Hybrid seed decline

Why do seeds saved from a superb hybrid crop give a disappointing and uneven harvest the following year?

A farmer plants an F1 hybrid maize and gets a field so uniform it looks printed: same height, same tasselling date, same ear. He keeps seed from the best cobs — the sensible thing to do, and what farmers did for ten thousand years before hybrids existed. The next season's field is ragged. Some plants are fine, many are small, the harvest date smears across three weeks, and the average yield has dropped.

The tempting explanations are all about the seed: it was sterile, or degraded, or deliberately crippled. But the seed germinated perfectly well, which is what makes the result confusing rather than merely disappointing. Something happened between the parent plant and the offspring that did not happen when his grandfather saved seed.

Start with what an F1 hybrid actually is. It is the cross of two inbred lines — populations self-pollinated for many generations until, at nearly every gene, both copies are identical. Each line on its own is feeble, which is the first clue that something interesting is going on.

Now cross them. Take a gene where line A carries only allele A and line B carries only allele b. Every seed of that cross receives A from one parent and b from the other, so every F1 plant is Ab at that gene — and the same argument runs at every gene where the two lines differ. Two consequences fall straight out, and they are different consequences that get bundled together as "hybrid vigour".

The first is uniformity. There is exactly one possible F1 genotype. Every plant in the field is a genetic clone of every other, so the only variation left is what the weather and soil impose. The field looks printed because it is printed.

The second is vigour. Each inbred line has accumulated, in the double dose, mildly harmful recessive alleles it could not purge. In the cross, each line's harmful copies are masked by the other line's functional ones. There is a long argument about whether that masking — dominance — is the whole story, or whether being heterozygous is itself advantageous at some loci; the dominance account, with contributions from gene interaction, is where most current evidence points, but it is not settled.

Now save that seed and let it self or intercross. At our single gene, the F1 was Ab. Its offspring are one quarter AA, one half Ab, one quarter bb — Mendel's ratio, and here it is doing the damage. Heterozygosity has gone from every plant to half of them, in one generation. If vigour tracks heterozygosity, then the F2 keeps about half the advantage the F1 had over its inbred parents, and loses the other half. That is a derived number, not a measured one, and it is roughly what breeders observe.

And the unevenness? Count the possibilities. With one segregating gene there are three F2 genotypes; with two, nine; with n, three to the power n. Ten segregating genes give 59,049 distinct combinations, and a real cross segregates at thousands. The F1 field had one genotype and the F2 field has effectively as many genotypes as it has plants. The raggedness is not decay. It is variety, arriving all at once.

Three things this account is careful not to say.

It does not say the seed is sterile or sabotaged. F1 seed is generally fully fertile, and nothing was engineered to prevent reproduction — the technology that would have done so was never brought to market. Failing to breed true is a property of being heterozygous, and heterozygosity is the thing the breeder was selling. The commercial convenience of a crop that must be repurchased is a consequence of the biology, not the reason for it.

It does not say the seed degraded in storage. Storage decline reduces germination and would hit an old open-pollinated variety equally.

And it is not regression to the mean, though the outcome resembles it. Regression explains why an exceptionally lucky measurement is followed by an ordinary one — the luck does not repeat. Here the parent was not lucky; the F1 genotype was constructed on purpose and would repeat exactly if the cross were made again. The drop is deterministic segregation, and unlike regression it is predictable in size before the seed is sown.

Which gives the test that would sink it. If the decline came from segregation and lost heterozygosity, then seed saved from a non-hybrid open-pollinated variety — already near enough homozygous, and reproducing a population rather than a genotype — should not show it. It does not, which is why seed-saving worked for millennia. And the account predicts most of the fall happens in the first generation only: heterozygosity goes one, one half, one quarter, so the F3 loses far less than the F2 did. If a farmer's F3 had collapsed as sharply again, this explanation would be wrong.

• Why inbred lines are feeble in the first place, and what that says about the load of harmful alleles a population carries.
• How synthetic and composite varieties try to hold some vigour while remaining saveable.

Every term the collection defines is gathered in the glossary.