Fifty years ago evolutionary biologists could only have dreamt of the richness and variety of genetic markers now routinely available. If asked what such technology might reveal, they might well have predicted a revolution in taxonomy, with the evolutionary history of species clearly laid out once and for all, given the power that must come from a diversity of marker information at many loci across a genome. This taxonomic revolution, although underway, is making slow progress.

Not for lack of marker information, however. Rather, it seems that as we build up information on the history of a taxon using different markers, we most often find not one history but many. When phylogeographic studies originally based on the nonrecombining mtDNA locus are extended to include nuclear markers, rather than the original findings being confirmed, often alternative histories (just as clearly supported) are found for each new nuclear locus, for example (Dolman and Moritz, 2006). In short, the marker revolution is not tidying up taxonomy but changing the way we think about the taxa we study. The taxonomic heresy of reticulate evolution is increasingly being applied across all five kingdoms of life.

Of course, this is what phylogeographers and palaeoclimatologists have been telling us for some time with regard to populations subject to repeated cycles of glaciation during the quaternary (Hewitt, 2000). In recent studies for two lizard populations, likely to have been separated during the last glacial maximum (Godinho et al, 2006), nuclear markers tell just such an evolutionary story, which we shall soon discuss further.

During glacial peaks, populations are broken up into independent units. Once gene flow is stopped, divergence between these units is not only likely but more or less guaranteed. If during glacial interludes suitable habitat becomes continuous once again, then these independent units have the opportunity to spread and meet. What happens next? This depends on the extent and nature of their divergence. At one extreme they may not recognise each other as potential mates, in which case we might describe individuals descended from different source units as belonging to different species. At another extreme descendants of different sources may freely reproduce, genetic lineages benefiting from the diversity of genetic backgrounds arising from the admixture of divergent gene pools. An intermediate (likely rare) scenario is that divergence is just so for the formation of a stable contact zone between descendents of different sources.

The fact that we observe many such contact zones across a wide range of taxa would argue that the game of ‘what happens next when divergent populations meet?’ has been played out many times over evolutionary history. With hindsight, it is not surprising that this admixture process has been a focus of those who wish to understand the evolutionary implications of Quaternary glaciation (Hewitt, 2000), as the quaternary cycle has caused some of the most dramatic and well-studied large-scale environmental changes on the planet. However, environmental change is, and always has been, everywhere, thus admixture events may be an important general component of the evolutionary process.

The evolutionary importance of admixture events is naturally open to debate. Fortunately, as is often the case in evolutionary biology, we have the choice of either spending our time debating in the absence of good evidence or searching in the evolutionary record for the evidence necessary to resolve the issue, and so there are tasks at hand to suit all tastes. Those who seek evidence might ask: Wouldn't it be useful to have a marker indicating admixture? In this, as in so many aspects of evolutionary biology, RA Fisher got there first. Recombination between DNA from different sources produces Fisher's junctions (see Figure 1) and, once created, a junction is inherited just like a point mutation (Fisher, 1953). In fact, a junction fits the description of a point mutation even better than a single-nucleotide polymorphism (SNP); a junction being the point between two nucleotides where the source state of the DNA changes. Junctions differ from the other genetic markers in the evolutionary biologist's toolbox because (1) they are created by recombination instead of mutation and (2) they cannot be detected without labels to identify the sources of their abutting DNA. This leads to both good news and bad.

Figure 1
figure 1

DNA from two sources are labelled in grey and black. An individual F1 by source can produce meiotic products recombinant by source. The point where DNA source changes in one of these products is a junction (Fisher, 1953). A single (or any odd number) of junctions in an intron implies flanking exons are of different source. When their products are spliced together, the result is a protein hybrid by source.

Full size image

First, the good news: for simplicity, consider junctions between only two sources of DNA. Not just any recombination event can produce these junctions: if two stretches of DNA, one from each source, find themselves lined up side by side during meiosis in a (tautologically) single individual, then and only then can a recombination event produce a junction. It follows that for a freely reproducing population with heterozygosity by source Hs over this interval of the genome, and per generation recombination rate r in the same interval, junctions will be produced at a rate rHs(Baird, 1995), analogous to the per generation mutation rate for traditional markers. Because junctions can only occur under this circumstance, observation of a junction is proof positive not only that descendants of two sources have met but also that they have produced fertile F1 crosses source X source (see Figure 1). The information arising from observation of a junction is two-fold; we are informed absolutely that admixture has occurred, and we also learn about the geographic history of contact between sources. For example, if individuals of two source populations meet only in one mountain pass, then all junction copies descending from a particular meiotic event must trace their ancestry back through the geography of dispersal to that mountain pass. This is a qualitative difference between junctions and traditional markers. While all DNA copies descending from a particular mutation event must also trace their ancestry back to that event, the event itself could have happened (within reason) anywhere in the geographic range of all potential ancestors.

