961. The higher classification of the Belostomatidae (Hemiptera). Annals of the Entomological Society of America  54: 644– 657. Google Scholar CrossRef Search ADS   Lauck DR. 1962. A monograph of the genus Belostoma (Hemiptera) Part I. Introduction to B. dentatum and subspinosum groups. Bulletin of the Chicago Academy of Sciences  11: 34– 81. Lauck DR. 1963. A monograph of the genus Belostoma (Hemiptera), Part. II. B. aurivillianum, stollii, testaceopallidum, dilatatum, and discretum groups. Bulletin of the Chicago Academy of Sciences  11: 82– 101. Lauck DR. 1964. A monograph of the genus Belostoma (Hemiptera) Part. III. B. triangulum, bergi, minor, bifoveolatum and flumineum groups. Bulletin of the Chicago Academy of Sciences  11: 102– 154. Lee C. 1991. Morphological and phylogenetic studies on the true water bugs. Nature & Life  21: 1– 183. Lewis PO. 2001. A likelihood approach to estimating phylogeny from discrete morphological character data. Systematic Biology  50: 913– 925. Google Scholar CrossRef Search ADS PubMed  Li M, Tian Y, Zhao Y, Bu W. 2012. Higher Level phylogeny and the first divergence time estimation of Heteroptera (Insecta: Hemiptera) based on multiple genes. Plos One  7: e32152. Google Scholar CrossRef Search ADS PubMed  Linnaeus C. 1758. Systema Naturae per Regna tria Naturae, Secundum Classes, Ordines, Genera, Species, cum Characteribus, Diferentiis, Synonymis, Locis . Editio Decima, Reformata, Holmiae: Laurentii Salvii. Lis JA, Ziaja DJ, Lis P. 2011. Recovery of mitochondrial DNA for systematic studies of Pentatomoidea (Hemiptera: Heteroptera): successful PCR on early 20th century dry museum specimens. Zootaxa  2748: 18– 28. Maddison DR, Maddison WP. 2005. MacClade 4: Analysis of phylogeny and character evolution. Version 4.08a . Available at: http://macclade.org. Maddison WP, Maddison DR. 2015. Mesquite: a modular system for evolutionary analysis. Version 3.04 . Available at: http://mesquiteproject.org Magalhães ILF, Santos AJ. 2012. Phylogenetic analysis of Micrathena and Chaetacis spiders (Araneae: Araneidae) reveals multiple origins of extreme sexual size dimorphism and long abdominal spines. Zoological Journal of the Linnean Society  166: 14– 53. Mahner M. 1993. Systema Cryptoceratorum Phylogeneticum (Insecta, Heteroptera). Zoologica  48: 1– 302. Malm T, Johanson KA. 2008. Revision of the New Caledonian endemic genus Gracilipsodes (Trichoptera: Leptoceridae: Grumichellini). Zoological Journal of the Linnean Society  153: 425– 452. Google Scholar CrossRef Search ADS   Martínez-Delclòs X, Nel A, Popov YA. 1995. Systematics and functional morphology of Iberonepa romerali n. gen. and sp., Belostomatidae from the Spanish Lower Cretaceous (Insecta, Heteroptera). Journal of Paleontology  69: 496– 508 Google Scholar CrossRef Search ADS   Mayr GL. 1853. Zwei neue wanzen aus Kordofan. Verhandlungen der Zoologisch-Botanischen Vereins in Wien  2: 14– 18. Menke AS. 1960. A taxonomic study of the genus Abedus Stål. University of California Press  16: 393– 40. Menke AS. 1965. A new South American toe biter (Hemiptera, Belostomatidae). Contributions in Science  89: 2– 4. Merrit RW, Cummins KW. 1996. Aquatic insects of North America, 3rd edn . Dubuque: Kendal/Hunt Publishing Company. Montandon AL. 1909. Nepidae et Belostomidae. Notes diverses et descriptions d’espèces nouvelles. Annales Historico-Naturales Musei Nationalis Hungarici  7: 59– 70. Montandon AL. 1911. Deux genres nouveaux d’hydrocorises. Annales Musei Nationalis Hungarici  9: 244– 250. Monte O. 1945. Baratas d’água. Chácaras e quintais  71: 454– 458. Moreira FFF, Barbosa JF, Ribeiro JRI, Alecrim VP. 2011. Checklist and distribution of semiaquatic and aquatic Heteroptera (Gerromorpha and Nepomorpha) occurring in Brazil. Zootaxa  2958: 1– 74. Müller J, Reisz RR. 2006. The phylogeny of early eurepties: comparing parsimony and Bayesian approaches to the investigation of a basal fossil clade. Systematic Biology  55: 503– 511. Google Scholar CrossRef Search ADS PubMed  Nel A, Paicheler JC. 1992. Les Heteroptera aquatiques fossiles, état actuel des connaissances (Heteroptera: Nepomorpha et Gerromorpha). Entomologica Gallica  3: 159– 182 Nel A, Waller A. 2006. A giant water bug from the Lower Cretaceous Crato Formation of Brazil (Heteroptera: Belostomatidae: Lethocerinae). Zootaxa  1220: 63– 68. Nel A. 1991. Analyses d’entomofaunes cénozoïques. Intérêts de la Paléoentomologie pour les Sciences de la Terre et de la Vie . Unpublished D. Phil Thesis, Université de Reims-Champagne-Ardenne. Nieser N, Melo AL. 1997. Os heterópteros aquáticos de Minas Gerais. Guia introdutório com chave de identificação para as espécies de Nepomorpha e Gerromorpha . Belo Horizonte: Ed. UFMG. Nieser N, Zettel H, Chen P-P. 2009. Notes on Laccotrephes Stål, 1866 with the description of a new species of the L. griseus group (Insecta: Heteroptera: Nepidae). Annalen des Naturhistorischen Museums in Wien  110B: 11– 20. Nieser N. 1975. The water bugs (Heteroptera: Nepomorpha) of the Guyana Region. Studies on the Fauna of Suriname and Other Guyanas  16: 88– 128. Nylander JAA. 2004. MrModeltest v2 . Program distributed by the author. Uppsala: Evolutionary Biology Centre, Uppsala University. Ohba S. 2011. Field observation of predation on a turtle by giant water-bug. Entomological Science  14: 364– 365. Google Scholar CrossRef Search ADS   Pellerano GN, De Carlo JM. 1985. Espermatecas de hemipteros acuáticos. Physis  43: 17– 22. Pendergrast JG. 1957. Studies on the reproductive organs of the Heteroptera with a consideration of their bearing on classification. Transactions of the Royal Entomological Society of London  109: 1– 63. Google Scholar CrossRef Search ADS   Perez-Goodwyn PJ. 2006. Taxonomic revision of the subfamily Lethocerinae Lauck & Menke (Heteroptera: Belostomatidae). Stuttgarter Beiträge zur Naturkunde Serie A (Biologie)  695: 1– 71. Pluot-Sigwalt D. 1986. Les glandes tégumentaires des coléoptères Scarabeidae: structure et diversité des canalicules. Annales de la Société Entomologique de France (Nouvelle série)  22: 163– 182. Pluot-Sigwalt D, Lis JA. 2008. Morphology of the spermatheca in the Cydnidae (Hemiptera: Heteroptera): bearing of its diversity on classification and phylogeny. European Journal of Entomology  105: 279– 312. Google Scholar CrossRef Search ADS   Poisson R. 1949. Hémiptères aquatiques. Exploration du Parc National Albert (Mission G. F. de Witte (1933–1935)  58: 1– 94. Poisson R. 1954. Hydrocorises (2e contribution) (Mission Lamotte et Roy, Juillet-Décembre 1951). Mémoires de l’Institut Français d’Afrique Noire  40: 359– 370. Poisson, RA. 1947. Hydrocorises du Cameroun. (Mission J. Carayon 1947). Revue Française d’Entomologie  15: 167– 177. Polhemus DA, Polhemus JT. 2013. Guide to the aquatic Heteroptera of Singapore and Peninsular Malaysia. X. Infraorder Nepomorpha-families Belostomatidae and Nepidae. The Raffles Bulletin of Zoology  61: 25– 45. Polhemus JT. 1995. Nomenclatural and synonymical notes on the genera Diplonychus Laporte and Appasus Amyot and Serville (Heteroptera: Belostomatidae). Proceedings of the Entomological Society of Washington  97: 649– 653. Popov YA. 1971. Historical development of Hemiptera infraorder Nepomorpha. Trudy Paleontologicheskogo Instituta AN SSSR  129: 1– 230 [in Russian]. Rambaut A, Suchard MA, Xie D, Drummond AJ. 2014. Tracer v1.6 . Available at: http://beast.bio.ed.ac.uk/Tracer Rambaut A. 2010. FigTree v1.3.1 . Edinburgh: Institute of Evolutionary Biology, University of Edinburgh. Ribeiro JRI, Meyin-a-Ebong SE, Le-Gall P, Guilbert E. 2014. A taxonomic synopsis of Limnogeton Mayr, 1853 (Insecta: Hemiptera: Heteroptera: Belostomatidae). Zootaxa  3779: 573– 584. Google Scholar CrossRef Search ADS PubMed  Ribeiro JRI. 2007. A review of the species of Belostoma Latreille, 1807 (Hemiptera: Heteroptera: Belostomatidae) from the four southeastern Brazilian States. Zootaxa  1477: 1– 70. Ronquist F, Teslenko M, van der Mark P, Ayres D, Darling A, Ohna SH, Larget B, Liu L, Suchard MA, Huelsenbeck JP. 2012. MrBayes 3.2: efficient Bayesian phylogenetic inference and model choice across a large model space. Systematic Biology  61: 539– 542. Google Scholar CrossRef Search ADS PubMed  Schlüter T. 1981. Fossile Insekten aus dem Jura/Kreide- Grenzbereich Südwest-Ägyptens. Berliner Geowissenschlaftliche Abhandlungen, A  32: 33– 52. Schmidt H, Strimmer K, Vingron M, von Haeseler A. 2002. TREE-PUZZLE: maximum likelihhod phylogenetic analysis using quartets and parallel computing. Bioinformatics  18: 502– 504. Google Scholar CrossRef Search ADS PubMed  Schnack JA, Estévez AL. 1990. On the taxonomic status of Abedus Stål (Hemiptera, Belostomatidae). Revista Brasileira de Entomologia  34: 267– 269. Schnack JA, Estévez AL. 2005. On the taxonomic status of the genus Horvathinia Montandon (Hemiptera: Belostomatidae). Zootaxa  1016: 21– 27. Google Scholar CrossRef Search ADS   Schuh RT, Slater JA. 1995. True bugs of the world (Hemiptera: Heteroptera) . New York: Cornell University Press. Scudder GGE. 1959. The female genitalia of the Heteroptera: morphology and bearing on classification. Transactions of the Royal Entomological Society of London  111: 405– 467. Google Scholar CrossRef Search ADS   Sereno PC. 2007. Logical basis for morphological characters in phylogenetics. Cladistics  23: 565– 587. Simon C, Buckley TR, Frati F, Stewart JB, Beckenbach AT. 2006. Incorporating molecular evolution into phylogenetic analysis, and a new compilation of conserved polymerase chain reaction primers for animal mitochondrial DNA. Annual Review of Ecology, Evolution, and Systematics  37: 545– 579. Google Scholar CrossRef Search ADS   Simon C, Frati F, Beckenbach A, Crespi B, Liu H, Flook P. 1994. Evolution, weighting, and phylogentic utility of mitochondrial gene sequences and a compilation of conserved polymerase chain reaction primers. Annals of the Entomological Society of America  87: 651– 701. Google Scholar CrossRef Search ADS   Smith VS, Page RDM, Johnson KP. 2004. Data incongruence and the problem of avian louse phylogeny. Zoologica Scripta  33: 239– 259. Google Scholar CrossRef Search ADS   Sokal RR, Rohlf FJ. 1995. Biometry, 3rd edn . New York: W.H. Freeman. Sorenson MD, Franzosa EA. 2007. TreeRot, version 3 . Boston: Boston University. Spinola M. 1850. Tavola Sinottica dei Generi Spettanti alla Classe degli insetti Artrodiginati, Hemiptera, Linn. Latr. - Rhyngota, Fab. - Rhynchota, Burm . Modena: R. D. Camera. Spooner D. 1938. The phylogeny of the hemiptera based on a study of the head capsule. Illinois Biological Monographs  16: 1– 102. Staddon BW, Thorne MJ. 1979. The methathoracic scent gland system in Hydrocorisae. Systematic Entomology  4: 239– 250. Google Scholar CrossRef Search ADS   Stål C. 1861. Nova methodus familias quasdam Hemipterorum disponendi. Öfversigt af Kungliga Vetenskapsakademiens Förhandlingar  18: 195– 212. Stål C. 1862. Hemiptera mexicana enumeravit speciesque novas descripsit (continuatio). Stettiner Entomologische Zeitung  23: 437– 462. Stål C. 1866. Hemiptera Africana 3 (1865): 1–200 . Holmiae: Nordstedtiana. Stamatakis A. 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics  22: 2688– 2690. Google Scholar CrossRef Search ADS PubMed  Swofford DL. 2002. PAUP: Phylogenetic analysis using parsimony, (and other methods), version 4.0 . Sunderland: Sinauer Associates. Takiya DM, Tran P, Dietrich CH, Moran NA. 2006. Co-cladogenesis spanning three phyla: Leafhoppers (Insecta: Hemiptera: Cicadellidae) and their dual bacterial symbionts. Molecular Ecology  15: 4175– 4191. Google Scholar CrossRef Search ADS PubMed  Tamura K, Dudley J, Nei M, Kumar S. 2007. MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) software version 4.0. Molecular Biology and Evolution  24: 1596– 1599. Google Scholar CrossRef Search ADS PubMed  Thomsen PF, Elias S, Gilbert MTP, Haile J, Munch K, Kuzmina S, Froese DG, Sher A, Holdaway RN, Willerslev E. 2009. Non-destructive sampling of ancient insect DNA. Plos One  4: e5048. Google Scholar CrossRef Search ADS PubMed  Uhler, PR . 1861. Descriptions of four species of Hemiptera collected by the North-Western Boundary Survery. Proceedings of the Academy of Natural Sciences of Philadelphia  1861: 284– 286 Viscarret MM, Bachmann AO. 1997. Estudio de la espermateca en especies argentinas del género Belostoma (Heteroptera: Belostomatidae). Revista de la Sociedad Entomológica Argentina  56: 39– 42. von Dohlen CD, Moran NA. 1995. Molecular phylogeny of the Homoptera: a paraphyletic taxon. Journal of Molecular Evolution  41: 211– 223. Google Scholar CrossRef Search ADS PubMed  Wilkinson M, Pisani D, Cotton JA, Corfe I. 2005. Measuring support and finding unsupported relationships in supertrees. Systematic Biology  54: 823– 831. Google Scholar CrossRef Search ADS PubMed  Yeates D. 1992. Why remove autapomorphies? Cladistics  8: 387– 389. Google Scholar CrossRef Search ADS   Zamboni JC. 2001. Contribution to the knowledge of the aquatic paleoentomofauna from Santana Formation (Araripe basin, Lower Cretaceous, northeast Brazil) with description of new taxa. Acta Geologica Leopoldensia  24: 129– 135. APPENDIX 1 Matrix of morphological characters for the phylogenetic analysis of Belostomatidae. Innaplicable characters are coded as ‘-’ and missing data as ‘?’. Fossil taxa are indicated with ‘(+)’. Eleven outgroup taxa (four genera) and 22 ingroup species representing all 13 genera of Belostomatidae was coded based on 104 delimited morphological characters. View largeDownload slide View largeDownload slide APPENDIX 2 Morphological characters and states for the phylogenetic analysis of Belostomatidae. Consistency (CI) and retention (RI) indices were estimated by considering the optimization done over combined morphological and molecular tree. Body Body, general aspect: (0) subcylindric; (1) flattened. CI = 1.00; RI = 1.00. Head 2. Head, degree of development in relation to largest width of pronotum, dorsal view: (0) distinctly narrower; (1) distinctly larger than anterior portion and largest width; (2) somewhat larger than anterior portion and narrower than largest width; (3) equal to anterior portion and narrower than largest width. CI = 0.60; RI = 0.87 (Figs 8, 9). 3. Frons, dorsal aspect: (0) acute; (1) rounded. CI = 0.50; RI = 0.75 (Figs 8, 9). 4. Frons rounded, degree of development: (0) somewhat smoothly rounded; (1) strongly curved. CI = 1.00; RI = 0 (Figs 8, 9). 5. Vertex, longitudinal carina, degree of development: (0) obviously developed; (1) absent (Fig. 9C). CI = 0.33; RI = 0.71. 6. Lateral eye margins, general aspect: (0) neither flushing with lateral margin of pronotum nor with frons; (1) flushing with at least one of those parts. CI = 0.25; RI = 0.0.63 (Figs 8, 9). 7. Lateral eye margins when flushing with with pronotum and/or frons, general aspect: (0) flushing with lateral margin of pronotum, protruding laterally; (1) flushing with frons, not protrunding laterally; (2) flushing with both. CI = 0.67; RI = 0.67 (Figs 8, 9). 8. Eyes, form: (0) rounded with their posterior borders curved; (1) somewhat projected laterally with angled borders. CI = 1.00; RI = 1.00 (Figs 8, 9). 9. Eyes rounded, degree of development: (0) poorly developed; (1) strongly developed. CI = 1.00; RI = 1.00 (Fig. 8). 10. Angled eyes, general aspect: (0) oblong anteriorly; (1) oblong laterally. CI = 0.33; RI = 0.50 (Fig. 8). 11. Directed eyes, degree of development in both lateral and frontal views: (0) completely globular, readily evident above of the frons; (1) poorly globular; (2) slightly planar, with curved borders; (3) slightly planar, with angled borders; (4) planar, with angled borders, streaming little with frons; (5) planar, with angled borders, readily streaming with frons. CI = 0.83; RI = 0.83 (Fig. 9). 12. Laterally produced eyes, degree of development of their area without cornea: (0) strongly developed in lateral view; (1) poorly developed in lateral view; (2) never developed. CI = 0.67; RI = 0.83. 13. Antennae, number of segments: (0) three; (1) four. CI = 1.00; RI = 1.00 (Fig. 11). 14. Antennae with four segments, aspect of third segment: (0) with one finger-like projection; (1) with a large expanded dorsal lobe. CI = 1.00; RI = 0 (Fig. 11). 