Gene drive organisms (GDOs) represent a paradigm shift relative to previous generations of genetically modified organisms (GMOs), including with respect to risks. To limit the potential for negative impacts, the rule so far has been to keep genetically modified organisms (GMOs) restricted to their release sites – such as keeping GM crop plants to their respective agricultural fields. The aim is to prevent the spread of GMOs into the wider environment, and to avoid or minimise the flow of modified genes into non-GM populations or wild relatives. In contrast, gene drive organisms are specifically designed to spread into the wild, often by actively modifying each subsequent generation, with the aim of altering, replacing or eliminating target populations and possibly entire species. Risk can be described as the severity of a potential harm multiplied by the likelihood of this harm. The overall risks will be underestimated if: a potential harm is not identified, its severity is not fully recognised, or its likelihood is misjudged. This emphasises the need for full comprehension of potential harms, which in turn requires full understanding of the matrix of biological interactions and physical settings in which gene drives will operate and have their effects. Reaching such an understanding is extraordinarily difficult. Thus, gene drive organisms not only bring new risks at various levels, but also serious challenges to risk assessment. I. Overarching issues and challenges relating to risks include: There is a high degree of complexity More on this... Gene drives and gene drive organisms affect and are affected by a multitude of interactions at various levels. These would include the levels of DNA and gene regulation, the genome as a whole, the internal and external environment of a cell or organism, including signalling pathways, metabolic pathways and feedback loops, and interactions at the population, and ecosystem levels. Such complexity makes predicting how gene drives will behave and their consequences very difficult. There is a wide spectrum of performance scenarios More on this... A gene drive might work as intended, or it might work differently under different conditions or stop working altogether after some time, for example due to gene silencing, build-up of resistance or the wide range of genetic variation in wild populations. Equally, impacts and consequences might be different under different conditions or natural settings. Once released self spreading gene drives cannot be halted or recalled nor can their modifications be reversed. More on this... Whilst ‘anti-drives’ have been conceptualised, such capacities do not exist for use in the wild as a risk management response should a released GDO cause harm. There is no consensus as to what constitutes a harm or a risk and who should decide on this. More on this... For example, is the reduction of any species a harm or only of those that are explicitly listed in a countries or agency’s ‘protection goals’? Is the ‘value’ of a species intrinsic or externally defined according to specific parameters? At a practical level: does the spread of genetic modifications, including transgenes or engineered gene drives, beyond the target population or the target species constitute a harm? Such gene movement can occur due to sexual vertical gene transfer or non-sexual horizontal gene transfer. Whilst some see such gene transfers and ‘gene flows’ as a problem and a harm in themselves, others do not. It is unclear what the evolutionary effects of engineered gene drives will be over space and time. More on this... As they are strong interventions or disruptions, responses seem likely. Evolutionary effects could occur within the genome of the target species, either in response to the gene drive element and its action, or in the gene drive construct itself. Or they could be at the co-evolutionary or ecosystem level, in response to the loss or alteration of the target species. It will be important to combine the ‘overarching issues and challenges’ above with all the areas of ‘effects and risks’ listed below to build the basis for a comprehensive risk assessment. In line with the Precautionary Principle, releases of GDOs should not be undertaken where there is the potential to cause serious harm and scientific evidence (that such harm will not occur) is lacking. II Risks - viewed through cause & effect logic One way of seeking to identify or understand risks is analysing chains of causes and effects. In the absence of wide-ranging data from open release experiments and long-term ecological studies this becomes a largely theoretical undertaking. In such circumstances, a ‘cause and effect’ approach to understanding risks or potential harms rests on the assumption that we know all the links and connections in a chain and how they will play out. Adverse effects will play out differently depending on the type of gene drive, the target species (e.g. specific mammal, insect, or plant) and the specific environmental settings. Harmful impacts are likely to be consequences of both direct effects (such as eliminating/’suppressing’ a population) and indirect effects (such as the reduction in a predator species resulting from such elimination/’suppression’). Harms could also result from cascading knock-on effects where there are greater numbers of steps in the causal chain, including branching chains of effects. In the context of the overarching issues described in the section