Introduction

(with respect to New World spine-throat grasshoppers): “…those relationships are unlikely to be resolved through the use of gross morphological characters such as those employed to date. Almost certainly it will be necessary to use molecular traits to group them properly.” — Perez-Gelabert & Otte (2000)

The North American component of the grasshopper subfamily Melanoplinae is numerically dominated by the genus Melanoplus Stål. Included among the over 230 species (Vickery & Kevan 1983, Otte 1995) are species such as the economically important migratory grasshopper M. sanguinipes (F.) and the now-extinct Rocky Mountain grasshopper, M. spretus (Walsh). Species are usually identified by distinctive male genitalic characters which, according to Knowles and Otte (2000), may have evolved rapidly by sexual selection, possibly contributing to an explosive diversification within the genus during the past 4 million years (My). Various schemes, not totally concordant, have been proposed for assigning taxa to species ‘series’ or ‘species groups’ (Table 1). Hebard defined 45 species-groups (Gurney 1960), but unfortunately these listings were never published in their entirety (Hebard 1917, 1919).

This paper, as part of our ongoing efforts to understand relationships within Melanoplus, expands on a previous molecular analysis (Chapco et al. 1999) by including additional taxa and genes. This study also assesses the extent of polymorphism within species and examines its possible confounding effects with respect to achieving phylogenetic resolution at lower taxonomic levels.

Materials and methods

Sequences were obtained from 64 individuals distributed among 32 Melanoplus species and members of two monotypic genera with suspected close connections to Melanoplus: Phoetaliotes nebrascensis (Thomas) and Bohemanella frigida (Boheman). Schistocerca gregaria (Forskål) and Locusta migratoria (L.) served as outgroup taxa.

Total DNA was extracted according to Chapco et al. (1992) and Philips & Simon (1995). Regions of four mitochondrial genes (cytb, COII, ND2 and COI) were amplified and sequenced, following procedures described in Litzenberger & Chapco (2001a,b).

The resulting sequences were easily aligned by visual inspection and imported into MacClade (Maddison & Maddison 1992). Phylogenetic relationships were inferred using maximum parsimony (MP), weighted parsimony (wMP) [according to Farris' (1969) iterative reweighting scheme], and the neighbor-joining (NJ) method (Saitou & Nei 1987), all available in the software package, PAUP, version 4.0b8 (Swofford 2001). For the NJ analysis, distances were estimated using the K2 transformation (Kimura 1980). [Application of Posada and Crandall's (1998) MODEL-TEST program revealed that the model “TIM + I + G” best fitted the data; however, the latter returned the same bootstrap topology as that based on K2.] For all analyses, the four sequences were treated as a combined unit, a maneuver that, as in all our previous studies (Chapco et al. 2001; Litzenberger & Chapco 2001a, b), resulted in improved resolution and bootstrap support when compared to the outcomes of single gene studies.

Levels of support for derived relationships were estimated through 1000 bootstrap replicates. Within species, nucleotide diversities were estimated using MEGA version 2.1 (Kumar et al. 2001).

Results and discussion

Approximately 1716 bases of the cytb, COII, COI and ND2 regions of mtDNA were sequenced (Appendix 1). For 10 museum specimens certain regions proved refractory, either to amplification or sequencing. Missing data were treated as unknown in parsimony analyses and ignored in pairwise comparisons for distance analyses. Across all genes, 680 sites were variable and 386 were phylogenetically informative. Maximum parsimony yielded 1435 equally parsimonious trees, each of step length 1438 and consistency index 39.2%: weighted parsimony recovered 240 equally parsimonious trees of length 316.4, with a somewhat higher consistency index value of 59.1%. There was general agreement between the MP and wMP topologies. The wMP majority-rule consensus tree is reproduced in Fig. 1, with MP and wMP bootstrap values placed on those branches that have at least 50% support using wMP, the latter method generally yielding higher values. Values (< 50%) for clades lacking bootstrap support are not shown. There is less agreement with the NJ tree for which resolution is generally poor (Fig. 2). Discussion will primarily refer to associations in Fig. 1, but where necessary, attention will be drawn to the NJ phylogram.

