Cladistics

From the data matrix (17 taxa × 27 characters) presented in Appendix 2, cladogram searches were carried out under the application of 2 different character weighting approaches: equal and successive weighting. The analysis using all characters, equally weighted, resulted in 6 most parsimonious cladograms with length = 68 steps, CI = 0.55, RI = 0.64. In the strict consensus tree (Fig. 1A), Heterostylum appears as monophyletic including 3 polytomic major clades, one formed by the Neotropical species: H. haemorrhoicum. (H. rufum (H. evenhuisi, H. maculipennis)), the second also with Neotropical members (H. ferrugineum (H. hirsutum, H. pallipes)) and a third one, including all the Nearctic species, with no internal resolution.

After successive weighting was applied, a better resolved topology was obtained and the analysis yielded a single most parsimonious cladogram (Fig. 1B) with length = 68 steps, CI = 0.55, RI = 0.64. This cladogram is identical to one of the 6 cladograms resulting under equal weighting analysis.

The results (both equal and successive weighting analyses) support the monophyly of Heterostylum based on 4 unambiguous synapomorphies: row of setae on the ocellar tubercle separating the anterior ocellus from the others (character 1); row of setae on the anterodorsal surface of femur III (character 10, state 1); veins M1 and R5 merging near or at the point where R4 arises (character 13, state 1); and sperm pump longer than spermathecae (character 25, state 1); and one ambiguous synapomorphy: head as wide as or wider than thorax (character 7, state 1).

Based on the cladogram from the successive weighting analysis (Fig. 1B), 2 major groups within the genus were found: one large clade (Clade B) including the Neotropical species (H. ferrugineum. (H. hirsutum, H. pallipes)) + (H. haemorrhoicum. (H. rufum. (H. evenhuisi, H. maculipennis))), the sister of a Nearctic clade (Clade A) with (H. helvolum, H. robustum) + H. deani + (H. croceum, H. engelhardti). With the acctran optimization, clade B appears supported by one synapomorphy, which is the presence of circular spots of different color on the tergites (character 19, state 1). However, this is misleading as, actually, only part of clade D presents this state. Therefore, the following corrected scenario for character 19 (pattern of coloration on abdominal tergites) in clade B is: presence of a central longitudinal stripe (state 2) ambiguously support the clade C (H. ferrugineum (H. hirsutum, H. pallipes)), while in clade D the loss of any mark (state 3) is observed in H. haemorrhoicum, and circular spots are present in the clade (H. rufum (H. evenhuisi, H. maculipennis)) This last clade is supported by one unambiguous synapomorphy, the presence of golden hairs on the mesonotum and dark brown hairs on the pleura (character 8, state 2).

Although no higher-level phylogeny that includes Heterostylum, Triploechus and Apiformyia together has been published to date, some authors have suggested and discussed their morphological affinities (Hall 1976; Hall & Evenhuis 1981; Yeates 2008). These affinities have been based mainly on the sinuous posterior eye margin, which is uncommon in Bombyliinae. Also, among the 3 genera, Apiformyia could possibly be recovered closer to the root in the phylogeny, because its female terminalia lacks the sand chamber and acanthophorite spines. In the present study, the outgroup was composed by A. australis and 3 species of Triploechus: 2 Andean (T. bellus and 71 heteronevrus) and one Nearctic (T. novus), besides Toxophora aurea, which was used to root the resulting trees. We did not include all 5 species of Triploechus in the analysis, but its monophyly was recovered using the 3 representative species. Furthermore, these preliminary data indicated a sister-group relationship between Triploechus and Heterostylum.

