« April 2007 | Main | June 2007 »

May 29, 2007

Coevolution and gene flow

Two species coevolve when changes in either lead to changes in the other. This includes "arms races" between species that compete with each other, but also interactions that benefit both species. "Gene flow" is the movement of genes from one population into another, of the same or related species. For example, some genes in modern cows seem to have come from mating with wild aurochs, before they went extinct. Gene flow often provides new genes; some may be useful to the recipient population. For example, pollen from transgenic sugar beets could transfer herbicide resistance (along with other crop genes) to related weed beets. More often, genes that were useful in the source environment may be harmful to the recipient population. Natural selection will tend to eliminate these, unless gene flow rates are too high. For example, if plants growing on toxic soil around an old mine are outnumbered by neighbors on nontoxic soil nearby, gene flow may swamp natural selection, preventing evolution of tolerance to toxic soil.

This week I'll discuss a review article on coevolution and then an experimental paper showing how gene flow can affect coevolution. The review is "Variable evolution" by Elizabeth Pennisi, published in the May 4 issue of Science. It discusses coevolution of wild parsnip with the webworms that eat them and coevolution of pine trees with birds and squirrels, among other topics.

Like most plants, wild parsnips produce natural insecticides that keep most insects from eating them. 150 years ago, webworms that eat parsnips followed them to America from Europe. Although The Origin of Species was published two years later, nobody at the time was paying attention to how webworms affected the evolution of wild parsnip. But, fortunately, botanists were collecting plant samples and preserving them in herbariums.

Samples of plants, soil, etc. collected over time are really valuable sources of information. They are often used for answering questions that couldn't even have been asked, much less answered, when the samples were collected. The long-term experiments at Rothamsted (UK) have collections of crop and soil samples going back over 150 years. Analyses of soil samples collected there before pH was even defined as a measure of acidity show gradual acidification of the soil with some fertilizers. Similarly, herbarium samples have shown how plants have changed in response to rising CO2.

It's too bad that such collections often get thrown away. That's what had happened to Hans Jenny's old soil samples, when a colleague went looking for them to figure out why cotton yields were decreasing in California. My former department even talked about getting rid of our herbarium, to make room for a new molecular biology lab. It's ironic, because molecular methods have greatly increased the amount of information we can get from old samples. For example, scientists at Rothamsted have used molecular methods to count fungal pathogen spores in wheat samples collected over decades. When I helped set up the UC Davis Long-Term Research on Agricultural Systems experiment, I collected soil samples using sterile sampling methods, so that DNA in the samples could someday be used to study the ecology and evolution of soil microbes in response to agricultural practices. Robert Norris collected and saved weed seeds so that future weed scientists could study their evolution.

But back to webworms. When Arthur Zangerl and May Berenbaum, both of the University of Illinois, analyzed old parsnip samples from herbaria, they found that toxin levels had increased soon (within 20 years) after the webworms arrived. Since then, the webworms have also evolved, so they are now mostly resistant to the parsnip's natural insecticides. Webworms in Europe haven't evolved as much resistance, because there they mostly eat a less toxic plant.

There's a lesson here for those interested in herbal remedies. Many of the chemicals that make plants useful as drugs (caffeine and aspirin, for example) are natural insecticides that tend to evolve rapidly. "Ancient wisdom" about the value of certain herbs in medicine may have been true at the time. But the plants have evolved since then, so they need careful testing again, using statistically valid methods, of course. Furthermore, the same species may have very different chemical properties in different places. Wild parsnips differ between America and Europe, but differences over much shorter distances can also be important.

For example, Pennisi's review discusses how coevolution of pine trees and birds varies across the western US. The birds open pine cones to get the seeds, which they eat. But they also hide a lot of seeds, some of which grow into new trees when the birds forget where they hid them. On balance, this is a good deal for the pines, so they have evolved cones that are easy to open. But not everywhere.

Where squirrels are common, as well as birds, coevolution has led to cones that are harder to open. Maybe squirrels have better memories, and return to eat most of the pine seeds they stash? Whatever the reason, we end up with what John Thompson called a "geographic mosaic of coevolution", a patchwork of areas where coevolution has, or does, vary in rate or even direction.

How does gene flow affect coevolution? That is the subject of a paper in the June issue of The American Naturalist. Samantha Forde, John Thompson, and Brendan Bohannan, of UC Santa Cruz and the University of Oregon, showed that "Gene Flow Reverses an Adaptive Cline in a Coevolving Host-Parasitoid Interaction." Does that sound a little more intimidating than "Variable Evolution"? Well, American Naturalist is a more specialized journal than Science.