Five large “clades”, labeled A to E for purposes of discussion, were recovered using parsimony (six clades: A, B, C′ to F′ using NJ). M. lakinus formed a separate clade basally to the above groups. While bootstrap support for these clusters varies from moderate to poor, the branching order among them is indeterminate. An examination of the NJ phylogram (Fig. 2) shows that the internodes separating these groups are quite short compared to many of the proximal branches. A similar phenomenon emerged in Knowles and Otte's (2000) study of a few species groups comprising mostly montane species of Melanoplus. Their interpretation was that evolution took place very rapidly within very short “opportunistic” periods, engendered by the numerous glaciation/deglaciation cycles during the Pleistocene. Statistically, their data suggest that diversification occurred sequentially, albeit rapidly, rather than in a single explosive radiation (Knowles 2000). We concur. The order, however, is not clearly resolved. Thus, we compared the likelihood of trees depicting other branching orders with the tree in Fig. 1 (represented as ABCDE: it has the highest likelihood) by the Kishino-Hasegawa (1989) test, and discovered that there are several trees (e.g. ACBDE) with likelihoods not significantly lower than that of ABCDE. However, not all permutations are statistically equally likely. Several trees had significantly lower likelihoods. For example, most trees in which clade E is displaced from its lower position to a more internal one (e.g., ABECD) are significantly less likely. Discovering branching order may in fact be difficult to attain because the number of phylogenetically meaningful nucleotide substitutions is expected to be small within the periods represented by the short internode lengths (Hoelzer & Melnick 1994). Sexual selection, in this case involving male genitalic traits, is proposed as a possible mechanism underlying rapid evolutionary change (see references in Knowles & Otte 2000).

It would seem that, given our parallel results, which are based on a wider sampling of species groups and taxa with wider distributions, the speculations of Knowles and Otte (2000) on the evolutionary significance of the Pleistocene and the role of sexual selection, are not restricted to those taxa occupying the “sky islands” and western North America, but might also apply to all species of the genus. Assuming that mitochondrial DNA changes are approximately clock-like, application of the molecular clock (calibrated at a rate of 1% per My per lineage, see references in Chapco et al. 2001) provides an estimate of 4.8 My (based on the most disparate pair of species, devastator and kennicotti) for the age of the genus: in rough agreement with the 3.6 My value of Knowles and Otte (2000). (If we use their rate of 1.15 %, then our estimate reduces to 4.1 My.)

While resolution at the higher levels is poor, this is generally not the case for many taxa within the major clusters. Phylogenetic relationships suggested by the molecular data are, in some cases, supportive of groupings proposed by early workers. There are, however, some notable deviations:

1) The clade (A), consisting of sanguinipes, devastator, bruneri, borealis, femurrubrum, fasciatus, dawsoni, gaspesiensis and madeleineae, is supported in 72% of bootstrap replicates (81% for MP). Helfer (1987) and Gurney & Brooks (1959) assigned the first four species to the species group Mexicanus. Interspersed among them are the remaining five species, the first three of which are assigned respectively to the species groups Femurrubrum, Fasciatus and Dawsoni. Since this is the largest clade, it will be dealt with in subsections.

1a) Sanguinipes and femurrubrum: The connection between this pair is extremely close (mean K2 distance, d = 1.15 ± 0.14 %) with four and three specimens of each species intermingled respectively (the implications of which are discussed later). A morphological trait that is used to define the Mexicanus group, a pronounced mesosternal hump (Gurney & Brooks 1959), may be of dubious phylogenetic value, given this result.