15. Antennae with four segments, aspect of fourth segment; (0) with two finger-like projections; (1) with one finger-like projection. CI = 1.00; RI = 1.00 (Fig. 11). 16. Antennae with four segments, form of fourth segment with only one finger-like projection: (0) short, dorsoventrally elongated; (1) more bulbous than prolongations of segments 2 and 3; (2) slightly more bulbous than prolongations of segments 2 and 3; (3) similar to prolongations of segments 2 and 3. CI = 0.75; RI = 0.86 (Fig. 11). 17. Antennae with four segments, form of segment 2 ventrally: (0) not flattened; (1) flattened, with short and broad projection. CI = 0.50; RI = 0 (Fig. 11C). 18. Antennae with four segments, form of segment 3 ventrally: (0) not flattened; (1) flattened, with short and broad projection. CI = 1.00; RI = 0 (Fig. 11C). 19. Antennae with four segments, length of segment 1 in relation to lateral prolongations of segments 2 and 3, and segment 4 in relation to segment 1: (0) all shorter; (1) shorter and equal; (2) equal and longer. CI = 0.67; RI = 0.50 (Fig. 11). 20. Rostrum, length of segment 1 in relation to its greatest thickness: (0) thicker than long; (1) longer than thick. CI = 0.33; RI = 0.80. 21. Frontogenal suture, aspect, dorsal view: (0) slightly convergent and opened distally; (1) distally closed. CI = 0.33; RI = 0.67 (Fig. 9C-D). Thorax 22. Pronotum, general aspect: (0) without visible lateral expansions; (1) with strongly developed lateral expansions. CI = 1.00; RI = 0. 23. Pronotum, degree of development of longitudinal carinae: (0) strongly evident; (1) somewhat or poorly evident. CI = 0.33; RI = 0.71. 24. Fovea, degree of development: (0) poorly developed; (1) conspicuously evident (Fig. 9C, D). CI = 1.00; RI = 1.00. 25. Prosternum, general aspect: (0) without elevation in ventral view; (1) with an elevation in lateral view. CI = 0.50; RI = 0.86. 26. Elevation on prosternum, general aspect: (0) widened; (1) slender as a carina. CI = 0.50; RI = 0.50. 27. Prosternal carina, degree of development: (0) elevated; (1) poorly elevated. CI = 0.50; RI = 0.75. There are some species of Appasus which have an elevated carina, although such a genus has been considered to have a carina slightly elevated (J. R. I. Ribeiro, in prep.). 28. Hemelytrum, degree of development of membrane: (0) differentiated from the elytra; (1) not differentiated from the elytra. CI = 1.00; RI = 0. 29. Hemelytrum, width of membrane and translucent margin combined in relation to greatest width of clavus: (0) distinctly larger than greatest width of clavus; (1) approximately equal to greatest width of clavus; (2) smaller than greatest width of clavus. CI = 1.00; RI = 1.00. 30. Hemelytrum, form of pruinose area: (0) as a long stripe; (1) as a rounded stripe; (2) as a narrow and short glabrous stripe far from margin of membrane; (3) as a rounded area. CI = 1.00; RI = 1.00. 31. Clavus, general aspect of the clamp: (0) with outer projection flattened, overlapping inner part; (1) with outer projection reduced and inner part as a hole; (2) with both inner and outer parts contiguous with each other, not overlapping; (3) with outer projection not flattened, overlapping inner part. CI = 0.60; RI = 0.82 (Fig. 14). 32. Microtrichiae on outer carina, degree of development: (0) with fewer than three rows along external margin; (1) with three rows along external margin, covering small portion in dorsal view; (2) with many rows (more than four), covering considerable portion in dorsal view. CI = 0.33; RI = 0.73 (Fig. 13). 33. Tile-like microtrichiae, shape: (0) rounded at apex; (1) strongly acute at apex. CI = 1.00; RI = 1.00 (Fig. 12). 34. Tile-like rounded microtrichiae, general aspect: (0) never toothed along the margin of its apex; (1) usually toothed along the margin of its apex. CI = 1.00; RI = 0 (Fig. 12). 35. Microtrichia of lateral border of hemelytrum, degree of development: (0) absent; (1) present. CI = 0.25; RI = 0.67 (Fig. 15). 36. Metacoxae, general aspect: (0) short and free; (1) firmly united with metapleuron. CI = 1.00; RI = 1.00. 37. Tarsi, number of segments: (0) all tarsi mono-segmented; (1) tarsi 2- or 3-segmented, rarely foretarsi 1-segmented. CI = 1.00; RI = 1.00. 38. Forefemur, number of grooves for reception of tibia: (0) one; (1) two; (2) without grooves. CI = 0.40; RI = 0.75. 39. Furrow of forefemur with one groove, general aspect: (0) narrow and profound; (1) wide and flat. CI = 0.50; RI = 0.80. 40. Furrows of forefemur with two grooves, general aspect: (0) symmetrical; (1) asymmetrical. CI = 1.00; RI = 0. 41. Furrows, general aspect: (0) with spurs inside; (1) without spurs but spines inside. CI = 1.00; RI = 1.00. 42. Foretarsus, number of segments: (0) three (often appearing 2-segmented externally); (1) two (externally usually appearing 1-segmented). CI = 0.50; RI = 0.88. 43. Foretarsus 2-segmented, degree of development of segment 1: (0) inconspicuous, poorly developed; (1) conspicuous, developed. CI = 1.00; RI = 0. 44. Foretarsus, degree of development of claws: (0) bearing only one long claw; (1) bearing one anterior claw and one vestigial to absent posterior claw; (2) bearing two very short, inconspicuous claws; (3) bearing both claws vestigial; (4) bearing two large equal or subequal claws. CI = 1.00; RI = 1.00. 45. Legs, general aspect taking into account middle and hind tibia and tarsus: (0) not different from each other; (1) completely different from each other, with mesothoracic legs long and slender and metathoracic legs wide in relation to its length. CI = 1.00; RI = 0. 46. Legs not different from each other, general aspect: (0) slender and cursorial, probably not having any of the legs modified for swimming; (1) broadly flattened, with swimming setae, being probably developed for swimming; (2) flattened, not broadly dilated, with swimming setae. CI = 0.67; RI = 0.93. 47. Hind trochanters, aspect of their outer margins: (0) with short hairs or bristles; (1) with a conspicuous spine directed laterally. CI = 1.00; RI = 0 (Fig. 11G). 48. Outer margin of hind trochanter, general aspect: (0) without hairs, naked, never carinated; (1) with hairs, carinated. CI = 1.00; RI = 1.00. 49. Posterior portion of hind tibiae, general aspect: (0) not subdivided in two parts; (1) divided in two parts, both separated by a sulcus. CI = 0.33; RI = 0.60. 50. Posterior portion of hind tibiae without sulcus, general aspect: (0) without a projection; (1) with a projection. CI = 0.20; RI = 0.50 (Fig. 11H–J). 51. Projection of posterior portion of hind tibiae without sulcus, general aspect: (0) with a poorly developed projection; (1) with a conspicuous projection. CI = 1.00; RI = 0 (Fig. 11H–J). Abdomen 52. Lateral margins in adults, aspect: (0) smooth; (1) not smooth but interrupted at the borders between the segments (scale-like abdomen). CI = 1.00; RI = 0. 53. Parasternites, degree of development: (0) visible; (1) not visible. CI = 1.00; RI = 1.00 (Fig. 7). 54. Pubescence of ventral laterotergites, degree of development: (0) absent; (1) present. CI = 1.00; RI = 1.00. 55. Pubescence of ventral laterotergite 3, degree of development: (0) attaining entire external margin; (1) not attaining entire external margin. CI = 0.33; RI = 0.60. 56. Pubescence of ventral laterotergite 4, degree of development: (0) attaining entire external margin; (1) not attaining external margin. CI = 1.00; RI = 1.00. 57. Sterna, aspect of abdominal sternites 3–7: (0) divided laterally by weak, suture-like folds into median and parasternites; (1) undivided laterally by weak folds. CI = 0.33; RI = 0.67. 58. Sterna, distribution of abdominal spiracles: (0) located on mesial margins of lateral plates close to lateral sulci; (1) located near centre of lateral plates, far removed from lateral sulci. CI = 0.50; RI = 0.80. 59. Lateral sulci, degree of development, ventral view: (0) terminating near proximal angles of right lateral plate 7; (1) terminating near left proximal angles of lateral plate 7. CI = 1.00; RI = 1.00. 60. Lateral lobes 7, general aspect: (0) proximal portions subdivided into sublateral and lateral plates; (1) proximal portions formed entirely or largely by lateral plates, with sublateral plates developed as minute triangular sclerite. CI = 1.00; RI = 1.00. 61. Pubescence of connexivum, degree of development: (0) not extending or extending posteriorly along less than half of genital operculum; (1) extending posteriorly along about half of or almost entire genital operculum. CI = 0.20; RI = 0.43. 62. Genital operculum of female, aspect of apex: (0) without setae and acutely rounded; (1) with setae nearly to external margin (bristle combs) and usually with small incised notch apically. CI = 1.00; RI = 1.00. 63. Genital operculum of female, general aspect: (0) ovipositor-like; (1) never ovipositor-like. CI = 1.00; RI = 1.00. 64. Setae on genital operculum of female, general aspect: (0) as one pair of spines; (1) fringed with hairs; (2) as one apical tuft; (3) as two apical tufts. CI = 1.00; RI = 1.00. 65. Paramere hooks, general aspects: (0) curl ventrally; (1) curl dorsally. CI = 1.00; RI = 1.00. This character was mentioned by Keffer (2004). 66. Parameres, general aspect: (0) tapering slowly at apex; (1) large basally, tapering abruptly near to apex. CI = 1.00; RI = 0 (Fig. 16A, B). 67. Apex of parameres, general aspect: (0) never bifurcated; (1) bifurcated. CI = 1.00; RI = 1.00. 68. Lateral arms of the basal plate, general aspect: (0) curling around the proximal end of the phallosoma and meeting in the midline to form a lamina ventralis; (1) continuous with lamina ventralis, 2-parted; (2) forming a distinct joint with the lamina ventralis. CI = 1.00; RI = 1.00. This character was mentioned by Keffer (2004). 69. Basal plate, general aspect of transverse bridge: (0) without membranous part among it; (1) not entire; (2) clearly jointed, entire. CI = 1.00; RI = 1.00 (Fig. 16C, D). 70. Gubernaculum, degree of development: (0) developed, with lamina ventralis extending posteriorly; (1) without evidence of either a central strut or a gubernaculum. CI = 1.00; RI = 1.00. This character was mentioned by Keffer (2004). 71. Phallosoma, degree of fusion with ventral diverticulum: (0) separate; (1) fused. CI = 1.00; RI = 1.00. 72. Phallosoma, aspect of dorsal portion: (0) straight or irregular, without a sulcus or a pair of dorsal arms; (1) dorsally slightly concave insinuating vestigial arms; (2) strongly bifurcate, with arms developed extending along ventral diverticulum. CI = 0.50; RI = 0.60 (e.g. Fig. 16C, E). 73. Dorsal arms, degree of development, dorsal view: (0) weakly developed; (1) developed, extending nearly to apex of ventral diverticulum; (2) well-developed, meeting at tip of ventral diverticulum and enclosing it. CI = 0.67; RI = 0.50 (Fig. 16E). 74. Dorsal arms, aspect of their margins, lateral view: (0) without projections; (1) each with low angular medial projection. CI = 1.00; RI = 1.00. 75. Dorsal arms, location related to ventral diverticulum: (0) on ventral diverticulum; (1) somewhat laterally directed; (2) completely laterally located. CI = 1.00; RI = 1.00 (Fig. 16E). 76. Phallotheca without a pair of elongate dorsal arms, degree of development: (0) not extending onto entire phallosoma [= vesica sensuKeffer (2004)]; (1) almost extending to apex of phallosoma. CI = 1.00; RI = 0. 77. Apex of phallosoma fused to ventral diverticulum, position of orifice: (0) postero-ventrally located; (1) dorsally located. CI = 0.50; RI = 0. 78. Orifice, degree of development: (0) small; (1) strongly developed. CI = 0.50; RI = 0. 79. Ventral diverticulum, general aspect: (0) distally curling dorsally; (1) distally curling ventrally. CI = 1.00; RI = 1.00. Keffer (2004) mentioned it as posterior diverticulum. 80. Ventral diverticulum, degree of sclerotization: (0) weakly sclerotized, flexible; (1) strongly sclerotized, rigid. CI = 1.00; RI = 1.00. 81. Ventral diverticulum fused with phallosoma, general aspect along its median portion: (0) conspicuous, largely expanded; (1) inconspicuous, not widened; (2) elongated with rounded apex; (3) quadrangular-shaped; (4) rounded; (5) conspicuously shaped; (6) laterally flattened. CI = 1.00; RI = 1.00 (Fig. 16C–E). 82. Ventral diverticulum strongly sclerotized, aspect in ventral view: (0) U-shaped, not dorsoventrally flattened and often bearing a pair of processes; (1) distally never U-shaped. CI = 1.00; RI = 1.00. 83. Ventral diverticulum fused with phallosoma, general aspect of its apex: (0) slightly concave, dorsoventrally flattened; (1) straight; (2) slightly convex; (3) strongly convex; (4) acute at apex. CI = 1.00; RI = 1.00. 84. Ventral diverticulum, degree of development of flanges, in ventral view: (0) absent; (1) present. CI = 0.50; RI = 0.75. 85. Ventral diverticulum, aspect of apex in ventral view: (0) bilobed; (1) contiguous, never bilobed. CI = 1.00, RI = 1.00 (Fig. 10A, B). 86. Apex of ventral diverticulum fused to aedeagus, general aspect: (0) without ventroapical protuberance; (1) with ventroapical protuberance. CI = 0.33; RI = 0.50 (Fig. 17). 87. Spines on apicoventral diverticulum, degree of development: (0) poorly developed; (1) coarsely developed; (2) strongly developed throughout protuberance. CI = 1.00; RI = 0 (Fig. 18). 88. Genitalic protuberances, general aspect: (0) tuberculous-shaped; (1) spiny-shaped. CI = 1.00; RI = 1.00 (Fig. 18). 89. Surface of apex of ventral diverticulum, general aspect: (0) with spines or tubercles; (1) without spines or tubercles. CI = 0.25; RI = 0.63. 90. Phallothecal plate (lateral view), degree of development related to ventral diverticulum: (0) poorly developed; (1) developed. CI = 1.00; RI = 1.00. 91. Phallothecal plate developed (lateral view), degree of development related to ventral diverticulum: (0) somewhat developed; (1) as long as ventral diverticulum. CI = 1.00; RI = 1.00. 92. Phallothecal plate somewhat developed (lateral view), general aspect: (0) directed downward; (1) fused or close to ventral diverticulum. CI = 1.00; RI = 1.00. 93. Pygophore, general aspect: (0) tapering evenly between basal portion and apical semitubular portion; (1) with abrupt sculptured shoulder between basal portion and apical semitubular portion. CI = 0.50; RI = 0.50. 94. Apex of second valvulae, general aspect: (0) without spine and/or protuberance; (1) with spine and/or protuberance. CI = 1.00; RI = 1.00. 95. Apex of second valvulae with spine and/or protuberance, degree of development: (0) with a protuberance; (1) with a protuberance and a conspicuous spine; (2) with just a spine. CI = 0.50; RI = 0.33. 96. Spermatheca, aspect of its basal part: (0) with distinct apodemes for muscle insertion distributed throughout; (1) with no distinct apodemes. CI = 0.33; RI = 0.75 (Fig. 19). 97. Basal part of spermatheca, general aspect: (0) without either ampulla or an area more inflated; (1) with either ampulla or an area more inflated. CI = 0.50; RI = 0.86 (Figs 20, 21). 98. Basal part of spermatheca, degree of development of the area more inflated: (0) with an area conspicuously more inflated; (1) with ampulla. CI = 1.00; RI = 1.00 (Figs 20, 21). Here we considered here this part to be associated to some extent with the opening of the spermatheca into the vagina. 99. Median vaginal area (below spermatheca), general aspect: (0) without pouch; (1) with a conspicuous pouch. CI = 1.00; RI = 1.00 (Fig. 21). 100. Respiratory siphon, degree of development: (0) long; (1) short. CI = 1.00; RI = 1.00. 101. Respiratory short siphon, general aspect: (0) thin; (1) broad. CI = 1.00; RI = 1.00. 102. Short air straps, aspect: (0) lanceolate, evenly covered with long hairs dorsally; (1) spatulate, bearing special transverse bands of pubescence dorsally. CI = 0.50; RI = 0.80. 103. Short air lanceolate straps, general aspect: (0) with somewhat uniform width along its extension; (1) large at apex. CI = 1.00; RI = 0. 104. Respiratory siphon, aspect: (0) non-retractile; (1) retractile. CI = 1.00; RI = 1.00. © 2017 The Linnean Society of London, Zoological Journal of the Linnean Society http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Zoological Journal of the Linnean Society Oxford University Press

Phylogenetic analysis and revision of subfamily classification of Belostomatidae genera (Insecta: Heteroptera: Nepomorpha)

Loading next page...