Biogeography

The origin of the Bombyliidae (in a broader sense, including Mythicomyiinae, as recently supported by Trautwein et al. 2010) was hypothesized to have taken place in the Middle Jurassic, based on congruence among biogeographical, molecular and fossil data (see Lamas & Nihei 2007), whereas the diversification of the Bombyliinae lineage has occurred since the Early Cretaceous (125 Mya), according to molecular dating (Wiegmann et al. 2003). Fossil records for Bombyliinae and Bombyliini, where Heterostylum, Triploechus, and Apiformyia have been allocated, are rare, and this scarcity has hampered to further discussions about the ancient diversification of these taxa and also of any other bombyliine group. For example, the oldest fossil records for Bombyliini are described from the Eocene epoch (Evenhuis 1997; Lamas & Nihei 2007). These fossil records are represented by Bombylius from Germany (Miocene), France (Oligocene), and the Baltic region (Eocene/Oligocene); and Dischistus from France (Oligocene). However, given its fragmentary nature, a literal reading of the fossil record should be regarded as misleading (Heads 2005), and although not yet found, it is still possible that pre-Cretaceous fossils of the Bombyliidae lineage will be discovered.

Based on the cladogram of Fig. 1B, the area cladogram derived for Heterostylum provides evidence for a spatial disjunction between the Nearctic and Neotropical clades (clades A and B) associated with the Mexican Transition Zone (MTZ, see Morrone 2004, 2006). The Nearctic clade ranges widely over the western and central USA and northern and central Mexico, with the MTZ as the southern boundary (Figs. 2 and 3). The Neotropical clade B extends throughout the Neotropical region (Figs. 2 and 3). It is interesting to note that the genus is absent towards northeastern USA and the south of Brazil and Paraguay, as well as in eastern North America and west of the Andes. On the other hand, some of the lacunae in the distributional pattern are probably due to deficient sampling efforts, as might be the case in Colombia, Guyana, Bolivia, Peru, and southern continental Central America.

Fig. 1.

A: Consensus tree of 6 most parsimonious trees resulting from equal weighting analysis, under acctran optimization (length = 74 steps, CI = 0.51, RI = 0.57). B: Single most parsimonious tree resulting from successive weighting analysis, under acctran optimization (length = 68 steps, CI = 0.55, RI = 0.64). Black circles represent unique changes; white circles represent multiple changes. Bootstrap values and Bremer support values are shown inside rectangles below branches, separated by “/”.

Although the origin of Heterostylum. and the early diversification of Bombyliini lineages are not elucidated, a separate examination of the phylogenetic and distributional pattern of Heterostylum can add valuable data about the spatial history of the tribe. The distributional pattern of Heterostylum shows a generic distribution congruent with the phylogenetic pattern, as the basalmost dichotomy diverging to Nearctic and Neotropical clades (Fig. 1B) is congruent with the distributional disjunction between the Nearctic and Neotropical clades (Figs. 2 and 3). Although in close geographical proximity, the Nearctic and Neotropical clades do not overlap each other. The disjunction of Heterostylum. is spatially coincident with the boundaries of MTZ. The biogeographical scenario presented in Fig. 3 is suggested based upon the phylogenetic hypothesis proposed here (Fig. 1B). Yet, in our analysis the monophyly of clade B had no Bremer support, whereas clade A had bootstrap (51%) and Bremer supports (2). Furthermore, the ‘Nearctic’ definition is used here in the broad sense, as the Nearctic clade extends over areas within the Nearctic region and also southward through the MTZ (see Morrone 2004, 2006).

A number of biogeographers have been focusing their studies on the MTZ, correlating and explaining its biota with both Nearctic and Neotropical influences. Croizat (1958) suggested a node in southern Mexico as a major node regarding patterns on the intercontinental scales. Halffter (1964, 1987, and others) pioneered biogeographical studies on insects in the MTZ. In fact, Gonzalo Halffter was the most experienced authority on the MTZ history and biogeography. He regarded the complexity of MTZ as potentially responsible for the scarcity of biogeographical studies in that area (Halffter 1964). The Nearctic range of Heterostylum is congruent with the insect distributional patterns pointed out by Halffter (1987), wherein the species occurring north of the Transmexican Volcanic Belt have more affinities with the Nearctic fauna than with the Neotropical one.

