Natural hybridization in heliconiine butterflies: the species boundary as a continuum

Background To understand speciation and the maintenance of taxa as separate entities, we need information about natural hybridization and gene flow among species. Results Interspecific hybrids occur regularly in Heliconius and Eueides (Lepidoptera: Nymphalidae) in the wild: 26–29% of the species of Heliconiina are involved, depending on species concept employed. Hybridization is, however, rare on a per-individual basis. For one well-studied case of species hybridizing in parapatric contact (Heliconius erato and H. himera), phenotypically detectable hybrids form around 10% of the population, but for species in sympatry hybrids usually form less than 0.05% of individuals. There is a roughly exponential decline with genetic distance in the numbers of natural hybrids in collections, both between and within species, suggesting a simple "exponential failure law" of compatibility as found in some prokaryotes. Conclusion Hybridization between species of Heliconius appears to be a natural phenomenon; there is no evidence that it has been enhanced by recent human habitat disturbance. In some well-studied cases, backcrossing occurs in the field and fertile backcrosses have been verified in insectaries, which indicates that introgression is likely, and recent molecular work shows that alleles at some but not all loci are exchanged between pairs of sympatric, hybridizing species. Molecular clock dating suggests that gene exchange may continue for more than 3 million years after speciation. In addition, one species, H. heurippa, appears to have formed as a result of hybrid speciation. Introgression may often contribute to adaptive evolution as well as sometimes to speciation itself, via hybrid speciation. Geographic races and species that coexist in sympatry therefore form part of a continuum in terms of hybridization rates or probability of gene flow. This finding concurs with the view that processes leading to speciation are continuous, rather than sudden, and that they are the same as those operating within species, rather than requiring special punctuated effects or complete allopatry. Although not qualitatively distinct from geographic races, nor "real" in terms of phylogenetic species concepts or the biological species concept, hybridizing species of Heliconius are stably distinct in sympatry, and remain useful groups for predicting morphological, ecological, behavioural and genetic characteristics.

shape, broad orange anal streak on the upperside forewing and the yellow costal streak on the hindwing all suggest isabella. The procula characteristics include rounded, indistinct forewing spots, hindwing yellow costal streak shortened to a pale basal spot (procula has a white spot here), intervenal submarginal hindwing spots on the underside that show up on the upperside through the wing tip, but not at the base, the intervenal submarginal underside hindwing spots that are rounded, and not so flat or as near the edge as in isabella, and the orange cell streak of the forewing underside that is mainly costal, rather than anal as in isabella. Finally, the male hybrid 7 from SE Brazil is clearly related to Eueides pavana, but has very orange forewings and lacks the prominent intervenal black rays on the hindwing. It seems fairly clear that it is a hybrid with E. vibilia vibilia, the males of which have orange forewing bars.
Unlike Heliconius, Eueides are drab and not favoured by collectors, and the similarity of their colour patterns may mean that hybridization in other species has been missed.
Their colour patterns also make it difficult to decide which species are parents without further information. We predict that hybridization within Eueides is rather commoner than current records demonstrate, because of the difficulty of detecting hybrids between such similar species.
Hybrids involving "silvaniform" Heliconius (hybrid nos.  Hybrids between and other Heliconius are readily recognized because their mimetic patterns become very disturbed. The yellow and brown ithomiine mimicry rings to which the silvaniforms belong are very different from the other crimson/orange and yellow heliconiine mimicry rings. Only two hybrids (nos. 21 & 22) have been found between any pair of ithomiine-mimicking silvaniform species, though this is likely an underestimate of the true level of hybridization between these similar appearing and closely related forms. In view of the confusion surrounding the species status of the silvaniforms, some of which are highly polymorphic, it would not be surprising if some of the many aberrant forms among silvaniforms [1] were not produced by additional cryptic hybridization.
The remainder of the hybrids in this group are either between the red-marked silvaniform subgroup species H. elevatus and H. besckei and ithomiine-mimicking silvaniforms (nos. [14][15][16][23][24][25][26][27][28], or between silvaniform species and H. melpomene (nos. [8][9][10][11][12][13][17][18][19][20]. Inferences about parentage of these specimens are problematic, because the hybrids appear to be extremely rare, so that only a single specimen is known for most of the forms. For hybrid nos. 18-22 we have after careful consideration accepted the judgment of Brown [1] who reported such specimens in his review of silvaniform systematics (and to which we have added similar forms). Hybrid 11 is somewhat different from these others; it is a hybrid between a silvaniform and one of the two local rayed species, elevatus or melpomene. Its very rounded wing shape, the shape of the forewing bar, and other characteristics strongly suggest melpomene rather than elevatus, although the latter is more closely related to silvaniforms ( Fig. 1; [2]).