 
/lp/ou_press/phylogenetic-analysis-and-revision-of-subfamily-classification-of-QFYkUzCt4x
Publisher
The Linnean Society of London
Copyright
© 2017 The Linnean Society of London, Zoological Journal of the Linnean Society
ISSN
0024-4082
eISSN
1096-3642
D.O.I.
10.1093/zoolinnean/zlx041
Publisher site
See Article on Publisher Site

Abstract

Abstract Recent investigations into relationships among the cosmopolitan Belostomatidae have led to recognition of the Belostomatinae, Horvathiniinae and Lethocerinae clades. Here we investigate the relationships among the genera of this family (with Appasus, Benacus and Kirkaldyia now resurrected, and three fossils: Cratonepa, Iberonepa and Lethocerus vetus) by including modifications or clarifications of both somatic and genitalic characters (including some spermatheca traits), as well as multiple genes (COI, 18S rDNA and 16S rDNA). A comparative study of these genera yielded putative homology hypotheses coded as 104 morphological characters and 1829 aligned characters. A majority-rule tree utilizing the Bayesian method [the ‘master’ tree with highlighted clades estimated by maximum parsimony and maximum likelihood (ML)] is as follows: (Cratonepa, Iberonepa, (Benacus, Kirkaldyia, Lethocerus), (((Horvathinia, Hydrocyrius), Limnogeton), (((Abedus, Weberiella), Belostoma), (Appasus, Diplonychus))))). Lethocerinae and Belostomatinae (now with Horvathinia included) form two monophyletic groups. Weberiella appears as the sister group of Abedus, and Belostomatini had to be reformulated. In all trees, the monophyly of Diplonychinitrib. nov. (Appasus+Diplonychus) was recovered with high support indices. This study highlights a global tree based on combining the molecular and morphological data representing topologies from separate analyses, irrespective of the existence of missing data and the type of dataset. INTRODUCTION Water bugs or electric-light bugs are common and well-known insects from aquatic habitats throughout the world’s subtropical and tropical areas (Lauck & Menke, 1961; Merritt & Cummins, 1996). The family Belostomatidae Leach, 1815 comprises 11 genera and approximately 150 species, most of which are reported from the New World (Polhemus, 1995; Schuh & Slater, 1995; Perez–Goodwyn, 2006; Estévez & Ribeiro, 2011; Moreira et al., 2011) and are found in limnic habitats such as rivers, streams, lakes and ponds. They are ambush predators, often hunting by night, and representatives of some genera (e.g. LethocerusMayr, 1853) are known to capture a wide variety of prey, including woodpeckers (Hungerford, 1919) and turtles (Ohba, 2011). Spooner (1938) intuitively proposed that Belostomatidae and Nepidae Latreille, 1802 are closely related, by indicating morphological similarity. Also, the conspicuous phallus of representatives of these two families, with the apex fully sclerotized, is a condition uniquely found among Nepoidea, so that this was highlighted as evidence of a relationship between them by Dupuis (1955). Using cladistic methods, Mahner (1993), Hebsgaard, Andersen & Damgaard (2004) and Hua et al. (2009) have suggested Belostomatidae and Nepidae as sister groups. This monophyletic group was defined by the non-retractable condition of the air straps and their origin from the eighth abdominal segment. Lauck & Menke (1961) intuitively subdivided Belostomatidae into three subfamilies based on aspects of male genitalia morphology and in a more critical analysis of other characteristics, such as the condition of the abdominal sternites divided or not by a suture and the degree of development of tarsal claws of the first pair of legs. Mahner (1993) undertook the most comprehensive phylogenetic analysis of the generic and suprageneric relationships within Belostomatidae based on morphology. Mahner also showed that the family is monophyletic, being supported by the presence of a metepistern quite developed in nymphs and a hairy conspicuous stripe visible on the connexivum of adults. Finally, based on combined analysis (with 16S+28S rDNA) by Hebsgaard et al. (2004), Belostomatidae is supported by the following two unambiguously optimized morphological changes: (1) conical metacoxae, firmly united with metapleuron; and (2) hind tibiae flattened, with swimming hairs. Lauck & Menke’s (1961) subfamily classification of Belostomatidae has been adopted by many authors (De Carlo, 1966; Nieser, 1975; Mahner, 1993; Schuh & Slater, 1995; Schnack & Estévez, 2005; Perez-Goodwyn, 2006; Estévez & Ribeiro, 2011; Moreira et al., 2011), and only one subfamily has been subjected to revision (Lethocerinae: Perez-Goodwyn, 2006). Belostomatinae is a widespread subfamily with the highest number of described genera among the family Belostomatidae. AbedusStål, 1862 is restricted to the Nearctic Region, Mexico, where there is the largest number of representatives, and to Central America. BelostomaLatreille, 1807, the most diverse genus, occurs in the New World, with the largest number of representatives in South America (Lauck & Menke, 1961; Moreira et al., 2011). AppasusAmyot & Serville, 1843 and Diplonychus Laporte, 1833 are distributed throughout Southeast Asia, Africa, India and Australia. HydrocyriusSpinola, 1850 is restricted to the African continent, and little is known about its biology. LimnogetonMayr, 1853, restricted to north-eastern Africa, comprises apparently strict predators of snails (Voelker apudSchuh & Slater, 1995). Its members bear smooth profemora, lacking sulci, and are thought to be the only representatives of Belostomatidae which do not have middle and posterior legs modified for swimming (Ribeiro et al. 2014). WeberiellaDe Carlo, 1966 is monobasic, constituted by Weberiella rhomboides (Menke, 1965), and it is recorded from the states of Amazonas, Mato Grosso and Rondônia in Brazil, as well as French Guyana and Venezuela (Menke, 1965; De Carlo, 1966; Estévez & Ribeiro, 2011). The record of this species from Roraima, Brazil, is incorrect (F. F. F. Moreira, pers. comm.), and its members exclusively inhabit the surface film of freshwaters (kinon by Fittkau, 1977). Horvathiniinae is a monotypic subfamily (with HorvathiniaMontandon, 1911), and its members are restricted to the southeast and central south of South America (Lauck & Menke, 1961). The genus comprises only two species, Horvathinia lentiDe Carlo, 1957 and H. pelocoroidesMontandon, 1911 (Schnack & Estévez, 2005). In addition, only a few representatives of these species have been collected or observed in their natural habitat, and almost nothing has been published about their biology (see Armúa-de-Reyes, Schnack & Estévez, 2005). Lethocerinae sensu stricto, now reformulated as a subfamily comprising the genera BenacusStål, 1861, KirkaldyiaMontandon, 1909 and LethocerusMayr, 1853 (according to Perez-Goodwyn, 2006), is a cosmopolitan subfamily, being better represented in the Neotropical Region. In this subfamily, Lethocerus maximusDe Carlo, 1938 is known to be the largest of all Heteroptera, and can attain at least 110.0 mm (Monte, 1945). Having undertaken his phylogenetic analysis of suprageneric relationships within Belostomatidae based on morphology, Mahner (1993) documented the generic recognition of several putatively monophyletic groups. In addition, Mahner’s (1993) phylogeny supports the monophyly of the subfamilies created by Lauck & Menke (1961), and many features of the external morphology and male genitalia erected by these authors were important synapomorphies in his analysis. The author also indicated that Horvathiniinae and Belostomatinae are sister groups, sharing the fusion of phallosoma and ventral diverticulum. Mahner created Belostomatini within Belostomatinae, a tribe consisting of Abedus and Belostoma, based on the presence of 2-segmented tarsi bearing only one visible claw. However, Mahner did not include Appasus, Benacus and Kirkaldyia in his analysis, invalid or unavailable genera at that time, and considered Belostoma and Weberiella as genera without putative synapomorphies. As Mahner’s phylogeny neither included such resurrected genera nor used molecular data as additional information, the relationship among Belostomatidae genera has been rendered incomplete and outdated. Moreover, Mahner did not present a matrix of characters nor consequently the distribution of their states he proposed, making it very difficult to reproduce his analysis. According to Andersen (1995), Mahner gave up the use of computer algorithms, because the characters he proposed were ‘reliable’ enough that these computational efforts would be superfluous. Here, for the first time, we present phylogenetic relationships among the genera of Belostomatidae based on characters of external morphology, spermatheca, and male and female genitalia, utilizing analysis of portions of the 18S and 16S rDNA genes and the mitochondrial gene COI. The goal was to conduct groups of different analyses, either with each gene separately (‘separated analysis’ of De Queiroz, Donoghue & Kim, 1995), with only morphological data, with the genes combined in the same matrix or with all available data (‘simultaneous analysis’ of De Queiroz et al., 1995). As the taxonomy of Belostomatidae is still in an incipient state regarding the study of females, we used some characters based on the morphology of the spermatheca. Since we found evidence to support Appasus and Diplonychus as a monophyletic group, the new tribe Diplonychini was established to accommodate that clade. As various data sources were used to obtain different phylogenetic relationships, we tested the hypothesis that the quality and the kind of information did not change the results. MATERIAL AND METHODS Taxon sampling and terminology Species exemplars of all 13 Belostomatidae genera were selected to estimate the genus-level phylogeny of belostomatids, including three fossil taxa and four monobasic genera. Type-species were selected when available, but otherwise preference was given to species recently collected in ethanol for the molecular analyses. Additional species were included for larger genera, especially those belonging to Belostoma, which are harder to characterize morphologically and some of which are apparently para- or polyphyletic. Outgroup taxa included were selected to test the monophyly of Belostomatidae. Eleven outgroup taxa representing four Nepidae genera were used: NepaLinnaeus, 1758, LaccotrephesStål, 1866, CurictaStål, 1861 and RanatraFabricius, 1790. These outgroups have been given as closely related to Belostomatidae in previous phylogenetic analyses (Mahner, 1993; Hua et al., 2009; Li et al., 2012; Brożek, 2014). Acronyms of the institutions that loaned or donated the specimens are listed in Table 1 together with a list of material studied. Table 1. Species studied for the phylogenetic analysis of Belostomatidae. The origin (country and region) and depository are provided for each species. Underlined localities refer to voucher specimens. An asterisk indicates no available information for that particular taxon in at least one data partition Taxa  Depositories†  Localities  BELOSTOMATIDAE  Abedus dilatatus (Say)  MNHN  MEXICO  Abedus indentatus (Haldeman)  LEBIP  MEXICO: Ciudad de México (Mpio. Temascaltepec)  Abedus signoreti Mayr*  LEBIP  MEXICO: Ciudad de México (Mpio. Luvianos), Baja California  Abedus ovatus (Stål)  LEBIP  MEXICO: Michoacán  Appasus ampliatus (Montandon)  MNHN  IVORY COAST: Foro-Foro; CENTRAL AFRICAN REPUBLIC: Lamaboké; SENEGAL: Kedougou  Appasus capensis (Mayr)  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga  Appasus grassei (Poisson)  MNHN  IVORY COAST: Lamto; SENEGAL: Dakar  Appasus japonicus Vuillefroy  LEBIP  CHINA: Gan Chouen Fou, Xiao Bei Lake; JAPAN (Sayo, Hyogo)  Appasus major (Esaki)*  MNHN/LEBIP  CHINA: Mandchourie; JAPAN: Sayo  Appasus nepoides (Fabricius)  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga; SENEGAL: Sangalkam; IVORY COAST: Adiopodoumé  Appasus procerus  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga; CAMEROON: Batouri; CENTRAL AFRICAN REPUBLIC: Lamaboké  Belostoma angustum Lauck*  LEBIP  BRAZIL: Rio Grande do Sul; URUGUAY: La Paloma  Belostoma cummingsi De Carlo  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma dilatatum (Dufour)  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma elongatum Montandon*  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma flumineum Say*  LEBIP  USA: Kansas  Belostoma harrisi Lauck  LEBIP  SURINAM: Paramaribo  Belostoma micantulum (Stål)  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma nicaeum Estévez & Polhemus  DZRJ  BRAZIL: Amazonas  Belostoma oxyurum (Dufour)  MNRJ  BRAZIL: Paraná  Belostoma plebejum (Stål)*  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma ribeiroi De Carlo*  LEBIP  BRAZIL: Mato Grosso do Sul  Belostoma testaceopallidum Latreille  DZRJ  BRAZIL: Rio de Janeiro, Santa Catarina  Benacus griseus (Say)*  MNHN  USA: New York; GUYANA  Diplonychus esakii Miyamoto & Lee*  LEBIP  REPUBLIC OF THE CONGO: Zanga-Di-Makodia; VIETNAM: Tainin  Diplonychus rusticus (Fabricius)*  LEBIP  REPUBLIC OF THE CONGO: Nduizi River; GenBank  Horvathinia pelocoroides Montandon  LEBIP/MLPA/MNRJ/ Nieser collection  BRAZIL: Minas Gerais, Mato Grosso, Rio Grande do Sul, Santa Catarina; PARAGUAY: Guaíra  Hydrocyrius colombiae Spinola  MNHN/MRAC  EGYPT; REPUBLIC OF THE CONGO: Myamirundi, Bas Ogoqué; CAMEROON: Yaoundé; SENEGAL: Saint Luis; MADAGASCAR: Farafangan; IVORY COAST: Adiopodoumé; SUDAN: Dogo  Hydrocyrius nanus Montandon  MNHN/MRAC  REPUBLIC OF THE CONGO; CAMEROON: Yaoundé; CENTRAL AFRICAN REPUBLIC: Lamaboké  Hydrocyrius punctatus Stål  MNHN  MADAGASCAR. Tananarive  Hydrocyrius rectus Mayr  MNHN/MRAC  CONGO: Brazzaville, Leopoldville  Hydrocyrius sp.  MNHN/LEBIP  MADAGASCAR: Tananarive; REPUBLIC OF THE CONGO: Zanga-Di-Makodia  Kirkaldyia deyrolli (Vuillefroy)  MNRJ/LEBIP  CHINA: Chekiang Province; JAPAN: Kumamoto, Tokyo, Alpes de Nikko; LAOS: Baudan  Lethocerus annulipes (Herrich-Schäffer)*  LEBIP  BRAZIL: Rio de Janeiro; Rio Grande do Sul  Lethocerus cordofanus Mayr  MNHN  SOMALIA: Glohar  Lethocerus delpontei De Carlo  LEBIP  BRAZIL: Rio Grande do Sul  Lethocerus maximus De Carlo  LEBIP  FRENCH GUYANA: Route de Pointe Combi  Limnogeton expansum Montandon  MACN/MNHN  REPUBLIC OF THE CONGO: Brazzaville; CAMEROON: Batouri, Nyaounderé; TANGANYIKA: Mlingano  Limnogeton fieberi Mayr  MNHN/MRAC  EGYPT; SUDAN; CAMEROON: Garoua  Limnogeton hedenborgi (Stål)  MNHN/MRAC  SENEGAL: Kolda; EGYPT; REPUBLIC OF THE CONGO; CAMEROON: Yaoundé; UGANDA: Victoria Nyanza  Limnogeton scutellatum Mayr  MNHN/MRAC  EGYPT; ETHIOPIA; CENTRAL AFRICAN REPUBLIC: Lamaboké; DEMOCRATIC REPUBLIC OF THE CONGO: Molindi River; REPUBLIC OF THE CONGO: Brazzaville; ZAMBIA: Muliba  Weberiella rhomboides (Menke)*  INPA/DZRJ/LEBIP /MNHN  BRAZIL: Amazonas, Mato Grosso; FRENCH GUIANA: Piste Coralie  NEPIDAE  Curicta borellii Montandon  LEBIP  BRAZIL: Rio Grande do Sul  Curicta cf. pelleranoi  LEBIP  BRAZIL: Mato Grosso do Sul  Curicta volxemi (Montandon)*  LEBIP  BRAZIL: Mato Grosso do Sul  Laccotrephes japonensis (Scott)  LEBIP  JAPAN: Kumamoto  Laccotrephes pfeiferiae (Ferrari)  LEBIP  CHINA: Tianjin (Wuqing Country)  Laccotrephes sp.