What geological or biogeographical event could be associated with the basal divergence of Heterostylum (Figs 2 and 3)? The Transmexican Volcanic Belt has a recent origin, since its formation began during the Oligocene (Halffter 1987). Perhaps another ancient event in the Caribbean (and related to the formation of the Caribbean Plate) could be involved. Alternatively to the traditional northsouth Oligocene division across the Transmexican Volcanic Belt, Escalante et al. (2007) suggested the occurrence of an earlier Caribbean event during the Paleocene dividing the Mexican biota through an east-west division. These authors suggested a biogeographical relationship of eastern Mexico with the early supercontinent Gondwana, while its relationship with the adjacent areas (Central America and Caribbean) was not precisely discussed. The geological event argued as responsible for generating that pattern, the eastward movement of the Caribbean Plate, took place in the Paleocene (Escalante et al. 2007) and was related to the formation of the Proto-Greater Antilles during the Late Cretaceous. Nevertheless, we regard as equally valid the association of the east-west division of Mesoamerica with the early formation of the Proto-Greater Antilles. The Cretaceous Volcanic Arc formed during the Late Cretaceous and was deformed dynamically across the Paleogene while firmly accompanying the eastward movement of the Caribbean plate (Iturralde-Vinent & MacPhee 1999). The age of this geological event in dating the basal divergence of Heterostylum is congruent with the molecular dating of the subfamily Bombyliinae inferred by Wiegmann et al. (2003) as the Early Cretaceous (125 Mya).

In this context, the early formation of the Caribbean islands, derived from the Cretaceous-Paleogene Volcanic Arc, could explain the current distribution pattern of Heterostylum species throughout the Mesoamerican areas, with regard to both presence and absence of species. Heterostylum is highly diversified through the Caribbean islands, nearly absent on the Central American continental landmass, and reasonably diversified in the southern Nearctic (Mexico), but with no overlap between the Caribbean/Central American and the Nearctic species. The continuous eastward movement of the Cretaceous-Paleogene Volcanic Arc, since its formation in the Late Cretaceous and its dynamic conformation during the Paleogene until the current Caribbean was consolidated, paralleled the eastward movement of the Centroamerican Arc, a neighboring western plate bearing the early landmasses that formed the present continental region of Central America (Iturralde-Vinent & MacPhee 1999).

The basal diversification of the Neotropical clades C and D indicates a distributional fragmentation from the Caribbean areas (Figs. 4 and 5, see also Fig. 2). All species in clade C have a Caribbean distribution, and although most of the species have a widespread distribution in South American semiarid areas, this clade has not diversified outside the Caribbean (Fig. 2), i.e., there is no endemic species outside the Caribbean. With regard to clade D, the 2 species closest to the basal node (H. haemorrhoicum and H. rufum) occur in the Caribbean (Figs. 2 and 5). Similarly to some species in clade C, H. rufum, occurs in the Caribbean but is also widespread in South America, ranging through the Amazonia, Cerrado, Caatinga, to the Atlantic domains. However, unlike clade C, this other Neotropical clade is distinctive in having species occurring in the Atlantic Forest (Fig. 2). Both H. rufum and H. maculipennis occur in the Atlantic Forest, although only the latter actually occurs in humid environments of the Atlantic Forest (see discussion below). Again, unlike clade C, this clade has diversified into South America, with 3 endemic species in Brazil.

Fig. 2.

Area cladogram for Heterostylum (Legends: NEA, Southern Nearctic; AND, Andean region; CAR, Caribbean subregion; AMA, Amazonian subregion; CER, Cerrado; CAA, Caatinga; ATL, Atlantic Forest (=Paraná subregion); CHA, Chaco; STZ, South American Transition Zone; MTZ, Mexican Transition Zone).

