Mitochondrial DNA variation and GIS analysis confirm a secondary origin of geographical variation in the bushcricket Ephippiger ephippiger (Orthoptera: Tettigonioidea), and resurrect two subspecies

Introduction

The study of geographical variation and systematics has played a major role in the development of theories of speciation. Geographic variation can arise in a number of ways but, following from Mayr (1942), two major causes have been distinguished. Primary variation arises in situ because of contemporary or recent selection and drift. In contrast, secondary variation arises from historical vicariant events. Hewitt and colleagues have argued that most major hybrid zones in Europe reflect divergence during Quaternary ice ages (e.g. Barton & Hewitt 1985), with periodic expansion and contraction cycles accentuating and accumulating genetic differences (Hewitt 1988, 1989, 1996, 1999; see also Remington 1968; Taberlet et al. 1998).

Despite the common assumption that most hybrid zones result from secondary contact, steep coincident clines in organisms can have a primary origin, having arisen in response to exogenous selection across ecotones (Endler 1977; Harrison 1990). It is not possible to clearly distinguish primary or secondary origins of hybrid zones from an examination of their geographical context, because many areas of secondary contact are likely to occur across ecotones (Hewitt 1989). Also, theory suggests that the characteristics of primary and secondary clines will be identical (Endler 1977; Barton & Hewitt 1985; but see Durrett et al. 2000). Distinguishing primary and secondary origins of clines and hybrid zones remains a major, and relatively poorly met, aim of studies of geographical variation (Thorpe 1984; Harrison 1990, 1993; Willett et al. 1997; Schilthuizen 2000).

The bushcricket (katydid) Ephippigerephippiger (Feibig 1784) is a flightless tettigoniid with unusual patterns of geographical variation and, as a result, has been subject to a number of taxonomic surveys and revisions. The species is patchily distributed and highly variable in morphology and behaviour. This is most pronounced in southern France, from the Pyrenees to the Alps, where a number of species and subspecies have been described. Harz (1969), a standard taxonomic reference for European tettigoniids, recognized three species within the group under consideration here. E. ephippiger is a medium sized, predominantly green bushcricket with a monosyllabic calling song found in central Europe, with the subspecies E. e. vitium (Serville 1831) in western Europe and E. e.ephippiger (Fiebig 1784) in the east (an additional five subspecies have been described, Hartley & Warne 1984). E. cunii (Bolivar 1877) is small, almost black, with a polysyllabic calling song, found in Catalonia. E. cruciger (Fieber 1853, syn. E. bitterensis) is large, grey or green, usually with a melanized sulcus on the pronotum and a mono- or polysyllabic calling song, found around the Languedoc (see Duijm & Oudman 1983; Kidd & Ritichie 2000, for distribution maps). Recently, studies have reduced these forms in taxonomic status. Duijm & Oudman (1983), Duijm et al. (1983) and Hartley & Warne (1984), after analyses of morphology and copulation compatabilities, concluded that only the cunii form was valid (as a subspecies), and combined E. e. cruciger, E. e. vitium and E. e. ephippiger into a single highly variable ‘superspecies’ which they termed E. ephippiger diurnus Kruseman (referred to here as E. ephippiger for simplicity). Duijm and colleagues (Oudman et al. 1989, 1990; Duijm 1990) further questioned whether cunii was valid even as a subspecies. They found low genetic divergence between forms (Nei’s D, from four allozyme loci, was around 0.03 between forms). They also examined the geographical pattern of clines for the morphological traits used to define the various forms. Visual inspection implied the clines were usually noncoincident geographically and extremely wide. Some were 50–100 km wide, which compares with an average cline width for similar traits of 5–10 km in the Pyrenean hybrid zone of Chorthippus parallelus (Hewitt 1989; Butlin 1998), an organism which probably has a higher dispersal than E. ephippiger. Oudman et al. (1990) concluded that the geographical variation within E. ephippiger was unlikely to reflect secondary contact: ‘We see the E. ephippiger complex as a subspecies …  with sufficient gene flow to restrict differentiation. In the south of France the species has met conditions that have led to greater differentiation. This has on the one hand the character of shallow clines, perhaps along environmental gradients and probably adaptive, and on the other hand a chance character as the result of genetic drift and random dispersal effects.’Oudman et al. (1990) also concluded that no valid subspecies could be distinguished, the local forms being ‘without taxonomic status’. This interpretation has been adopted by other authorities (Ragge & Reynolds 1998; p. 71).