*  LEBIP  CHINA: Tianjin (Wuqing Country)  Nepa cinerea Linnaeus  LEBIP  FRANCE: Vayral  Nepa hoffmanni Esaki*  LEBIP  JAPAN: Hyogo  Ranatra brevicauda Montandon  LEBIP  BRAZIL: Mato Grosso do Sul  Ranatra chinensis Mayr*  LEBIP  JAPAN: Kumamoto  Ranatra heydeni Montandon*  LEBIP  BRAZIL: Mato Grosso do Sul  Ranatra robusta Montandon  LEBIP  BRAZIL: Mato Grosso do Sul; Rio Grande do Sul  Ranatra sattleri De Carlo  LEBIP  BRAZIL: Mato Grosso do Sul  Taxa  Depositories†  Localities  BELOSTOMATIDAE  Abedus dilatatus (Say)  MNHN  MEXICO  Abedus indentatus (Haldeman)  LEBIP  MEXICO: Ciudad de México (Mpio. Temascaltepec)  Abedus signoreti Mayr*  LEBIP  MEXICO: Ciudad de México (Mpio. Luvianos), Baja California  Abedus ovatus (Stål)  LEBIP  MEXICO: Michoacán  Appasus ampliatus (Montandon)  MNHN  IVORY COAST: Foro-Foro; CENTRAL AFRICAN REPUBLIC: Lamaboké; SENEGAL: Kedougou  Appasus capensis (Mayr)  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga  Appasus grassei (Poisson)  MNHN  IVORY COAST: Lamto; SENEGAL: Dakar  Appasus japonicus Vuillefroy  LEBIP  CHINA: Gan Chouen Fou, Xiao Bei Lake; JAPAN (Sayo, Hyogo)  Appasus major (Esaki)*  MNHN/LEBIP  CHINA: Mandchourie; JAPAN: Sayo  Appasus nepoides (Fabricius)  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga; SENEGAL: Sangalkam; IVORY COAST: Adiopodoumé  Appasus procerus  MNHN  DEMOCRATIC REPUBLIC OF THE CONGO: Katanga; CAMEROON: Batouri; CENTRAL AFRICAN REPUBLIC: Lamaboké  Belostoma angustum Lauck*  LEBIP  BRAZIL: Rio Grande do Sul; URUGUAY: La Paloma  Belostoma cummingsi De Carlo  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma dilatatum (Dufour)  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma elongatum Montandon*  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma flumineum Say*  LEBIP  USA: Kansas  Belostoma harrisi Lauck  LEBIP  SURINAM: Paramaribo  Belostoma micantulum (Stål)  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma nicaeum Estévez & Polhemus  DZRJ  BRAZIL: Amazonas  Belostoma oxyurum (Dufour)  MNRJ  BRAZIL: Paraná  Belostoma plebejum (Stål)*  LEBIP  BRAZIL: Rio Grande do Sul  Belostoma ribeiroi De Carlo*  LEBIP  BRAZIL: Mato Grosso do Sul  Belostoma testaceopallidum Latreille  DZRJ  BRAZIL: Rio de Janeiro, Santa Catarina  Benacus griseus (Say)*  MNHN  USA: New York; GUYANA  Diplonychus esakii Miyamoto & Lee*  LEBIP  REPUBLIC OF THE CONGO: Zanga-Di-Makodia; VIETNAM: Tainin  Diplonychus rusticus (Fabricius)*  LEBIP  REPUBLIC OF THE CONGO: Nduizi River; GenBank  Horvathinia pelocoroides Montandon  LEBIP/MLPA/MNRJ/ Nieser collection  BRAZIL: Minas Gerais, Mato Grosso, Rio Grande do Sul, Santa Catarina; PARAGUAY: Guaíra  Hydrocyrius colombiae Spinola  MNHN/MRAC  EGYPT; REPUBLIC OF THE CONGO: Myamirundi, Bas Ogoqué; CAMEROON: Yaoundé; SENEGAL: Saint Luis; MADAGASCAR: Farafangan; IVORY COAST: Adiopodoumé; SUDAN: Dogo  Hydrocyrius nanus Montandon  MNHN/MRAC  REPUBLIC OF THE CONGO; CAMEROON: Yaoundé; CENTRAL AFRICAN REPUBLIC: Lamaboké  Hydrocyrius punctatus Stål  MNHN  MADAGASCAR. Tananarive  Hydrocyrius rectus Mayr  MNHN/MRAC  CONGO: Brazzaville, Leopoldville  Hydrocyrius sp.  MNHN/LEBIP  MADAGASCAR: Tananarive; REPUBLIC OF THE CONGO: Zanga-Di-Makodia  Kirkaldyia deyrolli (Vuillefroy)  MNRJ/LEBIP  CHINA: Chekiang Province; JAPAN: Kumamoto, Tokyo, Alpes de Nikko; LAOS: Baudan  Lethocerus annulipes (Herrich-Schäffer)*  LEBIP  BRAZIL: Rio de Janeiro; Rio Grande do Sul  Lethocerus cordofanus Mayr  MNHN  SOMALIA: Glohar  Lethocerus delpontei De Carlo  LEBIP  BRAZIL: Rio Grande do Sul  Lethocerus maximus De Carlo  LEBIP  FRENCH GUYANA: Route de Pointe Combi  Limnogeton expansum Montandon  MACN/MNHN  REPUBLIC OF THE CONGO: Brazzaville; CAMEROON: Batouri, Nyaounderé; TANGANYIKA: Mlingano  Limnogeton fieberi Mayr  MNHN/MRAC  EGYPT; SUDAN; CAMEROON: Garoua  Limnogeton hedenborgi (Stål)  MNHN/MRAC  SENEGAL: Kolda; EGYPT; REPUBLIC OF THE CONGO; CAMEROON: Yaoundé; UGANDA: Victoria Nyanza  Limnogeton scutellatum Mayr  MNHN/MRAC  EGYPT; ETHIOPIA; CENTRAL AFRICAN REPUBLIC: Lamaboké; DEMOCRATIC REPUBLIC OF THE CONGO: Molindi River; REPUBLIC OF THE CONGO: Brazzaville; ZAMBIA: Muliba  Weberiella rhomboides (Menke)*  INPA/DZRJ/LEBIP /MNHN  BRAZIL: Amazonas, Mato Grosso; FRENCH GUIANA: Piste Coralie  NEPIDAE  Curicta borellii Montandon  LEBIP  BRAZIL: Rio Grande do Sul  Curicta cf. pelleranoi  LEBIP  BRAZIL: Mato Grosso do Sul  Curicta volxemi (Montandon)*  LEBIP  BRAZIL: Mato Grosso do Sul  Laccotrephes japonensis (Scott)  LEBIP  JAPAN: Kumamoto  Laccotrephes pfeiferiae (Ferrari)  LEBIP  CHINA: Tianjin (Wuqing Country)  Laccotrephes sp.*  LEBIP  CHINA: Tianjin (Wuqing Country)  Nepa cinerea Linnaeus  LEBIP  FRANCE: Vayral  Nepa hoffmanni Esaki*  LEBIP  JAPAN: Hyogo  Ranatra brevicauda Montandon  LEBIP  BRAZIL: Mato Grosso do Sul  Ranatra chinensis Mayr*  LEBIP  JAPAN: Kumamoto  Ranatra heydeni Montandon*  LEBIP  BRAZIL: Mato Grosso do Sul  Ranatra robusta Montandon  LEBIP  BRAZIL: Mato Grosso do Sul; Rio Grande do Sul  Ranatra sattleri De Carlo  LEBIP  BRAZIL: Mato Grosso do Sul  †List of specimen depositories is based on Arnett, Samuelson & Nishida (1993), except for DZRJ, LEBIP and MNRJ. DZRJ, Departamento de Zoologia, Instituto de Biologia, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil; LEBIP, Laboratório de Estudos da Biodiversidade do Pampa, São Gabriel, Universidade Federal do Pampa, Rio Grande do Sul, Brazil; MNRJ, Museu Nacional, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil. View Large Identification was based on the following authors: Menke (1960) for Abedus; Lauck (1962, 1963, 1964), Nieser (1975), Estévez & Polhemus (2001) and Ribeiro (2007) for Belostoma; Perez-Goodwyn (2006) for Benacus, Kirkaldyia and Lethocerus; Keffer (1996) for Curicta; Poisson (1949, 1954), Lee (1991) and Ribeiro et al. (2014) for Appasus, Hydrocyrius and Limnogeton; De Carlo (1964), Lansbury (1972), Nieser (1975) and Chen, Nieser & Ho (2004) for Ranatra; Lansbury (1972) and Keffer (2004) for Nepa; Polhemus & Polhemus (2013), Keffer (2004) and Nieser, Zettel & Chen (2009) for Lacotrephes. The terminology of wing parts follows Gorb & Perez-Goodwyn (2003), and of genitalia Dupuis (1955), Dupuis & Carvalho (1956), Scudder (1959), Lansbury (1972), Lalitha, Shyamasundari & Rao (1997), Ribeiro (2007) and Keffer (2004). Techniques for preparation of male and female genitalia were adapted from those of Ribeiro (2007). Dissected parts of the male and female genitalia are stored in microvials with glycerin. The spermatheca was studied by dissection of fresh or dried material according to the protocol proposed by Pluot-Sigwalt (1986, 2008). The spermatheca was examined after the abdomen had been removed from the specimen and placed into cold 10% KOH solution for 12–24 h, depending on the size of the insect. This treatment removes tissues and does not affect the cuticular intima of the ectodermal part of the internal genitalia. Following washing of the abdomen in distilled water and rinsing in 2% liquid detergent DECON90, the genital tract was transferred to glycerol. The organ was then dissected out and stained in noir chlorazol according to the method of Carayon (1969). Spermatheca samples were stored in vials with glycerol. The terminology used for spermatheca parts follows Pellerano & De Carlo (1985), Pluot-Sigwalt (2008) and Viscarret & Bachmann (1997). Many of the differences and peculiarities found therein are known to occur at genus level, which makes us believe that the morphology of the spermatheca is probably useful for understanding the evolution and systematics of these organisms (Pellerano & De Carlo, 1985; Viscarret & Bachmann, 1997). To give an accurate impression of structures such as antennae, brush-to-brush frictional surfaces of the hemelytra, the clavus–clavus clamp of hemelytra and spines on the ventral diverticulum of male genitalia, we obtained scanning electron micrographs of dorsal, lateral and ventral views of the aforementioned structures using a Hitachi scanning electron microscope. Preparations were critical-point-dried, mounted on holders and sputter-coated with gold–palladium (10 nm) when sufficient material was available. If not, such material was examined at 5 kV, without a coat of gold–palladium. DNA preparation Insect specimens collected in the field were placed directly in 95–100% ethanol and stored at −20 °C until processing. To amplify genes, genomic DNA was extracted from a single foreleg and associated muscles using a modified ethanol precipitation/resuspension protocol (Bender, Spierer & Hogness 1983) or the DNEasy tissue kit (Qiagen Inc.). In some cases, multiple individuals from the same species were extracted. All belostomatid vouchers were stored in 99.5% ethanol pro-analysis (PA) at −20 °C deposited at the LEBIP (Laboratório de Estudos da Biodiversidade do Pampa, Universidade Federal do Pampa, Rio Grande do Sul, Brazil), except for W. rhomboides, which was dried and is deposited at the MNHN (Museum National d’Histoire Naturelle, Paris, France). For amplification of genes from W. rhomboides, genomic DNA was extracted following protocols adapted from those of Gilbert et al. (2007), Thomsen et al. (2009) and Lis, Ziaja & Lis (2011), which allowed us to recover amplifiable DNA from dried museum specimens that were up to 10 years old. The specimens were pinned again and redeposited in the collection as vouchers. PCR, sequencing and alignment Modified primers based on Simon et al. (1994), von Dohlen & Moran (1995) and Simon et al. (2006) were used to amplify parts of the mitochondrial gene COI (~730 bp) and 16S rDNA (~480 bp). Nuclear 18S sequences were amplified using the primers Ns1 and Ns2a (Barker et al., 2002) (~600 bp) (see Table 2). Amplification was carried out in a 25-µL volume reaction, with 5 µL of Taq and Load Mastermix, 0.5 µL of each primer at 25 µM and 2 µL of extraction product. For 16S loci, all reactions were performed using HifiTaq DNA Polymerase – an enzyme mixture that greatly increases fidelity and amplification of genomic targets. For COI reactions, a ‘step-up procedure’ was used: 2 min at 95 °C, five amplification cycles to improve DNA stock of 1 min at 95 °C, 15 min at 45 °C and 1 min at 72 °C; followed by 35 cycles of 40 s at 95 °C, 1 min at 51 °C and 1 min at 72 °C, with a final extension at 72 °C for 5 min. The protocol for the 16S rDNA region was: 94 °C for 3 min and 80 °C for 20 min (hot start), followed by 35 cycles of 94 °C for 1 min, 50 °C for 1 min and 72 °C for 2 min, with a final extension at 72 °C for 7 min. The protocol for 18S rDNA was 94 °C for 2 min, 35 cycles of 94 °C for 30 s, 49 °C for 30 s and 72 °C for 30 s, with a final extension at 72 °C for 7 min. Negative controls (no template) were always run simultaneously with our PCR experiments. All reaction mixtures were discarded when any DNA appeared in the negative control. Table 2. Oligonucleotide primer sequences. Primers used in polymerase chain and sequencing reactions of COI, 18S and 16S Primer  Direction  Sequence 5′→ 3′  Alias  Reference  COI+  F  GGTCAACAAATCATAAAGATATTGG  LCO-1490  Folmer et al. (1994)   COI+a  F  GGAGGATTTGGAAATTGATTAGTTCC  C1-J-1718  Simon et al. (1994)   COI-  R  TAAACTTCAGGGTGACCAAAAAATCA  HCO-2198  Folmer et al. (1994)   COI+*  F  TGTAAAACGACGGCCAGTTTTCAACWAATCATAAAGATATTGG  LCO1490puc_t1  Cruaud et al. (2010)   COI+a*  F  TGTAAAACGACGGCCAGTTTTCAACTAAYCATAARGATATYGG  LCO1490Hem1_t1  Germain et al. (2013)   COI-*  R  CAGGAAACAGCTATGACTAAACTTCWGGRTGWCCAAARAATCA  HCO2198puc-t1  Cruaud et al. (2010)   COI-a*  R  CAGGAAACAGCTATGACTAAACYTCAGGATGACCAAAAAAYCA  HCO2198Hem2-t1  Germain et al. (2013)   COI-b*  R  CAGGAAACAGCTATGACTAAACYTCDGGATGBCCAAARAATCA  HCO2198Hem1-t1  Germain et al. (2013)   18S+  F  GTAGTCATATGCTTGTCTC  Ns1  Barker et al. (2002)   18S-  R  CGCGGCTGCTGGCACCAGACTTGC  Ns2a  Barker et al. (2002)   16S+  F  CCGGTYTGAACTCARATCA  LR-J-12887  Takiya et al. (2006)   16S-  R  CRMCTGTTTAWCAAAAACAT  LR-N-13398  Takiya et al. (2006)   16S-b  R  TAAGTGTGCAAAGGTAGC  16S-Lepto-F  Malm & Johanson (2008)   Primer  Direction  Sequence 5′→ 3′  Alias  Reference  COI+  F  GGTCAACAAATCATAAAGATATTGG  LCO-1490  Folmer et al. (1994)   COI+a  F  GGAGGATTTGGAAATTGATTAGTTCC  C1-J-1718  Simon et al. (1994)   COI-  R  TAAACTTCAGGGTGACCAAAAAATCA  HCO-2198  Folmer et al. (1994)   COI+*  F  TGTAAAACGACGGCCAGTTTTCAACWAATCATAAAGATATTGG  LCO1490puc_t1  Cruaud et al. (2010)   COI+a*  F  TGTAAAACGACGGCCAGTTTTCAACTAAYCATAARGATATYGG  LCO1490Hem1_t1  Germain et al. (2013)   COI-*  R  CAGGAAACAGCTATGACTAAACTTCWGGRTGWCCAAARAATCA  HCO2198puc-t1  Cruaud et al. (2010)   COI-a*  R  CAGGAAACAGCTATGACTAAACYTCAGGATGACCAAAAAAYCA  HCO2198Hem2-t1  Germain et al. (2013)   COI-b*  R  CAGGAAACAGCTATGACTAAACYTCDGGATGBCCAAARAATCA  HCO2198Hem1-t1  Germain et al. (2013)   18S+  F  GTAGTCATATGCTTGTCTC  Ns1  Barker et al. (2002)   18S-  R  CGCGGCTGCTGGCACCAGACTTGC  Ns2a  Barker et al. (2002)   16S+  F  CCGGTYTGAACTCARATCA  LR-J-12887  Takiya et al. (2006)   16S-  R  CRMCTGTTTAWCAAAAACAT  LR-N-13398  Takiya et al. (2006)   16S-b  R  TAAGTGTGCAAAGGTAGC  16S-Lepto-F  Malm & Johanson (2008)   *In some species of Ranatra and Laccotrephes COI sequences were amplified with these published primers. View Large Sequences were sent to the Genoscope (Centre National de Séquençage, Evry). The sequences obtained were read and cleaned with CodonCode (CodonCode Corporation, 2002–2015). All sequences were aligned using MUSCLE (Edgard, 2004), and MAFFT v.7 (Katoh & Standley, 2013) for non-coding sequences. In all regions, gaps were checked manually and treated as missing data. This procedure yielded approximately 1830 bp per taxon, although sequences for some taxa were not complete (Table 1). All Belostomatidae and Nepidae sequences are original accessions, except Abedus brevicepsStål, 1862, Curicta scorpioStål, 1862 and Nepa apiculata Uhler, 1862, which were taken from GenBank. NCBI accession numbers and references are summarized in Table 3. Table 3. Material examined with accession numbers for belostomatid genes. A dash (–) indicates that a region was not sequenced for that species because no PCR product could be obtained and hyphen (-) indicates sequences from different PCR products Taxon  GenBank acession(s): COI, 18S, 16S  BELOSTOMATIDAE  Abedus indentatus (Haldeman)  KY320459, KY389074, KY389116  Abedus signoreti Mayr  KY320460 - KY320461, KY389073 - KY389075, KY389110 - KY389122  Appasus japonicus Vuillefroy  KY320473, –, KY389104  Appasus major (Esaki)  KY320474, KY389067, KY389098 - KY389137  Belostoma angustum Lauck  KY320477, KY389084, –  Belostoma cummingsi De Carlo  KY320478, –, KY389115  Belostoma elongatum Montandon  –, KY389076, –  Belostoma flumineum Say  KY320475, KY389085, KY389121 - KY389127 - KY389132  Belostoma plebejum (Stål)  KY320476 - KY320479, KY389086 - KY389087, KY389097  Belostoma ribeiroi De Carlo  KY320462, KY389079, KY389103 - KY389109  Diplonychus esakii Miyamoto & Lee  KY320464 - KY320465, KY389068 - KY389069, KY389113 - KY389119 - KY389125 - KY389131  Diplonychus rusticus (Fabricius)  KY320466, KY389088, KY389136  Horvathinia pelocoroides Montandon  KY320467, KY389080, KY389100  Hydrocyrius colombiae Spinola  KY320481 - KY320482, KY389090, KY389139  Hydrocyrius sp.  