Alternative Biogeographic Scenarios

Besides the taxon-area cladograms (Fig. 2), we performed a DIVA analysis to test and refine the results obtained and discussed above. DIVA is an event-based analysis and can recognize biogeographical processes (vicariance, dispersal and dispersion) which possibly originated the current taxon distribution. The DIVA analysis performed using the phylogenetic hypothesis and distributional data of Heterostylum produced a similar interpretation but it also derived alternative scenarios (Figs. 6A–D), which will be discussed herein. In a relaxed analysis, not constraining the maximum number of areas of the ancestral distribution (maxareas = 7) (see methodology description above), DIVA resulted in a biogeographical scenario (Scenario 1, Fig. 6A) with ancient origin and diversification, wherein the ancestral distribution is inferred to be wide (= sum of all areas, ABCDEFG) with the occurrence of subsequent vicariant events. Sympatric distributions (overlapping) between terminal taxa and between clades were resolved by assuming at least 10 ad hoc events, however, the DIVA algorithm cannot correctly distinguish whether dispersion or dispersal occurred, i.e., simple range expansion or a true dispersal process. This correction can be made ‘manually’, post-algorithm. Moreover, DIVA may often produce some unrealistic optimizations for some nodes, as seen in Scenario 1 (Fig. 6A), when inferring the ancestral distribution BG (Caribbean + Atlantic Forest) for the node of terminals 9 to 12. This disjunct ancestral distribution is spurious. DIVA has some pitfalls and inherent problems, and many of them were raised and well discussed in Kondandaramaiah (2010). For this reason, one should be very careful when carrying out a DIVA analysis, especially in interpreting its results (the optimized scenarios).

Alternative (and more realistic) scenarios were optimized constraining maxareas on DIVA, as in Figs. 6B and 6C. Both scenarios applied maxareas = 4, and these 2 scenarios differ only in the ancestral distribution assumed for the basalmost node: ABC (Nearctic + Caribbean + Amazon) in Scenario 2, and AB (Nearctic + Caribbean) in Scenario 3. The former (Fig. 6B) proposes an ancient vicariant hypothesis for Heterostylum as in Scenario 1, although not completely widespread. In contrast, Scenario 3 proposes a northern origin for Heterostylum with a north-to-south dispersal - ist hypothesis. Both scenarios assume widespread distribution of the terminals by models involving events of dispersion (e.g., terminals 10 and 11 in Scenario 2) and dispersal (e.g., node of clade D in Scenario 2).

Interestingly, DIVA analyses did not infer any scenario with the basalmost distribution without including the Nearctic region. Meanwhile, another serious inherent problem is that DIVA is not able to infer any event in the basalmost node other than vicariance. This is to say that at this node, dispersion or dispersal events will never be inferred by DIVA analyses. (For that, there should be the addition of taxa intermittently joined to the root.) Hence, to regard such events inherently ignored by DIVA, we can manually include additional ad hoc events to the basalmost node, as in Scenario 4, where the ancestral distribution ABC would have been reached after a dispersal event (+A) with subsequent divergence into A and BC. Observe that this is the same as presented in Scenario 2, but contrasting in the explanation for the ancient distribution and the associated event (vicariance in Fig. 6B, dispersal in Fig. 6D). As a consequence, while Scenario 2 supports an ancient vicariant scenario with a widespread ancestral distribution, Scenario 4 supports a southern origin with a south-to-north dispersalist scenario. Did Heterostylum Species Diversify Mostly within Forested Areas?

Fig. 3.

Geographical distribution of the genus Heterostylum; with the disjunctional basal dichotomy separating the Nearctic and Neotropical clades (clades A and B, respectively).

The species distributions maps (Figs. 7 and 8) and the area cladogram (Fig. 2) provide interesting information about species distributions and endemicity. It is well known that bombyliids often inhabit drier environments, although this would not seem to be the case for Heterostylum species by a mere glance at the maps. Furthermore, it is quite curious that no endemic species occur in the semi-arid areas of South America east of the Andes, namely the Cerrado, Caatinga, and Chaco (although H. duocolor is known only from the type-locality in the Chaco). However, these superficial impressions are misleading.

Fig. 4.

Sequence of diversification of the Neotropical clade C.

Fig. 5.

Sequence of diversification of the Neotropical clade D.

Fig. 6.

