Monday, June 22, 2009

Do Hummingbirds Speciate faster?

I think that the evidence from the two studies below (here and here) suggests that they just might. (I'll show a third, and perhaps strongest case, from Selasphorus in a future post).

Next posts are going to look at some non-hummingbirds that I'm working on.

Re-assessing species limits for the Escudo Hummingbird (Amazilia handleyi)

Two weeks ago, I was fortunate to be invited to give a seminar to the Panama Audubon Society. I chose to talk about how molecular systematics can help inform questions of species limits. One of the examples that I used was the case of the Escudo Hummingbird, Amazila handleyi. The Escudo Hummingbird looks a lot like the Rufous-tailed Hummingbird (an Amazilia common in the lowlands of Central America that ranges from southeast Mexico to northwest Ecuador) but on steroids. Above is a photo of specimens taken by Michael Lelevier who is leading our collaboration on the phylogeography of Rufous-tailed Hummingbirds (D are Escudo Hummingbirds). The Escudo Hummingbird is endemic to Isla Escudo, a small island off the coast of Caribbean Veraguas, near the Bocas del Toro archipelago. There's a map below. The Escudo Hummingbird is quite clearly related to Rufous-tailed Hummingbirds, and the first scientific specimens of this taxon date to 1962. Notice that I'm calling it a taxon not a species, because that is the debate. Although Alexander Wetmore described the Escudo Hummingbird as a distinct species in 1963, noting both the size and coloration differences, subsequently this taxon has been down-graded to a sub-species of the Rufous-tailed hummingbird. Today, the American Ornithologists' Union (the official checklist for North and Central America) follows that classification (i.e. Amazilia tzacatl handleyi); the Panama Audubon Society maintains their own checklist, and considers the Escudo Hummingbird a full species (Amazilia handleyi).

As I pointed out to the Audubon folks, the Escudo Hummingbird
is the most physically distinctive member of the Rufous-tailed group, but not the most genetically distinctive (at least using mitochondrial DNA [mtDNA]). Within the entire complex, genetic distances are modest (less than 2%, or about a million years). Deep splits include the split between Central America and South America. There is also a relatively distinctive mitochondrial clade found most common on Isla Coiba but also fr
om one individual in Pacific Costa Rica. In contrast the Escudo Hummingbird is a little nub on a mass of closely related individuals from the Caribbean lowlands of southeastern Mexico all the way down to central Panama.

Isla Escudo is believed to have been connected to the Veraguas mainland during the Pleistocene, when sea levels were lower. The estimates that I have found in the scientific literature date its isolation to the Holocene (8900 years ago, to be precise...we'll get to these estimates later). So, my thinking up to the seminar was that the mtDNA data agreed with the isolation of a small founder population about 10,000 years ago. Do what you want with that, but unlikely to be a reproductively isolated population, which is my criterion for defining a species (there others, but I won't bore you with that debate).

But after the seminar I got to thinking about it a little more and I was bothered by something. You see, the I presented a figure of mitochondrial haplotypes (particular genetic sequences from a fragment of mtDNA) from the Bocas mainland, Isla Colon (in the Bocas archipelago) and Escudo. To the right is that image. Isla Colon and the Bocas mainland share
haplotypes, which is what we would expect given our understanding of the isolation of the Bocas archipelago: namely that they were also isolated during the
Holocene (albeit a bit earlier, i.e. as early as six thousand years and as recently as one thousand). Two populations share haplotypes for two reasons. The most obvious is gene flow. But two populations can not be currently genetically in contact, but still share haplotypes because their time in isolation is quite recent. For
example, if we sent one million people to Mars to establish a Martian colony, they would share our genes for considerable time, even if no rocketships went back and forth between
Mars and Earth. Only the slow accumulation of independent mutations would cause DNA sequences to be different.

