New evidence for hybridization and introgression is unsettling
The idea that hybridization and introgression among species occurs and may be evolutionarily important has long been acceptable to botanists: in plants clear examples of hybridization and introgression existed before the advent of molecular genetics. The base of the whole tree of life is now known to be much more like a web than a tree, due to abundant horizontal transfer among prokaryotes. However, such ideas have found little favour with zoologists until recently.
Advances in science can be unsettling. They can be fought against, sometimes irrationally, by those seeking to uphold the status quo ante. Many evolutionary biologists either think that species are “reproductively isolated,” or adhere to phylogenetic or cladistic species concepts with underlying assumptions that deny reticulation since speciation. Hybridization and the possibility that genes might often flow (or ‘introgress’) between species challenge these modes of thought. To those with such convictions, introgression is only rarely involved in adaptive evolution or speciation.
Such feelings still exist among many fellow evolutionary biologists. In the last year, I’ve visited different universities around the USA and discussed recent genomic evidence for introgression in Heliconius butterflies. Many people have commented after my talks that they find the whole idea of such leaky species in animals disturbing.
Heliconius is not the only animal group to show introgression: a variety of other groups such as Darwin’s finches, cichlid fish, Littorina winkles, and among insects Rhagoletis, Zeiraphera, the Drosophila pseudoobscura-persimilis species pair, Hawaiian Drosophila, and Timema, have all lent support to the idea that species boundaries are weaker than we thought even 10 years ago. Increasingly, reviews by Michael Arnold and others have suggested that regular patterns of hybridization and introgression are much more likely than we used to think, especially among recently derived radiations of species. But Heliconius is maybe one of the icons of these new ideas.
Andrew Brower knows the genus Heliconius as well as many of us. Nonetheless, he has recently voiced extreme skepticism to our work in two recent diatribes (Brower 2011, 2012). Brower is disturbed by all this new evidence upending his view of species. As he points out, “… if [homoploid hybrid speciation] has occurred multiple times just within this charismatic genus of butterflies, then speciation theory would seem to be headed for a paradigm shift.”
It would be interesting to know the history of Brower’s 2012 article, and what the reviewers said about it. The article was slipped into the “Reviews” section of the main biological journal of the Royal Society of London, essentially the journal in which Isaac Newton published. Beneath a modest title and apparently neutral abstract, the article is not a balanced review at all. It contains a detailed, point-by-point dismissal of much Heliconius genetic and evolutionary work done by many different laboratories worldwide (except his own) over the last 20 years.
I may have unwittingly triggered Brower’s 2012 article by my reaction to his 2011 piece. I wrote him a personal email, outlining some of the many errors he had made. Undaunted, he has aired a revised set of his views in public, and presumably would like any new errors he has made to be made public also. In this and the next post, I shall attempt to do this.
First, let’s briefly survey evidence for natural hybridization and introgression among species of heliconiine butterflies, that it can lead to the sharing of adaptive colour patterns used in Müllerian mimicry, and that it is involved in speciation. Then we can discuss whether Brower’s critiques have any validity (in the next post).
I take his criticisms personally because the conclusions attacked refer to many major findings of our group since the 1990s, and to the findings of a number of ex-students and ex-postdocs.
I also take this personally because Brower helped to sabotage in 2001 an earlier attempt to publish details of hybrid specimens. The data was documented from museums and private collections worldwide, and had been gathered since the mid-1990s. Brower’s signed review helped to prevent publication of the data set until the late 2000s (Mallet et al. 2007). A major argument he used then, and which he still does against the same data set, is you can’t trust museum specimens: they are either fraudulent or misinterpreted. (Another argument he used then was that in any case they aren’t hybrids at all, but he seems to have dropped that particular idea for now). Even if some of the many specimens in the data set are fraudulent (unlikely, and anyway a possibility discussed carefully in the original paper), and even if he didn’t like the conclusions, it seems to me that to use these grounds to suppress detailed documentation of these interesting hybrids was inexcusable.
Evidence for hybridization, introgression and hybrid speciation in Heliconius
1) Hybridization in nature. Hybrids between Heliconius species in nature are usually common on a per species basis (Mallet et al. 2007), although rare on a per individual basis (mostly less that 1 in 1000 individuals of any species). As well as F1 crosses, F2 and backcross phenotypes are produced in nature. F1 females are usually sterile, but F1 males can be used in backcrosses. This applies to the whole melpomene-silvaniform group, within which it is possible to cross and backcross, enabling transfer of colour patterns in the laboratory, across that whole group of ~15 species (see suppl. info. in Mallet et al. 2007).
These facts alone suggest that some genes must flow between species, and that, although many loci may find themselves to be incompatible, it would not be surprising that sometimes transferred genes or groups of genes can be advantageous in their new contexts. Although hybridization at a rate of 1/1,000 or 1/10,000 seems rare, it would still be much more important than mutation in introducing new variants into populations. When multilocus adaptations are involved, which have already been tested within one species, the importance of (multi-)gene flow could be even greater.
