Rewriting The Mosquito

AMalaria is carried by only a handful of the roughly 3,500 known mosquito species, and in sub-Saharan Africa most transmission is attributed to three closely related members of the Anopheles gambiae complex. Control has traditionally meant attacking the insect where it is easiest to reach: treated bed nets, indoor spraying, the draining of standing water. Those measures were responsible for a substantial fall in deaths between 2000 and 2015. The fall then stalled, and insecticide resistance is generally blamed. Against that background, a proposal first demonstrated in a laboratory in 2015 has drawn attention out of all proportion to the number of insects involved. A gene drive is a stretch of engineered DNA that copies itself onto the partner chromosome in the germline, so that an insect carrying a single copy passes it to almost all of its offspring rather than to half of them. A sequence that reduces fertility, which ordinary inheritance would erase within a few generations, can in principle spread through a population instead of vanishing from it.

BDmitri Tanaka, a population geneticist at the Tallow Institute of Vector Biology, has spent a decade asking how far that principle survives contact with real insects. Between 2016 and 2022 her group ran 24 large cages, each seeded with 600 mosquitoes of which 60 carried a drive aimed at a gene required for female fertility. In 18 of the cages no eggs at all were produced by the eleventh generation; in the remaining 6 the collapse stopped short and numbers recovered. Tanaka's interest lies in the six. Sequencing showed that in every one of them the target had acquired small deletions which left the gene working but no longer recognisable to the drive: resistance alleles, in the vocabulary of the field, thrown up by the cell's own repair machinery rather than by anything the insects themselves did. She stresses that the rate at which such alleles appear, and not the speed of the initial spread, is what decides whether a release succeeds.

CThe account yields a prediction, and Tanaka has tested it. If failure comes from repair rather than from selection acting on the insects, a drive aimed at a sequence that cannot tolerate alteration should fail far less often, because any repaired copy would itself leave the insect sterile. Her group therefore moved the target into a region of the doublesex gene in which even a single altered letter renders a female infertile. Across 12 further cages the populations collapsed without exception, and no working resistance allele was recovered. The result is regularly cited as the moment at which the technology stopped being speculative. Tanaka herself is careful about what it shows: her cages held a few hundred insects mixing freely in one fixed volume, and she presents the finding as a demonstration that a single obstacle can be engineered around, not as an estimate of what would happen across a landscape.

DKwabena Osei, a field entomologist at the Volta Basin Vector Unit, replies that a cage measures the cage. Populations in the Sahel, he points out, are not one pool but many: in the dry season breeding sites shrink to scattered puddles, and his team's marking studies put the median journey of an adult female at 380 metres, though a small minority are carried much further on the wind. In simulations built on those movement data, a drive released at one site reached fewer than half of the puddles he modelled within three years, and only where releases were repeated at several sites did coverage approach the level earlier projections had assumed. Osei concedes two things. His dispersal figures come from a single river basin in an unusually wet year, and his simulations assume that carrying the drive costs a mosquito nothing in the field, an assumption he calls generous rather than proven.

EThe argument has not stayed in the journals. A framework issued in 2023 by an international health body requires that any first release be reversible in practice as well as in principle, and several governments have asked for evidence that a drive can be halted before they will license a trial. Tanaka observes, awkwardly for her own case, that nothing in her cage results speaks to that requirement, since a construct which spreads dependably is by that very property harder to withdraw. Osei answers that the two matters cannot be kept apart in the field, because regulators treat patchiness as a safety feature while researchers treat it as a failure, and the same map is being read in opposite directions. Both object to the habit of reporting the generation at which a cage population disappeared while saying nothing about the cages that recovered.

FA settlement of a kind is taking shape in which the two accounts are assigned to different questions rather than ranked against each other. Where the issue is whether a construct can be built that resists repair, cage work answers it, and the answer is now largely yes; where the issue is whether a released construct would reach the insects that matter, the structure of the landscape governs the outcome and no cage can stand in for it. The unresolved question is how long a suppressed population stays suppressed once the puddles around it begin to refill, and no trial yet approved runs long enough to find out. What both researchers reject is the assumption behind the early enthusiasm, that a technique which works on a species in a room will work on a species in a country.