Conflict Inside the Genome

Something in your genome copies itself, and it does not need to do you any good

Take the word apart first. All of an organism's DNA together is its genome, and only a small slice of that is instructions for making proteins. About half of the human genome is what a class of sequences called transposable elements has left behind. These sequences differ wildly from one another and share exactly one property: they can increase their own copy number. Some make an RNA copy and write it back into DNA somewhere else, like running off a photocopy; others simply cut themselves out and paste themselves in elsewhere.

The numbers are more extreme than you would guess. One family in humans, called Alu, runs to over a million copies and makes up more than a tenth of all our DNA. Maize is more extreme still: about eighty-five percent of its genome is built of such sequences — the researcher who stared at those speckled kernels later received a Nobel Prize, and the plant she was looking at is, in the main, made of things that move.

One sentence is dangerously easy to say here, and it is wrong: they do not "want" to spread. Nothing wants anything. The logic actually running is dull. If a stretch of sequence happens to carry a property that gets itself copied, then in the next generation there are more copies of it than of things without that property. It needs no benefit at all for the individual carrying it; it only needs to copy itself faster than it gets cleared out. This is the replicator logic from the piece on the gene's-eye view in its barest form — selection never settles accounts on who benefits, only on who leaves more copies.

And the host often pays. Land in the middle of a working gene and that gene is wrecked; several human genetic diseases start exactly that way. Fortunately most copies are long dead and can no longer move — of that million Alu copies only a handful of source elements can still generate new insertions, and of the family called LINE-1 only some eighty to a hundred copies in the whole genome remain active.

Fifty-fifty is a rule that has to be maintained, not a natural default

You get one full set of chromosomes from each parent. When sperm or eggs are made the two sets have to separate and each gamete carries away only one — so at any given position, the father's copy and the mother's copy each have a fifty percent chance of being passed on. That "half" is the most basic accounting in genetics, so basic that we never ask what makes it hold.

Some things can change the ratio. House mice carry a stretch of chromosome (researchers call it the t-haplotype) that a carrier male passes to more than ninety percent of his offspring instead of half.

The method is not dignified. The male makes both kinds of sperm, with and without it; the version carrying it poisons the whole batch, and the antidote is present only in the half that carries it. It does not swim faster — the others were sabotaged. A stretch of chromosome in fruit flies uses much the same trick.

The cost is out in the open. Such stretches usually drag a lethal defect along: embryos that inherit two copies do not survive, or carrier males are entirely sterile. Judged by whether the mouse population as a whole comes out ahead, this is a pure loss. But selection does not settle the group's books — inside each carrier it lifts its own transmission from fifty percent to ninety, faster than the deaths it causes drag it down, so it persists, and it is found in wild mouse populations to this day.

The phenomenon is called meiotic drive, and there is a maternal version: making an egg, meiosis yields four products and only one becomes the egg while the other three are discarded — "who gets that slot" is by its nature a channel that can be biased, and maize has a chromosome structure that takes it more often.

Once you know this, the question inverts: not why anything cheats, but why almost nothing does.

Some of your genes run on the father's copy alone, because the two copies are fighting over the same thing

There is a small stretch of human chromosome 15 that causes trouble when it is missing — and which trouble depends on whether the missing copy came from the father or the mother. Missing from the paternal chromosome, the child has one condition (weak sucking in infancy, then appetite that is extremely hard to control); missing from the maternal one, a condition with entirely different symptoms. Same size of gap, two outcomes.

So cells can tell which parent each copy came from. As sperm and eggs form, certain regions get chemical marks — the same marking system as in the piece on epigenetics and the limits of inheritance — and the marked copy stays switched off. This is genomic imprinting, and only one or two hundred genes in humans and in mice do it: a very small set.

The functions of that small set cluster tightly: most of them bear on how much a fetus draws from its mother. Mice have a growth-promoting gene where normally only the paternal copy works. Break the paternal copy and the pups are born at about sixty percent of normal weight; break the maternal one and nothing changes, since it was switched off anyway. On the restraining side, it is often the maternal copy that is expressed.

Haig's explanation runs like this. In a species where a mother may bear litters by different males, the paternal copy inside one fetus need not be present in her future offspring at all, whereas her own copy has a fifty percent chance of being present in every child she has. So "draw a bit more from the mother" scores differently on the two ledgers: drawing more depletes her and lowers her future offspring count, a cost the maternal copy bears and the paternal copy largely does not. The result is that variants letting a fetus take more are more readily retained when they come down the father's line, and variants restraining the taking are more readily retained down the mother's. Two copies of one gene get pushed into two different settings.

This is not settled. It explains the growth-related imprinted genes beautifully, but others sit in the brain with unclear functions, and rival hypotheses are in play. The argument is still open.

Put the three sections together: a genome holds self-copying stretches, cheats that bias their own transmission, and two parental copies pulling against each other — and it still runs as "one individual". That is not free. Germ cells carry a dedicated set of small RNAs holding transposable elements down, and recombination each generation breaks the chromosomes up so a cheat cannot drag its accomplices along. As the piece on the major evolutionary transitions put it, individuality is something suppression produces rather than something given — and the inside of a genome is the smallest example of that sentence.