1b) Devastator and sanguinipes: In the previous study (Chapco et al. 1999) these two species, although close phylogenetically, were not as close as sanguinipes and femurrubrum, a somewhat surprising result given the ability of devastator and sanguinipes to hybridize (Orr et al. 1994). Our examination of additional specimens provides a more comprehensive picture. In particular, we included two specimens of each of three recognized subspecies (Gurney & Brooks 1959) of M. sanguinipes: sanguinipes, vulturnus and defectus, respectively occuring in the northern US and Canada, the US southeast and the US southwest. Subspecies sanguinipes had been used in our 1999 paper. In this study, subspecies sanguinipes clusters with subspecies vulturnus and femurrubrum, whereas subspecies defectus clusters with two specimens of devastator, a species that also occurs in the US southwest. Two other specimens of devastator occupy unresolved positions within clade A.

1c) Gaspesiensis, madeleineae and borealis: Both gaspesiensis and madeleineae are thought to have arisen in refugia during the Wisconsin glaciation period, or possibly in earlier glacial times, from ancestors of the now-widespread M. borealis (Vickery 1987, 1989). These interrelationships were studied and discussed fully in Chapco & Litzenberger (2002) and the addition of more taxa here leaves their relative positions unchanged.

1d) Dawsoni: Both Helfer (1987) and Scudder (1898) placed this species, along with gladstoni, in the species group/series Dawsoni, although Scudder regarded the latter as a hodgepodge of taxa. In our previous study, dawsoni and gladstoni were far removed from each other; they remain so with the inclusion of additional genes.

1e) Fasciatus and borealis: Scudder (1898) placed these species in the series Fasciatus, described by him as “not very homogeneous”. Helfer assigned fasciatus and borealis to separate groups. Given the poor resolution within some parts of clade A, it is difficult to ascertain whether there is a direct linkage between them.

2) Recovered within the moderately supported (79%) clade B are packardii, foedus, angustipennis, gladstoni and aspasmus. Species packardii and foedus are difficult to distinguish morphologically, except for certain aedeagal features (Brooks 1958); both are assigned to the Packardii species group (Helfer 1987). Molecular support for connecting the two remains very strong (97%) with the inclusion of more sequence. As with sanguinipes and femurrubrum, there is an overlap of specimens. The remaining three species belong to the Angustipennis, Dawsoni and Marginatus species groups respectively (Table 1). The direct connection between the pair (packardii, foedus) and angustipennis, discovered earlier (Chapco et al. 1999), is sustained with the addition of more data. However, an association between gladstoni and aspasmus, which had not emerged previously, is indicated here, albeit with 68% bootstrap support (87% using NJ).

3) Clade C, consisting of bivittatus, keeleri, differentialis, calidus, franciscanus and littoralis, emerged in parsimony searches; however, bootstrap support was below 50%, a result which may be due to the absence of some sequences for the last three species (Appendix 1). Species bivittatus, keeleri, calidus and differentialis belong to four separate species groups or series (Table 1), whereas, to our knowledge, franciscanus and littoralis are unassigned. Some connections within clade C, however, are suggested in the older literature. Similarities between franciscanus and members of the Immunis species group, to which calidus belongs, were noted by Gurney (1960). Roberts (1942) pointed out common features between littoralis and differentialis, a linkage that is moderately supported (72%) by our molecular data. In the previous investigation (Chapco et al. 1999), differentialis was very strongly tied to bivittatus, but with the inclusion of more species the association, while still present, is more tenuous. Another pair, bivittatus and yarrowi, characterized by large body size, is united in the group Bivittatus (Table 1). Their relationship, nevertheless, has to be regarded as unknown, given the poor bootstrap support for clades C and E (the latter containing yarrowi). Interestingly, neighbor-joining does link yarrowi with franciscanus (80% bootstrap support) within clade C, but this may be an artifact reflecting the lack of complete sequence for franciscanus (Appendix 1).