2) Evidence from gene flow using Isolation-Migration (IM) coalescent algorithms. Coalescent analyses of small numbers of some (but not all) loci sequenced using traditional PCR-Sanger methods suggest gene flow between cydno, melpomene, and related Heliconius (Bull et al. 2006, Kronforst et al. 2006, Mavárez et al. 2006, Kronforst 2008).
3) Heliconius heurippa — a hybrid species. The colour pattern of a putative hybrid species, Heliconius heurippa, can be produced in the laboratory via crosses between Heliconius cydno cordula and H. melpomene melpomene, which overlap nearby. Heliconius heurippa appears to be a cydno-lineage species that obtained its red forewing colour pattern and underside markings from the H. melpomene with which it still overlaps (Mavárez et al. 2006).
Populations of H. cydno cordula and H. melpomene melpomene overlap a few tens of km to the North, in San Cristóbal, Venezuela. This population shows abundant evidence of past hybridization, with some specimens of melpomene having polymorphic colour pattern elements from cydno, and some specimens of cydno showing polymorphic elements of melpomene (Mallet & Mavárez 2003, Mavárez et al. 2006).
Bizarrely, Brower interpreted a single specimen from among these polymorphic forms in San Cristóbal as a new species in the cydno/timareta species group (Brower 2012, Fig. 1.14), but this is surely a very unparsimonious idea, given the multiple other hybrid phenotypes also present at the site (Mallet & Mavárez 2003). This Venezuelan site thus provides exactly the sort of intermediate hybrid population (i.e. a red-banded colour polymorphism in cydno) required as a first step towards establishment of a fixed red-banded hybrid taxon, such as H. heurippa (Mavárez et al. 2006).
An earlier critique of the hybridization hypothesis by Michael Turelli & Jerry Coyne was submitted to Nature, but unfortunately, in my view, remained unpublished. Nonetheless I still believe that the Mavárez et al. paper had the correct interpretation. The most parsimonious explanation is that the colour pattern of H. heurippa was obtained via a cross between a yellow banded cydno and a red-banded melpomene. H. heurippa and H. melpomene continue to hybridize occasionally; I have examined a wild-caught hybrid (Mallet et al. 2007, no. 97). The most suitable races of each parental species to produce the heurippa phenotype are also those that happen to overlap or nearly overlap currently with heurippa. The Turelli & Coyne critique didn’t really address the central idea that the unique colour pattern of H. heurippa is parsimoniously explained as a hybrid between colour patterns which occur locally.
Camilo Salazar later published a paper suggesting, on the base of extensive sequencing, that introgression of colour pattern in H. heurippa involved the gene kinesin (Salazar et al. 2010). Kinesin is now known to be very close to, but is probably not at the actual location of the colour pattern regulatory switches, which seem largely to affect the expression of optix in the pupal wing (Reed et al. 2011, Heliconius Genome Consortium 2012, Pardo-Diaz et al. 2012).
Personally, I thought that the Coyne & Turelli critique missed something they could have argued. This was not that hybridization was not involved in the origin of heurippa‘s colour pattern (it likely was), but that the resultant taxon may not be a “good” species since it is not known to overlap with one of its putative parents. H. cydno cordula occurs nearby, but only at a small distance North of the known range of H. heurippa on the Eastern slopes of the Andes. H. heurippa might then be only a geographic race or “semispecies” of the cydno/timareta superspecies. Guerilla activity has hitherto prohibited exploration of the probable contact zone between heurippa and this cydno race, near Yopal, Colombia. Nonetheless, Camilo Salazar in Mavárez et al. (2006) showed, using courtship tests, that heurippa is partially reproductively isolated from both parents by its hybrid colour pattern, and therefore that one could justify the species label on that ground. As pointed out elsewhere, similar critiques of non-overlap of parental and offspring hybrid taxa could be levelled against most other animal or plant examples of homoploid hybrid species; they too overlap only rarely with both putative parents (Mallet 2007). A recently revealed exception to this pattern is Heliconius elevatus, which overlaps extensively with both putative parents H. pardalinus and H. melpomene across the entire Amazon basin (Heliconius Genome Consortium 2012, and see point (6) below).
Next-generation genomic studies have strengthened these preliminary forays, and reveal even more evidence for “promiscuous” gene flow than we thought.
4) RAD resequencing across the whole genome. Where “postman-patterned” H. timareta overlaps with “postman-patterned” H. melpomene in NE Peru, so-called ABBA-BABA nucleotide sites in RAD tags show genome-wide evidence of flux (excess of ABBA sites over BABA sites) across the species boundary (NB, timareta is another member of the cydno group lineage) (see part (b) of Fig. 1 below. The adjacent “rayed” race of H. melpomene used as a control, which does not overlap with H. timareta in Peru, shows virtually no ABBA-BABA excess. This provides critical evidence as the races of H. melpomene are separated by only a few tens of km across a narrow hybrid zone, and the Fst between postman and rayed melpomene is minuscule, as expected for spatial structuring at this geographic distance (Heliconius Genome Consortium 2012).