We have re-examined E. ephippiger from southern France from two perspectives. Geographical Information Systems (GIS) facilitate the detailed examination of spatially patterned data. We employed GIS to examine concordance between interpolated trait clines and ecotones, to ask if the broad, nonconcordant clines are coincident with ecotones, as would be expected if they result from primary selection. Additionally, GIS has been used to try to identify clusters in the geographical variability, which might reflect secondary origins, separate from or additional to the primary variation. If geographical clusters truly reflect secondary subdivisions within the species, we would expect these to be supported by similar patterns in genetic differentiation. A study of random amplified polymorphic DNA (RAPD) markers broadly separated the cunii form from cruciger and vitium (Ritchie et al. 1997), but this survey was geographically limited. Here we report restriction fragment length polymorphism (RFLP) analysis of mitochondrial DNA (mtDNA) using more extensive samples. Partial Mantel test analyses (Thorpe 1996) of matrices of genetic distance compared with geographical distance, ‘environmental distance’ and potential ice age refugial origins imply that most genetic variation reflects refugia, though significant partitioning is also associated with environmental variation.

Background: Geographical Information Systems analysis of geographical variation inE. ephippiger

Full details of our GIS analysis of geographical variation in E. ephippiger are available elsewhere (Kidd & Ritchie 2000, 2001). Briefly, data was compiled on geographical variation in morphological, allozyme and behavioural traits, as well as abiotic environmental parameters. Individual and multivariate trait surfaces were interpolated, compared with environmental data, and examined for covariation and concordance. Environmental features explained little other than body size of the organism. In contrast to the interpretation of these data by Oudman et al. (1990), some concordance between patches was observed in the interpolated surfaces, and fairly sharp, coincident transitions in some traits were detected. Figure 1 indicates the centre of a relatively steep multivariate cline, which broadly follows the path of the Aude river. Multivariate discriminant analyses (with resampled null models) confirmed that the regions either side of the cline represent significant clusters within a complex pattern of geographical variability. Kidd & Ritchie (2000) concluded that the cline represented a ‘cryptic hybrid zone’ where some traits, notably genitalia, probably reflected secondary contact between forms originating from western (Iberian) and an unidentified eastern refugia. This was broadly but imperfectly coincident with the vitium and cunii forms. The cruciger form was more difficult to fit into this reconstruction. Some traits, notably the frequency of some allozymes and the third principal component, broadly matched the distribution of crickets identified as cruciger, but this was around both sides of the eastern end of the potential hybrid zone, near Narbonne (around the mouth of the Aude). Kidd & Ritchie (2000) suggested three possibilities; cruciger was not valid, cruciger was a hybrid form, or cruciger was from a distinct western refugium.

Materials and methods

Results

mtDNA phylogeography

A neighbour-joining tree based on the mtDNA RFLPs is not well resolved (Fig. 2). This is probably not surprising given the high number of samples from regions of hybridization. Despite this, some clades are well supported. The deepest division within Ephippiger ephippiger almost perfectly separates individuals identified as vitium from the remainder of the samples (Fig. 2). This division fairly closely corresponds with the location of the putative Kidd & Ritchie contact zone (compare Fig. 3). The only samples in the ‘incorrect’ clades are one from Lodeve, which is from the area of transition between the vitium and other forms, and one from the Col du Pourtalet. The ‘pure’ vitium clade is reasonably well supported.

There are other striking features of this tree. The deepest node within the nonvitium samples separate off a branch containing cunii samples from the Segre river valley (the Cerdagne region of the Pyrenees) from the remainder of the cunii and cruciger samples. A second node separates off cunii samples from the Têt river valley (the Conflent region). The remainder of the tree includes samples from a third Pyrenean valley, that of the river Aude (the Capcir region) and samples from the Languedoc. This branch, therefore, contains samples identified as cunii and nearly all the cruciger samples. The Aude valley is the northernmost of the eastern Pyrenean valleys sampled here (north of the main watershed), and leads down into the Languedoc.

Discussion

The large morphological and behavioural variability within Ephippiger ephippiger has lead to a confusing series of taxonomic classifications and revisions (Harz 1969; Hartley & Warne 1984). Previous studies concluded that none of the described species or subspecies were taxonomically valid, and that the variation reflected primary differentiation (Oudman et al. 1989, 1990; Duijm 1990; but see Grandcolas 1987). Our current analyses lead us to conclude that the major source of variation is historical isolation during recent glaciations, followed by secondary contact between eastern and western forms, as has commonly shaped subspecific variation in European flora and fauna (Taberlet et al. 1998; Hewitt 1999; Vogel et al. 1999). However, in partial Mantel analyses, genetic variability is correlated with environmental contrasts even after allowing for the influence of refugia, and significant environmental influences are also detectable on body size (Kidd & Ritchie 2000). Both secondary contact and environmental selection have therefore influenced the pattern of genetic variability, though secondary contact has had the greater effect.

