Darwin as a Role Model for Aging

I turned 70 years old yesterday. I woke up a bit earlier than usual, and so at 7 AM I set out on a 7-kilometer run. I’m slower than when I was younger, of course, but happy to be alive and reasonably fit. As one of my grandfathers liked to say, “I’m in right good shape for the shape that I’m in.”  But I’m also increasingly cognizant of the declining time I have remaining, which reminds me that I still need to get some things done, like writing a book or two.

It’s often said that most scientists do their best work when they are young, and that might be a good rule of thumb. But some people get a lot done in their later years. To mention a favorite example, Charles Darwin published The Origin of Species when he was already 50 years old. Moreover, that 500-page masterpiece was a mere abstract of the book that he had intended to write if he hadn’t been rushed to publish by Wallace’s independent discovery of natural selection. In the Introduction to The Origin, Darwin wrote:

My work is now nearly finished; but as it will take me two or three more years to complete it, and as my health is far from strong, I have been urged to publish this Abstract. I have more especially been induced to do this, as Mr. Wallace, who is now studying the natural history of the Malay archipelago, has arrived at almost exactly the same general conclusions that I have on the origin of species … This Abstract, which I now publish, must necessarily be imperfect. I cannot here give references and authorities for my several statements; and I must trust to the reader reposing some confidence in my accuracy … I can here give only the general conclusions at which I have arrived, with a few facts in illustration, but which, I hope, in most cases will suffice. No one can feel more sensible than I do of the necessity of hereafter publishing in detail all the facts, with references, on which my conclusions have been grounded; and I hope in a future work to do this.

And true to his word, Darwin plowed ahead. He wrote another 10 books that were published between 1862, when he was 53 years old, and 1881, a few months before his death at the age of 73. That total does not include revisions that led to several later editions of The Origin. His later books include work on the various adaptations of plants: three on the means by which plants attract pollinators and reproduce (1862, 1876, 1877), two on the movement of plants as they climb and track the sun (1875, 1880), and one on carnivorous plants like the Venus flytrap (1875). Some of these books report new findings from experiments that Darwin performed at Down House, including studies on how plants manage to bend toward the light as they grow, and on the effects of outcrossing versus self-fertilization on their fitness.

Among Darwin’s other books is a massive two-volume tome on The Variation of Animals and Plants Under Domestication (1868), in which he compiled and interpreted the evidence that he gathered from his many correspondents concerning the influence of heredity on all sorts of organismal traits. While that evidence clearly supported the role of heredity in evolution, alas, Darwin’s own theories concerning the mechanism of heredity did not stand the test of time. It took the rediscovery in 1900 of Gregor Mendel’s previously ignored experiments with pea plants, along with new studies of chromosomes and flies by Walter Sutton and Thomas Hunt Morgan, for the modern understanding of genetic inheritance to begin to take form.  

Two of Darwin’s other books focused on humans—a topic that he neglected in The Origin, presumably to avoid a sensitive topic that wasn’t necessary for understanding his broader findings and theories. The Descent of Man (1871) introduced the idea of sexual selection, whereby by traits that might seem to hinder an organism’s survival, such as the tail of a male peacock, could evolve if they gave the individual an advantage in attracting or competing for mates. My own favorite among Darwin’s later works is The Expression of Emotions in Man and Animals (1872), in which he presented evidence that people are not so unique among animals, even when it comes to our most human feelings and behaviors.

Darwin’s final book was on earthworms (1881) and how their activities recycle vegetation, thereby contributing to the formation of soils, something he had studied decades earlier. Despite the lowly subject, the book proved to be a surprising best-seller by the standards of the day. In her engaging two-volume biography of Darwin (2002), Janet Browne speculated that he might even have returned to the study of earthworms in anticipation of soon joining them in the local church’s graveyard.

The quantity, quality, and impact of Darwin’s work are unmatchable. But I still have work to do, and Darwin is an inspiring role model for remaining inquisitive, engaged, and productive … whatever the distance to the end.

One of the figures from Darwin’s book on The Expression of Emotions in Man and Animals, this one showing “Dog approaching another dog with hostile intentions.”

