NEW YORK (GenomeWeb) — Birds appear to have rapidly evolved around the time of the mass extinction 66 million years ago that included the dinosaurs, giving rise to the 10,500 bird species we know today, according to results released today from a large-scale international effort to construct an avian tree of life, based on genomic sequence data.
Other key findings from the project include the idea that vocal learning in birds evolved independently at least twice and can provide clues to speech development in humans; that crocodilians, some of the closest living relatives of birds, have extremely slowly evolving genomes, allowing researchers to reconstruct the genome of their ancient ancestor; and that flamingoes and doves are surprisingly closely related, despite their different morphology.
These and other results from the Avian Phylogenomics Consortium were published in almost 30 papers in Science and other journals today. Based on a comparison of the genomes of 48 modern bird species, representing almost all major branches, and with additional help from the genomes of three crocodilians, the researchers constructed a phylogenetic tree that provides new insights into the evolution of birds and their traits that distinguish them from other vertebrates.
The new avian family tree allows researchers to tackle "everything from the basic question of what makes a bird, how birds relate to each other,[to] the genetic basis of specific bird traits, like flight, feathers, vision, olfaction, [and] sexual selection," Thomas Gilbert of the Natural History Museum of Denmark and Curtin University in Australia, and one of the consortium leaders, noted in a press teleconference held yesterday to recap the findings. Ultimately, he said, the tree will provide a better understanding of dinosaurs, from which modern birds directly descend.
The four-year project involved more than 200 scientists at 80 institutions in 20 countries, and related studies involved scientists at more than 140 institutions. Besides Gilbert, the consortium's leaders are Guojie Zhang of the National Genebank at BGI in China and the University of Copenhagen and Erich Jarvis of Duke University and the Howard Hughes Medical Institute.
The 45 bird species sequenced for the project — the chicken, turkey, and zebra finch genomes were already available — came from museum collections, zoos, and research institutions, and most of the sequencing was performed at BGI. The data analysis, which involved the comparison of 14,000 coding and non-coding regions across the genomes, required the development of new computational techniques, an effort led by Tandy Warnow and her student Siavash Mirarab, both at the University of Illinois Urbana-Champaign. Several computing centers, including the Texas Advanced Computing Center, the Munich Supercomputing Center, and the San Diego Supercomputing Center, crunched the data, which took 400 years of CPU time.
Funding for the project came from various sources. The consortium has not estimated the overall cost, calling it "a big number." BGI provided support for most of the genome sequencing, and supercomputing centers for the data analysis. Individual investigators contributed funds that came from the National Science Foundation, the Howard Hughes Medical Institute, and universities.
The two major consortium papers, both published in Science, focus on the phylogeny of birds and comparative genomics. Other papers, published in Genome Biology, GigaScience, BMC Genomics, and other journals, delve into avian biology, such as vocal learning and teeth loss; data analysis methods; and the genomes of crocodilians. At least 12 other studies are still under review and will be released in the coming months. During this week's teleconference, several consortium members discussed some of the projects and their findings.
Birds fall into three broad evolutionary groups, the Paleognathae, which include large flightless birds like the ostrich and the emu; the Galloanseres, which contains chickens and ducks; and the Neoaves, which comprises all other birds, such as songbirds, woodpeckers, owls, birds of prey, many waterbirds, hummingbirds, cuckoos, and doves. The consortium's work focused on the Neoaves, which include over 90 percent of all bird species. Three orders from the Paleognathae were missing from the analysis.
According to Gilbert, it has been difficult to work out the evolutionary relationships between Neoaves species because depending on which genetic markers were used, the relationships changed. "Given how fundamental this is to so many other studies, we proposed to do this at the biggest level and sequence the genomes of at least one bird species per avian order, and then try and use this to reconstruct the phylogeny," he explained.
According to Jarvis, the lead author of the phylogeny paper, the new avian tree of life is "the biggest DNA sequence tree ever generated."
One important finding is that birds rapidly evolved into many different species around the time of the extinction of dinosaurs 66 million years ago. "This solves, we hope, a long-standing controversy around whether this radiation or diversification of bird species occurred before or after the mass extinction of dinosaurs," Jarvis said. "We're suggesting it occurred right at that time."