KY320483, KY389089, –  Benacus griseus (Say)  –, –, –  Kirkaldyia deyrolli (Vuillefroy)  KY320484, KY389091, KY389106 - KY389112 - KY389118  Lethocerus annulipes (Herrich-Schäffer)  KY320486 - KY320487, KY389093, KY389133 - KY389138  Lethocerus delpontei De Carlo  –, KY389070 - KY389094, KY389128  Limnogeton expansum Montandon  KY320488, KY389095, KY389124 - KY389130 - KY389135  Weberiella rhomboides (Menke)  –, –, KY389102  NEPIDAE  Curicta borellii Montandon  KY320480, –, KY389141  Curicta cf. pelleranoi  –, KY389078, KY389123  Curicta volxemi (Montandon)  KY320463, KY389077, KY389129 - KY389134  Laccotrephes japonensis (Scott)  KY320485, KY389071, KY389108  Laccotrephes pfeiferiae (Ferrari)  KY320469, KY389092, KY389101  Laccotrephes sp.  KY320468, –, –  Nepa cinerea Linnaeus  KY320489, KY389096, KY389140  Nepa hoffmanni Esaki  KY320470, KY389072, KY389107  Ranatra brevicauda Montandon  KY320471, KY389083, KY389117  Ranatra chinensis Mayr  KY320490, –, KY389114 - KY389120 - KY389126  Ranatra heydeni Montandon  KY320491, KY389081, KY389099 - KY389105  Ranatra robusta Montandon  KY320472, KY389082, KY389111  Taxon  GenBank acession(s): COI, 18S, 16S  BELOSTOMATIDAE  Abedus indentatus (Haldeman)  KY320459, KY389074, KY389116  Abedus signoreti Mayr  KY320460 - KY320461, KY389073 - KY389075, KY389110 - KY389122  Appasus japonicus Vuillefroy  KY320473, –, KY389104  Appasus major (Esaki)  KY320474, KY389067, KY389098 - KY389137  Belostoma angustum Lauck  KY320477, KY389084, –  Belostoma cummingsi De Carlo  KY320478, –, KY389115  Belostoma elongatum Montandon  –, KY389076, –  Belostoma flumineum Say  KY320475, KY389085, KY389121 - KY389127 - KY389132  Belostoma plebejum (Stål)  KY320476 - KY320479, KY389086 - KY389087, KY389097  Belostoma ribeiroi De Carlo  KY320462, KY389079, KY389103 - KY389109  Diplonychus esakii Miyamoto & Lee  KY320464 - KY320465, KY389068 - KY389069, KY389113 - KY389119 - KY389125 - KY389131  Diplonychus rusticus (Fabricius)  KY320466, KY389088, KY389136  Horvathinia pelocoroides Montandon  KY320467, KY389080, KY389100  Hydrocyrius colombiae Spinola  KY320481 - KY320482, KY389090, KY389139  Hydrocyrius sp.  KY320483, KY389089, –  Benacus griseus (Say)  –, –, –  Kirkaldyia deyrolli (Vuillefroy)  KY320484, KY389091, KY389106 - KY389112 - KY389118  Lethocerus annulipes (Herrich-Schäffer)  KY320486 - KY320487, KY389093, KY389133 - KY389138  Lethocerus delpontei De Carlo  –, KY389070 - KY389094, KY389128  Limnogeton expansum Montandon  KY320488, KY389095, KY389124 - KY389130 - KY389135  Weberiella rhomboides (Menke)  –, –, KY389102  NEPIDAE  Curicta borellii Montandon  KY320480, –, KY389141  Curicta cf. pelleranoi  –, KY389078, KY389123  Curicta volxemi (Montandon)  KY320463, KY389077, KY389129 - KY389134  Laccotrephes japonensis (Scott)  KY320485, KY389071, KY389108  Laccotrephes pfeiferiae (Ferrari)  KY320469, KY389092, KY389101  Laccotrephes sp.  KY320468, –, –  Nepa cinerea Linnaeus  KY320489, KY389096, KY389140  Nepa hoffmanni Esaki  KY320470, KY389072, KY389107  Ranatra brevicauda Montandon  KY320471, KY389083, KY389117  Ranatra chinensis Mayr  KY320490, –, KY389114 - KY389120 - KY389126  Ranatra heydeni Montandon  KY320491, KY389081, KY389099 - KY389105  Ranatra robusta Montandon  KY320472, KY389082, KY389111  View Large Phylogenetic analyses Morphological characters were identified based on their topographical identity before proposing hypotheses of primary homology (de Pinna, 1991) by defining the states in the data matrix. All characters were assigned equal weights and treated as unordered. Multistate characters were treated as unordered under Fitch parsimony. The program Mesquite v.3.04 (Maddison & Maddison, 2015) was used for matrix preparation. Analyses of the morphological dataset were conducted using the cladistic method, considered as the current paradigm for systematic research based on morphological characters (de Pinna, 1991). Most-parsimonious trees were derived under the maximum-parsimony criterion using exact searches (branch-and-bound algorithm) with the addition sequence=furthest. The parsimony inference (PI) analyses utilizing morphological data were carried out under the equal weighting scheme (as for all the other analyses here). Because non-applicable characters are treated as missing data, we instructed the software to collapse zero-length branches (Brazeau, 2011). Multistate taxa were counted as uncertainty. All algorithms were implemented in PAUP* 4.0a152 (Swofford, 2002). Character states were optimized a posteriori, using the ACCTRAN criteria (de Pinna, 1991). Character states were scored as dashes (-) if inapplicable and as question marks (?) if ambiguous or missing (Appendix 1). Some of the characters we analysed were based on those of Lee (1991), Mahner (1993) and Keffer (2004), as well as Gorb & Perez-Goodwyn’s (2003) ideas. Character coding was contingent (Forey & Kitching, 2000) and the character statements were formulated following Sereno (2007). Character optimizations were reconstructed in MacClade (Maddison & Maddison, 2005) over trees based on morphological and total evidence analyses. Only unambiguously optimized characters are discussed in the text. Character optimizations were made both over morphological consensus and over combined morphological and molecular analyses, such that if any character optimization over their trees was different, both are treated in the Discussion. Branch support was assessed by 1000 non-parametric bootstrap replicates (Felsenstein, 1985) using heuristic searches, where each search was conducted using random additional sequences with branch-swapping by tree-bisection-reconnection (TBR) and ten replicates and hold=10. Apart from the parsimony analyses, we analysed our matrix of morphological characters using Bayesian inference (BI) of phylogeny using a likelihood model suitable for morphological data. We did so because this approach permits branch lengths to be estimated and helps in the resolution of clades in the presence of high levels of homoplasy, given that the likelihood approach takes the aforementioned branch lengths into consideration. Thus, by combining both approaches we could test whether our parsimony analysis could be under the influence of long-branch attraction (Magalhães & Santos, 2012). For that, two analyses were performed, one under the Mkv model and one under the MkvΓ model, which allows rate variation amongst characters to follow a gamma distribution. These are the variations of the ML models of Lewis (2001) for discrete morphological data (datatype=standard), implemented in MrBayes 3.2 (Ronquist et al., 2012). In the first model, we assumed equal rates of character change. The second one accounted for rate variation across characters (lset rates=gamma), i.e. Mkv model, for discrete morphological data allowing rate heterogeneity among characters. Flat priors were adopted here. For each of these analyses, four independent runs with four chains each were run for five million generations, sampling trees every 500 generations. Ten per cent of the trees were discarded as burn-in and the remaining were used to calculate the posterior probabilities. The best Bayesian topology was selected by comparing the harmonic mean of the log-likelihood of each of the two aforementioned analyses, as a way of estimating the marginal likelihood. The analysis with the harmonic mean closest to 0 is preferred. Also, we calculated twice the difference between log marginal likelihoods (Bayes factor) of these two BI analyses (Müller & Reisz, 2006). As a rule of thumb, a Bayes factor of greater than ten usually indicates strong support (Kass & Raftery, 1995). With this in mind, parsimony-uninformative autapomorphic characters were included in the data matrix as proposed by Yeates (1992), which made it possible to estimate all branch lengths (Lewis, 2001). All PI analyses utilizing molecular data were also run using PAUP* 4.0a152. Phylogenetic analyses were performed on each gene separately and on a combined molecular dataset. Combination in part of the molecular datasets was supported by 1000 replicate incongruence length difference (ILD) tests (Farris et al. 1994) (each with ten random addition replicates, with uninformative and constant characters excluded) of the partitions COI vs. 18S (P = 0.68). COI vs. 16S (P = 0.001) and mitochondrial vs. nuclear (p = 0.001) were not supported. All gaps were treated as missing data. Models of molecular evolution for use in ML and BI analyses were estimated by an Akaike Information Criterion (AIC) test (Akaike, 1974) using MrModeltest 2.3 (Nylander, 2004). As measures of phylogenetic content, the g1 (skewness) statistics of 1000 random trees and the percentage of clades supported by > 50% parsimony bootstrap among all resolved clades in the strict consensus trees were calculated (Table 4). Also, likelihood mapping was conducted by examining 10000 quartets using an approximate likelihood function based on the selected model parameters for each molecule. Models of molecular evolution are given in Table 4 for each molecule and codon position. Table 4. Descriptive statistics on sequence data. Models of molecular evolution were selected by AIC for each molecule and codon position in protein encoding genes. Per cent pairwise divergence (%PD) between Belostomatidae taxa is given. Skewness (g1) was calculated based on 1000 random trees. Percentage of clades with parsimony bootstrap support > 50% is based on strict consensus trees (%BS) Gene  Model selected  Length (bp)  Number of taxa  Number of variable sites (%)  Number of informative sites (%)  %A  %C  %G  %T  PD  g1  %BS  COI  GTR+I+Γ  735  38  569 (77)  408 (56)  40  19  9  33  0–59  −0.74  45  COI pos1  HKY+Γ  245                      COI pos2  GTR+Γ  245                      COI pos3  GTR+I+Γ  245                      18S  SYM+I+Γ  606  32  272 (45)  197 (33)  25  25  25  25  0–12  −0.71  81  16S  HKY+I+Γ  488  45  248 (51)  220 (45)  49  12  6  33  0–27  −0.58  64  Combined    1829  56  1829 (100)  834 (46)  34  21  17  28  0–30  −0.60  78  Gene  Model selected  Length (bp)  Number of taxa  Number of variable sites (%)  Number of informative sites (%)  %A  %C  %G  %T  PD  g1  %BS  COI  GTR+I+Γ  735  38  569 (77)  408 (56)  40  19  9  33  0–59  −0.74  45  COI pos1  HKY+Γ  245                      COI pos2  GTR+Γ  245                      COI pos3  GTR+I+Γ  245                      18S  SYM+I+Γ  606  32  272 (45)  197 (33)  25  25  25  25  0–12  −0.71  81  16S  HKY+I+Γ  488  45  248 (51)  220 (45)  49  12  6  33  0–27  −0.58  64  Combined    1829  56  1829 (100)  834 (46)  34  21  17  28  0–30  −0.60  78  View Large Despite a lack of support for the homogeneity of the signal from mitochondrial and nuclear partitions, a combination of the molecular and morphological datasets was supported (1000 replicate ILD tests, P = 0.97). In fact, we also performed phylogenetic analyses on that combined unsupported molecular dataset, as well as on the combined molecular and morphological datasets. Parsimony-based tree searches were performed using TBR branch swapping with 1000 random addition replicates, keeping ten trees at each step (hold=10). With regard to the combined molecular and morphological dataset, we conducted 10000 random addition sequence replicates and TBR branch swapping, keeping 100 trees at each step. All analyses used a random starting tree. BI analyses were performed with MrBayes, including that with molecular and morphological data. Mitochondrial protein encoding genes (COI) were classified by codon position and a mixed model approach (five partitions for the molecular analysis and six for the combined molecular and morphological data) was conducted. Models chosen for use in Bayesian analyses of combined molecular dataset were: HKY+Γ for COI pos1, GTR+Γ for COI pos2, GTR+I+Γ for COI pos3, SYM+I+Γ for 18S and HKY+I+Γ for 16S. Assuming flat priors, we performed each analysis consisting of four independent runs of 5000000 generations, each with four chains, sampling topologies from every 500 generations. For the molecular and morphological data combined, four Markov chain Monte Carlo chains were run for 10000000 generations sampling topologies from every 100 generations. Convergence was checked with the average standard deviation of split frequencies, which approach 0.01 when the runs converge. The data before the convergence were analysed with the software Tracer 1.6 (Rambaut et al., 2014) and burned-in. Ten per cent of the trees were discarded as burn-in after examination by Tracer 1.6 and the remaining trees were used to calculate the posterior probabilities. Calculations were performed on the parallel computing Linux cluster developed at the MNHN (80 CPUs in 23 nodes, 4 Go Ram per node). Descriptive information on the molecules studied is summarized in Table 4. Likelihood mapping diagrams for each molecule are given in Figure 1. The ML analyses were conducted with RAxML (Stamatakis, 2006). The parameters were estimated over 200 runs to ensure adequate searching, and the best scoring tree was based on the support for the likelihood-derived topologies estimated by bootstrap resampling (MLB) calculated using 1000 replicates. Figure 1. View largeDownload slide Likelihood mapping diagrams. Diagrams for each gene region based on 10000 quartets. Each point represents a four-taxon statement, with those quartets that are well resolved towards a certain topology falling in the three corners (high phylogenetic signal), while those in the central triangle are those that do not strongly support any of the three possible relationship statements (low phylogenetic signal). Figure 1. View largeDownload slide Likelihood mapping diagrams. Diagrams for each gene region based on 10000 quartets. Each point represents a four-taxon statement, with those quartets that are well resolved towards a certain topology falling in the three corners (high phylogenetic