4) Clade D, comprised of infantilis, alpinus, aridus, Phoetaliotes nebrascensis (Thomas) and scudderi, is weakly supported (55%); there are, however, some noteworthy internal connections supported by high bootstrap values. The previously recovered association (Chapco et al. 1999) involving alpinus and infantilis is preserved here with 100% bootstrap support, justifying their assignment to the same species group, Infantilis (Table 1). Scudder (1898) placed alpinus and infantilis, along with confusus and keeleri, in the series Collinus, a conglomerate not supported by our data. The connection between Phoetaliotes and aridus is extremely strong (99%) and yet there is nothing in the literature to suggest such a relationship or, indeed, that there should be an intimate association between Phoetaliotes and Melanoplus itself. If additional data were to sustain the internal position of Phoetaliotes, it would seem reasonable either to rename this species Melanoplus nebrascensis or to regard Melanoplus as a paraphyletic taxon. Since Phoetaliotes has only one distinctive feature, a disproportionately large head relative to the thorax (Scudder 1898, Helfer 1987), the former view is favored. The species scudderi is part of clade D, but its association is not strongly maintained.

5) Within weakly supported (59%) clade E, several interesting associations emerge. There is a moderately strong (83%) relationship between confusus and occidentalis, belonging respectively to their nominate species groups. Occidentalis also includes lakinus, but the latter is outside E (see below). There is an extremely tight affiliation (100 %) between marginatus and microtatus, consistent with their placement within the species group Marginatus (Rentz 1978). Marginatus includes aspasmus, but clearly the latter is far removed (Fig. 1). On the basis of relatively little differentiation of male genitalia, Marginatus is regarded as the most primitive of all Melanoplus species groups (Rentz 1978), a point that is discussed further below. Another highly supported cluster consists of Bohemanella frigida, kennicotti, and bowditchi. Over the past 50 y the first species has been assigned to the Holarctic tribe Podismini or, as M. frigidus Boheman, to the Nearctic tribe Melanoplini (Vickery 1987, 1989). A previous molecular investigation (Litzenberger & Chapco 2001a) presents and discusses evidence for the latter point of view. With the inclusion of more species of Melanoplus in this study, we were able to ascertain the relative position of frigidus within the genus and it would appear that its nearest relatives are kennicotti and bowditchi. The last two species belong to different species groups and frigidus, to our knowledge, has never been assigned to one. It is possible that bowditchi's involvement may be spurious since we were only able to obtain sequences for two genes (Appendix 1). Both frigida and kennicotti are boreal species with overlapping ranges (Vickery 1987); however, there is nothing else in the early writings that would have predicted a connection between them.

From what was stated above regarding the “primitiveness” of the species group Marginatus, one might expect the latter to be basal to all species of Melanoplus. Species of Marginatus, along with the other taxa in clade E are certainly basal to clades A to D, but with rather weak support (52%), although, as pointed out above, moving E from its present position significantly lowers the likelihood. In any case, basal to all species (with 87% support) is a species that is not part of Marginatus, M. lakinus (Scudder). Statistically, moving lakinus from its basal position to a more internal one significantly lowered the likelihood, except in one instance in which lakinus and cluster E were reversed. The p-level, however, was just above the 5% mark at 7%. In the study by Knowles and Otte (2000) lakinus also emerged as basal to Melanoplus, depending on the method of tree construction and choice of outgroup.

If present results are confirmed with the inclusion of additional taxa and other sequences, in particular those of nuclear genes, the validity of many species groups or series will have to be questioned. Monophyly of certain other species groups of Melanoplus was similarly disputed by Knowles and Otte (2000). Such taxonomic units may be useful from an identification point of view, but it should not be expected that they necessarily reflect phylogeny. In their molecular phylogenetic analysis of the order Orthoptera, Rowell and Flook (1998) pointed out that if a reliable molecular phylogeny exists, but is discordant with the outcome of morphological studies, the most likely explanation is that many anatomical traits relied upon by taxonomists are subject to convergent evolution. Assuming our analysis accurately mirrors phylogeny, then convergence could easily explain the joining together of, for example, bivittatus and yarrowi or the various members in Scudder's (1898) Collinus species series. We have already alluded to the possible role of sexual selection underlying changes in genitalic characters that may have accompanied (caused?) the massive radiation in the early history of Melanoplus. Presumably, these changes took place during the short internodes depicting bouts of rapid evolution (Fig. 2).

