EDIT (23 June 2015): PLOS Biology has published a condensed version of this blog-conversation.
Over at the Dynamic Ecology blog, Jeremy Fox asked me some interesting questions about the history, philosophy, and science of the E. coli long-term evolution experiment. Perhaps mistakenly—in terms of time management, not my interest!—I agreed to try to answer them … though over what time frame, I’m not sure. Anyhow, here is Jeremy’s first question followed by my (very) short and (too) long answers.
- When you first started the LTEE, did you consider it to be a low risk or high risk experiment? Because I could see arguing both ways. In some ways, it’s low risk, because one can imagine lots of different possible outcomes, all of which would be interesting if they occurred. But in other ways, it’s high risk–I imagine that many of the interesting outcomes (including those that actually occurred!) would’ve seemed unlikely, if indeed they’d even occurred to you at all. Or did you not worry much about the range of possible outcomes because the experiment was basically a lottery ticket? “This’ll be cheap and not much work, let’s just do it and see what happens. Something really cool might happen, but if it turns out boring that’s ok because it wasn’t a big investment.”
The short answer: Life was good, and I wasn’t thinking about risk. Or as they say about investing: it’s better to be lucky than smart!
The long, non-linear* answer: I’d already had success with some shorter duration, more traditionally designed experiments (e.g., Lenski, 1988), and so it wasn’t a total shot in the dark—that is, I knew the LTEE would yield data. I also knew, though, that it was an unusually abstract, open-ended, and non-traditional experiment, and that it might not appeal to some people for those reasons. But I loved (and still do) the seemingly simple (but in reality complex) questions, issues, and hypotheses that motivated the LTEE.
I never thought of the LTEE project as a “lottery ticket”, but some follow-up work that grew out of it had that feel.** And, oddly enough, there was one lottery-ticket aspect of the research early on, although that reflected a lack of preparation rather than a well-conceived feature.***
Maybe I was overly confident, but I’d also say that I was pretty sure the outcomes—whatever they might be—would be “cool.” The questions were intriguing, and there hadn’t been many, if any, previous attempts to answer them quite so directly. Data would be forthcoming, and even if the results weren’t definitive, I felt there would be some interest in trying to interpret whatever data emerged.**** Plus, I knew enough about what would happen—based on the experiments I had already done—that I was confident that the data and analyses would be informative with respect to at least some of my questions. Also, the use of microbes to study evolution in action was still uncommon, so the novelty of the approach would ensure some interest among my colleagues—although let me emphasize that Lin Chao, Dan Dykhuizen, Barry Hall, and Bruce Levin, among others, had already demonstrated the power of using microbes for experimental studies of evolutionary questions.*****
I should also say, in case it’s not obvious, that I had no idea or intention that the experiment would continue for anywhere near as long as it has lasted—nor that it might, I now hope, be running long after I’m gone. I had previously performed some experiments that lasted several hundred generations, and as I saw the dynamics and thought about the math behind the dynamics, I realized that over those time scales I might be seeing the effects of only one or two beneficial mutations as they swept to fixation. That hardly seemed satisfactory for experiments to explore the structure of the fitness landscape. So I decided the experiment should run for 2,000 generations, over which time I expected there would be at least several fixations of beneficial mutations in each population (and I was right), and that would deserve calling it long-term. That would take a little less than a year, given the 100-fold dilution and 6.6 generations of re-growth each day.
Of course, propagating the lines for 2,000 generations was one thing—running the competitions to measure fitness, analyzing the data, writing the paper, responding to reviews, all that took longer. So while the experiment began in February 1988, the first paper (Lenski et al., 1991) was not submitted until August 1989, resubmitted September 1990, accepted that November, and finally published in December 1991. Meanwhile, the LTEE itself continued and the generations ticked by. The baseline work of keeping the populations going is not that onerous—yes, somebody has to attend to the transfers every day, but once a lab team reaches a moderate size, it’s not too hard to arrange. And I lived next to the campus in Irvine, so it wasn’t hard for me to come in on the weekends and holidays … and my wife still loves me, and my kids recognized my face ;>)
You also wondered whether some of the interesting possible and actual outcomes had occurred to me when I started. Definitely not! I had made a strategic decision to make the environment of the LTEE very simple in order to eliminate, or at least reduce, certain complications (especially frequency-dependent interactions and clonal interference). And while I think my planning kept these complications from getting out of hand, the tension between the simplicity of the experimental design and all the complications has definitely been part of its interest. That tension, along with time, the evolutionary potential of the bacteria, and the smart, talented, creative** and hard-working students and colleagues have made the LTEE what I call “the experiment that keeps on giving.”
