Researchers at LSU, Baylor College of Medicine and Rice University have published a paper that sheds light on how traits in living organisms, which are characteristics such as tail length in dogs or egg size in birds, are related to one another and to organisms' underlying genetic code. Morgan Kelly, an assistant professor in the LSU Department of Biological Sciences who studies ecology, evolution and conservation of marine invertebrates, is a co-author on the paper.
To learn more about trait correlation, we asked Morgan and her co-author Julia Saltz, an assistant professor in BioSciences at Rice, to tell us about the insights they gleaned while writing "Trait Correlations in the Genomics Era," a Trends in Ecology & Evolution review paper.
College of Science: What are trait correlations? Why are they interesting or why do we care about them?
Morgan: Trait correlations occur whenever particular combinations of traits are more common in a population than others. For example, let's take the traits 'egg size' and 'egg number.' These two traits are correlated if some individuals lay a few, large eggs, while others have many small ones, but it is rare to lay many eggs that are also large.
One of the trait correlations we work on in my lab is the correlation between heat tolerance and fecundity in the tide pool copepod Tigriopus calfiornicus. We find that as copepods evolve greater heat tolerance, they also evolve reduced birth rates. This has important implications for species' responses to climate change, because it means that even if species are able to adapt to rising temperatures, this adaptation might slow their population growth rates if the cost of being more tolerant is investing less energy in having babies. This work was done with LSU undergraduate co-authors, including Vidal Villea.
Trait correlations are interesting for a couple of reasons. First, understanding trait correlations can help us to understand how evolution works - why some evolutionary paths might be more common than others. Second, trait correlations can be important to understanding many human diseases, which often manifest as a whole set of symptoms that cluster together.
Julia: One thing about trait correlations that is interesting is that they have relevance to different questions in different fields. The example Morgan provided is from life history theory, where the questions are about how organisms maximize fitness in the face of tradeoffs. Just like you can't spend all your money and save it all for the future, animals can't produce an infinity of giant eggs.
In animal behavior, correlations among behaviors (a commonly-observed one is activity and aggression) are more of a surprise, because intuitively it seems like animals should be able to have any combination of values for these traits (for example high aggression and low activity), but instead we see correlations. Here, we are often asking whether these behavioral correlations represent constraints, like the life history tradeoffs, or something else.
In human disease, researchers are often interested in co-morbidities, or correlations among diseases. Often, if you have disease A, you are at greater risk for disease B. Here, the goal is to identify the genetic causes of co-morbidity as a window into the etiology of these diseases.
College of Science: What motivated you to write this review paper?
Morgan: Dr. Saltz and I read Lande and Arnold’s paper titled "The Measurement of Selection on Correlated Characters" in a class we took together in graduate school. This paper had a really important impact on the field of evolutionary biology, because it made the prediction that traits that are correlated may not always be able to evolve independently. When the paper was written (in 1983), no organism’s genome had been fully sequenced. Now, 30 years later, it is common to use genomics to investigate trait correlations. Dr. Saltz and I wondered what new information had been learned from these techniques. Had any of this new information changed scientific understanding of trait correlations?
Julia: Our next update will be due in 2043!
College of Science: Why is it important to understand the genetic basis of trait correlations? How has new technology to sequence genomes changed the study of trait correlations?
Morgan: Before DNA sequencing, scientists could observe that a given set of traits were correlated, but they couldn't identify the actual DNA sequence differences that were causing variation in those traits. Now that whole genome sequencing is relatively common, scientists have been able to identify the actual genes and regions of the genome where variation in DNA sequences leads to variation in the traits they're studying. This has led to the discovery of a number of surprising examples where variation in a single gene produces variation in multiple (sometimes seemingly unrelated) traits.
College of Science: What is your favorite trait correlation in a species?
Morgan: My favorite example (which I learned from Dr. Saltz) is the T gene in dogs. There are 17 dog breeds that have naturally short tails. The version of the T gene that produces short tails has a mutation that disrupts the ability of the T protein to bind to DNA. Dog breeds with short tails never have two copies of this gene, which suggests that any dog that inherits two copies of this gene (1/4 of a typical litter in these breeds) must die in utero. Thus, a version of a gene that was favored by breeders also has a critical effect on the animal's viability.
Julia: My favorite was one that we didn't include in the paper because no one has investigated its genetic basis. It is the genetic correlation in some bluebirds between helping at the nest (instead of leaving to start their own nest) and being helped later by their own offspring. Theory predicts a strong genetic correlation, but Anne Charmantier and colleagues actually measured it in wild animals. The genetic correlation was indistinguishable from a perfect correlation of r=1! I remember gasping aloud when I read that, which I probably shouldn't admit because that's kind of weird.
College of Science: What were some of the main takeaways from your review paper? Did you gain any insights in writing this paper that you didn't have before?
Morgan: The basic predictions laid out by Lande and Arnold have not been changed by new genetic information. However, at the time they wrote their paper, it was common to think of trait correlations mainly as an obstacle to evolution. (For example, natural selection might favor having many eggs AND having large eggs, but you can't have both). We were surprised by the number of examples where it is clear that natural selection has favored trait correlations, and as a result has favored genetic mechanisms that allow particular trait combinations to be inherited together.
College of Science: In your paper, your write that DNA mutations causing these correlated traits can be both pleiotropic and linked in different ways. What does this mean?
Julia: A mutation is pleiotropic if it affects more than one trait. For example, if you're a stickleback fish, you may have a mutation in the Eda gene that has pleiotropic effects on growth rate and shoaling behavior: individuals with the mutation both grow faster and shoal less, compared to individuals without the mutation.
Linkage refers to how close together mutations are on a chromosome. When mutations are very close together, if you inherit one of the mutations from a parent you almost certainly inherit the other one too. It's like inheriting china - if you get the dishes, you'll probably get the coffee cups, too. In this case, if a mutation that affects growth rate is very close to a mutation that affects shoaling behavior, then fish who grow faster will also shoal less, because they have both mutations.
Linkage and pleiotropy are often described as "either/or." If scientists observe a correlation between two traits, it's either due to one pleiotropic mutation, or two linked ones. But we found evidence that most trait correlations are caused by many mutations, not just one or two. One mutation could be pleiotropic, and others could be linked, and they could all contribute to producing the same trait correlation. There's no reason that a pleiotropic mutation can't be in linkage with other mutations (pleiotropic or not!) So we emphasize that these aren't really mutually-exclusive options. They are just different ways that mutations can produce trait correlations, and they could all be happening at the same time.
College of Science: What was the most rewarding aspect of working on this research? What was the most challenging aspect?
Morgan: The most challenging part of this project was attempting to synthesize a large body of literature in only 3,000 words. The most rewarding was our collaboration. Dr. Saltz has always been one of my favorite people to talk to about science, and so writing this paper was a great excuse to get to do this on a regular basis!
Julia: Yes, there's a huge and diffuse literature on trait correlations, so it was hard to get to generalizations without worrying that we had missed something (although we probably missed a lot). But, that's also how review papers benefit the community - we did all that work so you don't have to!
Read the paper: Trait Correlations in the Genomics Era.