Microbes evolve; flies evolve and learn
"Can plants predict the future?" asked one of my Crop Ecology lectures at UC Davis. Yes, they can. Plants use decreasing daylength to predict oncoming winter, and flower early enough to finish seed development before it gets too cold. Some plants detect early signs of drying soil and reduce their own water use, saving water in the soil for later.(Davies & Zhang. 1991, Bano, et al. 1993) Others detect "distress signals" from neighbors under insect attack, turning on chemical defenses in anticipation (Karban, et al. 2004).
But these anticipatory responses do not require learning: a beneficial change in individual behavior in response to individual experience. An alfalfa plant will never learn that the farmer always irrigates before it actually runs out of water. At least, I assume it won't. An evolving alfalfa population is a different story. Over a few generations under irrigation, genotypes that reduce their water use as the soil starts to dry (thereby reducing their growth) will be out-competed by genotypes that keep using water and growing.
Like plants, microbes can predict the future. As in plants, this trait can evolve. As they pass through the gut, bacteria typically see lactose before they see maltose. So they have evolved to "anticipate" maltose availability, turning genes for maltose use on as soon as they are exposed to lactose. After 500 generations of evolution on lactose without maltose, however, the bacteria have lost lose this anticipatory response, so that they turn maltose genes on only when they actually see maltose.(Mitchell, et al. 2009) The title of the news story in Nature about this work asked whether microbes can "learn from history", but this is clearly not an example of individual cells modifying their responses to lactose based on their individual experience.
Individual insects can learn. But is learning always a good thing? Aimee Dunlap, a grad student in my department working with David Stephens, just published a paper in Proceedings of the Royal Society exploring the conditions under which natural selection will favor learning (Dunlap & Stephens. 2009).
She reasoned that, if the behavior that maximizes fitness doesn't change over generations, then natural selection should turn that behavior into an instinct, rather than something that has to be learned. If fitness-maximizing behavior changes over generations, learning may increase fitness -- unless there is so much random change over an individual's lifetime that experience is a poor guide.
Dunlap used a system devised by Mery and Kawecki(2002) to test these hypotheses with fruitflies. The flies had to choose between laying eggs in a dish with pineapple- or orange-flavored medium. In the first, "experience" phase, one dish also had quinine, which the flies dislike. They won't lay eggs in a dish with quinine, but would they learn, and avoid that dish in the future, even when presented without quinine?
Dunlap let her fruitfly populations evolve for 30 generations in two different ways. One population experiences an "experience is reliable" treatment: any eggs laid in the type of dish (orange or pineapple) that previously had the quinine were discarded. For example, if a fly experienced pineapple+quinine early in life, any eggs it laid in quinine-free pineapple later were discarded. Only those laid in quinine-free orange made it into the next generation. In alternate generations, however, quinine would be paired with orange. So an evolutionary increase in instinctual preference for orange or pineapple wouldn't help. Instead, this population increased its ability to learn: whichever type dish had quinine at first was avoided in the future. In the "random" treatment, eggs from one dish were discarded at random, regardless of whether than dish had previously had quinine. 30 generations of this treatment caused no evolutionary change in learning ability. If the relationship between quinine and dish type was random, but eggs were raised from the orange dish, the populations evolved an instinctual preference for orange, and vice versa for pineapple. The authors concluded that, in addition to past emphasis on constant versus changing environments, we need to distinguish between "the reliability of experience [which selects for ability to learn], and underlying uncertainty about the appropriate action [which selects against ability to learn]."
I wonder whether these results could be used to explain differences in learning ability among species. There's a squirrel in our neighborhood that has trained my wife and me to give her peanuts. If all humans were a friendly source of food, that would presumably select against fear of humans, without selecting for learning. But if some humans consistently chased squirrels while others consistently supplied peanuts, would that select (over many squirrel generations) for ability to learn to tell us apart, and to remember which were friendly?
Bano A, Dorffling K, Bettin D, Hahn H. 1993. Abscisic acid and cytokinins as possible root-to-shoot signals in xylem sap of rice plants in drying soil. Aust. J. Plant Physiol. 20 : 109-15
Davies WJ, Zhang J. 1991. Root signals and the regulation of growth and development of plants in drying soil. Annu. Rev. Plant Physiol. Plant Mol. Biol. 42 : 55-76
Dunlap AS, Stephens DW. 2009. Components of change in the evolution of learning and unlearned preference. Proceedings of the Royal Society B: Biological Sciences. 276 : 3201-8
Karban R, Huntzinger M, McCall AC. 2004. The specificity of eavesdropping on sagebrush by other plants. Ecology. 85 : 1846-52
Mery F, Kawecki TJ. 2002. Experimental evolution of learning ability in fruit flies. Proceedings of the National Academy of Sciences of the United States of America. 99 : 14274-9
Mitchell A, Romano GH, Groisman B, Yona A, Dekel E, et al. 2009. Adaptive prediction of environmental changes by microorganisms. Nature. 460 : 220-4