Will it work

The outcome of any release of gene drive organisms would be subject to a complex interplay of processes at different levels, encompassing molecular and cellular biology, the physiology and behaviour of the whole organism, the interactions of the organism with the ecosystem, geography and weather, population biology and multiple evolutionary processes. As such it is possible – even likely - that unanticipated phenomena will occur, which alongside the other factors discussed below, could undermine the efficacy of engineered gene drives.

Imagine that a gene drive was employed to try and eliminate a non-native population of mice from an island where they were impacting biodiversity. Gierus and colleagues propose attempting this by using a gene drive system to spread genes encoding the DNA cutter CRISPR-Cas9, in this case engineered to inactivate an essential female fertility gene. Imagine the gene drive spread the CRISPR-Cas9 genes through almost all the mouse population, but before the mice became extinct a mutation arose making the targeted fertility gene resistant to inactivation. This mutation then spreads rapidly due to natural selection and the mouse population recovers, leaving the island inhabited by transgenic mice all carrying the gene drive. What would be the long-term effects of modifying the mouse population in this way? And given that mice stow away, could the gene drive spread to other populations?

As a result of fitness costs, natural selection can act against a gene drive spreading, or even completely eliminate the gene drive. Because gene drives are designed to spread, natural selection must generally act in combination with other factors to cause a gene drive to fail.

• 'Intended' fitness costs : Gene drives that aim to ‘suppress’ or eliminate wild populations would in some circumstances bring a fitness cost by design, for example by rendering females sterile. Gene drives that aim to ‘suppress’ or eliminate wild populations, for example by rendering females sterile, carry an inherent fitness cost.[ 2.1 ]
• Unintended fitness costs: Genetic modification can result in unintended side effects that render modified organisms less fit than the wild type. The design of gene drives is often based on the assumption that such effects will be insignificant – and that therefore that at least some gene drive organisms (e.g. males) will be competitive enough to mate and spread the drive. [ 2.2 ]

• Mutations or natural heterogeneity at target sequence: Many gene drive systems that rely on the DNA cutter CRISPR-Cas9 are vulnerable to the emergence of mutations at the target sequence that prevent cutting. In the case of homing gene drives such mutations can prevent the insertion of the gene drive element into the genome, and so render organisms carrying the mutations resistant to the drive. Pre-existing natural genetic variation at the target site could also render a gene drive ineffective through the same mechanism. [ 2.3 ]
• Mutations in gene drive element: Mutations could also appear in the gene drive element itself. For example mutations in the guide RNA sequences that target the CRISPR-Cas9 could render the DNA cutting less efficient, disrupting the action of the gene drive. Such mutations could also lead to off target DNA cutting – in the case of a homing gene drive this could cause the gene drive element to insert at unpredicted locations in the host genome, with consequences that are hard to predict.

Experiments with mice show that polyandry (i.e. females mating with more than one male) counteracts the spread of the T-haplotype, a natural selfish genetic element, and similar effects would be expected with gene drives based on the T-haplotype. Theoretical studies indicate that behaviours that encourage mating with siblings or ‘that cluster drive-free individuals’ would also interfere with the spread of gene drives.

Dispersal rates are influenced by a host of factors ranging from physiology and behaviour to geography and weather, and it is possible that fitness costs would cause gene drive organisms to disperse less well than their wild type counterparts.

Modelling shows that '…competing species or predators can facilitate drive-based suppression, albeit at the cost of an increased rate of drive loss outcomes.’

For example, in the case of mosquitos, the lifecycle is dependent on the presence of bodies of water for the development of larvae. Populations are therefore drastically reduced during prolonged dry periods, which facilitates replacement of gene drive modified populations by more fertile unmodified populations.

Geography and ecology combine to create complex population structures, with denser or sparser populations in different areas, as well as limitations on inter-breeding between different regions. Models suggest that such ‘real world’ complexity could allow islands of un-modified populations to persist, which could then re-colonise areas where a gene drive had eradicated the target species. Thus there could be perpetual cycles of invasion of un-modified populations by gene drive organisms, leading to localised extinction of the target, followed by re-colonisation of the area by unmodified organisms. This is sometimes referred to as ‘chasing dynamics’ or as a ‘rock/paper/scissors’ effect.

For example, if mosquitoes were modified to make them resistant to malaria infection, could the malaria parasite evolve to evade this resistance? If some mosquito vectors were eliminated, could malaria parasites adapt to spread via different vectors, as they have done in the past. And if mosquito species that specialise in feeding on humans were eliminated, would evolution drive other mosquito species to specialise in the same way? [ 2.5 ]These are just some examples of a range of possible outcomes in which the promised benefits to human health might be counteracted. As with other aspects of this technology, the complexity of the real world could bring outcomes that have not yet been imagined.

Many of the gene drive systems demonstrated in the lab (e.g. in Drosophila species and some mosquitoes) are forms of the homing CRISPR gene drive design, however, efforts to apply the same design in other species have either failed (e.g. in plants) or showed very limited effectiveness (e.g. in mice). [ 2.6 ]