We often see trends across landscapes in coevolution between species. These "clines" are hard to interpret. Is some physical factor affecting the evolution of one or both species? Have two parts ("populations") of a species that evolved separately for a while come back into contact as warming climate re-opened a mountain pass? Or is gene flow important, for one or both species?

Forde and colleagues set up an experiment in which these factors could be controlled. They used the gut (and lab) bacterium E. coli and a virus, T7, that infects it. (Although lab strains of E. coli aren't very dangerous, someone needs to find a bacteria that's as easy to work with as E. coli, but can't grow in humans!) The bacteria can evolve resistance to the virus, but then the virus can evolve so that it attacks the "resistant" bacteria, and so on. Both species evolve faster in food-rich environments, because a higher growth rate leads to larger population size. If one in a million is a mutant, then the more bacteria, the more different mutants natural selection will have to select among.

After letting bacteria and viruses evolve together in rich, medium and poor environments (flasks differing in nutrient supply), they tested viruses from the rich environment, to see how well they infected bacteria from the three environments. The bacteria from the low-growth environment hadn't evolved as much resistance as those from the high-growth environment, so they were easier prey for the virus. Imagine a nuclear submarine attacking a WWI battleship.


That was without gene flow. But when they allowed a few of the bacteria and viruses to move between flasks, the trend reversed. With gene flow, viruses from the high-growth environment did worse on bacteria from the low-growth environment than on those from their own environment. Gene flow would have transferred resistance genes to bacteria in the low-growth environment, and apparently the bacteria from that environment had some additional differences to which the high-growth virus was not adapted.

The paper has additional data on the infection success of viruses from the low-growth environment, and on the effects of gene flow among environments that all have the same growth rates.

Can we generalize from these results to coevolution with gene flow in natural landscapes? There may be valid theoretical arguments one way or the other. But I would apply the same criterion proposed in previous posts: I would like to see whether other labs, doing similar experiments but with different species, get similar results. The great thing about microbes is that multigeneration evolution experiments can be done in days or weeks rather than years.

May 20, 2007

Rapid evolution of beneficial infections

Given my location halfway between the Twin Cities of Minneapolis and St. Paul, and my childish love of clever acronyms, I sometimes wish I'd named this blog This Week In Natural Selection. But then I suppose I'd have to review a pair of closely related papers each week. I'm going to do that this week, anyway.

This week's twins were both published in PLoS Biology, so both are freely available on-line. Both have new data on bacteria that infect insects. Both help us understand the conditions under which infecting bacteria evolve to be beneficial, rather than harmful. Finally, both disprove, again, the popular idea that any evolutionary change big enough to matter (except antibiotic resistance, which a creationist commenter once claimed always involves "horizontal transfer" of genes among bacteria, even though resistance often evolves in bacteria in a closed container all descended from a single cell) always involves lots of genes and takes millions of years. Evolution is our present and future, not just our past.

The first paper is "From parasite to mutualist: rapid evolution of Wolbachia in natural populations of Drosophila" by Andrew Weeks, my former colleague Michael Turelli, and three others (Universities of Melbourne, California at Davis, and Texas).

The popular idea that pathogens (microbes that cause disease) always evolve into mutualists (which help their hosts) "because otherwise both species will eventually go extinct" is wrong. Natural selection is never guided by future benefits, any more than a river will run uphill if that's the shortest route to the ocean.

On the other hand, there are some conditions that favor the evolution of mutualism. Bacteria or viruses that can easily spread from a very sick (or dead) host to a new host will often become more deadly over the course of evolution, as discussed in a previous post. But a microbe that only reproduces when its host does may evolve in a way that increases the host's chances of reproducing. (This may not be true if each host individual has several competing strains of the microbe. In this case "It may be better to keep alive the goose that lays the golden eggs than to kill it. But this argument depends on the assumption that, if you do not kill the golden goose, no one else will either: that is, it assumes that the host is infected by a single clone of symbionts." Maynard Smith 1989 Nature 341:284-285.)

Wolbachia are bacteria that infect various insects, including the fruit-fly Drosophila. Twenty years ago, female fruit-flies from California that were infected with Wolbachia laid more eggs if they were "cured" of their infection using antibiotics. So Wolbachia was harmful to fruit-flies, then. But because Wolbachia is transmitted mainly in eggs, the authors thought they might evolve to be helpful instead. So they collected fruit-flies from several locations and measured the effects of Wolbachia on egg production, using the same antibiotic method. Sure enough, curing the infection now reduces egg production.

