A recent article in the October 29th issue of Nature by Richard Lenski and his group at Michigan State (Genome evolution and adaptation in a long-term experiment with Escherichia coli, Barrick et al., Nature, 29 October 2009, p 1243-1247) has attracted a lot of attention. And the attention was merited, because this paper was the latest report on a long-lasting series of experiments in bacteria meant to address some evolutionary questions.
From an original ancestral E. coli strain, the investigators maintained 12 separate populations in culture medium over twenty years, with glucose as a limiting nutrient. The idea was to look at genomic and adaptive evolutionary changes under less than optimal conditions. In what sounds like a documentation, not to mention storage, nightmare, they transferred a sample of each culture into fresh medium every day, periodically freezing selected samples for later sequencing and analysis. For the study reported in this Nature paper, they sequenced the genomes of clones from generations 2,000, 5,000, 10,000, 15,000, 20,000 and 40,000 and compared these to the ancestral genome sequence, documenting the changes they found. They also compared the reproductive fitness of each population at different time periods (fitness is defined by population counts of competitors plated together on agar, as described in BMC Evolutionary Biology 2002, 2:19.)
This is interesting but there are a few points that are important to consider if, as is inevitable, the experiment is extended to evolution in general. Especially because everyone will use this to prove their selective scenarios.
First, this is not really a test of natural selection! Despite the wording of the paper and news statements, this is a test of artificial selection. The usual distinction is that in artificial selection the farmer chooses the parents based on their having traits he wants more of, while natural selection screens traits that make more of themselves for whatever reason. In this case, the investigator does let the bacteria struggle it out on their own, which is like natural selection. However, the selective agent (limited glucose) is controlled by the investigator, who also tightly controls the rest of the environment.
This is certainly, and positively, a case in which new technology enables a more precise look at genes and their dynamics than was possible before. The controlled and limited nature of the experiment makes it far more interpretable than real life, analogous in its own way to the difference between artificial and natural selection. But it is closer to a set-up test of evolutionary (population genetics) theory than of evolution itself.
To the extent that DNA sequence, rather than other non-sequence-based aspects of DNA (if those are relevant), is the basis of evolution and adaptation, we at least see the entirety of the sequences that have arisen, except to the extent that rare mutations and sequencing errors are hard to distinguish.
On the other hand, to relate this to real-life outside the lab one has to be careful. In the human case, and this will certainly be widely cited as being relevant to humans, there is little that is in fact relevant. 20,000 bacterial generations would correspond to 400,000 years (2 to 4 times the age of our species!), and 40,000 bacterial generation is 800,000 years -- the length of time, roughly, since our ancestral species Homo erectus (depending on what species-reconstruction you prefer).
Not the population size and structure, the environments, nor the ability of genotypes to replace each other rapidly are similar to our species' history. We have had agriculture for only 500 generations, for example.
This is very interesting, high quality, illuminating work. But as with any oversimplified story, we need to beware of over-interpretation. Bacteria are interesting and the dominant form of life. But we're not bugs.