A profound loss in the LTEE family

I am heartbroken to report that Devin Lake passed away suddenly a few days ago at the age of 30. Devin was a good friend to all who knew him, always eager to help in any way that he could, and an outstanding young scientist with a bright future.

Devin grew up in DeWitt, Michigan, and first visited my lab as a high-school student getting a tour of our long-term evolution experiment (LTEE).  He came to MSU and majored in physics as an undergraduate, while also working closely with Neerja Hajela as an assistant in my lab. Devin decided to stay at MSU, doing double duty as a graduate student and lab manager after Neerja retired (see photo below).

Devin recently completed his Ph.D. with a dual degree in Integrative Biology and the Ecology, Evolution, and Behavior Program. In 2022, he received the Ralph Evans Award from the Department of Microbiology and Molecular Genetics for his contributions to the understanding of microbial evolution, and in 2025 he received the award for outstanding dissertation research in the Ecology, Evolution, and Behavior Program. Devin’s brilliant work included computer simulations of evolving bacterial populations that beautifully captured key features of the experiment that he had first encountered as a high-school student.

With Jeff Barrick’s return to MSU, along with the LTEE, Devin joined the team as a postdoc and lab manager. Devin was helping people in the new and old labs to the end.

I, and all of his friends old and new, will miss Devin dearly. It’s a great loss to science, and an even greater loss to those of us who knew him as a wonderful young man. May his memory be a blessing, and may he rest in peace.

Neerja anoints Devin as the Keeper of the LTEE

STEPS To It

I’m happy to announce the release of a new software program, STEPS, which stands for Serially Transferred Evolving Population Simulator. Using STEPS, one can simulate the dynamics of the E. coli Long-Term Evolution Experiment (LTEE) or any other asexual microbial populations evolving in a serial transfer regime, where the cells are periodically diluted into fresh medium and then regrow.

One can monitor each population’s fitness trajectory, the number of accumulated mutations, and more. One can also manipulate the number of replicate populations, the dilution factor, the final population size supported by the culture medium, mutation rates, distribution of mutation effects, and more. The figure below shows trajectories for average fitness and accumulated mutations for a run with parameters similar to the LTEE.

The STEPS program can be run two different ways. The easiest way to get started is with the web-based version, which uses a graphical interface. There’s also a command-line version that has more options and is better suited for larger runs with more populations, more generations, and higher mutation rates that require more lineages to be tracked. Both versions use the same underlying computational machinery.

STEPS was developed by Devin Lake (doctoral student in EEB), Zachary Matson (former CS undergrad), Minako Izutsu (former postdoc), and me. We hope you enjoy STEPS and find it useful in your teaching, research, or both.

Devin, Zach, and I also wrote a User Manual that explains: the context and purpose of STEPS (Chapter 1); the use of the web-based version including numerous exercises with figures (Chapter 2); the setup and full set of options available in the command-line version (Chapter 3); and the mechanics of the simulations (Chapter 4). I think that educators, in particular, will find the exercises in Chapter 2 valuable for classroom and/or lab-based courses on evolution. Even researchers with extensive experience (evolution, computation, or both) may find it helpful to start with these web-based exercises before advancing to the command-line version.

Here are the links to get started:

• STEPS User Manual:  https://zenodo.org/records/16690374
• Web-based version:  https://steps-portal.pages.dev/
• Command-line version:  https://github.com/zachmatson/STEPS

Screenshot from the STEPS portal showing the menu options and trajectories for average fitness and accumulated mutations using the default parameters and the randomization seed = 606. The run took under 10 seconds. As explained in the User Manual, the runs will take longer when (optional) neutral and deleterious mutations are included, because more lineages must be tracked, although these additional mutations often have little or no effect on the fitness trajectories.

Homeward Bound

With apologies to Rhymin’ Simon

Some Experiments Work, and Some Don’t

This coming Monday, February 24th, will be the 37th birthday of the Long-Term Evolution Experiment (LTEE) with E. coli. Happy birthday to the 12 lines! I hope you will keep on evolving for many, many more years.