Does this prove that the Wolbachia have evolved so that they are now beneficial rather than harmful? Not quite. Suppose Wolbachia have both beneficial and harmful effects, such as consuming energy inside the host but also making some vitamin. If the hosts (fruit-flies) evolved so that they lost the ability to make that vitamin themselves, then curing them of bacteria could reduce their egg production, but they might still be worse off than their uninfected ancestors (de Mazancourt et al. 2005 J. Ecol.). In this case, however, the authors showed that it was the Wolbachia that have evolved, not the fruit-flies.

Even the evolved Wolbachia can reduce egg production, however. When an infected male mates with an uninfected female fruit-fly, many of her eggs fail to hatch. This "you're either with us or against us" strategy helped Wolbachia spread rapidly in California.

This week's second paper is "Aphid thermal tolerance is governed by a point mutation in bacterial symbionts" by Helen Dunbar and colleagues at the University of Arizona.

Whether aphids infected with the bacterium Buchnera are better or worse off than their uninfected ancestors is not known, as far as I can tell. But now, at least, they depend on them for essential nutrients.

Dunbar et al. looked at a Buchnera mutation, seen both in lab and field, that reduces a biochemical response to high temperature. With high temperature exposure, the mutants apparently died, which greatly reduced the reproduction of their aphid host. Under cool temperatures, however, aphids containing the mutant Buchnera reproduced earlier and laid more eggs total. The mutant bacteria are fairly common in aphids in the field, probably reflecting back-and-forth evolution with varying temperature exposure in time and space.

With respect to the gene studied, the bacteria and the aphids have a shared interest in adapting to the temperatures to which they are exposed in the field. This contrasts with the more complex case of Wolbachia, where bacteria and host have both shared interests and conflicting interests. Are there other genes in Buchnera where a mutation could benefit bacteria while hurting their host? Or has this symbiosis evolved to the point where there are few such conflicts of interest? These aren't the first interesting papers I've seen on Wolbachia and Buchnera, and they probably won't be the last.

May 18, 2007

Helpful cheaters?

Paul Rainey has a very interesting essay in the April 5 issue of Nature. Much of what we know about "cheating" in bacteria that form floating mats comes from his research, including collaboration with Michael Travisano, recently hired here at University of Minnesota. See my earlier post, "how disturbed are cheaters", for background on this system. Although cheaters that don't invest in the goop that holds floating mats together can result in mats breaking up and sinking, Rainey's new essay suggests that a similar form of cheating may have contributed to the evolution of multicellular life.

Here's the problem. As long as cells reproduce independently, natural selection will favor anything that helps individuals reproduce more in a given generation, even if all the cells would have more descendants, over the long term, if they all cooperated. It's a tragedy of the microbial commons.

If selection operated on groups, rather than individual cells, that would favor mechanisms to suppress cheating. Humans can and do impose group selection (see Darwinian Agriculture), but group selection strong enough to overcome individual selection is thought to be rare in nature.

However, Rainey points out that cheaters don't just disrupt groups of floating cells. They also tend to swim away, where they could potentially found new groups. For the new groups to be most successful, they would need to float, so most individuals in a group would need to produce goop. Therefore, cheaters that have lost the goop gene, thought mutation, wouldn't form successful groups.

But what if, instead of losing the gene altogether, they just turn it off long enough to swim away? If they then turn it back on again before reproducing (by dividing), they could form a new floating group. Therefore, a genotype that keeps most cells in the goop-producing mode (keeping mats floating), but with a few cells not making goop (swimming away and founding new groups), might out-compete genotypes with no cheaters at all. Rainey suggests that the swimming cheaters are analogous to germ cells (sperm or eggs), whereas the cooperative goop-producers are more like somatic (nonreproductive) cells.

Could the division of labor between reproductive and nonreproductive cells, a key to multicellular life, really have arisen by a similar process? I have no idea. I look forward to reading responses to Rainey's essay from other scientists.

I will point out that nothing in the essay suggests that most forms of "cheating" are beneficial. What about rhizobium bacteria that infect bean roots but don't provide their host with nitrogen (my specialty), bees that cut through the base of a flower to take nectar without pollinating, or people who cheat on taxes? If any of these forms of cheating are beneficial to anyone other than the cheater, the reasons would have to be very different from those proposed for Rainey's floating bacteria.

P.B. Rainey 2007. Unity from conflict. Nature 446:616.
P.B. Rainey and M. Travisano. 1998. Adaptive radiation in a heterogeneous environment. Nature 394:69-72.
Rainey PB, Rainey K. (2003) Evolution of cooperation and conflict in experimental bacterial populations. Nature 425: 72-74.

May 14, 2007

Who are you?