But what of other morphological traits, commonly used to identify species, that we suspect of having evolved convergently? If morphological changes and speciation are decoupled phenomena, as some would suggest (e.g., Larson 1989), then one reasonable scenario is that nongenitalic morphological traits evolved somewhere along the various lineages postradiation. If we consider, for instance, species pairs Bohemanella and kennicotti, gladstoni and aspasmus or aridus and Phoetaliotes in the NJ phylogram, it will be noticed that each pair is connected by relatively long branches which could allow sufficient opportunity for the accumulation of independent morphological changes. Such changes may have resulted in convergence (e.g., in species group Bivittatus) or simply autapomorphy (e.g., in Phoetaliotes). In contrast, others (e.g., Arnqvist & Thornhill 1998) have argued that genitalic and nongenitalic characters are correlated and influenced by a common array of genes. In this case we would have to conclude that at least some of the nongenitalic traits may have also evolved during the short bursts of speciation. These different viewpoints are admittedly based on different sets of organisms. There is, therefore, all the more reason to undertake detailed analyses of both genitalic and nongenitalic traits in Melanoplus, perhaps with the aid of scanning electron microscopy (Dakin 1987) and more modern multivariate techniques such as landmark-based morphometrics (Bookstein 1991). Traits could then be mapped onto a reliable phylogeny to ascertain where changes took place and whether the two types of characters are interrelated.

Table 2 lists, in increasing order, nucleotide diversities for those species with sample sizes > 1. Additional sequenced specimens are included to provide as large a sample as possible. Many factors can influence diversity levels, such as population size and time since divergence, modulated perhaps by selection and gene flow. Although M. sanguinipes is probably the most widely distributed species and, at times, an outbreak species (Vickery & Kevan 1983), its diversity is at the lower end of the range. In terms of time of origin, application of the molecular clock would suggest that, on average, the “alleles” for this species shared a common ancestor about 490,000 y ago. This relatively short time may possibly account for the absence of lineage sorting that accompanied the evolution of sanguinipes and femurrubrum from their common, presumably polymorphic, ancestor. Such “sorting out” may well have already occurred during events leading to the evolution of the monophyletic, somewhat older taxa, dawsoni, Bohemanella and confusus, which are two to three times as diverse as sanguinipes. Another monophyletic species is bivittatus, with a diversity only slightly larger than that of sanguinipes. It remains a possibility, however, that bivittatus may prove not to be monophyletic, with the inclusion of additional specimens of other species in clade C.

The species infantilis (the least diverse and monophyletic) and devastator (most diverse and not monophyletic) are somewhat enigmatic. A recent population increase might possibly account for the observed coalescence of specimens of M. infantilis, although this species has not, at least in agricultural times, exhibited outbreak tendencies. With respect to devastator, there would seem to be two groups. The first, consisting of devastator 1 and 2, clusters with subspecies defectus of M. sanguinipes (mean distance of 0.80 ± 0.16 %) and the second, consisting of devastator 3 and 4, is basal to all species, although unresolved, within the large cluster A. The large nucleotide diversity of devastator is mostly the result of the extraordinary branch length of devastator 3 and of the internode connecting the latter and devastator 4 to their root. This pair is clearly much older, whereas devastator 1 and 2 are more recently evolved, sharing a common ancestor with the two defectus subspecies of sanguinipes about 400,000 y ago. As with sanguinipes and femurrubrum, there may not have been sufficient time for lineage sorting. The situation involving packardii and foedus is similar, with three “alleles” of packardii having fairly old lineages and another sharing a common ancestor with foedus.

The North American southwest has been proposed as a major place of origin for the tribe Melanoplini (Rehn 1954, 1958; Vickery 1987), given the distribution of extant taxa. In this study, lakinus is a southwestern species and emerges basally to all Melanoplus. Within clades B and D, southwestern species aspasmus and aridus occupy basal positions. This may also be true of devastator (3 and 4) with respect to clade A and the pair (marginatus, microtatus) with respect to a small clade within E. If these indications are confirmed with further study, then we would have to conclude that there must have been several incursions from this region of North America.