I started the LTEE in 1988 while I was on the faculty at UC-Irvine. The LTEE moved with me to MSU in late 1991, where it reached the milestone of 75,000 generations. In May 2022, the LTEE moved to UT-Austin, where it continues in the able hands of Jeff Barrick and his team and has now passed 80,000 generations. I think it’s fair to say that the LTEE has worked out pretty well.

Another evolution experiment that worked well was just published in Science by Michael Barnett, Lena Meister, and Paul Rainey. Titled “Experimental Evolution of Evolvability,”  they show that bacteria can evolve to become more adept at adapting to changing conditions.

One way to do that is by increasing a cell’s mutation rate across its entire genome. In fact, that has happened in several LTEE populations, though in most cases the hypermutability was later reduced or even reversed. Genome-wide hypermutability is a double-edged sword, because random mutations may break other important functions before finding a solution to the new circumstances.

A better solution, in some scenarios, would be to mutate only those “local” bits of the genome that encode the functions that must change to fit the new conditions. Localized hypermutability might suggest some foresight, but that’s not really so. If populations of microbes have experienced similar changes repeatedly during their evolutionary history, then a lineage that evolved a more mutable local sequence in a relevant gene could be more likely to persist.

We know from molecular biology that some sequences — for example, homopolymeric runs like AAAAAA — are much more mutable than others. And we know from comparative studies that some microbes possess localized hypermutability in a subset of their genes that are important for dealing with unpredictable aspects of their environment. Imagine, for example, a protein that is required for transmission between hosts, but which makes the cell vulnerable within a host. This scenario would favor a lineage that has the capacity to inactivate that specific protein at a high rate and then to reactivate it at a high rate. The new study by Barnett et al. is the first to demonstrate this process experimentally. They did so with Pseudomonas fluorescens by selecting for “repeated phenotypic transitions between … the mat-forming, cellulose-overproducing CEL⁺ type and the mat-colonizing, non-cellulose-producing CEL^– type.”

This nifty new result reminds me of a conceptually similar experiment that Paul Sniegowski and I did way back in the 1990s, but which did not work out so nicely. Paul was a postdoc in my lab, and he discovered that some of the LTEE populations had evolved genome-wide hypermutability. (Paul later joined the faculty at Penn and, last year, became President of Earlham College.) Paul and I were also examining the evidence concerning the randomness of mutations in light of the possibility of so-called “directed” mutations; and I had recently collaborated with Richard Moxon and Paul Rainey on a review article that discussed the evidence and evolutionary hypothesis for the emergence of localized hypermutability in what we called “contingency genes.”  So, Paul Sniegowski and I set out to see if we could evolve a brand-new contingency gene.

It’s been a long time, and I may misremember some details. But to a first approximation, we sought to do the same experiment as Barnett et al., except using E. coli and two alternating environments appropriate to the biology of that species. In particular, in the course of my earlier work on the coevolution of E. coli and phage T4, I had learned that mutations that confer resistance to T4 infection also make the mutants more sensitive to the antibiotic novobiocin. This collateral sensitivity occurs because (i) phage T4 infects by adsorbing to the lipopolysaccharide (LPS) core of the E. coli cell envelope; (ii) the mutations that confer T4 resistance change the structure of the LPS core; (iii) novobiocin is a hydrophobic compound; (iv) the altered LPS core impacts the hydrophobicity of the cell envelope; and (v) that change allows novobiocin to enter T4-resistant cells at a much higher rate.

Given these points, Paul and I reasoned that we could propagate lines in a regime that alternated each day between exposure to T4 and novobiocin. Each round would impose lethal selection, and so we expected most lines to go extinct. But if a lineage happened to become resistant to one or the other killer by a mutation that also happened to increase the mutation rate in a gene encoding the relevant step in LPS synthesis, it would be more likely to survive the future back-and-forth challenges. Makes sense, right?

Given the lethality of the selection against sensitive cells, and the resulting high likelihood of extinction, Paul reasoned he would need a very large experiment. I forget the numbers, but he set up many tens or even hundreds of replicate lineages to start out.

