Complete population collapse of malaria vector Anopheles gambiae in cages is achieved using a gene drive that targets doublesex.Supplementary informationThe online version of this article (doi:10.1038/nbt.4245) contains supplementary material, which is available to authorized users.

Gene drives have enormous potential for the control of insect populations of medical and agricultural relevance. By preferentially biasing their own inheritance, gene drives can rapidly introduce genetic traits even if these confer a negative fitness effect on the population. We have recently developed gene drives based on CRISPR nuclease constructs that are designed to disrupt key genes essential for female fertility in the malaria mosquito. The construct copies itself and the associated genetic disruption from one homologous chromosome to another during gamete formation, a process called homing that ensures the majority of offspring inherit the drive. Such drives have the potential to cause long-lasting, sustainable population suppression, though they are also expected to impose a large selection pressure for resistance in the mosquito. One of these population suppression gene drives showed rapid invasion of a caged population over 4 generations, establishing proof of principle for this technology. In order to assess the potential for the emergence of resistance to the gene drive in this population we allowed it to run for 25 generations and monitored the frequency of the gene drive over time. Following the initial increase of the gene drive we observed a gradual decrease in its frequency that was accompanied by the spread of small, nuclease-induced mutations at the target gene that are resistant to further cleavage and restore its functionality. Such mutations showed rates of increase consistent with positive selection in the face of the gene drive. Our findings represent the first documented example of selection for resistance to a synthetic gene drive and lead to important design recommendations and considerations in order to mitigate for resistance in future gene drive applications.

Homing-based gene drives use a germline source of nuclease to copy themselves at specific target sites in a genome and bias their inheritance. Such gene drives can be designed to spread and deliberately suppress populations of malaria mosquitoes by impairing female fertility. However, strong unintended fitness costs of the drive and a propensity to generate resistant mutations can limit a gene drive’s potential to spread. Alternative germline regulatory sequences in the drive element confer improved fecundity of carrier individuals and reduced propensity for target site resistance. This is explained by reduced rates of end-joining repair of DNA breaks from parentally deposited nuclease in the embryo, which can produce heritable mutations that reduce gene drive penetrance. We tracked the generation and selection of resistant mutations over the course of a gene drive invasion of a population. Improved gene drives show faster invasion dynamics, increased suppressive effect and later onset of target site resistance. Our results show that regulation of nuclease expression is as important as the choice of target site when developing a robust homing-based gene drive for population suppression.

CRISPR-Cas9 nuclease-based gene drives rely on inducing chromosomal breaks in the germline that are repaired in ways that lead to a biased inheritance of the drive. Gene drives designed to impair female fertility can suppress populations of the mosquito vector of malaria. However, strong unintended fitness costs, due to ectopic nuclease expression, and high levels of resistant mutations, limited the potential of the first generation of gene drives to spread.Here we show that changes to regulatory sequences in the drive element, designed to contain nuclease expression to the germline, confer improved fecundity over previous versions and generate drastically lower rates of target site resistance. We employed a genetic screen to show that this effect is explained by reduced rates of end-joining repair of DNA breaks at the target site caused by deposited nuclease in the embryo.Highlighting the impact of deposited Cas9, many of the mutations arising from this source of nuclease activity in the embryo are heritable, thereby having the potential to generate resistant target sites that reduce the penetrance of the gene drive.Finally, in cage invasion experiments these gene drives show improved invasion dynamics compared to first generation drives, resulting in greater than 90% suppression of the reproductive output and a delay in the emergence of target site resistance, even at a resistance-prone target sequence. We shed light on the dynamics of generation and selection of resistant alleles in a population by tracking, longitudinally, the frequency of resistant alleles in the face of an invading gene drive. Our results illustrate important considerations for future gene drive design and should expedite the development of gene drives robust to resistance. Endonuclease-based homing gene drivesGene drives based on site-specific endonucleases were first proposed over 15 years ago [3] and recent advances in CRISPR technology have led to several demonstrations that this endonuclease, which is easy to reprogram to recognise a genomic site of choice, can be repurposed as a gene drive [4,5].The premise is that the endonuclease is sufficiently specific to recognise a DNA target sequence within a region of interest and the gene encoding the endonuclease is inserted within this target sequence on the chromosome, thereby rendering it immune to further cleavage. When a chromosome containing the endonuclease is paired with a chromosome containing the wild type target site, the site is cleaved to create a double stranded break (DSB) that can be repaired, either through simple 'cut and shut' nonhomologous end-joining (NHEJ) or through homology-directed repair (HDR). HDR involves strand invasion from the broken strand into regions of immediate homology on the intact chromosome, and synthesis across the intervening region to repair the gap. In the arrangement described this can lead to copying of the endonuclease, and its associated allele, from one chromosome to another in a process referred to as 'homing'. If homing takes place in the ge...

(Supplementary Fig. S8) confirming that most of the genes required for the late stages are transcribed before meiosis, as previously observed also in other model organisms and human spermatogenesis 24,55,56 .