Visitors per day: 42 (one of my favorite numbers)
Longest visit: 52 minutes
Only 26% use Internet Explorer (obviously a sophisticated bunch!)
Leading source of incoming links: Terry Tao's "What's new" math site (wow!) and Carl Zimmer's "The Loom"
Second-most-common language: Usually Portuguese (equal numbers from Portugal and Brazil) or Finnish
Islands: Islas Canarias (Spain), Iceland, Ireland, UK, New Zealand (both islands), Honshu (Japan), Dominican Republic, Australia? Kama`ainas too busy with real surfing?
Under-represented continents: Africa (cradle of human evolution) and Antarctica
Favorite compliment: "er zu einer wirklich raren Spezies gehört: ein Science Blog (fast) ausschließlich über Science (*gasp*)"

Evolution of babysitting in bluebirds

Major transitions in evolution have often involved loss of independence, as discussed last week. Most female bees work to increase their mother's reproduction, rather than laying eggs themselves. Less extreme examples of helping others reproduce are known in some animals. "Kin selection" favors helping relatives, if the cost of helping is less than the benefit to the one helped, times their relatedness to the helper. This is known as Hamilton's Rule. As Haldane put it, "I would jump into a river to save two brothers or eight cousins." "Cost" and "benefit" are measured in number of offspring and "relatedness" is relative to one's usual competitors. If surrounded by cousins, Hamilton's Rule would lead to helping only siblings.

For helping behavior to have evolved, there must have been genetic variation in helpfulness. This week's paper shows that this is still true for western bluebirds in Oregon.

Anne Charmantier (anne.charmantier@cefe.cnrs.fr), Amber Keyser, and Daniel Promislow, of Oxford and the University of Georgia, published "First evidence for heritable variation in cooperative breeding behavior" online in Proceedings of the Royal Society.

1593 bluebirds were observed at nest boxes by volunteers with the Prescott Bluebird Recovery Project. They recorded which birds helped feed or defend babies other than their own and which birds received such help. Only a small fraction of breeding pairs got help. 70% of helpers were sons of the pair helped, 16% were brothers, and 6% were daughters.

Because they had a seven-generation family tree for these birds, they were able to determine whether helping is inherited. It was, with a heritability of 76%. The authors also mention the possibility that inheritance could be cultural rather than genetic, if individuals copy relatives they see helping.

With such high heritability, they note that "rapid microevolutionary changes in the extent of cooperative breeding are possible." But microevolution in which direction? If the cost of helping is low (because helpers could not have found their own breeding territory), then kin selection favors helping relatives to reproduce. Helping may also enhance individual survival, if those that don't help are chased away and have nowhere to go (Trends in Ecology and Evolution 17:15-21).

On the other hand, if unoccupied breeding territories are available, then individual reproduction will have a higher payoff than helping. In that case, high heritability could lead to a rapid evolutionary decrease in helping. Rapid evolution back and forth over a few generations may cause little long-term change.

May 6, 2007

How disturbed are most cheaters, really?

Yesterday, my wife asked, "why are there so many theoretical papers in evolutionary biology?" I suggested one reason may be that evolutionary theory is better developed, in the sense of making accurate predictions, than theory in much of biology. This week's paper, comparing results from an evolution experiment to predictions of a mathematical model, is a good example.

The paper is about the evolution of cooperation. This is a hot topic and also my own area of research. Humans enforce cooperation, to varying extents. For example, we often punish cheaters, those who try to benefit from cooperative activities of others without contributing anything themselves. Human cheaters are mostly pretty stupid -- don't even think about plagiarizing this blog for a term paper! -- but what about cheaters with no brains at all?

Cooperation among cells is essential to multicellular life. Adult plants and animals are made up of billions of cells, which cooperate in various ways. Only sex cells (sperm, eggs, pollen, etc.) have direct descendants in future generations. If one of your lung cells starts reproducing on its own, we call it cancer. The evolution of multicellular life must have required, at some point, that most cells in a group give up individual reproduction. But why would natural selection ever favor anything giving up reproduction? Remember that natural selection can't look ahead and see long-term benefits, any more than a river can flow uphill towards the ocean.

The evolution of multicellularity is an example of what John Maynard Smith and Eors Szathmary called "The Major Transitions in Evolution." Another example is free-living algae giving up independent reproduction to become chloroplasts. These transitions are actually much harder to explain than the origin of "irreducibly complex" structures (which always turn out to be achievable through a series of steps), because individual selection tends to undermine group benefit. But we are starting to understand how such transitions can happen.

This week's paper is "Cooperation peaks at intermediate disturbance" by Michael Brockhurst, Angus Buckling, and Andy Gardner, from Liverpool, Oxford, and Edinburgh, published in Current Biology (vol. 17 p. 761-765). They developed a mathematical model of the effects of disturbance on the evolution of cooperation in bacteria, then did an experiment to test the model's predictions.

