Predicting Phenotypes in a Changing Climate
Carolyn M. Beans · BioScience · 2017
Statistical models merging genetic and environmental data sets suggest that the balsam poplar (Populus balsamifera), shown left near Jericho, Vermont, and right near Banff, Alberta, Canada, may be maladapted to the future climate conditions projected for the northernmost part of its range. Photographs: Stephen Keller, University of Vermont. Evolutionary ecologist Stephen Keller, of the University of Vermont, wants to walk into a stand of balsam poplars (Populus balsamifera), sequence the genome of a tree, and know precisely when it will break dormancy each spring. He would also like to know when it would break dormancy in another location or a future climate. He cannot do those things—not yet. But today, researchers are closer than ever before to answering this longstanding biological question: How can you predict an organism's phenotype from its genes and its environment? A century ago, phenotype largely meant morphological traits such as height, coat color, or the shape of a seed. The definition now includes behavioral and physiological characteristics. Keller studies the timing of when a tree breaks dormancy because, like many phenotypic traits, it is hugely important for survival and reproduction. “You don’t want to start too early because you might get nipped by a late spring frost and lose the resources you expended on making your leaves. But if you start too late, then you miss out on some of the growing season,” Keller explains. Scientists have long recognized the societal benefits that predicting phenotypes could offer. Farmers could identify the location where a crop variety would grow best. Doctors could explain why one gene carrier falls ill whereas another remains healthy. Now, in an age of rapidly changing climate, Keller and others are racing to solve this puzzle in time to predict whether and how species will survive under future environmental conditions. Many challenges remain, but with unprecedented access to genomics, environmental data, and computational resources, the potential to predict phenotypes in future climates looks increasingly promising. Last year, the National Science Foundation (NSF) included “Understanding the Rules of Life: Predicting Phenotype” in a list of key research directions. The proposed agenda stated that “the universally recognized biggest gap in biological knowledge is our inability to predict an organism's observable characteristics—its phenotype—from what we know about its genetics and environment.” “People from Mendel on have been trying to associate how genes are responsible for producing phenotypes,” says molecular endocrinologist David Durica, of the University of Oklahoma. Durica is a member of the Animal Genome to Phenome Research Coordination Network, a collaboration of scientists in diverse study areas, all focused on how genes translate into phenotypes. There is a deceptively simple equation scientists often use to explain this translation. The phenotype (P) is the result of genes (G), the environment (E), and the way genes are influenced by the environment (GXE). Scientists have made great strides in elucidating the complex genetic component of that equation. Within the time span of his own career, Durica says, “We’ve been able to go from a single-gene approach to looking at entire populations of genes.” But predicting exactly how genes will respond to different environments still confounds scientists. Researchers often try to tackle the problem with a “common garden.” They place individuals with different genotypes in the same place so trait differences can be assigned to genotype rather than environment. By repeating this setup in one or more other locations, they can measure how the same genes affect phenotypes differently in different environments. But understanding how a gene affects seed size in a cool versus a warm climate does not necessarily predict what happens at a temperature in between. “That's the limitation right now,” says Keller. “We need a better understanding of what the G-by-E relationship looks like for most traits through a continuous environmental gradient and not just at one common garden or two common gardens.” Scientists who want to predict how a species will respond to a changing climate often look for clues in its current range. If the species thrives in a warm, balmy region, it may fare equally well in areas projected to have similar climates in the future. But in reality, species are divided into populations—pockets of individuals adapted to unique environments in their own corners of the range. The genes that make one population well suited to a future climate may be at low frequency or completely lacking in another. Keller is tackling this problem by building statistical models that merge genetic and environmental datasets. These models pinpoint the genes underlying local adaptation and also predict where in the species range there is likely to be a gene–environment mismatch under future climate conditions. “The field has been working toward this direction for a long time,” says Keller. “Once genome-wide sequencing became accessible to a broader variety of ecologically relevant species, then people started wanting to identify the portions of the genome that are responsible for the local adaptive variation that evolutionary ecologists have always been interested in.” Keller sampled DNA from over 400 balsam poplars from 31 populations scattered across a range stretching from Newfoundland west to Alaska and south into the Rocky Mountains. He then sequenced portions of the genome that he knew were related to plant flowering times. He downloaded data on environmental variables such as mean summer temperature and summer precipitation from WorldClim, a freely available global climate database. Keller then plugged that genetic and climate data into statistical models that identify the relationships between gene frequencies and environmental variables. The models could show, for example, whether the frequency of a particular gene increases as populations experience increasingly cold winters. Keller's models are complex. They account for correlations and interactions among environmental variables. And they can handle nonlinear relationships between gene frequencies and the environment. Annual rainfall, for example, may have no effect on the frequency of a drought-tolerance gene until a forest becomes so dry that only individuals carrying a particular gene variant survive. Keller reported in Ecology Letters in January 2015 that these models showed a strong association