"Birds are dinosaurs," said Ed Braun, a researcher at the University of Florida and a consortium member. "They are the one lineage of dinosaurs that made it through the mass extinction" and gave rise to all modern Neoaves species, most within a 10 to 15 million year time period.
Interestingly, the vocal learning trait appears to have evolved independently at least twice in Neoaves — it is present in songbirds, parrots, and hummingbirds. The trait involves a number of genes that are expressed in a special region of the bird brain. Jarvis and his colleagues found that these genes are differentially expressed both in birds capable of vocal learning and in the speech area of the human brain but not in birds that do not have this trait or in non-human primates. Going forward, they plan to study the trait further in vocal learning birds, which can serve as model organisms to better understand speech disorders in humans.
The analysis of the family tree also revealed that the common ancestor of core landbirds was probably a top-level or apex predator, and that several lineages subsequently lost the predator trait.
A surprising new finding in the family tree was that flamingoes and grebes, both waterbirds, are closely related to doves — they all belong to the Columbea branch of Neoaves. "There really is no morphological characteristic that unites them," said Braun, "so this is a real surprise."
Other than family relationships between birds, the project also revealed what makes birds different from other vertebrates, said Zhang, the lead author of the comparative genomics paper. For example, a feature of all bird genomes is their small size, about a gigabase or a third of that of mammalian genomes.
Birds appear to have gotten rid of much of the repetitive DNA that mammalian genomes harbor, and have lost a large number of genes —they have 30 percent fewer genes than the human genome. Many of the missing genes have essential functions in humans, including reproduction and skeleton formation, and their loss has had "very important effects on the evolution of many distinct phenotypes of birds," such as the loss of teeth, Zhang said. The fact that birds often only carry one version of a gene of which two homologs exist in humans makes them good model organism for the study of the function of these genes, he added.
In general, birds serve as model organism for the study of vertebrate embryonic development, he said— unlike mammals, which develop in the womb, birds can be easily studied in their eggs — which has implications for human development as well.
Birds also appear to have a highly conserved genome structure, which "has stayed remarkably unchanged for more than 100 million years," Zhang said. In addition, their molecular evolution rate is much slower than that of mammals.
However, the closest living relatives of birds — crocodilians — have an even slower evolutionary rate, according to an analysis of the genomes of three representatives, the American alligator, the saltwater crocodile, and the Indian gharial, which helped the researchers assemble the bird family tree.
Among the major vertebrate groups, "crocodilians show the absolute slowest rate," said Braun, allowing researchers to look for ancient features in their genomes that provide glimpses of the common ancestor of crocodiles, birds, and dinosaurs.
Constructing species trees
To come up with the avian family tree, the consortium constructed not one but two separate species trees, which required new computational tools. According to Warnow, the two trees, which are nearly identical, are "based on completely different ways of computing an evolutionary tree."
To build one of them, the ExaML TENT species tree, the researchers used DNA from different genomic regions of the 48 species, put them into a matrix, and applied a statistical technique called maximum likelihood. This approach is computationally very intensive and the researchers developed a new method that reduced the analysis time from two years to one month, she said.
The second approach, which resulted in the MP-EST* TENT species tree, first computed a separate tree for each of 14,000 genomic regions and then combined those into a single tree. The problem with this method is that trees based on individual genes or regions can look very different, largely as a result of a phenomenon called "incomplete lineage sorting," ILS, where parts of a genome evolve differently from other parts over time. ILS has been "known to confound the analysis in many different species," Warnow said, and there has been "very substantial debate in the community" on how to analyze the discordant gene trees.
To get around this problem, her team developed a technique called statistical binning that improves species trees when the gene trees are heterogeneous. "It actually gives better gene trees, so you can combine them into a more accurate species tree," she said.
Because of the new approach, the two species trees looked very similar in the end, which "gave us great confidence in what we were finding," she said. The new technique will be useful for the analysis of any other phylogenomic datasets, she said, and her team plans to apply it in insect, mammal, and plant evolutionary genomics projects.
Going forward, the consortium plans to sequence additional bird genomes, and to study the function of highly conserved elements in the 48 genomes and their contributions to morphological differences. "We're not stopping here," Jarvis said.