We generally believe that avian mtDNA mutates at about 1% per million years. Therefore, two sequences that differ by 2% likely had a comm
on ancestor about 1 million year ago because each lineage was independently mutating at 1% per 1,000,000 years. So, the difference between 5200 years (the estimate for the age of isolation of Isla Colon) and Isla Escudo (8900 years) is essentially meaningless for mtDNA differences. So, given similar immigration patterns from the mainland, Isla Colon and Isla Escudo should have similar genetic profiles relative to the mainland, but they don't.

Isla Escudo is much further from the mainland than Isla Colon (17.6 km vs. 1.5), and is much smaller (4.3 vs. 59.0 sq. km.), so I simply chalked up the dif
ferences to les
s frequent immigration from the mainland and smaller population size (new mutations become fixed in smaller populations more quickly than in larger ones). But, population geneticists have gotten new tools to explore these issues in a much more robust manner. The ability to sequence massive amounts of DNA is one breakthrough that is revolutionizing systematic ornithology. Equally important is the use of Markov Chain Monte Carlo simulation techniques in super-fast computers that allow us to routinely do complex calculations of the probability of certain population genetics parameters that we couldn't even have imagined 15 years ago.

So, I used IM which is a software developed to do just that to estimate gene flow between Escudo and Isla Colon and the Bocas mainland. Not surprisingly, the result was no gene flow. IM allows the researcher to jointly estimate migration rates (i.e. gene flow) and divergence times between two closely-related populations. However, with limited data, precision on both the estimate of divergence time and gene flow are limited. But, because these analyses are Bayesian in nature, peeking at your data to refine analyses is OK. So I tur
ned off gene flow on a second run of the IM, and got some results that I should have realized before hand: the split between Escudo Hummingbirds and other Rufous-tailed Hummingbirds is much older than the Holocene isolation of Isla Escudo.

In fact, the split seems to be about 140,000 years ago. The
95% confidence interval on the estimate is 40,000 – 355,000 years ago; i.e. there is no way that the mtDNA differences date to the Holocene isolation of Es
cudo, they are an order of magnitude older (or our estimate of mtDNA mutation in Amazilia is an order of magnitude too low [quite unlikely]. To the right I show the graph of estimated divergence time (x-axis) vs. probability (y-axis). Note the mass of probability around 150,000 years ago.

And, the IM model assumes migration since divergence begins. So, Escudo Hummingbirds had at least 30,000 years (and probably more like 130,000 years) to exchange genes with other Rufous-tails before become isolated on tiny Escudo Island. Shockingly, I think that the evidence accumulated points not to the Holocene isolation of typical Caribbean Rufous-tailed Hummingbirds, but the preservation of a much older lineages as a relictual population found no where else.

During the late Pleistocene sea-levels waxed and waned. At lows, such as those from the Last Glacial Maximum (around 18,000 years ago) until the start of the Holocene (around 10,000 years ago) what ever was the range of the ancestor of modern Escudo Hummingbirds should have come contact with Rufous-tailed Hummingbirds and if they were the same species, should have exchanged copies of mtDNA. The small size of Escudo means that evidence of that contact might likely have vanished during the last 10,000 years. However, if that occurred, we should have observed Escudo haplotypes in mainland individuals (even in only 1 or 2 individuals). We did not. In fact we have sequenced over 90 birds from mainland Central America, and none have the Escudo haplotype.

It's time to revisit the biogeography of Escudo. In 2003, our lab at STRI published a paper in the Condor (Gonzalez et al. 2003) on Bay Wrens. They observed two unique, but distant haplotypes on Escudo Island (the Bay Wren on Escudo is quite distinctive from the Central American Bay Wrens). Both Coiba and the Pearl Islands are living museums of prehistoric avian diversity in Panama (the Coiba Spinetail exists no where else on Earth, and the nearest conspecific of the White-fringed Antwren on the Pearl Islands occurs in northern South America). In the meantime, it's time to give the Escudo Hummingbirds the distinction that it deserves, as a full species.