between how warm an area is and which variant of a gene known as GIGANTEA is most common there. Previous studies suggested that this gene may control bud flush in poplars. Keller's models suggest that the gene may be locally adapted in different areas, allowing individual populations to tune their bud-flush times to the particular temperatures in their portions of the range. But this gene–environment relationship may break down in some regions under future climate conditions. Keller's models suggest that there will be a major mismatch between GIGANTEA gene variants and temperature in the northern edge of the range, especially in northwestern North America. For poplars to continue to thrive in this area, the present-day populations would need to evolve, or better-suited populations would have to expand northward. Ecologist Matthew Lau, a postdoctoral fellow at Harvard University's Harvard Forest, wants to make the link between genes and ecosystems. Three years ago, he joined a multi-university team that is using ants as a model system for predicting how ecosystems in the eastern United States will respond to climate change. “We want to know how ecosystems tick, and genomics is a powerful tool for providing a window into that,” says Lau. Genomics can also offer insights into unseen ecosystems, he says, “ecosystems of the future that will be coming into existence under a changing climate.” Researchers tested how ant communities responded to warming chambers at Harvard Forest, in Massachusetts, that mimicked the effect of projected climate change. Photographs: Shannon Pelini, Bowling Green State University. Others in his group already showed that warming temperatures affect the success of individual ant species, as well as how those species interact with one another. The team set up a series of warming chambers at Harvard Forest, in Massachusetts, and Duke Forest, in North Carolina, that elevated the forest-floor temperature to mimic a range of climate-change scenarios. They placed nesting boxes inside the chambers and then tracked individual ant species as they moved into, occupied, and disappeared from boxes. In an October 2016 paper in Science Advances with lead author Sarah Diamond, of Case Western Reserve University, the team reported that, in many cases, temperature had a direct effect on a species’ ability to colonize and maintain a nest. But temperature also indirectly affected species through its effect on other potentially competing species. Lau has good reason to suspect that genetics underlie these findings. In a study his group published in March 2016 in BMC Genomics with lead author John Stanton-Geddes, of the University of Vermont, researchers exposed two ant species to 12 different temperatures. By studying the ants’ transcriptomes across this gradient, they discovered that at least 2 percent of all genes in the ants’ genomes changed expression in response to temperature. Lau is building on these results with a genome-wide search for the genetic underpinnings of climate adaptation in six ant species in the genus Aphaenogaster. He and the Harvard Forest team worked with researchers at the Eli and Edythe L. Broad Institute of Massachusetts Institute of Technology and Harvard to sequence the entire genomes of these species. By identifying where genetic sequences vary among species living in different portions of eastern North America, Lau hopes to identify the genes that are key for coping with varying climate conditions and also possibly the genes that guide behaviors such as seed dispersal and interactions with other ant species. “It could point the way to finding genetic variants of ants, as well as other insect species, that are signatures of what their response is going to be to climate change,” he says. Understanding the ant response will also lead to better predictions of how climate change will affect key ecosystem services ants provide, such as turning over soil, decomposing waste, and dispersing seeds. Environmental physiologist Donal Manahan, of the University of Southern California, is searching for what he calls the “winners” of ocean acidification—individuals within a species with phenotypes that could, in the short term, help them survive a new climate and, in the long term, help the species as a whole adapt. His focus is the Pacific oyster, Crassostrea gigas, a commercial species commonly raised in aquaculture. In recent years, oyster hatcheries along the West Coast of the United States have reported failures of oyster populations at the larval stage, which many researchers attribute to increasing ocean acidification. Manahan reported in the ICES Journal of Marine Science in 2017 that increased acidity affected larval shell length differently in different Pacific oyster families. In one family, there was no effect. In others, there was as much as a 28 percent decrease. The fact that one family was less affected than others suggests that commercial aquaculture could target this trait in future breeding programs. Predicting how species will respond to climate-change scenarios cannot be done by experimenting on wild-type organisms taken out of the ocean—or a lake or a tree, says Manahan. He argues that to really understand the genetic and genetic-by-environment part of the equation, researchers need to use model organisms. “It better have a genome. It better have pedigreed genetic lines,” he says. When Manahan began studying the Pacific oyster two decades ago with quantitative geneticist Dennis Hedgecock, also of University of Southern California, these lines did not exist. “And can you buy them off the shelf?” says Manahan. “No. You have to make them, and that involves decades of marine-animal breeding.” In fact, prior to this team's oyster work, there were no marine models. “Certainly, we have marine organisms that have genomes. The sea urchin is a good example. But there are no genetic pedigreed lines, so you can’t do genetic crosses and make different phenotypes and then ask, ‘How did that work?’” Scientists are launching genome-wide searches for the keys to climate adaptation in six ant species from eastern North America, including Aphaenogaster picea, shown here. Photograph: ©Aaron M. Ellison. Now, with the marine model established, Manahan and Hedgecock can combine the tools of genetics with biochemistry and physiology to continue the search for the mechanisms that may allow Pacific oyster populations to adapt to a changing climate. “Trying to predict the future from today's biology is really hard. But if you have these kinds of tools, you can find the needle in the haystack,” Manahan says. Capturing