Evidence supporting a vicariant origin of the differentiation comes from the consistently high Mantel regression coefficients associated with the refugial contrast matrices (Table 1), which is also reflected in the deepest node within the mtDNA relationships separating the samples into eastern and western forms (2, 3). The multivariate transition identified by Kidd & Ritchie (2000, 2001) lies within the same geographical region as the division in mtDNA, but the two are not exactly coincident (1, 3). The division in mtDNA occurs to the north of the multivariate transition, somewhere south of Lodeve and west of Montpellier. This lack of coincidence could be due to primary selection patterning differential introgression of traits (see below), although cytoplasmic DNA introgression through hybrid zones is not uncommon (Shaw et al. 1990; Whittemore & Schaal 1991).

Vitium, therefore, represents a form that probably originates from a refugium to the east of the study area. Many eastern European subspecies are thought to have originated in the Balkans or Caucasus (Hewitt 1996, 1999; Taberlet et al. 1998), or perhaps in the Carpathians or Ural mountains (Lagercrantz & Ryman 1990). Vitium presumably occupies most of central Europe now, and extends as far west as the central Pyrenees. One sample from here (from site 2) falls within the western clade of the mtDNA tree (Fig. 3). E. ephippiger from this area has been claimed to represent yet another subspecies, E. moralesagacinoi (Harz 1969), but it seems more likely that there is another rapid transition between the eastern and western forms here (Oudman et al. 1990).

The western form, comprising nearly all the samples identified as cunii or cruciger, has geographically structured mtDNA variation. Samples from around the Pyrenean watershed form distinct groups, with one from each of the three main valleys (Cerdagne, Capcir and Conflent). This provides a striking contrast with previous conclusions of extensive gene flow based on allozyme, morphological and RAPD markers (Oudman et al. 1990; Ritchie et al. 1997). The two deepest nodes within the western form separate off individuals from Cerdagne and Conflent. The remainder of the samples comprise a group including the northernmost cunii and almost all the cruciger samples, and geographically corresponds to Capcir and the Languedoc. Clearly, there is evidence of a shared mtDNA history to these samples, despite their extensive range (over 150 km). The distinctness of the cunii from different valleys, and the presence of cruciger within this clade, disrupts cunii as a genetic form. The 3-refugium matrix in the Mantel test was designed to test the distinctness of cruciger, and it always provided the greatest partial regression coefficient. So there is some evidence to support the existence of a distinct cruciger genotype, but the historical processes underlying the relationship between cunii and cruciger are obscure.

We consider there to be two possible explanations whereby cruciger may represent a distinct form. The first is that cruciger was confined to a third refugium east of the Pyrenees, possibly around the Mediterranean coast. Cruciger would then have been subject to hybridization, first with the northernmost samples of cunii expanding from a Spanish refugium, then latterly with vitium expanding from an eastern refugium. Cruciger could be facing extinction via introgression due to this pincer movement. Most reconstructions of Quaternary glacial refugia in southern Europe only include major ones in southern Spain, Italy and the Balkans (Taberlet et al. 1998; Hewitt 1999). Huntley & Birks 1983 (see also Vogel et al. 1999) suggest that there was another refugium around the Alpes Maritimes, and there is a suggestion of Pleistocene coastal Mediterranean refugia including one around the mouth of the Rhône (Comes & Abbott 1998), but no other phylogenetic studies support the presence of distinct genotypes from around this area (Comes & Kadereit 1998).

The second scenario is that cunii and cruciger are from partially distinct southern refugia, both in Iberia, but have a long history of hybridization and introgression during repeated expansion/contraction cycles, with the Languedoc being most recently invaded by a form already having undergone hybridization with northern cunii. Repeated hybridization could explain the imprecision of many of the taxonomic traits. This scenario is perhaps more likely given a very high level of Iberian endemism within the ephippigerinae in general (Harz 1969; Gangwere & Morales Agacino 1970). Also, several other studies have found evidence for multiple refugia or variation within Iberia (Cooper et al. 1995; Comes & Abbott 2000).