Similarly, the general darkness and the faint pale spotting in the forewing apex suggest that H. numata aurora, rather than another species from the area, is the silvaniform parent. Two hybrids involving a silvaniform and H. elevatus (nos. [14][15] have among the more dubious parentages on our list; we have seen only photos of these specimens, but we are almost certain they are hybrids. The only alternative would be that they were caused by rare mutations or variability within H. elevatus. If hybrids, they must represent backcrosses, because both are most similar to H. elevatus. Hybrid evidence for no. 15 is provided only by the a smeary red/brown forewing band in the position where H. elevatus is normally melanic; this is typical for certain hybrids with red-banded H. melpomene (for example with ethilla or cydno), but in this case it may be more likely from a more closely related silvaniform in which broken forewing bands often contain brown. Hybrid 14 is more unusual, having very undulate distal wing margins (typical of silvaniforms such as hecale) as well as an extraordinary (for this area and mimetic group) broken yellow forewing band (again typical of a number of silvaniforms). Unfortunately, we know of no collections of elevatus or other Heliconius from Puerto Inírida with which to compare this solitary specimen, so it remains enigmatic. These two hybrids are included as "possibles", because it would not be at all surprising that H. elevatus, included within the silvaniform group (see Fig. 1), hybridizes with other members of the group. In Finally, hybrids between Heliconius ethilla narcaea and H. besckei are known from a number of locations in S.E. Brazil (nos. [23][24][25][26][27][28]. There can be no doubt from the general silvaniform pattern, coupled with the orange forewing bar and the submarginal pale loop, characteristic of besckei, on the hindwing underside, that these forms are indeed interspecific hybrids between ethilla and besckei. The hybrids are of both sexes, and nos. 23-27 are relatively homogeneous, suggesting that they are all F 1 hybrids. Hybrid no. 28, however, shows a number of traits more similar to besckei, particularly dark forewings with very reduced orange-brown markings, and the yellow longditudinal forewing streak characteristic of besckei. The equivalent yellow forewing streak in melpomene is a recessive trait [3]; if this is also true in besckei, the specimen could only be produced in a backcross to besckei.
To date, we have no molecular evidence for interspecific hybridization within the silvaniforms or between silvaniforms and the melpomene/cydno group. However there is god laboratory evidence that crosses within the silvaniforms, and between the silvaniforms and melpomene group are possible. Lawrence E. Gilbert has produced unforced hybrids between Heliconius ismenius (a silvaniform) and other members of the silvaniform or melpomene groups (see [4]: Plate 1B; [5]: Table 30 The key evidence for hybridization between these two species is intermediacy in pattern. Usually, they are recognized by the presence of both a white (or yellow) forewing band, as in cydno, and a red outer band, as in melpomene. There are usually many other intermediate characteristics; for example, in F 1 hybrids and some backcrosses (e.g. hybrid no. 29), the paired brown underside marks on the hindwing are often reduced as compared to H. cydno, where these marks are normally found, and the pale hindwing bar of some races of cydno is absent, or expressed as a faint "shadow" (e.g. no. 45). We also have good laboratory evidence for the genetics of hybridization in this group. Hybrids and backcrosses between melpomene and cydno have been produced in Liverpool in the 1970s (Brakefield, unpublished), in Texas [4,[6][7][8], in Colombia [6,9], in Panama [10][11][12], and in France (Jean-Pierre Vesco pers. comm., Additional File 1). In the laboratory, F1 hybrids are normally produced by a female cydno x male melpomene; the reciprocal cross seems much more difficult ( [11]; L. Gilbert, M. Linares, pers. obs.). Hybrids can be of either sex, but mated female hybrids typically produce no eggs [11]. Males, however, are fertile, and can be backcrossed in either direction. Thus genes (except for mitochrondrial genes, or genes on the W chromosome, because females are heterogametic) are readily transferrable from one species to another, and molecular evidence for natural hybrids in the field and introgression of some nuclear genes has now been obtained [9,13,14].
Most are apparently backcrosses to H. cydno, judging from the very broad, cydno-like pale forewing band, ringed by a narrow red outer strip (nos. [29][30]32). This phenotype is similar to that found in backcross hybrids between Amazonian and extra-Amazonian races of Heliconius melpomene [3,15]; here the colour pattern genotype is apparently homozygous for the pale-band allele N N , i.e. N N N N B-, and this is readily reproduced in laboratory hybrids [12]. In contrast, inferred F 1 hybrids (no. N N N B B-genotypes, i.e. heterozygotes for the N locus [12]. Similar genotypes can be produced in crosses between races of melpomene: heterozygotes at locus N have a broader red band and a more or less suppressed pale band on the forewing [3,15]. The genetics of these colour pattern differences in H. melpomene are therefore well understood and have been mapped in H. melpomene, H. numata, H.

31) represent
cydno and H. pachinus [16][17][18][19]. The remaining specimen in this group is no. 97. This was collected at a well-known locality for H. heurippa (itself a stable species probably deriving originally from a H. cydno x melpomene hybridization [9]) as well as for H. m. melpomene. Laboratory-reared F 1 hybrids between these species look extremely similar to red-banded H. melpomene melpomene (Linares, unpub.). See also hybrid no. 92, which is between genetic background, and can be reproduced in laboratory crosses [8].
