Gene drives: the case for caution and the case for trying

Every trait you carry had roughly a coin-flip chance of being passed on. You have two copies of each gene, a sperm or egg gets one, and over generations that arithmetic keeps most new mutations rare. A gene drive breaks the coin flip. It is a genetic element that copies itself onto the matching chromosome, so that nearly every offspring inherits it. Give it enough generations and it spreads through a population even if it does the animal carrying it no good at all. Especially then, in fact, since harmful drives are the useful ones if what you want is fewer mosquitoes.

Natural versions have existed for millions of years. The engineered kind became practical in 2014 and 2015, when researchers realised that CRISPR made the copying step easy. The first demonstrations were in yeast, fruit flies and then mosquitoes, and the field has argued about itself ever since.

The mechanism, briefly

A CRISPR drive is a cassette of DNA inserted into a chosen gene. The cassette encodes the Cas9 enzyme and a guide RNA that points Cas9 at the exact spot on the chromosome where the cassette sits.

In an animal carrying one engineered and one normal chromosome, Cas9 cuts the normal one. The cell repairs the break, and if it uses the intact chromosome as a template, it copies the whole cassette across. Now both chromosomes carry the drive, and essentially all the offspring inherit it rather than half of them.

Two broad designs exist. A suppression drive targets a gene needed for fertility or female development, so the population shrinks as the drive spreads. A modification drive carries cargo, for instance genes for antibodies that block the malaria parasite inside the mosquito, so the population survives but stops transmitting.

The obvious failure mode arrived early. Sometimes the cell repairs the cut sloppily instead, leaving a mutated target site that Cas9 can no longer recognise. Such resistant sequences spread fast, because carrying them is better than carrying a drive that makes you sterile. Early mosquito drives failed this way within a few generations.

The 2018 answer was to pick a target that cannot easily change. A team at Imperial College London aimed a drive at a region of doublesex, the gene that controls female development in Anopheles gambiae. Mutations there leave females sterile and intersex, so resistant versions do not spread. In cages, the drive reached every mosquito within about seven to eleven generations and the populations collapsed (doi:10.1038/nbt.4245). In 2021 the same approach was tested in large indoor cages in Italy, with swarming, mating and varied conditions closer to real life. The populations were suppressed there too, without resistance emerging.

Cages are not villages. But as laboratory results go, these are not ambiguous.

Why malaria

Malaria still kills an enormous number of people, most of them African children under five; the World Health Organization's annual world malaria report has put annual deaths at well over half a million for years, with no sustained decline since around 2015. Insecticide-treated nets and indoor spraying produced large gains in the 2000s. Then mosquitoes evolved resistance to the insecticides, parasites evolved resistance to the main drugs, and the curve flattened. The two approved vaccines help but are only moderately effective and need several doses.

So the argument for gene drives is not that they are elegant. It is that the existing tools are losing ground and a drive, in principle, needs no cold chain, no clinic and no annual campaign. One of the species responsible, Anopheles gambiae, is also a plausible target: it specialises in biting humans, and no one has shown that anything depends on it specifically.

How it is actually being done

The project furthest along is Target Malaria, a research consortium led from Imperial College with partners in Burkina Faso, Ghana, Mali and Uganda, funded mainly by philanthropic sources. Its public plan has always been staged, with no gene drive at the start.

Stage one was a release of genetically modified but non-gene-drive male mosquitoes, sterile and unable to bite, in the village of Bana in Burkina Faso in July 2019. A few thousand were released alongside unmodified males for comparison, then recaptured to measure how far they flew and how long they survived. This was the first release of a genetically modified mosquito in Africa. Scientifically it was modest. Institutionally it was the point: it put a national regulator, a national research institute and a village through the full sequence of approval, consent and monitoring, with a strain that could not persist.

Stage two involves a self-limiting male-bias strain, which skews offspring towards males without driving. By early 2025 that strain had been imported into Uganda under national approvals, and dossiers for releases were in preparation elsewhere. A gene drive release has not been applied for.

Separately, the company Oxitec has been releasing self-limiting mosquitoes, in which female offspring do not survive, for years against dengue-carrying Aedes aegypti in Brazil and elsewhere, and from May 2024 against the invasive urban malaria vector Anopheles stephensi in Djibouti. These are not gene drives. They disappear from the population unless you keep releasing them, and they need repeated releases and a production facility, which is exactly the kind of recurring cost a drive is meant to avoid. They are a useful halfway house: real releases, real regulators, reversible if things go wrong.

