Speciation

Two monkeyflowers of western North America, one pink and visited mostly by bumblebees, one red and visited mostly by hummingbirds, can be crossed by hand to give hybrid seed. Yet the barriers measured between them were estimated to block, on average, 99.87% of gene flow. What holds them apart, and what makes them two species?

Answering that takes three steps. First you need to say what a species is, which turns out to depend on the evidence you have. Then you need to name the barriers that stop genes moving and learn to add them up. Last, you need to ask where and how the barriers arose, and keep those two questions apart.

The idea: a species criterion is a tool for a job

A lineage is a line of ancestors and descendants traced through time. A metapopulation is a set of local populations linked by occasional migration. The main species concepts agree that a species is a metapopulation lineage that evolves separately from others. The disagreements come when you try to recognize one.

The biological species criterion, associated with Ernst Mayr, asks whether populations are reproductively isolated: do they interbreed in nature and leave fertile offspring? It suits living, sexual organisms that meet. It fails for asexual organisms, which do not interbreed at all, and for fossils, which cannot be tested.

The ecological criterion separates species by their niche, the set of conditions and resources a population uses. The phylogenetic criterion comes in two versions. One asks that a species be diagnosable by fixed character differences. The other asks that it be monophyletic: a group made up of an ancestor and all its descendants, read from a tree as in reading phylogenetic trees. Morphological criteria use differences in form, which is what paleontologists and museum taxonomists work with.

Think of the criteria as different instruments pointed at the same object. A thermometer, an infrared camera and a hand on the forehead each detect a fever, and none of them is the fever. Kevin de Queiroz argued in 2007 that the species criteria work this way. Each detects a property, such as intrinsic isolation, a distinct niche, fixed differences or monophyly, that a lineage acquires as it separates. The properties arise at different times, so for young species the criteria disagree. In his unified concept they are lines of evidence, and the more of them agree, the stronger the case that two lineages are evolving separately.

Barriers before and after the zygote

A reproductive barrier is anything that reduces gene flow between populations. Barriers are sorted by when they act relative to the zygote, the fertilized egg.

Prezygotic barriers act before a zygote forms:

• Ecogeographic: the populations live in different places or habitats and rarely meet.
• Temporal: they breed at different times of day or year.
• Behavioral: courtship, song, scent or mate preference differ. For plants, pollinator preference plays this role.
• Mechanical: reproductive structures do not fit.
• Gametic: egg and sperm, or pollen and stigma, are incompatible, so fertilization fails.

Postzygotic barriers act after a hybrid zygote forms. In hybrid inviability hybrids fail to develop or survive. In hybrid sterility they survive but cannot reproduce. Sterility often traces to chromosomes that pair poorly at meiosis, covered in chromosomes and meiosis and chromosome changes.

Each barrier is measured with its own method, against its own control. Ecogeographic isolation comes from range maps and the overlap between them. Pollinator isolation comes from watching pollinators in a mixed stand. Gametic and postzygotic barriers come from controlled crosses, scored against crosses within a species. A barrier strength (RI) is therefore a fraction with a definite denominator, not a general impression.

Two cautions follow. Not every pair of species makes sterile hybrids. In a 2019 study, Heliconius elevatus and H. pardalinus butleri, butterflies that live together in the Amazon, rarely mated in captivity but produced fertile hybrids. Second, barriers are not built for a purpose. Selection acts on individuals; isolation often appears as a side effect of other changes.

Adding barriers up: the monkeyflower ledger

Barriers act in sequence. A pollinator barrier can only act on plants that grow close enough to share pollinators, and hybrid fertility only matters for hybrid seed that was formed. So each barrier works on the gene flow left over by the ones before it. This is the method Ramsey, Bradshaw and Schemske used in 2003 for Mimulus lewisii and M. cardinalis.

Worked problem. These numbers are illustrative, in the style of that study. Three barriers act in this order between two plant species. Each RI is the fraction of the remaining gene flow that barrier blocks. Assume the barriers act in sequence and independently.

• Ecogeographic, RI1 = 0.60
• Pollinator, RI2 = 0.90
• Hybrid fertility, RI3 = 0.50

1. First barrier. Its absolute contribution is AC1 = 0.60. Gene flow left: 1 - 0.60 = 0.40.
2. Second barrier. It acts on the 0.40 left: AC2 = 0.90 x 0.40 = 0.36. Total so far: 0.60 + 0.36 = 0.96.
3. Third barrier. It acts on the 0.04 left: AC3 = 0.50 x 0.04 = 0.02. Total isolation: 0.96 + 0.02 = 0.98.
4. Check. Remaining gene flow is the product of what each barrier lets through: 0.40 x 0.10 x 0.50 = 0.02, which equals 1 - 0.98.

total isolation = 1 - (1 - RI1) x (1 - RI2) x (1 - RI3)

5. Order matters for credit. Swap the first two barriers and the pollinator barrier gets AC = 0.90, the ecogeographic barrier 0.60 x 0.10 = 0.06, and hybrid fertility again 0.02. The total is still 0.98, but the credit moves. Here the early, weaker barrier gets more credit than the later, stronger one, because the later one has less gene flow left to block, so credit depends on position as well as strength.

