Evolution and Arithmetic

Chapter 16: Evolutionary Systems

← Chapter 16: Evolutionary SystemsTwo Systems, Hiding in One Bacterium →

Not One Letter at a Time

For most of this book, "mutation" has meant one specific thing: a single letter in the genetic code changing to another letter, at random, with no preference for where it lands or what it does once it's there. That is the mechanism this book has concentrated on, and for good reason: it is the one classical evolutionary theory leans on most heavily for producing new material. For decades, most of the genome was assumed to be doing very little — vast stretches of "junk DNA," present in enormous quantity, contributing nothing to the organism, existing mostly as harmless padding. A random point mutation landing somewhere in that padding was assumed to be free to drift, neither punished nor rewarded, until by chance it landed on something that mattered and, more rarely still, improved it. That picture still matters, and this book isn't retiring it — later sections of this chapter will come back to why nothing else on offer can fully replace it. But it is no longer the only mechanism on the table, and it hasn't been for some time.

Starting a few decades ago, and accelerating sharply once genome sequencing became cheap, researchers found a series of other mechanisms — documented, operating in living organisms today — that can produce evolutionary change far faster, and by a completely different route, than a single letter flipping somewhere in a sea of unused DNA. This chapter is about what those mechanisms are, what they have been shown to do, and — this is the part to hold onto going in — what they still can't do, however impressive any single case looks.

Three categories do most of the work in what follows.

The first is large-scale genetic change: mutations that move, copy or insert substantial chunks of DNA at once, rather than a single letter. The best-studied source of such change is the transposon, or "jumping gene": a stretch of DNA that can copy itself, or cut itself out, and insert itself somewhere else in the genome. Barbara McClintock discovered transposons in maize in the 1940s, and transposons and their decayed remains turn out to make up nearly half of the human genome. They are there, as far as anyone can tell, mainly because they are good at copying themselves, not because they do the organism any good. When one lands somewhere new, the usual result is nothing at all, or damage — a gene broken in two, a documented cause of disease. But a landing can also do things that matter for evolution. Many transposons carry their own "on" switch, and landing beside a gene can switch that gene on in a new place or at a new time. Their many near-identical copies give the cell's repair machinery matching landmarks, so that the DNA lying between two copies can be doubled, or cut out, in a single step. Now and then part of a transposon ends up being read as an extra piece of an existing protein, though the cell mostly skips such pieces, since they usually do more harm than good. And occasionally a transposon's own machinery is captured and put to permanent use: the enzyme that shuffles antibody genes, giving the immune system its near-limitless variety, is a domesticated transposase, taken over wholesale from the "cut and paste" equipment transposons use to jump around the genome.

The second is the regulatory toolkit — a large, ancient set of "switch" genes, shared across strikingly different animals, that don't build any particular structure themselves but control where, when, and how much other genes get used. Hox genes control body-segment identity in everything from insects to vertebrates. Pax6 is the master switch for building an eye, so deeply conserved that a mouse's version of the gene, inserted into a fruit fly, will trigger eye formation there. Tellingly, the eye that forms is a fly's eye, not a mouse's: the switch says "build an eye here," and the fly's own machinery decides what kind. Present-day cases show this kind of regulatory change producing fast, repeated evolutionary novelty, and the stickleback fish, met in Chapter 7, is the clearest. After colonizing freshwater lakes, stickleback populations have repeatedly lost their pelvic skeleton. In each case studied, this traces to the deletion of a single regulatory switch controlling a gene, Pitx1, that is itself unchanged and still does its other jobs elsewhere in the body — and the same deletion has arisen independently in population after population.

The third is modular construction at the level of the protein itself — building something new not by writing a new sequence letter by letter, but by recombining existing, functional pieces into new combinations. Many proteins are built from independent structural and functional units called domains, and a process called exon shuffling can mix and match these domains between otherwise unrelated genes, producing a working protein that has never existed before out of parts that already worked somewhere else. This is a different kind of event from either a point mutation or a transposon jump, because what's being varied isn't a sequence of letters at all — it's a set of pre-tested building blocks, tried in new arrangements.

Alongside these three, a handful of other mechanisms deserve naming, even though they do a more limited kind of work. Sexual recombination reshuffles existing genetic variation into new combinations every generation, without requiring any new mutation at all. Horizontal gene transfer, and its more distant cousin hybridization, let an organism acquire a gene that some entirely separate lineage already built and tested — Tibetans, for instance, carry a high-altitude adaptation gene picked up through ancient interbreeding with Denisovans, not through any mutation of their own. Whole-genome duplication is the large-scale version of ordinary gene duplication, doubling an organism's entire genetic toolkit at once rather than one gene at a time, and it is thought to have happened twice near the root of the vertebrate lineage. And endosymbiosis is the most extreme case of all: rather than building a new capability piece by piece, an organism can acquire one wholesale, fully formed, by absorbing another organism that already had it — mitochondria and chloroplasts being the textbook case, once free-living bacteria, now permanent, essential residents inside every complex cell that has them. What these four have in common is that each moves existing material from one place to another. They can change what a given organism has, but they add nothing to what exists across life as a whole, and for that reason they get no further separate treatment in this chapter.

Here is the general shape of the answer this chapter will arrive at, stated plainly before working through the details. With one likely exception — the regulatory toolkit, which gets a much closer look later on — none of these mechanisms produce new sequence that wasn't already present somewhere. They rearrange. They redistribute. They borrow, duplicate, relocate, and recombine material that, in one form or another, already existed. A shuffled protein domain had a function before it was shuffled. A duplicated gene is, at the moment of duplication, simply two copies of something that worked. A horizontally transferred gene was built by some other organism's history. Even a newly switched-on regulatory gene is switching on machinery that was sitting there, fully formed, doing something else. Every one of these processes can give the organism carrying it something new — a new capability, a new structure, a new way of surviving a particular environment — without requiring a new sequence to be written.

That is not a small caveat. It is close to the whole reason for this chapter. If none of these mechanisms can produce original genetic material, then all of them, however useful, sit downstream of a harder question that is still unanswered: where did the material being rearranged come from in the first place? That is why point mutation, and the vast reserve of raw sequence classical theory calls junk DNA, haven't been retired by any of this. On everything covered so far, they remain close to the only proposed route by which something with no prior existence in the genome could come into being at all. Biologists have a name for that route when it succeeds: de novo gene birth, a new gene arising from stretches of DNA that previously coded for nothing. It has been documented, and it is the process Chapter 8 put to the test. This chapter's task, once the toolkit has had its full hearing, is to ask whether that is still the only source of new material — and if it is, how much of life's history these newer mechanisms can explain, and how much still has to be carried by the much older, much slower process this book has already spent so many pages putting numbers to.

Last revised 2026-11-19

← Chapter 16: Evolutionary SystemsTwo Systems, Hiding in One Bacterium →