Where it is being argued

At the Convention on Biological Diversity, mostly, under the heading of synthetic biology.

In 2016 at the Cancún meeting, a coalition of civil-society organisations and some governments called for a moratorium on releasing gene drive organisms. It was not adopted. In 2018 in Sharm El-Sheikh the parties settled on a compromise that has shaped the debate since: a call to apply a precautionary approach, to conduct risk assessment case by case, and to seek the prior informed consent of potentially affected indigenous peoples and local communities before any release. In 2022 in Montreal, a renewed push for a moratorium again failed, and the parties instead set up a process of regular horizon scanning and assessment of developments in synthetic biology. At the Cali meeting in late 2024, parties to the Cartagena Protocol welcomed additional voluntary guidance on assessing the risks of gene drive organisms, over objections from groups who considered it insufficiently precautionary.

The pattern repeats: no moratorium, no clear permission, more process. The one firm commitment, consent from affected communities, is the hardest to operationalise. A mosquito does not stop at a border, so who counts as affected? The village where the release happens, the district, the country, every country the species occupies? There is no agreed answer, and the question is not merely procedural. It decides who holds the veto.

That dispute belongs to a family of arguments about who benefits from biotechnology and on what terms, the same family as the fight over genetic sequence data traced in our piece on digital sequence information and the Cali Fund.

The case for caution

It is not recallable in the ordinary sense. Once self-propagating organisms are out, you cannot collect them. Proposed reversal drives and immunising drives exist on paper and in the laboratory. Nobody has shown one working in the wild, because nobody has put a drive in the wild.

Ecological knowledge is thin. A. gambiae is a complex of closely related species that exchange genes. Suppressing one could open a niche for another, possibly one that bites humans less, possibly one that bites them more and resists the usual controls. Models help; models of ecosystems have a poor record.

Mosquitoes travel. A drive released in one country can reach its neighbours, which have their own regulators and did not consent. International biosafety law was written for crops in fields, not for self-spreading insects.

Precedent matters more than this case. The malaria application is the most sympathetic one imaginable. Approve the method here and the next proposals will be for agricultural pests, invasive rodents on islands and commercially inconvenient weeds, where the balance of risk and benefit looks very different. Agricultural biotechnology has a long record of arriving as a humanitarian promise and leaving as a commercial product, a pattern visible in everything from golden rice to the pharmaceutical crops I wrote about in the Ventria trial in Peru.

Consent at scale is unsolved. Village-level engagement, which Target Malaria has invested in heavily, is necessary and insufficient. Nobody has a mechanism for the consent of a region.

The case for trying

The counterfactual is not safety. Hundreds of thousands of deaths a year, concentrated in small children, is the status quo. Delay has a body count, even if it is never attributed to a decision.

The resistance problem is getting worse. Pyrethroid resistance is widespread, parasite resistance to artemisinin has been detected in East Africa, and new chemistry arrives slowly. Betting everything on incremental improvement of existing tools is also a bet.

The science has answered its first objection. Resistance to the drive itself, the reason early designs failed, was solved by choosing a constrained target. That is a real result, not a press release.

The staging is genuine. Years of non-drive releases, contained studies and regulatory groundwork before any drive application is roughly what a critic in 2016 would have demanded.

Research is not release. A moratorium on laboratory work would push the science somewhere with less scrutiny. The expertise is no longer concentrated in a few northern universities, as the spread of synthetic biology capacity described in synthetic biology's turn to Asia makes plain.

What I think a careful reader should hold

Two things at once. The malaria case is strong and the precedent it sets is genuinely dangerous, and neither cancels the other.

The questions worth tracking are not "is this safe" in the abstract. They are narrower. Who decides, when the organism crosses borders the decision does not? Is there a tested way to stop a drive, not merely a published design? Who pays if a release goes wrong, and under what liability rules? And does the staged, consent-heavy process Target Malaria has followed survive contact with the second applicant, the one in a hurry, with a commercial product and a less sympathetic target?

A field release of a gene drive has not happened. When one is proposed, those four questions will matter more than the cage data, and they are not scientific questions at all.