In the current issue, Godinho et al (2006) present strong evidence for junctions between DNA descending from two populations of lizards likely to have been separated during the last glacial maximum, and currently in contact in a narrow hybrid zone in a valley of the mountains of Iberia. The junctions are located within an intron of the β-fibrinogen gene, mutations of which are associated with arterial disease in mammals. A mutation in an intron might be held of little interest; however, a junction in an intron implies the flanking exons are from different sources (see Figure 1), and this is potentially of great importance: if the flanking exons diverged during isolation, then an intron junction arising after secondary contact of source populations means exon products are spliced together to form proteins never previously seen in either source population (see Figure 1). While the creation of novel variation is one of the intuitive explanations for the ubiquity of recombination, when recombination causes such junctions it has real variational meat to work with, and evolutionary biologists have what seems to be a tractable system to describe: the epistatic interactions between exon alleles within the same gene can be visualised in terms of the tertiary structure of a single final protein product. This is a simpler undertaking than understanding the causes of epistasis between different genes whose products interact in a biochemical pathway.

Godinho et al propose a parsimonious explanation for allozyme variants found in contact zones but apparently absent from their source populations – such variants are the expected outcome of intragenic junctions in the DNA encoding the allozymes.

Allozymes were one of the first types of markers widely used to survey natural populations, and allozyme variants restricted to contact zones were found to be such a general phenomenon that a word was coined to describe them: hybrizymes. If Godinho et al's suggestion proves correct this word was well coined, because then these allozyme variants are indeed hybrid by source at the level of the genes that produce them. Further, if hybrizymes are novel proteins created by junctions between diverged sources of DNA, then a direct effect of source admixture has already been reported by numerous studies of natural populations.

The bad news about junctions stems from the second property that distinguishes them from other genetic markers: junctions cannot be detected without labels to identify the sources of their abutting DNA. When I submitted a manuscript in 1993 describing an approach to estimate the time of initial contact between source populations using the frequency spectrum of distances between junctions arising, an anonymous reviewer suggested that the theory part of the manuscript was worthwhile (Baird, 1995), but its application was ‘like proposing a method for measuring the rotational velocity of flying saucers.’ At the time, this was fair comment – we simply did not have the molecular tools to detect junctions with any accuracy. What is needed is a very dense distribution of mutational markers diagnostic of DNA source. Times move quickly however, and by 1998 Loren Rieseberg's lab had developed exactly such a data set, allowing a ballpark estimate of the number of generations necessary to produce a hybrid species (Ungerer et al, 1998). A more recent beautiful example of use of the patterns arising from junctions is the study of Helicobacter pylori living in the human intestine (Falush et al, 2003b), made possible by MCMC estimation of junction locations (Falush et al, 2003a). The approach has since been applied to humans themselves (Smith et al, 2004), including, at last, estimation of time of initial contact between source populations.

In principle, the problem of detecting junctions is similar to identifying homologous SNPs. Because an SNP can have only four states, identity by state is a poor indication of identity by descent. Mountain et al (2002) suggested using the genetic state of microsatellites flanking an SNP to resolve this issue, with the compound marker being called an SNPSTR (SNP-serial tandem repeat, pronounced ‘snipster’). In effect, an observable junction is always a compound marker: a junction flanked by (at least) two mutational markers informative about their source. The putative junctions in the Iberian lizard study are single strand conformation polymorphism (SSCP)–junction–SSCP compound markers, the SSCPs having been shown by sequencing to reveal SNPs. The compound marker thus suffers the weakness of SNP homoplasy. The potential power of SNPSTR and juxtaposed microsatellite (Estoup et al, 1999) analyses suggests that STR–junction–STR compound markers would be powerful tools in the analysis of admixture events, and the Iberian group have already put this suggestion into practice with promising results (N Ferrand, personal communication). It is amusing to invent names for such compound markers. JUNCSTRs is unattractive, Why not simply ‘Fishers’? I think Fisher would be pleased that his markers of admixture are finally coming of age.