After a few days, though, most or all of the lineages had survived. But how? Paul tested the evolved cells for their susceptibilities to T4 and novobiocin, and he got an unexpected result — the lines had become simultaneously resistant to both T4 and novobiocin!

We then realized that the reasoning behind our experimental design had been faulty. While mutations that disrupt the LPS core affect sensitivity to novobiocin, the cellular target of that antibiotic is a different macromolecule, namely the DNA gyrase that is required for genome replication. What had evidently happened, therefore, was sequential selection for double mutants that first became resistant to T4 by mutations in genes impacting the LPS core and then resistant to novobiocin by mutations in the DNA gyrase. The bacteria did not need localized hypermutability to solve the alternating environments that we imposed.

Edited to add:  Paul Sniegowski and I were hoping to find a quick and easy route to building a contingency gene. Something like this, perhaps—that among, say, 1000 mutants that became T4 resistant, maybe a few would have, say, a new 4-bp sequence like AAGA. And then one of those that became novobiocin resistant might have an AAAA sequence, at which point some slippage and frameshifts would start happening. That was our thinking. The results of Barnett et al. were more subtle and complex than what we imagined, and required a lot more persistence, ingenuity, and insight to understand. Bravo!

Two lessons: Some experiments just don’t work out. And you get what you select for — in other words, evolution usually finds the simplest solution that is available to the organism, even if you were hoping for something else.

So perhaps a third lesson is in order: Persistence often pays off, as exemplified by both the LTEE and the elegant and sustained work on evolvability by Barnett, Meister, and Rainey.

Note:  I know that many scientists, and especially early-career scientists, are concerned by other issues at this time. However, for many of us, one of the joys of science is to immerse ourselves in thinking about research and education. I offer this post in that spirit.

How Not to Start a Microbial Evolution Experiment

Amir Mani (University of Chicago Medical School) wrote me yesterday about a paper he had read, and about which he was a bit skeptical. The paper reports a striking case of rapid parallel evolution in an experiment with bacteria. I’m not going to identify the paper, as I have no wish to criticize the authors’ work.

However, I realized that I could rework my response in a way that I hope will be helpful to anyone thinking about starting an evolution experiment with microbes. (For those who might be analyzing and writing up the results of such an experiment, be sure to address this issue in your methods, and take it into account when interpreting and presenting your results.)

In general, your experiment will be more powerful if you have replicate evolving populations. And having replicate populations is essential if you want to say anything about the repeatability of evolution (i.e., parallelism).

However, the devil is in the details—specifically, whether the replicate populations are truly independent.

I’ll begin by explaining how to set up an experiment the wrong way, because I think it seems simpler, easier, and more intuitive if you want to get started quickly and/or haven’t thought deeply about how you will interpret the results of your evolution experiment.

The wrong way to start:  Take your ancestral strain from the freezer. Streak for a single colony and use it to inoculate a culture of the ancestral strain. Then split or transfer aliquots of that ancestral culture into your N replicate populations, which you will then propagate under the conditions of your experiment.

In that case, if the ancestral culture happened to have produced a single mutation that would be advantageous in the new selection regime, and if that mutation happened early enough during the growth of the ancestral culture, then you might well see the exact same mutation quickly spread through many or all of the replicate populations as they evolve. In essence, you would be rediscovering the “jackpot” effect in Luria and Delbruck’s classic 1943 paper on the randomness of mutations. While they won the Nobel prize for that and related work, alas, this would be the wrong way to start your evolution experiment.

The right way to start:  The correct approach would be to grow the ancestral culture, as before, but then plate from that culture for single colonies. Each colony results from the outgrowth of a single cell, and no mutation can move from one colony to another. (If you happen to work with a highly motile organism, perhaps you should plate for single colonies on separate agar plates, since even a motile microbe won’t be able to move between plates.) You would then choose N colonies at random and start each of the N replicate populations from a different single colony. Therefore, there is no possibility that derived mutations found in multiple evolved lines will be “identical by descent” (i.e., derived from the same mutational event).