signal), while those in the central triangle are those that do not strongly support any of the three possible relationship statements (low phylogenetic signal). Support for nodes on the obtained trees were thus evaluated by non-parametric parsimony bootstrapping (PB) (Felsenstein, 1985), MLB and Bayesian clade probability (BCP) (as recovered from the majority-rule consensus of the 36000 or 360000 post-burn-in sampled trees). Bootstrap searches were conducted in general with 1000 pseudoreplicates each with ten random addition replicates and TBR branch swapping. For morphological and combined data, however, we used maxtrees=100000. For molecular and morphological data combined, partitioned Bremer decay (Bremer, 1994) (PBD) indices were also calculated based on one of the most parsimonious trees with the aid of TreeRot V.3 (Sorenson & Franzosa, 2007). The initial unconstrained and each constrained search were conducted with 10000 random addition replicates, keeping 100 trees at each step, and TBR branch swapping. As trees may differ in topology, we considered the MrBayes tree to be the ‘master’ tree, giving the figure its topology and branch lengths. The same splits were highlighted in the master tree. Support values from PI and ML bootstrap analyses and PBD indices (if applicable) were combined with the posteriors from the Bayesian tree. Topological agreement: support for classification changes We used parametric bootstrap analysis to test the monophyly of the Belostomatinae containing H. pelocoroides highlighted by some genes and morphological data. This technique is preferable to other methods of tree comparison in that it is less prone to type II statistical error (Goldman, Anderson & Rodrigo, 2000). Based on Smith, Page & Johnson’s (2004) procedure, model trees were estimated from the ML analyses with taxa constrained to be compatible with a hypothesis of monophyly for both the Belostomatinae and the Lethocerinae, and for the Belostomatinae taxa alone (loosest constraint possible). For each model tree (with branch length), we used the model of sequence evolution and parameters estimated to generate 500 simulated datasets of the same size as the original. Two heuristic searches were conducted on each replicate dataset using NNI branch swapping: once to find the overall optimal tree and again to find the best tree compatible with the constraint used to generate the model tree. Scores of these likelihood trees were then used to construct an expected distribution of likelihood differences under the null hypothesis being tested. Significance of the test statistic (the difference in log-likelihood values between the constrained and optimal trees) was assessed by direct comparison with the expected distribution. Data incongruence by evaluating the occurrence of unsupported and conflicting clades To indirectly assess resolution issues, some parts of the obtained phylogenies and the unsupported combination of COI and 16S partitions and mitochondrial and nuclear datasets, we used an alternative view of the correspondences between relationships among trees (by associating objective measures that quantify these correspondences) to examine the impact of using different partition schemes and methods of analysis, as well as the existence of missing data. We considered the nature of the data used for obtaining characters, as well as whether information for a particular taxon was available (e.g. a fossil specimen). We used the average number of conflicts between each of the proposed relationships (i.e. number of nodes that are contradicted by any other relationships involved by these nodes divided by the total number of nodes in the tree) and other different trees upon different treatments (not shown), by using the same logical approach when making the calculation of conflicts between super-trees created by Wilkinson et al. (2005). Two trees are understood to conflict if they assert logically contradictory relationships so that the tree clade cannot be present in any tree that includes the relationships in the input tree. Analyses were conducted among all different partitions, methods of analysis and taking into account the presence of missing data. A particular advantage of this approach is that it can be used to compare topologies with taxon sets that do not entirely match. In addition, not only did we adopt here the idea of lack of support, but also that tree clades conflicting with any input tree are more objectionable. Support values were obtained with stsupport (Wilkinson et al., 2005). Using such support for conflict among trees from different data sets as a means of assessing the independence between datasets can be seen as a test of heterogeneity (De Queiroz et al., 1995). Missing data were identified because of the degree of conservation of ancient exemplars of W. rhomboides, some problems in amplifying gene sequences of a small number of taxa, and the addition of fossil data, so that some gene sequences and morphological characters were missing from the analyses (see Table 1). To address the problem of the difference among the numbers of unsupported tree clades per tree in each treatment (i.e. using all available data or only those taxa with sequences of good quality) when different datasets could not be used, we performed a two-way ANOVA (Sokal & Rohlf, 1995). The number of conflicts per tree was the response variable, and datasets (Factor I) and quality of those data (Factor II) were used as factors, under the null hypothesis that the different datasets and the presence of missing data, with different methods of analysis (the replications) used to infer phylogeny (i.e. PI, ML and BI), did not affect the existence of phylogenetic signal and, in turn, the genus-level relationships in Belostomatidae. A failure to reject H0 suggests a strong congruence among such treatments. RESULTS Morphological analysis of Belostomatidae A morphological character matrix (Appendix 1) including 11 outgroup taxa (four genera) and 22 in-group species representing all 13 genera of Belostomatidae was coded based on 104 delimited morphological characters. Of these morphological characters, 16 were autapomorphies and 23 were multistate. The characters and their consistency and retention indices (CI and RI, respectively) are listed in Appendix 2. The PI analysis resulted in 42 most-parsimonious trees of 201 steps each and CI = 0.71, RI = 0.90 and RC = 0.64 (RC is the rescaled consistency index, which is obtained by multiplying the CI by the RI). A strict consensus among them was found by the branch-and-bound algorithm, as shown in Figure 2A, and is shown with unambiguous characters optimized. Conversely, Figure 2B–E are selected trees which have different topologies with regard to Belostomatinae (and position of Horvathinia). Figure 2. View largeDownload slide Morphological analysis. A, consensus of 42 most-parsimonious trees (length = 201, CI = 0.71, RI = 0.90, RC = 0.64) with non-ambiguous characters optimized; B, part of other most-parsimonious tree indicating Horvathinia as sister group of Diplonychinitrib. nov.; C, part of other most-parsimonious tree indicating Horvathinia as sister group of Belostomatini; D, part of other most-parsimonious tree indicating Hydrocyrius + Limnogeton as sister group of Horvathinia + Diplonychinitrib. nov. + Belostomatini; E, part of other most-parsimonious tree indicating Hydrocyrius + Limnogeton as sister group of Horvathinia + Diplonychinitrib. nov. Arrow indicates the node with main changes throughout the 42 most-parsimonious trees. (+), Fossils. Outgroups were omitted. Figure 2. View largeDownload slide Morphological analysis. A, consensus of 42 most-parsimonious trees (length = 201, CI = 0.71, RI = 0.90, RC = 0.64) with non-ambiguous characters optimized; B, part of other most-parsimonious tree indicating Horvathinia as sister group of Diplonychinitrib. nov.; C, part of other most-parsimonious tree indicating Horvathinia as sister group of Belostomatini; D, part of other most-parsimonious tree indicating Hydrocyrius + Limnogeton as sister group of Horvathinia + Diplonychinitrib. nov. + Belostomatini; E, part of other most-parsimonious tree indicating Hydrocyrius + Limnogeton as sister group of Horvathinia + Diplonychinitrib. nov. Arrow indicates the node with main changes throughout the 42 most-parsimonious trees. (+), Fossils. Outgroups were omitted. The monophyly of Belostomatidae s.l. (and as defined here: with Cratonepa enigmatica included) was not refuted in the PI analysis (PB = 96%) (Figs 2, 3). With low support, however, Belostomatinae is monophyletic (PB = 51%) with two well-defined clades, Belostomatini (Abedus+Belostoma and now W. rhomboides) (PB < 50%) and Diplonychini trib. nov. (Appasus+Diplonychus) (PB = 93%). Also, Lethocerinae was recovered as the sister group of Belostomatinae in the consensus tree, now with Benacus griseus and Kirkaldyia deyrolli resurrected and included, as well as the fossil taxa Lethocerus vetus and Iberonepa romerali. The main differences among the trees (Fig. 2B–E), however, are the sister group relationships of Horvathinia pelocoroides with Belostomatini and Diplonychini trib. nov., while Limnogeton and Hydrocyrius may be a monophyletic group and a sister group of Diplonychini trib. nov. + Horvathinia + Belostomatini. The support values are indicated in the preferred Bayesian topology (Fig. 3) referred below. Figure 3. View largeDownload slide Morphological analysis: Bayesian master tree (harmonic mean in log-likelihood = −787.68) selected by comparing the harmonic mean of the log-likelihood of trees obtained by the addition or not of the gamma parameter to the model. Support is based on Bayesian clade probability (BCP), maximum likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/MLB/PB. Clades are coloured according to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. ‘-’, Support values less than 50%. (+) with dashed lines, Fossils. Outgroups were omitted. Figure 3. View largeDownload slide Morphological analysis: Bayesian master tree (harmonic mean in log-likelihood = −787.68) selected by comparing the harmonic mean of the log-likelihood of trees obtained by the addition or not of the gamma parameter to the model. Support is based on Bayesian clade probability (BCP), maximum likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/MLB/PB. Clades are coloured according to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. ‘-’, Support values less than 50%. (+) with dashed lines, Fossils. Outgroups were omitted. The preferred Bayesian topology (Fig. 3) was selected by comparing the harmonic mean of the log-likelihood of each of the two aforementioned analyses presented in the Material and Methods. In this study, the addition of the gamma parameter to the model improved its fit to the data. The value of the harmonic mean of the Bayesian analysis without the gamma parameter was −799.71, whereas that of the analysis with the gamma parameter was −787.68. Because the latter value is closer to 0, the tree derived from the Bayesian analysis with gamma was preferred. The Bayes factor was approximately 24, which means that the difference between the harmonic means of the two analyses was significant. Some differences were found among the tree from BI analysis including gamma and those from parsimony analysis. The preferred BI analysis yielded trees less similar to those obtained through PI analysis with regard to the subfamily Lethocerinae (Fig. 3). Lethocerinae appear to be a paraphyletic taxon in relation to Belostomatinae, which was recovered as monophyletic only in the consensus tree based on parsimony (see above). Its strange configuration in the Bayesian trees, however, should be viewed cautiously, because it is probably an artefact of missing characters – the fossil specimen of L. vetus, which made it impossible to code characters from the wings and male genitalia for this species. In contrast, H. pelocoroides is recovered as the sister group to Diplonychini trib. nov. (BCP = 78%, MLB < 50%, PB < 50%), similar to some most-parsimonious trees with the almost same topology (e.g. Fig. 2B). One of those most-parsimonious trees (Fig. 2B) is slightly similar to that obtained by BI analysis. Another difference found here concerns that topology which deems Horvathinia as the sister group of Belostomatini, while Limnogeton is a sister group of Hydrocyrius + Diplonychini trib. nov. + Horvathinia + Belostomatini (Fig. 2C). Overall, the morphological data suggest the transfer of Horvathinia to Belostomatinae. The monophyly of Belostomatidae (BCP = 100%, MLB = 95%, PB = 96%), Belostomatinae (BCP = 88%, MLB < 50%, PB = 51%) and Belostomatini (BCP = 64%, MLB < 50%, PB < 50%) is not refuted in any of our analyses (i.e. PI and preferred BI and ML analyses) (Fig. 3), whereas Diplonychini trib. nov. was recovered as monophyletic across almost all analyses (BCP = 79%, MLB = 53%, PB = 93%), except ML. Molecular analyses of Belostomatidae The complete molecular dataset contained 1829 aligned characters, all of which were variable; 834 (about 46%) were parsimoniously informative. Within the COI gene fragment, 569 (77%) characters were variable, and 408 of these (56%) were parsimony informative, this being the partition with the largest fraction of variable sites and parsimony-informative sites (Table 4). As measures of phylogenetic content, COI and 18S contained higher values of g1 (skewness) of 1000 random trees (−0.74 and −0.71, respectively), while only 18S showed a higher number of the percentage of clades (81%) supported by > 50% parsimony bootstrap among all resolved clades in the strict consensus trees (Table 4). Likelihood mapping was conducted by examining 10000 quartets using an approximate likelihood function based on the model parameters selected for each molecule. The final distribution of points mapped into regions of the triangle, in each partition, reveals that COI and 18S fragments performed somewhat better than 16S, with a little over 83% of the quartets mapped into the strongly tree-like regions. Likelihood mapping diagrams for each molecule are given in Figure 1. Heuristic parsimony searches resulted in two most-parsimonious trees for COI (length = 1764, CI = 0.32, RI = 0.52, RC = 0.17), two for 16S (length = 687, CI = 0.48, RI = 0.78, RC = 0.37) and 144 for 18S (length = 327, CI = 0.64, RI = 0.78, RC = 0.50). Parsimony analyses still recovered 20655 most-parsimonious trees for the combined molecular dataset (length = 2,817, CI = 0.39, RI = 0.54, RC = 0.21) (Supporting Information: clades presented in the strict consensus compatible with the Bayesian consensus and ML method are highlighted in Figs S1–S3). The present phylogenetic analysis based on three gene regions provides