Do not add the raw strengths. Here they sum to 2.00, which is meaningless as a fraction of gene flow. And the multiplication itself rests on the assumptions above. Sobel and Chen (2014) proposed a unified linear index of barrier strength because earlier indices used different null expectations and denominators.

Ecogeographic RI 0.60 Pollinator RI 0.90 Hybrid fertility RI 0.50 Order the barriers act

The ledger, in the order the barriers act

Barrier	RI	Gene flow left before	AC = RI × left

Total isolation = 1 − ∏(1 − RI)

0.98

Gene flow remaining

0.02

Sum of ACs (check)

0.98

Raw sum of the RIs (not a fraction)

2.00

In this sketch: three barriers act in sequence and independently, and each RI is the fraction of the remaining gene flow that barrier blocks. Real combinations can depart from this. The numbers are illustrative, in the style of the monkeyflower study.

Each barrier works on the gene flow left by the ones before it, so total isolation does not depend on the order, but the credit each barrier gets does. Never add the raw strengths.

In the real monkeyflower study, ecogeographic isolation was estimated at 58.7% from herbarium records, 97.6% of pollinator foraging bouts in a mixed population stayed on one species, and greenhouse crosses gave nearly 50% fewer seeds than conspecific controls. Combined, the measured barriers were estimated to prevent an average of 99.87% of gene flow for that pair. Those numbers belong to those two species; they are not a ranking of barriers across all plants.

Geography and mechanism are separate questions

Two questions are easy to merge and should be kept apart. The first is geography: where were the populations while they diverged? The second is mechanism: what process built the barriers?

On geography, allopatric speciation happens in populations that are physically separated. Separation can come by vicariance, when a barrier such as a new sea or mountain range divides one population, or by dispersal, when a few individuals found a new population somewhere else. Sympatric speciation happens without a geographic barrier, within one area.

The Isthmus of Panama, which closed roughly 3 to 3.5 million years ago by the usual estimate (some recent studies argue for a much earlier closure), is a classic vicariance event. It divided marine animals into Caribbean and Pacific halves. Nancy Knowlton and colleagues compared seven Alpheus snapping-shrimp pairs split by it, and the comparison indicated that isolation was staggered rather than simultaneous. A 2009 follow-up found that ancestral variation explained much of the spread in mitochondrial distances, and that two pairs had probably diverged well before the final closure. One event did not split every pair at one moment.

On mechanism, ecological speciation is the evolution of isolation as a by-product of adaptation to different environments by divergent selection, selection favoring different traits in different places (see selection in ecological context). Other mechanisms include drift in small populations (drift and gene flow) and, in plants especially, polyploidy, a doubling of chromosome sets that can isolate a new form in one generation.

Ecological speciation is a mechanism, not a third geographic mode. It can act in allopatry or in sympatry. In three British Columbia lakes, Rundle, Schluter and colleagues found that benthic and limnetic sticklebacks were strongly isolated in mating trials, while fish of the same form from different lakes, which had evolved independently, were not. Isolation followed ecology, which is what divergent selection predicts, but that result alone does not tell you where the forms first diverged.

Coexistence does not establish sympatric origin either. Two species living together today could have diverged there, or diverged elsewhere and met later, called secondary contact. For crater-lake cichlids in Nicaragua, Kautt, Machado-Schiaffino and Meyer (2016) compared explicit demographic models with genome-wide data. Their analyses supported divergence beginning in sympatry, but also admixture from source populations into both lakes shortly before diversification.

Design problem

Two cichlid species live together in one small crater lake. Design a study to decide whether they diverged in the lake (sympatric speciation) or diverged elsewhere and arrived separately (secondary contact).

One way to do it

Start by saying what will not count: present coexistence fits both histories, so it proves nothing.

1. Sample widely. Collect both species in the crater lake and every plausible source population in nearby large lakes and rivers.
2. Genotype genome-wide. Use thousands of markers (for example RADseq) so that ancestry can be traced across the genome, not from one gene.
3. Ask who is related to whom. If the two lake species are each other’s closest relatives, a single colonization with divergence in the lake is plausible. If each groups with a different outside population, separate arrivals are more likely.
4. Compare explicit models. Simulate divergence with gene flow from the start against isolation followed by contact, and compare how well each fits the data.
5. Test for admixture. Use admixture statistics, remembering that drift in one population can hide an admixture signal. Alleles can enter from outside even when divergence happens in the lake, as Kautt and colleagues found.
6. Use geology. The age of the crater sets an upper bound on how long divergence in the lake could have taken.

State the conclusion as support for one model over another, not as proof, and name the histories you could not test.