I will explain the experimental system first, to make the discussion less abstract. They worked with bacteria already shown to evolve the ability to make goop that sticks cells together into a "biofilm." Biofilms are important in disease and may resemble an early stage in the evolution of multicellular life. In liquid culture, bacteria benefit from being in a biofilm, because they float on the surface, where there is more oxygen. (The Exploratorium, in San Francisco, used to have a great demonstration of this. A flask of light-emitting bacteria glowed mainly at the surface. Pushing a button mixed oxygen into the flask, so the glow spread all the way down.) Mutants that don't produce goop are "cheaters"; they benefit from being in the biofilm, without paying the cost of making goop. A flask full of cheaters doesn't grow well -- most cells don't get enough oxygen -- but a few cheaters in a biofilm will tend to out-compete cooperators, because they use all their resources for their own reproduction, instead of for making goop. In the long run, who will win, cheaters or cooperators?

"It depends on disturbance", suggested the authors. Their model included two different effects of disturbance on cooperation. Each disturbance kills most of the cells in the flask. After a disturbance, the cells in the flask tend to be more related to each other than before. (Imagine a nuclear war that destroys every US state except West Virginia.) So each group is mostly cooperators or mostly cheaters. Cooperating groups grow faster than cheating groups, so disturbance should favor the evolution of cooperation. This is an example of "kin selection": a gene for cooperation can spread if it causes behavior increasing the survival of other copies of the gene, usually in close relatives.

But wait! Frequent disturbance kills so many cells that there may not be enough to form a biofilm. (Ten people who never cheat on their taxes may not have as good a school as ten thousand people, even if some of them cheat.) Combining these two effects of disturbance, the model predicted that cooperation would be most common at medium rates of disturbance.

When they did the experiment, that's exactly what they found.


The paper includes a nice discussion of their assumptions and of other possible interpretations of their results. For example, the least-disturbed cultures tended to run out of nutrients. This hurt all the bacteria, but maybe especially the cooperators, due to the cost of making goop. Cooperators might have done a little better, in the least-disturbed system, if nutrients were replaced more often. It might be possible to improve this experiment, but the authors have a better suggestion: try similar experiments with different kinds of bacterial cooperation. For example, Buckling has previously studied bacteria that cooperate by releasing molecules that help all bacteria (even cheaters that don't make the molecule) to take up iron. I look forward to more papers on this topic.

May 2, 2007


What did our early ancestors and related species eat? Different data seemed to give different answers. This week's paper may have helped to solve this mystery.

Isotope data suggest that tropical grasses were a big part of the diet of the hominins Australopithecus africanus and Paranthropus robustus. These grasses have CO2-concentrating C4 photosynthesis. As a result, they have a little more of the rare carbon-13 isotope, and a little less C12, relative to most other plants. So do the fossil teeth of these early human relatives, as if they ate these grasses. But the shape of their teeth, and wear patterns, are wrong if they mostly ate grass leaves or animals that ate grass. What about roots, or underground storage organs? These are an important food for some human foragers today, especially in dry climates. If our early relatives mostly ate these "USOs", then the isotope ratios in their teeth should be like those of other species with a similar diet. Mole rats, for example.

This week's paper is "The isotopic ecology of African mole rats informs hypotheses on the evolution of human diet", by Justin Yeakel and colleagues at UC Santa Cruz and the University of Pretoria, South Africa, published online in Proceedings of the Royal Society.

They collected skulls of five mole rat species in South Africa. They then analyzed their teeth for C13 and the more common isotope C12. To compare modern samples with those from the Pleistocene, they had to correct for the decrease in the percent C13 in the atmosphere since then. This is due to our burning so much C12-rich coal and oil. Only two of five mole rat species had isotope ratios like those of fossil hominins. The other three apparently didn't eat much grass (leaves or roots).

But isotope ratios in fossil mole rat teeth were similar to those of fossil hominins from the same area. The data suggest that tropical grasses (perhaps their underground storage organs) were an important part of the diet of both mole rats and hominins, but not 100% for either.

These extinct hominins weren't our direct ancestors. But they branched off our family tree close enough to human ancestors that their diets could be similar.

I liked this paper because it's like much of science: filling in pieces of the puzzle, rather than one breakthrough that gives a final answer. Also, it reminds me of research at my old Long-Term Research on Agricultural Systems project at UC Davis, where researchers have used isotope ratios in soil organic matter to see how much of it came from recent corn crops (more C13) versus previous alfalfa and tomatoes (less C13).