further support for the classification changes proposed above for morphological characters. All molecular trees support the monophyly of Belostomatidae s.s. (i.e. Lethocerinae+Belostomatinae) (BCP = 69%, PB ≤ 50%, MLB = 100% for COI; BCP = 100%, PB = 65%, MLB = 100% for 18S; BCP = 100%, PB ≤ 50%, MLB = 100% for 16S), as well as of the Lethocerinae (BCP = 100%, PB = 100%, MLB = 100% for all gene partitions). However, only the 16S rDNA data support the monophyly of Belostomatinae (BCP = 100%, PB = 64%, MLB = 100%) with all methods (Supporting Information: Fig. S3). Conversely, COI and 16S rDNA support the monophyly of Belostomatini with all methods (BCP = 100%, PB = 100%, MLB = 100% for COI; BCP = 85%, PB = 60%, MLB = 100% for 16S). Interestingly, none of the molecular trees based on the separated gene regions supports the Diplonychini trib. nov., and COI gives low congruence. In this case Diplonychus is close to the base of the Belostomatinae, whereas Appasus is within the clade consisting of Appasus + Belostomatini. Finally, Horvathinia and Lethocerinae are sister groups (BCP = 95%, PB ≤ 50%, MLB = 100%) based on only the 18S fragment, and not all methods used here support such a clade (only BI and ML). With all other fragments, instead, Horvathinia comprises a clade, which includes all belostomatines, either close to the African genera Limnogeton and Hydrocyrius or close to the base of the Belostomatinae. Congruence amongst the molecular phylogeny with all datasets from PI, ML and BI is much higher (Fig. 4). BI analysis of the combined molecular dataset recovered the monophyly of the Diplonychini trib. nov. (BCP = 100%, PB = 58%, MLB = 100%), Belostoma (BCP = 99%, PB = 51%, MLB = 100%) and Belostomatinae (BCP = 100%, PB = 70%, MLB = 100%) (now with Horvathinia included), almost all with high levels of clade support. Lethocerinae were recovered as a monophyletic subfamily (BCP = 100%, PB = 99%, MLB = 100%) across all tree reconstruction methods, as well as Lethocerus (BCP = 100%, PB = 95%, MLB = 100%). Unfortunately, Benacus griseus was not included in those analyses because of problems in amplifying its gene sequences. Besides, all but PI trees support Belostomatidae as monophyletic (BCP = 98%, PB < 50%, MLB = 100%) as well, with low bootstrapping values instead. Additionally, PI, ML and BI supports the transfer of W. rhomboides to Belostomatini (BCP = 94%, PB = 80%, MLB = 100%). Conversely, neither Belostoma and Abedus nor Diplonychini trib. nov. + Belostomatini were recovered as a monophyletic group in the strict consensus of PI analysis. Figure 4. View largeDownload slide Molecular analysis: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the combined molecular dataset (five partitions: pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. Support is based on Bayesian clade probability (BCP), maximum-likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/PB/MLB. Clades are coloured according to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. (+), Fossils. Figure 4. View largeDownload slide Molecular analysis: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the combined molecular dataset (five partitions: pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. Support is based on Bayesian clade probability (BCP), maximum-likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/PB/MLB. Clades are coloured according to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. (+), Fossils. Morphological and molecular dataset analyses PI analysis yielded 25 most-parsimonious trees using the combined molecular and morphological dataset (length = 2592, CI = 0.45, RI = 0.56, RC = 0.25), and clades present in the strict consensus compatible with the Bayesian consensus and ML inference are highlighted in Figure 5. Partitioned Bremer analysis totalled 2592 characters and showed the majority of the signal coming from the first codon position of COI and 16S (COI pos1, 39.6%; 16S, 24.7%). Morphology and the third codon position of COI were minimal in their support (8.2 and 6.6%, respectively). Interestingly, for most well-supported clades, morphology is the partition agreeing with COI pos1 and nuclear partition 18S by showing positive values of decay indices. Figure 5. View largeDownload slide Combined analysis: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the combined morphological and molecular dataset (six partitions: morphology, pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. Support is based on Bayesian clade probability (BCP), maximum-likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/MLB/PB. Partitioned Bremer indices are shown as graphs near to respective clades, with the x-axis representing partitions: morphology, 16S, COI and 18S, and y-axis support ranging from −3.6 to 2.4. Clades are coloured accordingly to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. (+) with dashed lines, Fossils. Figure 5. View largeDownload slide Combined analysis: Bayesian consensus phylogram. Mixed-model Bayesian analysis of the combined morphological and molecular dataset (six partitions: morphology, pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Thicker clades refer to those also found by maximum-likelihood and maximum-parsimony analyses. Support is based on Bayesian clade probability (BCP), maximum-likelihood bootstrap (MLB) and parsimony boostrap (PB). Percentages are given near to the branches as BCP/MLB/PB. Partitioned Bremer indices are shown as graphs near to respective clades, with the x-axis representing partitions: morphology, 16S, COI and 18S, and y-axis support ranging from −3.6 to 2.4. Clades are coloured accordingly to the proposed classification. The coloured names refer to the taxa affected by the present classification proposal. (+) with dashed lines, Fossils. Belostomatidae s.l. was recovered as monophyletic across all analyses (BCP = 100%, MLB = 96%, PB = 100%, PBDmorphology = 2.3), and C. enigmatica is now thought to be included as a member of this family together with I. romerali. Although the phylogenetic analyses performed poorly with respect to clade posterior probabilities, parsimony bootstrap values and Bremer supports, Lethocerinae was recovered as a monophyletic subfamily (BCP = 58%, MLB < 50%, PB = 63%, PBDmorphology = 2.3) across all tree reconstruction methods. We recovered the position of H. pelocoroides as being the sister group of Hydrocyrius (BCP = 66%, MLB < 50%, PB = 91%, PBDmorphology = 2.3), and Limnogeton was recovered as the sister group of H. pelocoroides + Hydrocyrius (BCP = 51%, MLB < 50%, PB = 73%, PBDmorphology = 2.2) only in the BI and PI analyses. The monophyly of Belostomatinae was corroborated here (BCP = 61%, MLB = 60%, PB = 74%, PBDmorphology = 2.3) across all methods, despite performing poorly with respect to clade posterior probabilities and ML and parsimony bootstrap values. Instead, Belostomatini (BCP = 76%, MLB = 81%, PB = 85%, PBDmorphology = 2.3) and Diplonychini trib. nov. (BCP = 96%, MLB = 98%, PB = 89%, PBDmorphology = 2.2) were recovered as monophyletic and sister groups with slightly higher support values across all methods. Overall, congruence amongst the molecular and morphological datasets obtained from BI, PI and ML was much higher, despite the existence of the strange group comprising Horvathinia + Hydrocyrius + Limnogeton (Fig. 5). Topological agreement with morphological analysis: support for classification changes Parametric bootstrap analysis of the molecular data clearly rejects the null hypotheses of non-monophyly for both the Lethocerinae and the Belostomatinae (containing Horvathinia) clades predicted by some genes and morphology (P < 0.001, Figs 4–6), except for the 18S dataset (P = 0.99) (Supporting Information: Figs S4–S11). Under the constraint of Belostomatinae monophyly, the loosest constraint possible in this analysis, the log-likelihood difference between the simulated datasets with and without the constraint ranges from +100 to +300. However, the observed difference in log-likelihood for the actual dataset was +909. Thus, the probability of observing a difference of this magnitude (if the null hypothesis were true) is considerably less than 1%. BI, PI and ML trees share many common nodes, and the number of unsupported tree clades conflicting with all input trees seems to show strong congruence among different partitions when one considers the absence of some or all of the relevant data (F = 1.423, P = 0.944) (Tables 5–6). On average, 79.6% of the variation among the trees is due to differences among the matrices with or without some taxa, and only 5.4% of the variation is due to differences among the datasets. Table 5. Number of unsupported clades conflicting with all input trees per tree obtained by stsupport (Wilkinson et al., 2005), taking into account maximum-parsimony (PA), maximum-likelihood (ML) and Bayesian analysis (BI) approaches and the type of data partition. Figures in parentheses are U = number of unsupported clades conflicting with all relevant input trees and TC = number of tree clades, respectively   Quality of the data    With all data  Incomplete data and fossils absent  Partitions  PA vs. ML + BI  ML vs. BI + PA  BI vs. ML + PA  PA vs. ML + BI  ML vs. BI + PA  BI vs. ML + PA  COI  0.30 (10/33)  0.11 (4/35)  0.07 (2/30)  0.39 (7/18)  0.11 (2/18)  0.06 (1/17)  18S rDNA  0.05 (1/20)  0.08 (2/26)  0.08 (2/26)  0 (0/20)  0.13 (3/24)  0 (0/21)  16S rDNA  0.17 (6/35)  0 (0/35)  0 (0/34)  0.14 (5/35)  0 (0/35)  0 (0/33)  Morphology  0.22 (4/18)  0.14 (3/21)  0 (0/20)  0 (0/15)  0.21 (4/19)  0 (0/18)  Molecular dataset  0.11 (4/36)  0.04 (2/46)  0.04 (2/46)  0.22 (5/23)  0.14 (5/35)  0 (0/32)  Molecular + morphology  0.11 (4/38)  0.31 (16/51)  0.12 (5/43)  0.29 (6/21)  0.06 (2/36)  0.03 (1/34)    Quality of the data    With all data  Incomplete data and fossils absent  Partitions  PA vs. ML + BI  ML vs. BI + PA  BI vs. ML + PA  PA vs. ML + BI  ML vs. BI + PA  BI vs. ML + PA  COI  0.30 (10/33)  0.11 (4/35)  0.07 (2/30)  0.39 (7/18)  0.11 (2/18)  0.06 (1/17)  18S rDNA  0.05 (1/20)  0.08 (2/26)  0.08 (2/26)  0 (0/20)  0.13 (3/24)  0 (0/21)  16S rDNA  0.17 (6/35)  0 (0/35)  0 (0/34)  0.14 (5/35)  0 (0/35)  0 (0/33)  Morphology  0.22 (4/18)  0.14 (3/21)  0 (0/20)  0 (0/15)  0.21 (4/19)  0 (0/18)  Molecular dataset  0.11 (4/36)  0.04 (2/46)  0.04 (2/46)  0.22 (5/23)  0.14 (5/35)  0 (0/32)  Molecular + morphology  0.11 (4/38)  0.31 (16/51)  0.12 (5/43)  0.29 (6/21)  0.06 (2/36)  0.03 (1/34)  View Large Table 6. Two-way ANOVA. The number of unsupported tree clades conflicting with all input trees was the response variable and datasets and quality of those data were used as factors, under the null hypothesis that the different dataset and the presence of missing data, with different methods of analysis used to infer phylogeny, did not affect the existence of phylogenetic signal and, in turn, the genus-level relationships in Belostomatidae. A failure to reject H0 suggests strong congruence among such treatments. Abbreviations: d.f. = degree of freedom; SS = sum of squares; MS = mean of squares Source  d.f.  SS  MS  F-value  P-value  Quality of data  5  0.074  0.015  1.235  0.324  Partitions  1  0.001  0.001  0.083  0.776  Interaction  5  0.014  0.003  0.233  0.944  Residuals  24  0.288  0.012  –  –  Source  d.f.  SS  MS  F-value  P-value  Quality of data  5  0.074  0.015  1.235  0.324  Partitions  1  0.001  0.001  0.083  0.776  Interaction  5  0.014  0.003  0.233  0.944  Residuals  24  0.288  0.012  –  –  View Large TAXONOMY Since we found here evidence to support Appasus and Diplonychus as a monophyletic group, the new tribe Diplonychini is herein described to accommodate that clade. Family Belostomatidae Leach, 1815 Subfamily Belostomatinae Leach, 1815 Tribe Diplonychini trib. nov. Type genus: Diplonychus Laporte, 1833 Diagnosis: Diplonychini differs from other belostomatine suprageneric groups by the frons rounded or curved in dorsal view (Fig. 9D, G, H), as well as the following male genitalia features: (1) the transverse bridge of basal plate of male genitalia clearly jointed and entire (Fig. 16B); (2) plate of phallotheca as long as ventral diverticulum. Likewise, females of Diplonychini differ from the others by the presence of an ampulla located at the basal part of spermatheca (Fig. 21A–C). Distribution: Africa, Australia, East Indies, southern Asia (Estévez & Ribeiro, 2011). Description: Measurements. – Total length (from apex of head to apex of abdomen at rest): from 8.2 to 27.7 mm. General coloration. – Almost uniformly brown. External morphology. – Body ovate with wings usually covering abdomen. Frons rounded or curved (Fig. 9D, G, H); vertex without median longitudinal carina (Fig. 9D); antennae with segments 2 and 3 not flattened ventrally, with fourth segment similar to or slightly more bulbous than prolongations of segments 2 and 3 (Fig. 11B); frontogenal suture slightly convergent and opened distally (Fig. 9D). Pronotum without longitudinal median carina (e.g. Fig. 9D); prosternal keel usually poorly elevated, except for some Appasus species; hemelytra with rounded pruinose area; clamp of clavus with its outer projection overlapping inner part, always far from the margin of hemelytra (Fig. 14A); its outer carina with three rows of microtrichiae along its external margin, covering small portion in dorsal view (Fig. 13A, B); tile-like microtrichiae rounded at apex, never toothed along the margin of its apex (Fig. 12B); foretarsi with two segments, externally usually appearing one-segmented, with segment 1 conspicuous, with two symmetrical grooves; claws vestigial; hind trochanters carinated, with short hairs or bristles along outer margins. Pilosity developed, covering half of connexivum, slightly constricted between spiracles, extending posteriorly along about half of or almost entire genital operculum; pubescence of ventral laterotergites 3 and 4 not attaining entire external margin; air straps lanceolate, with somewhat uniform width along its extension. Male genitalia: Phallosoma fused to ventral diverticulum; arms of phallosoma well developed, extending nearly to apex of ventral diverticulum, enclosing ventral diverticulum in some Appasus (Fig. 16E), somewhat laterally directed; orifice strongly developed, dorsally located on apex of phallosoma; ventral diverticulum contiguous, never bilobed, with its apex without ventroapical protuberance, not showing spines or tubercles in ventral view; transverse bridge of basal plate of male genitalia clearly jointed and entire (Fig. 16B); plate of phallotheca somewhat developed, fused to or close to ventral diverticulum. Female genitalia: Operculum of females with two tufts of setae on apex; apex of second valvulae with an inconspicuous spine; basal part of spermatheca