Natural hybridization within the melpomene/cydno group can rarely be confirmed using molecular markers, although good molecular evidence for introgression between the species now exists in Costa Rica and Panama [13,14]. The rarity of natural hybrids normally makes collecting fresh hybrid material difficult. However, in one unusual site at San Cristobal, Venezuela, hybrids form about 8% of the population ( [9]; Additional File 1). Microsatellite markers suggest that most of the hybrid phenotypes are late-generation backcrosses, but the normally distinct mtDNA markers are found in the "wrong" species, at this site only [9].  (Gilbert, 1991). Kapan's study of H. cydno polymorphism in W. Ecuador [24] led to the marking of 2513 cydno and 90 melpomene, as well as two backcross hybrids most easily confusable with melpomene (nos. 39-40). Overall, these field studies have therefore scanned 3521 cydno/pachinus and 400 melpomene from zones of sympatry, and only two hybrids have resulted. The fraction of hybrids in natural sympatric populations is therefore of the order of 0.05%. Because Kapan [24] concentrated his efforts in habitats more suitable for H. cydno, it is probable that the frequency of hybrids may be somewhat higher in areas of 50:50 overlap.
Heliconius himera is the sister taxon to H. erato, and its distribution in dry forest of the Huancabamba depression of northern Ecuador and southern Peru abuts with that of H. erato, which occurs in Ecuador both west and east of the Andes, and also east of the Andes in Peru. 57 hybrids are known from all three contact zones: in W. Ecuador with H. erato cyrbia, in the lower Marañon of E. Peru with H. e. lativitta, and in the upper Río Mayo drainage with H. e. favorinus. In spite of the existence of these parapatric zones of contact, comparable to the hybrid zones between geographic races of H. erato, we regard H. himera operationally as a separate species because hybrids are always rare compared with the parental forms, unlike in the more classic hybrid zones where H. erato subspecies meet [5].
The best studied hybrid zone is that in W. Ecuador with H. e. cyrbia. We know of 52 hybrid specimens from this area ( Table 1, nos. 101-152; see also [25,27,30]). Hybrids in the centre of the narrow hybrid zone make up ~10% of the combined population of the two species, about evenly split between F 1 s and backcross phenotypes, the remaining 90% being parental [30]. Mitochondrial DNA and allozyme differences are retained in most "pure" individuals from the hybrid zone: the lack of morphological intergradation is paralleled by only occasional introgression throughout large portions of the genome [28]. Crosses in the laboratory occur in both directions, and the viability and fertility of hybrids is indistinguishable from that of parental species [29].
Natural hybrids are readily recreated in the laboratory, and the genetics of the colour pattern differences between the two species is known and has been mapped [31][32][33].
Instead, the absence of a randomly-mating hybrid swarm in the centre of the hybrid zone must be due to two major groups of factors. Firstly, strong mate choice expressed in both laboratory [29] and field [30] ensures that mating is 92-95% assortative. Secondly, there is as yet unidentified ecological selection against hybrids, presumably a mixture of predation on rare hybrids with poor warning signals, and environmental selection due to temperature and humidity [29,30,34]. from the Andes, downriver from Bagua, Amazonas. This hybrid is, judging from existing laboratory crosses with erato and himera [31,33], an F 1 . Hybrid frequencies must again be low: a collection in 1984-1986 of 7 erato lativitta and 77 himera from within 20 km of the overlap revealed no hybrids, even though individuals of parental species were found at the same sites [26].
Only one putative hybrid, presumably an F 1 , is known between H. erato and H. charithonia, from southern Mexico (no. 158). This form was previously recorded as an aberration of H. erato [35]. The expression of the narrow split yellow forewing band shape is more prominent on the underside than on the upperside; this feature is also true of heterozygotes of the gene Sd in many interracial crosses between Amazonian yellow-banded and extra-Amazonian red-banded H. erato. In interracial crosses, the Sd gene also reduces the expression of the yellow hindwing bar of erato [3,15] [36]. These two species were until recently recognized as geographic races of a single species, H. charithonia. However, H. charithonia is quite homogeneous both in allozymes and mtDNA sequence throughout its range from N. Peru to Central America, Florida and the Caribbean; and strongly divergent from H. peruvianus [36].
The two forms are predominantly parapatric: H. peruvianus is a Müllerian mimic of Elzunia pavonii and is restricted with its co-mimic to drier parts of W. Ecuador, S. from Costa Rica south to Peru. Heliconius hortense and H. clysonymus are normally recognized as separate species, and differ primarily in wing shape, although they have similar colour pattern and ecology. Both species are most abundant at 800m-1500m above sea level, and H. clysonymus is not known to be in contact with H. hortense across the lowland plains of central Nicaragua [37]. Molecular studies (Fig. 1) and the presumably homologous colour patterns between the two species suggest that the two fall within the H. erato group, including H. telesiphe and H. hecalesia. Because hecalesia mimics Tithorea and other ithomiine models, while hortense/clysonymus have their own striking red-and-black, non-mimetic, but purely heliconiine colour pattern, specimens 160 and 161 are unmistakably hybrids of known parentage, even though supporting laboratory data does not yet exist.