*Hey, that’s what footnotes are for, right?
**I’ve thought that way about some follow-on work that uses the LTEE lines, but not about the project as a whole. Here are a couple of examples of “lottery tickets” that people suggested to me, and that won big. A former postdoc Paul Sniegowski, now on the faculty at Penn, wanted to know whether the actual mutation rate itself might be evolving in the LTEE populations. Bingo! Several lines evolved hypermutability and so, curiously enough, the first mutations we ever mapped affected the mutation rate itself (Sniegowski et al., 1997). Another example: Dominique Schneider is a molecular microbiologist in Grenoble, and we’ve collaborated for over 15 years. He thought we should look at whether DNA topology—the physical supercoiling inside the cell—might have changed in the LTEE lines. Well, I thought to myself, why would it change? But Dom’s lab will do all the work, so sure, why not look? And it turns out, sure enough, that DNA supercoiling changed repeatedly in the LTEE lines (Crozat et al., 2005), and it even led us to discover a gene not previously known to affect supercoiling (Crozat et al., 2010). There’s a lesson here, by the way—work with people who are smarter, who have different interests, and who have different skills than oneself.
***I actually started two versions of the LTEE—not one experiment with two proper treatments, but two separate experiments that differed in terms of both the starting strain and the environment. Unlike the successful LTEE, I hadn’t done any previous evolution experiments with the other ancestral strain and environment. Anyhow, I soon stopped the other version when the populations evolved a phenotype that made it very difficult to work with them. In brief, the populations evolved to make pinprick-sized colonies that were next-to-impossible to count in the assays we use to measure fitness. Who needed that headache! So, in a way, I guess I had two lottery tickets: I hadn’t done the relevant prior work for one of them, whereas the one that paid off was actually a pretty strategic gamble.
****I was at UC Irvine when I started the LTEE, and Michael Rose was one of my colleagues there. His work on the evolution of aging—postponed senescence—in fruit flies (e.g., Rose 1984) was an inspiration in terms of the importance and power of long experiments. We also spent a lot of time discussing fitness landscapes, the alternative perspectives of Sewall Wight and R. A. Fisher about the dynamics on those landscapes, and what experiments might tell us. Michael didn’t design, direct, or do the lab work for the first LTEE paper, but he helped me clarify my thinking and write the first paper on the LTEE (Lenski et al., 1991). Perhaps more importantly, his interest in the questions and issues made me realize that other smart people would also be interested.
*****I used to complain, mostly in jest, that “Evolutionary biologists say I’m asking the right questions, but studying the wrong organism, and microbiologists tell me I’m studying the right organism but asking the wrong questions.” I got that sort of response occasionally, but many people from both fields were very interested and encouraging. For example, I remember David Wake telling me, after one of my first talks about the LTEE, how much he liked both the questions and the approach.
Lenski, R. E. 1988. Experimental studies of pleiotropy and epistasis in Escherichia coli. II. Compensation for maladaptive pleiotropic effects associated with resistance to virus T4. Evolution 42: 433-440.
Lenski, R. E., M. R. Rose, S. C. Simpson, and S. C. Tadler. 1991. Long-term experimental evolution in Escherichia coli. I. Adaptation and divergence during 2,000 generations. American Naturalist 138: 1315-1341.
Crozat, E., N. Philippe, R. E. Lenski, J. Geiselmann, and D. Schneider. 2005. Long-term experimental evolution in Escherichia coli. XII. DNA topology as a key target of selection. Genetics 169: 523-532.
Crozat, E., C. Winkworth, J. Gaffé, P. F. Hallin, M. A. Riley, R. E. Lenski, and D. Schneider. 2010. Parallel genetic and phenotypic evolution of DNA superhelicity in experimental populations of Escherichia coli. Molecular Biology and Evolution 27:2113-2128.