Although the main genetic divisions reflect secondary contact, we also detect evidence supporting primary selection influencing Ephippiger. This is most apparent in body size and body ratios (Kidd & Ritchie 2000). However, the partial Mantel tests show that, after allowing for refugia, environmental contrasts are still correlated with variation in mtDNA. Environmental conditions are unlikely to select directly on mtDNA, it is more likely that indirect selection influences the introgression of genotypes including mtDNA. Cunii is probably more cold adapted or robust to inclement environmental conditions than vitium. Hybrid zones with a broad front are likely to act as semi-permeable barriers, with recombination separating out genotypes under direct selection from linked genetic variation (Barton & Gale 1993), but patchy, mosaic or intermittent episodes of hybridization may produce more opportunity for persistence of associations across different types of genetic markers (Harrison & Rand 1989; Rieseberg & Wendel 1993; Arnold 1997).

Our studies provide a good example of the value of an integrative approach, involving techniques from geographical analysis as well as spatial statistical analysis and molecular phylogeography, in unravelling the pattern and historical processes contributing to geographical variation. Conventional population genetic methods for estimating substructure within species allow little more than rejection of the hypotheses that populations are panmictic or show isolation by distance (Neigel 1997; Bossart & Prowell 1998; Whitlock & McCauley 1998). Techniques are required which allow detection of more realistic historical scenarios. GIS allows independent identification of clusters in population structure, and phylogeographic and Mantel tests, or hierarchical analysis of F_ST or its derivatives can assess the validity of these externally derived predictors of clustering (Kidd & Ritchie 2001).

These analyses have resurrected two forms of the bushcricket E. ephippiger from a taxonomic hinterland, and we conclude that vitium is a valid subspecies. Whether cruciger and cunii warrant separate subspecies status is more problematical and will rely on further analyses. The Mantel tests support their distinction, and several of the trait surfaces of Kidd & Ritchie (2000) show clusters around Narbonne. In contrast, the mtDNA relationships imply that cruciger mtDNA is a derived form of the cunii genotype, though this might be confused by hybridization between the northern cunii and cruciger forms. It is difficult to predict the effect of hybridization on phylogenetic trees (Arnold 1997).

Finally, what lessons might E. ephippiger have for systematics? We think it appropriate to designate at least two forms of E. ephippiger as subspecies because they hybridize in nature, yet are distinguishable on multiple traits throughout the majority of their ranges. Some species definitions, particularly the Phylogenetic (Cracraft 1989) and, depending on details of hybrid populations, the Cluster species definitions (Mallet 1995) might justify the elevation of these forms to species status. We do not accept this due to the obvious hybridization and lack of obvious incompatibilities between the forms. However, does the systematic term ‘subspecies’ have genuine biological validity? Mayr (1942, p. 106) thought not: ‘The taxonomist is an orderly person whose task it is to assign every specimen to a definite category (or museum drawer!). This necessary process of pigeonholing has led to the erroneous belief … that subspecies are clear-cut units which can easily be separated from one another.… But subspecies intergrade almost unnoticeably in nearly all the cases in which there is distributional continuity’. This reflects Mayr’s (possibly essentialist, Mallet et al. 1998) view that species were the only ‘true’ biological categories within organisms. The increasing body of evidence that hybrid zones result from historical vicariance episodes between forms that have accumulated multiple concordant genetic differences over a time scale possibly stretching as far back as the Pliocene, must support a valid general concept of the subspecies (Avise & Ball 1990).

Acknowledgements

Matthijs Duijm and Leendert Oudman provided very generous help during the initial stages of this project. Bill Black, Hans Peter Comes, Colin Hartley, and Roger Thorpe provided advice, and several people helped with fieldwork. Klaus-Gerhardt Heller kindly supplied the specimen of E. perforatus, and Roger Butlin and Jeff Graves gave advice on the manuscript. The work was funded by the NERC, UK, via a research fellowship and grant to MGR.

Mike Ritchie completed his PhD with Godfrey Hewitt, studying the Chorthippus parallelus hybrid zone in the Pyrenees. There he encountered Ephippiger ephippiger, and carried out a postdoctoral fellowship in Hewitt’s laboratory studying the behavioural variation of this species. Dave Kidd has collaborated with MGR in applying GIS methods to the study of geographical variation in E. ephippiger. Jenny Gleason studies molecular evolution and quantitative genetics, and collaborated with MGR to develop the mtDNA RFLP analysis of E. ephippiger.