Banking the World's Seeds

AThe idea that a crop variety can be rescued from extinction simply by putting its seed into a freezer has an appealing finality about it, and it underpins a global network holding something in the order of seven million samples. The arithmetic behind the network is genuinely impressive: a sample dried to five per cent moisture and stored at minus eighteen degrees will, for most species, still germinate after several decades, while duplicate copies lodged in a second country insure against war, fire and administrative collapse. What the arithmetic conceals is that a banked sample is not a museum object but a living population on pause. Seed loses vigour, slowly but without exception, so every accession must eventually be taken out, sown, grown to maturity and harvested again. That operation, known as regeneration, is where the awkward questions begin. Curators must decide how many plants to sow, how to keep pollen from neighbouring plots out, and which fraction of the resulting harvest goes back into store. Each cycle passes the variety through a limited number of parent plants, and whatever is not represented among those parents does not come back.

BThe most systematic attempt to measure the cost of that bottleneck comes from Ingrid Sorhaug, a population geneticist at the Rodal Centre for Crop Diversity, who contends that the threat to banked diversity lies inside the routine rather than in any dramatic failure of the freezers. Between 1978 and 2014 her institution regenerated forty-two accessions of rye and fodder beet, both of which depend on pollen carried from other plants, and archived a portion of every harvest. Sequencing the archived material in reverse allowed her group to count what had gone. Alleles carried by fewer than five per cent of the plants in the founding sample had vanished from sixty per cent of accessions after only three cycles, whereas alleles common in the founding sample survived essentially intact. The losses tracked one variable closely, which was how many parents each cycle had drawn on. Accessions regenerated from at least two hundred plants lost little, while those regenerated from around eighty, a figure curators reach for when greenhouse space is short, lost a great deal.

CSorhaug is careful to say what her figures do not show. Drift on this scale is a problem of outcrossing crops, in which every generation reshuffles the population; wheat and barley, which fertilise themselves, pass through regeneration with far less erosion, and the great cereal collections are therefore in better condition than her headline number suggests. Nor is the damage spread evenly across a genome, since it falls overwhelmingly on rare variants, which is precisely why it went unnoticed for so long: a curator comparing plant height, ripening date or grain colour before and after a cycle would see nothing wrong at all. The difficulty is that rare variants are the ones a breeder eventually needs, because resistance to a disease that has not yet arrived tends to be carried by a handful of plants rather than by the crowd. Her recommendation is unglamorous and expensive: bigger regeneration plots, controlled pollination, and a written record of how many parents each cycle actually used. Whether a national programme can afford all three at once is a question she leaves to the ministries that write the budgets.

DNot everyone accepts that this is where the effort should go. Emeka Oduya, who directs the Kirundo Agricultural Station, dismisses the drift literature as a bookkeeping worry and maintains that the great failure of seed banking is institutional rather than molecular. An audit he published in 2019 found that, across eleven banks in one regional network, just six per cent of accessions had been requested by any breeder in the preceding decade, while a third of the collections had not had their germination rates checked within the interval their own protocols required. A variety that nobody grows, he argues, is being conserved in name only. Sorhaug does not challenge the audit, and concedes that use matters, but replies that an accession never asked for still retains its option value, whereas an allele lost in a greenhouse is gone for every future request. His evidence, moreover, rests on a single network of modest size, and demand from breeders is an imperfect measure of what a collection is worth.

EBoth arguments assume, in any case, a seed that tolerates drying, and a substantial minority of plants produce nothing of the sort. Those of cocoa, mango, avocado and most oaks are killed by the very treatment that preserves a grain of wheat, so their diversity has to be maintained as living trees in field collections, exposed to drought, disease and, in one much-cited case, a hurricane that destroyed a Caribbean cocoa plantation in 2004. Cryopreservation offers a way round the problem by freezing excised embryonic tissue in liquid nitrogen at minus one hundred and ninety-six degrees, yet a workable protocol exists for only a small proportion of these species, and the cost of each accession runs to roughly eight times that of conventional storage. The picture that emerges is not one of a system failing but of a system whose weakest points sit where they are hardest to see. Freezers rarely break down. Nor, on the evidence of the past forty years, do the institutions that operate them collapse as often as the founders of the network once feared. What erodes is the fine structure of a collection, cycle by cycle, in species that outcross, and what is missing altogether is much of what will not sit in a freezer in the first place.