without distinct apodemes, clearly with the presence of an ampulla; median vagina area below spermatheca without pouch (Fig. 21A–C). Taxonomic notes: In general, members of the new tribe Diplonychini share with W. rhomboides the prosternal carina poorly elevated (‘prosternal keel rounded’ according to Estévez & Ribeiro, 2011: 51) (character 27: 0 > 1) and with H. pelocoroides the surface of apex of ventral diverticulum (in ventral view) without spines or tubercles (character 89: 0 > 1). Our findings support Diplonychini trib. nov. as the sister group of the tribe Belostomatini by the following unambiguous homoplastic synapomorphies: (1) clamp of clavus with its outer projection not flattened, overlapping the inner part (similar to Fig. 14A) (character 31: 2 > 3); and (2) outer carina of the clamp with three rows of microtrichiae along its external margin (character 32: 2 > 1) (Fig. 13A, B). The new tribe Diplonychini shares with other members of the subfamily Belostomatinae the foretarsi bearing both claws vestigial (character 44: 0 > 3), phallosoma fused with ventral diverticulum (character 71: 0 > 1), ventral diverticulum contiguous, never bilobed (character 85: 0 > 1) (Fig. 16E), and apex of second valvulae with a spine and/or protuberance (character 94: 0 > 1). All morphological character states cited above are exclusive to this subfamily (Fig. 6). Figure 6. View largeDownload slide Combined analysis: Bayesian consensus phylogram of Figure 5. Mixed-model Bayesian analysis of the combined morphological and molecular dataset (six partitions: morphology, pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Optimized tree with non-ambiguous characters, except for the clades including Hydrocyrius + Horvathinia + Limnogeton and Belostoma ribeiroi + B. plebejum. (+), Fossils. Outgroups were omitted. Figure 6. View largeDownload slide Combined analysis: Bayesian consensus phylogram of Figure 5. Mixed-model Bayesian analysis of the combined morphological and molecular dataset (six partitions: morphology, pos1 = HKY+Γ, pos2 = GTR+Γ, pos3 = GTR+I+Γ, 18S = SYM+I+Γ, 16S = HKY+I+Γ). Optimized tree with non-ambiguous characters, except for the clades including Hydrocyrius + Horvathinia + Limnogeton and Belostoma ribeiroi + B. plebejum. (+), Fossils. Outgroups were omitted. DISCUSSION Support for classification changes in the composition of groups The present study is the first higher-level phylogenetic study to include all Belostomatidae genera. Lee (1991) concurred with the decision made by Lauck & Menke (1961) of separating Belostomatidae into the Lethocerinae, Horvathiniinae and Belostomatinae, highlighting the existence of further characters shared by Lethocerus (Lethocerinae) and Limnogeton (Belostomatinae), and also by Horvathinia (Horvathiniinae) and Hydrocyrius (Belostomatinae). The results of all the datasets do not corroborate the restriction of Belostomatidae to the subfamilies Lethocerinae, Horvathiniinae and Belostomatinae as previously proposed by Mahner and earlier by Lee, and by the intuitive phylogeny produced by Lauck & Menke (1961), but rather suggest the inclusion of H. pelocoroides within the subfamily Belostomatinae instead of comprising the monobasic subfamily Horvathiniinae related to the subfamily Belostomatinae (Figs 2–6). All analyses support the inclusion of C. enigmatica in Belostomatidae s.l. and the inclusion of I. romerali in Belostomatidae s.s., based at least on moderately high clade support (combined dataset, BCP = 100%, MLB = 96%, PB = 100%, PBDmorphology = 2.3 and BCP = 81%, MLB = 63%, PB = 70%, PBDmorphology = 2.3, respectively) with the use of the combined morphological and molecular datasets (Fig. 6). The present study provided evidence and assigned C. enigmatica to Belostomatidae based on the following putative synapomorphy of Mahner: the siphon retracted into the abdomen (character 104: 0 > 1) (Figs 2, 3A, 4, 6). As indicated by Jattiot et al. (2012), the abdominal sternites divided into parasternites and median sternites (cited herein as character 53: 1 > 0) (Fig. 7A) is a condition also considered by Mahner (1993) as a putative synapomorphy. Our results, however, considered such a character as a plesiomorphy for Belostomatidae. The present study also provides further evidence for the monophyly of Belostomatidae s.l., based on the following unambiguous non-homoplastic synapomorphies: (1) eyes somewhat projected laterally with angled borders (character 8: 0 > 1) (Figs 8, 9); and (2) tarsi 2- or 3-segmented, rarely 1-segmented (character 37: 0 > 1). Figure 7. View largeDownload slide General aspect of abdomen. A, ventral laterotergites (= connexivum), parasternites and sternites, ventral view. B, subcylindric body, transversal view. C, connexivum and sternites, ventral view. Parasternites are not visible. D, flattened body, transversal view. Adapted from Mahner (1993). con, connexivum; gop, genital operculum; lfs, like-fold suture; par, parasternites; spi, spiracles; ste, sternites. Figure 7. View largeDownload slide General aspect of abdomen. A, ventral laterotergites (= connexivum), parasternites and sternites, ventral view. B, subcylindric body, transversal view. C, connexivum and sternites, ventral view. Parasternites are not visible. D, flattened body, transversal view. Adapted from Mahner (1993). con, connexivum; gop, genital operculum; lfs, like-fold suture; par, parasternites; spi, spiracles; ste, sternites. Figure 8. View largeDownload slide Head and pronotum, dorsal view. A, Ranatra sp. [female]. B, Lethocerus collosicus [male]. C, Abedus ovatus [male]. D, Belostoma bifoveolatum [male]. E, Belostoma dentatum [female]. F, Hydrocyrius sp. [male]. cfg, distally closed frontogenal suture; ofg, opened frontogenal suture. Figure 8. View largeDownload slide Head and pronotum, dorsal view. A, Ranatra sp. [female]. B, Lethocerus collosicus [male]. C, Abedus ovatus [male]. D, Belostoma bifoveolatum [male]. E, Belostoma dentatum [female]. F, Hydrocyrius sp. [male]. cfg, distally closed frontogenal suture; ofg, opened frontogenal suture. Figure 9. View largeDownload slide Head and pronotum, dorsal view. A–E, general aspects: A, Horvathinia pelocoroides [male]. B, Nepa hoffmani [male]. C, Weberiella rhomboides [female]. D, Diplonychus sp. [female]. E, Hydrocyrius colombiae [male]. F–I, comparison among the species with streamlined and rounded eyes: F, Limnogeton fieberi [female] (streamlined eyes). G, Appasus sp. [female] (streamlined eyes). H, Diplonychus esakii [female] (streamlined eyes). I, Weberiella rhomboides [male] (rounded eyes). car, developed longitudinal carina; cfg, distally closed frontogenal suture; fov, fovea; ofg, opened frontogenal suture; ver, vertex. Figure 9. View largeDownload slide Head and pronotum, dorsal view. A–E, general aspects: A, Horvathinia pelocoroides [male]. B, Nepa hoffmani [male]. C, Weberiella rhomboides [female]. D, Diplonychus sp. [female]. E, Hydrocyrius colombiae [male]. F–I, comparison among the species with streamlined and rounded eyes: F, Limnogeton fieberi [female] (streamlined eyes). G, Appasus sp. [female] (streamlined eyes). H, Diplonychus esakii [female] (streamlined eyes). I, Weberiella rhomboides [male] (rounded eyes). car, developed longitudinal carina; cfg, distally closed frontogenal suture; fov, fovea; ofg, opened frontogenal suture; ver, vertex. On the other hand, Hebsgaard et al. (2004) considered the presence of a metacoxa conical, firmly united with metapleuron (character 36: 0 > 1), as well as the hind tibiae flattened, with swimming hairs (considered here as characters 45 and 46), synapomorphies for the clade Belostomatidae. These characters were therefore thought to be characteristic of Belostomatidae s.l. according to the aforementioned authors; however, this result is in conflict with the present study. The condition of the metacoxae and hind tibiae was herein not recovered in C. enigmatica (Jattiot et al., 2012), but rather in Belostomatidae s.s. as follows (Figs 2A, 6): (1) metacoxae firmly united with metapleuron as an unambiguous non-homoplastic synapomorphy; (2) middle and hind tibia and tarsus completely different from each other, with metathoracic legs wide in relation to its length (character 45: 0 > 1) as an unambiguous non-homoplastic apomorphy of I. romerali, either with the use of morphological or combined morphological and molecular datasets. In fact, for those authors C. enigmatica remained a taxon of enigmatic position within the Nepoidea, with several plesiomorphic characters that imply that it could be a basal sister group of Nepidae, or even the sister group of all the Nepoidea. Curiuosly, the simple, slender and cursorial middle and hind legs, probably not modified for swimming, are unambiguously a homoplastic apomorphy in the consensus tree based on the morphological dataset because it supported Limnogeton as a reversal (character 46: 2 > 0) (Fig. 2A). The biology of these aquatic predators was probably different from those of Cenozoic and Recent Nepidae, based on the absence of an elongate siphon that stopped these insects breathing under water deeper than its length. According to Jattiot et al. (2012), the non-flattened hind tibiae of Cratonepa suggest that this species had a means of locomotion similar to that of the recent Nepidae, walking on the bottom of a pond or on emergent vegetation. In fact, the elongation of the tibiae and the tarsi was not considered in the present study but the specialized condition of the middle and hind tibiae and tarsi broadly flattened with swimming setae (character 46: 0 > 1) shared by lethocerines is reported here as an ambiguous synapomorphic trait with either the morphological or the combined morphological and molecular datasets. Likewise, Mahner (1993), in his phylogeny, proposed the presence of pubescence along ventral laterotergites (character 54: 0 > 1) as a synapomorphy shared by all belostomatids. The present analysis gives evidence for this, having reported it instead as an ambiguous non-homoplastic synapormorphy of Belostomatidae s.l. with either the morphological or the combined morphological and molecular datasets. Popov (1971) considered the Stygeonepinae to be a subfamily of Belostomatidae. Schlüter (1981), however, placed Stygeonepa among the Nepidae although he gave no reason for doing so and Nel (1991) restated that no precise reason exists for such an allocation. As stated by Martínez-Delclòs, Nel & Popov (1995), the present study shows that I. romerali (a Stygeonepinae) is a Belostomatidae (considered herein Belostomatidae s.s.) (BCP = 81%, MLB = 63%, PB = 70%, PBDmorphology = 2.3, in combined morphological and molecular datasets) and not proximal to the Nepidae based on the metacoxae being firmly united with metapleuron, although ML analysis did not give more than 70% support for such a clade using combined gene partitions. Only with the morphological dataset (Fig. 2A), the posterior portion of its hind tibiae, divided in two parts, both separated by a sulcus (character 49: 0 > 1), was Iberonepa grouped together with members of the Lethocerinae, as reported in the consensus tree. Also, this trait appears to arise independently in the Belostomatinae (Hydrocyrius). Likewise, in analysis of the combined morphological and molecular datasets (Fig. 6) the aforementioned condition is also homoplastic (convergent) and either ambiguously supports Belostomatidae s.l. or it is an unambiguous synapomorphy of Hydrocyrius. Finally, there is confusion promoted by Mahner (1993) regarding the existence of a third pair of valvulae in Mahner’s Belostomatidae. According to the author, Belostomatidae are defined by the absence of these valvulae, but we are inclined to accept the opinion of Lalitha et al. (1997) that the third pair of valvifers and valvulae both are fused so that only one plate can be observed. Thus, Mahner possibly incorrectly reported this trait and it is not considered in this study. Lee (1991) continued to characterize the subfamily Belostomatinae based on the symplesiomorphic character state of the absence of divisions in the abdominal sternites. Mahner (1993) probably assigned neither W. rhomboides nor H. pelocoroides to Belostomatinae because of the supposed lack of synapomorphies and misunderstandings concerning some apomorphic states of the abdomen; however, the present analyses based either on morphological or on combined morphological and molecular datasets gives evidence for the monophyly of such a subfamily (with H. pelocoroides and W. rhomboides included), based on 20 morphological synapomorphies (when optimized only on trees based on the morphological dataset), of which six are unambiguous (Fig. 2A), and 20 morphological synapomorphies (when optimized on the tree based on the combined morphological and molecular datasets), of which six are unambiguous (Fig. 6). Excluding the homoplasies, the non-homoplastic synapomorphies are as follows: (1) foretarsus bearing both claws vestigial (character 44: 0 > 3); (2) phallosoma fused with ventral diverticulum (character 71: 0 > 1); (3) ventral diverticulum contiguous, never bilobed (character 85: 0 > 1) (Fig. 10A, B); (4) genitalic protuberances spiny-shaped (character 88: 0 > 1) (Fig. 18); (5) phallothecal plate developed related to the ventral diverticulum, in lateral view (character 90: 0 > 1); and (6) apex of second valvulae with spine and/or protuberance (character 94: 0 > 1). All the aforementioned characters supported that clade in both the morphological and the combined morphological and molecular datasets. Analyses of the morphological, combined molecular dataset and the morphological + molecular dataset presented here recover Horvathinia + Belostomatinae with moderate to high support with all methods (i.e. PI, ML and BI) (Figs 3–6). It therefore seems plausible to transfer H. pelocoroides to Belostomatinae, although nothing is known about brooding behaviour in this genus. Horvathinia is supported here by the abdominal sternites 3–7 divided laterally by weak, suture-like folds into median- and parasternites (character 57: 1 > 0), always a homoplasy (reversal) according to our results based on the morphological and the combined morphological and molecular datasets. This study hence is in conflict with previous hypotheses (Lauck & Menke, 1961; Lee, 1991; Mahner, 1993) that define Lauck & Menke’s Belostomatinae as having abdominal sternites 3–7 undivided laterally by weak folds (character 57: 0 > 1). The present study highlights that this trait ambiguously supports Belostomatidae s.s. as a homoplastic synapomorphy with either morphological or combined morphological and molecular datasets. However, note that this result should be viewed cautiously, because it is probably an artefact of missing characters (see below) in I.