When I was in my 20’s, out of college and largely floundering, my dad lent me a paperback copy of Stephen Jay Gould’s Wonderful Life. I had occasionally enjoyed Gould’s column in Natural History, but I lacked the background to understand much beyond that (my undergrad was in political science). Wonderful Life got me interested in evolution, and I started reading other popular books, including more of Gould’s, Dawkins, Simon Conway Morris (the anti-Gould), etc., and pretty soon I found myself in Chris Parkinson‘s lab working on a master’s degree. This is all just to say that this particular book had a big influence on my life, and my first year in Montana was punctuated (see what I did there?) by a trip to Yoho National Park to see the Burgess Shale, the nominal topic of Wonderful Life.

One of the central arguments of Wonderful Life (and others of Gould’s works) was that the outcome of evolution is inherently unpredictable. Contingency, which can be interpreted as true randomness, stochasticity, or sensitive dependence on initial conditions, plays a large role in Gould’s view. Rarely at a loss for a good metaphor, Gould claimed that if “life’s tape” were rewound to some arbitrary time in the past and played again, the outcome would almost certainly be different from the first run.

This idea of ‘rewinding the tape of life’ has become a mainstay of discussions about evolutionary processes. Many of these discussions revolve around the question of how contingent and how deterministic are evolutionary outcomes, and this was the topic of two AbSciCon sessions chaired by Betul Kacar and Rika Anderson.

When Gould first posed his question about contingency, the tools available to look for answers were basically all retrospective. Richard Lenski had just begun his famous Long Term Evolution Experiment, and most of the genomic tools we use to look at the outcomes of experimental evolution simply didn’t exist. Since then, the fields of experimental microbial evolution and genomics have exploded, providing an unprecedented ability to observe evolution in real time, or, in Gould’s terminology, run the tape (or multiple tapes) forward. Starting replicate experimental populations with a single clone or genotype and subjecting them to (as near as possible) identical conditions allows investigators to separate deterministic outcomes (those that happen in all or nearly all populations) from contingent outcomes (those that happen in one or a few).

In the session ‘Chance and necessity: from molecules and viruses to cells and populations II,’ Betul Kacar, an alumna of the NASA Postdoctoral Program (NPP) now at Harvard, described her work resurrecting ancient proteins in modern bacteria. Using ancestral character state reconstructions based on modern microbes, she inferred the amino acid sequence of a bacterial elongation factor protein (EF-Tu) from a 700 million year-old ancestor of E. coli. Then she replaced the EF-Tu gene in a modern strain of E. coli with a DNA sequence coding for the ancient protein. In eight replicate populations, she allowed the genetically engineered bacteria to evolve for 2000 generations (about a year). The ancient protein initially caused the engineered bacteria to suffer a decrease in fitness compared to the wild-type (un-engineered) strain, but by 1500 generations the bacteria in all eight populations not only increased in fitness but exceeded that of the wild-type. In seven of the eight populations, the evolutionary changes included an insertion in a regulatory element that increased expression of the EF-Tu gene. In one population, though, the mutations that were fixed affected not the expression of EF-Tu but the amino acid sequences of several of the proteins with which it interacts. Dr. Kacar suggested that these results show the action of both deterministic and stochastic effects: the early evolutionary changes set a trajectory by chance, but once a lineage starts down a certain path, subsequent evolution is surprisingly similar in independently evolving populations.

In the Q&A period following her talk, Dr. Kacar was asked if she could use the same techniques to infer the genome sequence of a bacterial ancestor and simply replace the genome of a modern bacterium in its entirety, effectively resurrecting an ancestral bacterium. She must have been thinking about the years of work required to accomplish this with a single gene when she answered, “You’re not an experimentalist, are you?”

I met up with Dr. Kacar a bit later and asked about her future plans. She told me that she intends to expand this work to include even more ancient versions of the EF-Tu protein, as well as expanding into other genes, which she says will be “of particular interest to astrobiology.”

Rika Anderson, a current NPP fellow at the University of Illinois, Urbana-Champaign, reported on her work with the hot springs Archaean Sulfolobus acidocaldarius. These extremophiles thrive at scalding temperatures (up to 80 degrees Celsius) and in extremely acidic conditions (pH < 2). By sequencing the genomes of 50 S. acidocaldarius isolates from two different hot springs, Dr. Anderson was able to identify a set of core genes that are present in all of the isolates and a number of variable genes that are present in some and missing from others. Phylogenetic reconstructions showed genetic structure both within and between the springs. Within the core genes, synonymous single nucleotide polymorphisms were overrepresented (compared to those that change the amino acid sequence of the gene), suggesting that much of the primary divergence within the species was due to genetic drift. However, Dr. Anderson identified one hypervariable region of the genome in which the changes do appear to have been driven by natural selection, possibly as a result of an arms race between S. acidocaldarius and a virus. As with Betul Kacar’s work, these results suggest that evolution involves both deterministic (selection) and stochastic (drift) factors, although the role of chance appears (to me) to have played a larger role in S. acidocaldarius evolution than in Dr. Kacar’s experiments. Maybe this shouldn’t be terribly surprising, since I’m comparing evolution in carefully controlled laboratory conditions to potentially variable natural conditions.

Gavin Sherlock, from Stanford University, described a method his lab group has developed for tracking individual lineages at extremely high resolution during microbial evolution experiments. By inserting a unique DNA “bar code” into each of a half million yeast clones, the frequencies of these clones can be estimated accurately and cheaply, and changes in frequency can be used to measure even very small changes in fitness. By measuring changes in fitness across a large number of evolving lineages, Dr. Sherlock was able to estimate the distribution of fitness effects, a key parameter in many evolutionary models that has previously been difficult to measure. Around 20,000 of the 500,000 starting clones gained a beneficial mutation within 200 generations of evolution, and many of these mutations occurred within one particular signaling pathway, the RAS/cAMP/PKA pathway. Interestingly, the fitness effect of a mutation seems to be primarily a function of which gene is affected, with different mutations in the same gene usually having similar effects.

I had lunch with the Michod and Ratcliff lab groups and Eric Libby from the Santa Fe Institute. We found a good and surprisingly cheap sushi place close to the hotel (it’s hard to get good sushi in Missoula, so I tend to load up when I get the chance). For dinner, we went to Russian Tea Time, where Jenya Kroll enjoyed chatting with the owner in Russian. I’m sitting in the hotel lobby surrounded by Blackhawks jerseys…there’s to be a parade today in honor of the Stanley Cup victory.