If identical mutations arise in truly independent populations even when this correct procedure is followed, then that would indicate parallelism at the nucleotide level. That outcome certainly can and occasionally does happen, but it is not typical in most microbial evolution experiments. Such nucleotide-level parallelism, when it occurs, typically suggests that only one mutation (among all the sites in that gene, pathway, and genome) can produce the selected phenotype; that one genomic site is much more mutable than the others; or some combination of these two possibilities.

You can read a bit more about this issue in the context of the LTEE here, here, and here. See also this nice paper by David Stern, in which he coins the term “collateral evolution” to describe “evolution in independent lineages of alleles that are shared among populations” and hence identical by descent. (Full citations below.)

I would argue that collateral evolution is usually an unwanted artifact in microbial evolution experiments, one that results from a flawed experimental design as described above. (There are exceptions, such as this experiment designed to compare the contributions of starting variation and new mutations to rates of adaptation.) However, collateral evolution is unavoidable in experiments conducted using sexually reproducing animals and plants, where evolution depends largely on pre-existing (standing) genetic variation in the ancestral population, especially in the early generations.

• Luria, S. E., and M. Delbrück. 1943. Mutations of bacteria from virus sensitivity to virus resistance. Genetics 28, 491–511. [DOI: 10.1093/genetics/28.6.491]
• Woods, R., D. Schneider, C. L. Winkworth, M. A. Riley, and R. E. Lenski. 2006. Tests of parallel molecular evolution in a long-term experiment with Escherichia coli. Proceedings of the National Academy of Sciences, USA 103, 9107-9112.
• Lenski, R. E.  2017.  Convergence and divergence in a long-term experiment with bacteria.  American Naturalist 190, S57-S68. [DOI: 10.1086/691209]
• Lenski, R. E. 2023. Revisiting the design of the long-term evolution experiment with Escherichia coli. Journal of Molecular Evolution 91, 241–253. [DOI: 10.1007/s00239-023-10095-3]
• Stern, D. 2013. The genetic causes of convergent evolution. Nature Reviews Genetics 14, 751–764. [DOI: 10.1038/nrg3483]
• Izutsu, M., and R. E. Lenski. 2022. Experimental test of the contributions of initial variation and new mutations to adaptive evolution in a novel environment. Frontiers in Ecology and Evolution 10, 958406. [DOI: 10.3389/fevo.2022.958406]

Note:  I know that most scientists, and especially early-career scientists, are concerned by much larger issues at this time. However, for many of us, one of the joys of science is to immerse ourselves in thinking about research and education. I offer this post in that spirit.

A Small Correction

Since transferring the LTEE to Jeff Barrick’s lab at UT-Austin in 2022, we’ve been going over the old lab notebooks, making sure everything looks good. It turns out, though, that I made a small error when I started the LTEE back in 1988. I thought that transferring 10 ml into 10 ml was a hundred-fold dilution because there’s a 0 right there after each of the 1s, and 100 has two zeros. QED: a hundred-fold dilution. Right?

Well, it turns out I was a bit off. That’s only a two-fold dilution because, apparently, the correct way to do the math is 10 / (10 + 10) = 1/2. Who knew? New math, I guess. Anyhow, everyone in the lab thought I had figured it out, since I was the perfesser, and they just kept doing the same thing all these years. So instead of 75,000 generations, it was only something like 11,250 when we sent the stupid amazing LTEE to Taxes. Oh well, still a big number.

We also discovered another tiny error. You know, I always thought some sucker hard-working student came in and did the transfers on weekends and holidays. I never quite knew who it was, but I figured someone did the unpaid work transfers. Well, it turns out, not so much. OK, never. Fridays were ok at 40%, and Mondays were even better at 53%. On Tuesdays, we maxed out at 73%. Not bad! We trailed off a tad at 59% and 47% on Wednesdays and Thursdays.

Anyhow, after correcting for these tiny oversights, the LTEE had gone past 4,300 generations before we sent it down to Taxes. Speaking of Taxes, I hope I don’t get audited again this year. But I hear you can stall if you’re a big shot. Being a PI qualifies, right?