phenotypic measurements is often the most time-intensive variable in the genotype-to-phenotype equation. Enter the age of high-throughput phenotyping, or phenomics. Scientists can now use robots to precisely measure myriad traits in plants and even some model animals. In 2011, biochemist Argelia Lorence, of Arkansas State University, was the first to bring a plant-phenotyping robot to a university in the United States. “Robots usually don’t make mistakes,” says Lorence. “They are more precise. They don’t have any biases. If you and I are asked to jot down which group of plants is greener than others, we might have different results.” Lorence purchased the robot from the German-based company LemnaTec using NSF funding awarded to a statewide research consortium called the Arkansas Center for Plant-Powered Production. The instrument, known as Scanalyzer HTS, has a robotic arm fitted with four cameras that hover above potted plants in a lab, each snapping images in different wavelengths. A visible camera captures size, architecture, and color. A fluorescence camera captures chlorophyll fluorescence, a measure of plant health. A near-infrared camera measures water content. And a far-infrared camera measures leaf temperature. Lorence is measuring size, shape, color, and water content in maize seeds as part of the Genomes to Fields project, a nationwide effort to predict how unique corn varieties will grow in varying environments using common gardens in over a dozen states. Increasing ocean acidification threatens oyster hatcheries along the West Coast of the United States. Researchers at the University of Southern California are looking for the mechanisms that may enable the Pacific oyster (Crassostrea gigas), shown here as two 10-day-old larvae (approximately 150 μm in shell length), to adapt to a changing climate. Photograph: Scott L. Applebaum, University of Southern California. Through the NSF-funded Plant Imaging Consortium, the instrument is also available to researchers in Missouri in exchange for access to molecular imaging resources available there. And researchers throughout the United States can use it at cost. Other academic institutions, such as the University of Nebraska, now have their own plant-phenotyping devices. The University of Arizona's Maricopa Agricultural Center has one that captures images directly from an agricultural field. Arkansas State University's high-throughput phenotyping instrument helps researchers rapidly collect unbiased measurements of seeds and plants. Photograph: Lucia Acosta-Gamboa and Argelia Lorence, Arkansas State University. The scores of genetic, environmental, and phenotypic data required to predict phenotypes are only valuable if safely stored and effectively analyzed. The NSF recognized this issue in the mid-2000s. “They were putting hundreds of millions of dollars or more into generating genomic data in lots of small labs across the country,” says developmental biologist Parker Antin, of the University of Arizona. “Three people and a PCR [polymerase chain reaction] machine is a lab, and all of a sudden, they could generate lots of genomic data but had no place to deal with it. Sometimes, their files were too large to open on their computers.” CyVerse. The Project. (16 May 2017; www.cyverse.org/about) Diamond SE, et al. 2016. Climatic warming destabilizes forest ant communities. Science Advances 2 (art. e1600842). doi: 10.1126/sciadv.1600842 Fitzpatrick MC, Keller SR. 2014. Ecological genomics meets community-level modeling of biodiversity: Mapping the genomic landscape of current and future environmental adaptation. Ecology Letters 18: 1–16. et al. of in larvae under ocean acidification. ICES Journal of Marine Science et al. 2016. to in warm and ant species. BMC Genomics (art. In the NSF the a computational by the University of that plant scientists to and The to all and in the changed to CyVerse. Antin, the lead of says they now have from and to of Many make their data and the to it available to other The team wants to enable researchers to these diverse data sets to predict phenotypes. really in the of genotype to whether about plants or says one of our he are of different data able to search the data is They are a system to to files to make them more increasingly accessible and computational resources, as well as the already made by Keller, Lau, Manahan, and others, scientists still to predict phenotypes. For phenotypic measurements remains a “We need better to phenotype in populations that says Keller. this in by bud flush using data by the National and can measure the time the that a area from in to up in says Keller. it get you down to the of an individual with at the University of and State University, Keller is also known as that to tree in the the “The tree in spring it breaks bud and its which a change in the frequency with which it which the up Keller explains. In a study the the from dormancy to spring leaf at the physiological is also There are high-throughput that researchers about in transcriptomes and of small called that are by there is no high-throughput to look at the physiological of these genetic says Manahan. are good at morphological but trying to get into a and look at the of of interactions and what happens when you change a is much Stephen Keller, of the University of Vermont, is known as that when individual break dormancy each spring. Photograph: Stephen Keller, University of Vermont. Manahan also to the of phenotypes from different of biological make and do make is the of molecular Manahan. a that gene expression and often that in the of in change the But in a 2016 study published in the Journal of Marine and he and his reported that this did not They discovered that of across could predict which of oyster would grow the But the of a key in did not predict the And expression of the gene that for this key predict the of the “The are and why phenotypic is so says Manahan. you just that if gene expression that will change the of and that will change the of you might be is scientists to with the increasingly complex data sets required for tackling this can handle people who in and just know how to point and says you know you can in more and get a more out of The team a of from the to building a more computational and research a It has to be in the new that scientists are says not with increasingly computational scientists might be within of the puzzle of how genes and environment interact to predict a you know the genetic variation and you know the with the then you be able to predict its phenotype and its of adaptation in any says Keller. “That's the