The excess of ABBAs over BABAs used to infer gene flow is hard to explain by pre-existing population structure, which was one of Graham Coop’s suggestions (Coop 2012), because the local genome-wide bias exists right across many chromosomes in two species both of which have much broader distributions. Coop suggested also that this effect might be due to alignment bias. This sounds reasonable, because the alignments used a postman H. melpomene melpomene from Panama as a genomic reference sequence. But it is unlikely to be the explanation because we’re here comparing races in Peru that are much closer (smaller Fst) than either is to H. melpomene melpomene from Panama. There’s therefore no reason to expect a bias across the whole genome to just one of this pair of races Amazonian races, as found.
5) Sureselect resequencing of colour pattern regions. Fixed ABBA-BABA sites in colour pattern regions show extremely strong, in fact the genome-wide strongest, evidence for gene flow both between the Peruvian postman forms of melpomene and timareta, and also between Colombian rayed forms of melpomene and timareta (Heliconius Genome Consortium 2012). This reciprocal evidence for transfer, in exactly the expected direction(See (b) and (c) in Fig. 2 below), and almost exclusively within colour pattern divergence peaks (Fig. 2a, below) between the aforementioned races that meet in Peru, seems to us convincing evidence that colour patterns were transferred .
Graham Coop (Coop 2012), and Michael Turelli (question after a seminar, UC Davis in November 2012) have both suggested that this might instead be caused by ancestral population structure. But it’s hard to know how. This might happen, I suppose, if timareta speciated from melpomene “multiregionally,” so that each species inherited some of the same multiple, locally fixed colour patterns from their widespread and polytypic common ancestor. The need for this unknown mode of speciation is avoided, however, if gene flow among species takes place. There is already genome-wide evidence for polymorphic transfers (see (4) above). Therefore, exchanges of colour pattern regions will also take place from time to time. Then occasional fixation of such introduced ~10-50 kb colour-pattern determining genomic regions seems not at all improbable, given the kinds of mimetic selection expected.
6) Sureselect resequencing of colour pattern regions, contd. Resequencing also yielded phylogenetic evidence for wholesale transfer of two melpomene ray-pattern genomic regions to a Heliconius pardalinus-like ancestor of Heliconius elevatus (Fig. 2d, above), giving a very recently formed rayed species that, according to the rest of its genome is in fact nested within H. pardalinus (Heliconius Genome Consortium 2012). H. elevatus, together with its two putative parent species, today coexist in widespread sympatry throughout the Amazon basin. We hypothesise that this represents a hybrid speciation event triggered, at least in part, by the acquisition of this rayed colour pattern, which is unique among the ~9 spp. of the “silvaniform” group of Heliconius.
7) Whole genome resequencing. What we lacked in the Genome Consortium paper was a genome-wide RAD analysis of ABBA-BABA sites of the Colombian populations of timareta and melpomene, to match the SureSelect data from the colour pattern genomic regions. For the Peruvian populations we had both sets of data. The lack of a reciprocal design came about because the colour pattern data from Colombian timareta were added as an afterthought to the genome paper.
It seems worth mentioning that we’ve now carried out a reciprocal genome-wide analysis of the sort indicated. We used populations of cydno and melpomene in Panama (where the two are not mimetic, and did not exchange colour patterns), and of timareta and melpomene in Peru (where it is argued the two species have exchanged colour patterns). This time, instead of using RADs, we have Illumina whole genome sequence data to ~30-40x, for each of 4 individuals for each species and site. These intriguing new data suggest the exchange of very large fractions of the genome at both sites. This reciprocal whole-genome evidence from two sites provides very strong evidence for abundant genetic transfer in localities where Heliconius melpomene overlaps with H. cydno and H. timareta (Simon Martin et al., submitted).
Brower’s critique and its refutation
Faced with such a mountain of evidence, Brower has an uphill task to perform. The case for hybridization and introgression seems solid because dismantling any one piece of evidence does not cause the whole argument to fail. Undaunted, Brower proceeds to attack every single data item and inference that points to the foregoing conclusions. If Brower is correct, many separate workers across the world have conspired to slant their conclusions towards an idea for which there is, in fact, no evidence at all, according to Brower.
In the next post I’ll enumerate Brower’s criticisms and discuss them individually.
Brower AVZ. 2011. Genetica 138: 589.
Brower AVZ. 2012. Proc Roy Soc B 280 online.
Bull V et al. 2006. BMC Biol 4: 11.
Coop G. 2012. http://gcbias.org/2012/05/23/journal-tea-may-21st/
Heliconius Genome Consortium. 2012. Nature 487: 94.
Kronforst MR. 2008. BMC Evol Biol 8: 98.
Kronforst MR et al. 2006. Evolution 60: 1254.
Mallet J. 2007. Nature 446: 279.
Mallet J & Mavárez J. 2003. (In Mallet et al. 2007). See: http://www.biomedcentral.com/content/supplementary/1471-2148-7-28-s1/mavarez/cristobaltab.html
Mallet J et al. 2007. BMC Evol Biol 7: 28.
Mavárez J et al. 2006. Nature 441: 868.
Pardo-Diaz C et al. 2012. PLoS Genet 8: e1002752.
Reed RJ et al. 2011. Science 333: 1137.
Salazar C et al. 2010. PLoS Genet 6: e1000930.