Evolution and Arithmetic

Chapter 16: Evolutionary Systems

← Not One Letter at a TimeOne Program, Modified →

Two Systems, Hiding in One Bacterium

Chapter 7 told the story of Richard Lenski's flasks: twelve populations of E. coli, grown for years beside a food — citrate — that ordinary E. coli can't take in while oxygen is present. Only one flask ever learned to use it, and only after two changes in the right order: an easy first change, in place by around generation 20,000, and a much rarer second one, by generation 31,500, in which a stretch of DNA nearly three thousand letters long was duplicated and one copy of the citrate gate gene, citT, landed next to an "on" switch that works with oxygen around. That chapter also mentioned, in passing, an outside check that did the same thing in weeks. That check deserves a closer look now, because it shows the large-scale tools at work.

In 2016, three researchers at the University of Idaho — Dustin Van Hofwegen, Carolyn Hovde and Scott Minnich — asked what happens if citrate isn't a side dish next to plentiful glucose but nearly the only thing on the menu. They grew several strains of E. coli, some with particular genes deliberately removed, in flasks built around citrate, sometimes with a trace of another food to allow a few generations of growth before the real test began. Then they waited. For weeks the bacteria declined — by about nine in ten in some flasks, and nearly ninety-nine in a hundred in others — before citrate users appeared and the flasks recovered.

The result was dramatic by comparison. Rather than one success in twelve tries across thirty thousand generations, the setup produced citrate users forty-six separate, independent times, from every strain except the one deliberately missing citT itself. In the version where the cultures were refreshed once a week, they typically appeared within twelve to thirty generations. In the version left to run without a break, all but one appeared within forty days. The exception, a strain missing a gene that helps the cell recapture one of the by-products of using citrate, took 134 days — a small, useful demonstration of how much genetic background can matter to how long any of this takes.

What did the winners do? Sequencing found two kinds of large-scale change, and no single-letter mutation responsible for the breakthrough.

The first kind was copying. In the early citrate users, the stretch of DNA containing citT had been duplicated, and duplicated again, until some cells carried around four copies. More copies of a gene that barely works under oxygen means more of its product, and enough to start making a living. In some strains a second gene that helps with citrate use was multiplied the same way, inside a block of DNA some 140,000 letters long. That block was multiplied by the cell's own repair machinery, which normally mends broken DNA by matching it against a similar stretch elsewhere. Here it matched two near-identical stretches lying some distance apart, and looped the whole region between them into a second copy.

The second kind was switch capture. In one strain, a duplication fused citT's neighborhood to another gene's oxygen-proof "on" switch, much as had happened in Lenski's flask. In another, a jumping gene — a small transposon, over a thousand letters long — landed just in front of citT, and the switch it carried for its own use turned citT on.

Those are the two systems of this section's title, and neither can do its job anywhere, to any gene, on demand. The first, the cell's repair-and-recombination machinery, can only double or delete a stretch of DNA that happens to lie between two matching landmarks — repeated sequences already scattered at particular points in the genome. The second depends on a mobile element that already carries, in its own structure, the equipment to cut or copy itself and paste itself in a new place, and sometimes a switch that comes along as part of the package. Both depend on what's already sitting nearby, and on particular enzymes, built for this cutting, copying and pasting, being present and active.

Forty-six independent successes, all running through the same gene, can look as if the cell knew where to aim. It didn't need to. citT is the only gene E. coli has that can carry citrate into the cell — the strain without it never succeeded at all — so every success had to involve it. When there is only one door, everyone who gets out uses it. And the routes through that door differed: of the four winners the researchers sequenced, one had extra copies of citT, another had a jumping gene in front of it, and a third had borrowed another gene's switch. Copying, meanwhile, goes on all the time, all over the genome. Measurements in a close relative of E. coli show that in an ordinary growing culture, with nothing selecting for anything, roughly one cell in ten carries a duplicated stretch of DNA somewhere, and any particular gene is duplicated in anywhere from one cell in thirty to one in twenty thousand, depending on where it sits. In a flask holding hundreds of millions of cells, a great many already carried an extra copy of citT before the starvation began — along with extra copies of every other gene. Starvation didn't create them. It simply let the cells that happened to hold the one useful copy grow a little, while the rest starved.

Why was the outcome so much faster than in Lenski's flasks? One tempting answer is that starving bacteria mutate faster. Something like that does exist: under stress, E. coli can switch to sloppier, error-prone ways of repairing its DNA, and a stress protein called RpoS is involved in controlling this. The Idaho researchers expected it to matter, and included a strain lacking RpoS as a control that should fail. It didn't. That strain produced citrate users as readily as the others, and the researchers concluded that only gene copying and recombination were required.

The better explanation, offered by two bacterial geneticists in a commentary in the same journal, has to do with selection rather than mutation. Duplications are unstable. They arise often, but they are also lost often, as the extra copies recombine back out again. Under steady pressure — citrate the only food, day after day — a cell with a few extra copies of citT keeps growing a little while its neighbors starve, and its descendants with still more copies grow better still, until the lineage climbs to a working state. In Lenski's flasks, citrate was useful only for part of each day, after the glucose ran out, and each day ninety-nine cells in a hundred were thrown away. A fragile early duplication rarely survived long enough to be built on. Consistent with this, the Idaho team found that rapid adaptation also happened with regular transfers, as long as the gap between them was a week rather than a day. The machinery was the same in both experiments. What differed was how long, and how hard, selection was allowed to push.

So: yes, this can change things, and it can change them considerably faster than one letter at a time ever could. The open question — and it is a live one — is whether any of this creates a new function, or only speeds up access to one that already existed somewhere in the genome, waiting to be copied or switched on. The citrate case is about as clean an answer as this book is likely to find. Even at its most dramatic — forty-six independent successes, genes multiplied over and over, a jumping gene dropped into exactly the right place — what happened was never the invention of anything new. An existing gene, already working inside the cell under other conditions, was copied, and handed an existing kind of "on" switch. The researchers' own estimate of how rare the decisive combination is when nothing is selecting for it was about one cell in a hundred trillion. Under the right pressure, the large-scale tools can shuffle fast. What they have not been shown to do, in this case or any other covered so far, is build something that wasn't already there.

Last revised 2026-11-19

← Not One Letter at a TimeOne Program, Modified →