Event-based analysis by DIVA. A: Scenario 1 (ancient widespread origin): biogeographic scenario using maxareas = 7, required 10 events of dispersion/dispersal. For the node 11+12 we arbitrarily chose the ancestral distribution CG (Amazon + Atlantic Forest) within the 7 options resulting for the node. B: Scenario 2 (ancient restrict origin): biogeographic scenario using maxareas = 4, required 13 events of dispersion/dispersal. The basalmost ancestral distribution occurring in ABC (Nearctic + Caribbean + Amazon). C: Scenario 3 (northern origin): same optimization as Scenario 2, but differing in the basalmost ancestral distribution restricted to AB (Nearctic + Caribbean). D: Scenario 4 (southern origin): same optimization as Scenario 2, but including an additional event in the cladogram root unlikely to be inferred by DIVA: the dispersal of the ancestral occurring from BC to A, then becoming ABC. Total events required is 14. (Legends: numbers refers to species of Heterostylum; letters refers to areas: A, Nearctic; B, Caribbean; C, Amazon forest; D, Cerrado; E, Caatinga; F, Chado; G, Atlantic forest. For Scenarios 2, 3 and 4, DIVA analyses optimized alternative distributions for some nodes; please see the Supplementary Material)

Heterostylum rufum is widespread in the Caribbean and Amazonian subregions, and Cerrado, Caatinga and Atlantic Forest (Fig. 7). Heterostylum rufum certainly occurs in the Amazonian subregion, but in the Amazon Forest this appearance is artifactual for some localities (see Appendices 3 and 4). Some of the recorded localities within the Amazonian subregion are the following: 1) Serra Norte (Pará), in southeastern Pará, is known to have “campos rupestres” vegetation (a kind of Cerrado); 2) Alter do Chao (Pará, Rio Tapajás) has sandy beaches along the Tapajós River; 3) Imperatriz (Maranháo) has “mata dos cocais” vegetation, a transitional ecosystem between the Amazon Forest and the semi-arid Caatinga in northeastern Brazil; and 4) Porto Platon (Amapá) is covered by Cerrado vegetation. Similarly, H. rufum occurs in the Atlantic Forest (=Paraná subregion) but, in this case, in both humid and semi-arid environments: 1) the locality Sertãozinho (São Paulo) is covered mainly by Cerrado; 2) Conceição da Barra (Espirito Santo) is a coastal locality covered by “restinga”, a semi-arid vegetation bordering beaches; and 3) Marília (São Paulo) has semi-deciduous forest vegetation.

Fig. 7.

Geographical distribution of Heterostylum species: H. duocolor, H. evenhuisi, H. ferrugineum, H. hirsutum, H. maculipennis, and H. rufum.

Fig. 8.

Geographical distribution of Heterostylum species: H. croceum, H. deani, H. engelhardtii, H. haemorrhoicum, H. helvolum, H. pallipes and H. robustum.

The occurrence of H. ferrugineum in the Amazonian subregion is based on a record from Boa Vista (Roraima), which is covered by Cerrado vegetation. The remaining records are in the Caribbean and in the semi-arid diagonal (Cerrado + Caatinga + Chaco). In the Caatinga, the species is widespread, reaching the northeastern Brazilian coast (Ceará) and also extending up to the transition between the Caatinga and Atlantic Forest (Santa Rita-Bahia). Its absence from the Amazon Forest is conspicuous, such as in the sister-clade H. hirsutum + H. pallipes.

The records known for H. hirsutum (Fig. 7) are located separately in 2 disjunct areas (as far as we know it): one north, on the coastal regions of Venezuela and Colombia, and another south, on northwestern Argentinean Andes and southwestern Brazilian Chacoan area. Besides these confirmed records, the type-locality is mentioned only as Brazil (Cunha et al. 2007). It is also recorded in Paraguay, but unfortunately, this occurrence was not confirmed during our search (by examination of collections and literature review). Based on the reliable records, the known and confirmed distribution of this species is restricted to the Caribbean subregion, Chacoan subregion, and the South American Transition Zone (STZ).

Only one species of Heterostylum, H. duocolor, occurs exclusively in the Chacoan subregion, with a single locality so far recorded in the Argentinean Patagonia (Córdoba). This species was not included in the cladistic analysis because of the unavailability of material for study (the type is lost, and it has not been recognized by subsequent authors).

On the other hand, H. maculipennis is actually endemic to humid environments of the Atlantic Forest (e.g., the localities of Nova Friburgo and Itatiaia in the state of Rio de Janeiro, and Nova Teutonia in Santa Catarina), although it also occurs in coastal areas with comparatively dry vegetation (e.g., the localities of Praia Grande, state of São Paulo, and Angra dos Reis, state of Rio de Janeiro).