above, risks will not necessarily be easily identified or understood, especially as gene drives cannot readily be tested in the wild. Risk assessment will often view risks through the ‘cause and effect’ lense, which is valuable, but has its limitations. System level risks and their enhanced complexities are easily missed out, but equally important. They will be covered in section III. Examples of ‘cause and effect’ logic Note this is an indicative and non-exhaustive list. 1) Suppression & elimination Suppression and elimination of target organisms – if the intervention works as intended - will likely have intended and unintended effects, both potentially with a wide range of knock-on effects. These may include but are not limited to: food-web impacts: More on this... For example, loss of the target organism will impact any predators that rely on it as a food source. Potential target species such as mice, rabbits and insects, including mosquitoes (both in larval and adult stage) are all prey for other species. Knock-on effects of the removal of a target species may include reduced populations of a predator, or a predator suddenly switching to alternative prey, potentially including predation of species that were not previously prey. [1.1] host-parasite impacts: More on this... For example, target organisms may be the main host of a parasitic organism, such as the parasitic wasp Trichopria drosophilae in Drosophila suzukii in its home range.[1.2] Elimination of the host may cause behavioural changes in the parasitic organism, or its complete loss. harm to non-target organisms, and loss of ecosystem functions: More on this... For example, impacts on non-target organisms could occur due to shifts in the food-web or due to the loss of ecosystem functions and services provided by the target organism, such as pollination or decomposition. An example of an area where there is potential for non-obvious effects, is disruption to the contribution of mosquito larvae to nutrient recycling through consuming debris, with possible consequences for water quality. niche replacement: More on this... It is not known which species might fill the niche vacated by the target organism and how this will play out with respect to ecosystem interactions and impacts. In the agricultural context the suppression or elimination of a pest insect may for example result in the emergence of secondary pests, that then might become established as a primary pest.[1.3] Potential impacts of niche replacement include the arrival or spread of a more problematic or differently problematic organism than the target organism. wider knock-on effects over time and space (e.g. delayed domino effects): More on this... for example, unintended elimination or near elimination of a target organism in its home range, when it has originally been targeted as an invasive species outside of its home range (e.g. grey squirrels, rabbits, mice, certain insect pest). Rebound effects: More on this... If a target organism was ‘supressed’ or near eliminated for a period, then for some reason the population recovered, the re-emergence of the target organism in this new context organism would bring consequences. In the case of malaria mosquitoes, the re-emergence of a pathogen (such as the Malaria pathogen) can result in an increased severity of the disease in people that re-encounter the pathogen after its prolonged absence - see [1.4]. Chasing dynamics: More on this... “wherein the drive recurrently eliminates the population from local areas, which are then recolonized by wild-type individuals entering from other areas” [1.5], which may also have different genotypic or phenotypic characteristics. Such chasing dynamics have potential impacts on food chains for example, pathogen flow, or predator behaviour. 2) Modification & replacement Modification of a target organism/species or the replacement of wild populations with populations with altered traits will – if the intervention and gene drive action proceed as planned - have a wide range of potential knock-on effects. These may for example include: More on this... Altered behaviour, and/or altered niche occupation, impacting ecosystems unpredictably. Altered mating behaviour, leading to low level mating with wild populations (as previously observed for t-haplotype mice), resulting potentially in different sub-populations. Reduced diversity, unintentional collapse of populations or loss of species. Reduced fitness and general competitiveness, unintentional collapse of populations. Increased fitness and competitiveness compared to unmodified organism, outcompeting other species, which may be closely related or unrelated (see below). Potentially permanently incorporating active CRISPR/Cas into the genome of the target species with evolutionary consequences (see below). Host-pathogen impacts: Modification of the host organisms may intentionally or unintentionally change the interaction between pathogen and host (vector). Questions arise whether this could potentially result in a change of virulence due to adaptation pressures, make the pathogen seek out a different host, attract a different pathogen, etc., and what such changes would depend upon. 3) Increase in persistence & invasiveness Modification/replacement drives may potentially – depending on their performance– result in increased persistence and invasiveness of the altered target organism. This may for example be due to enhanced competitiveness and fitness of the gene drive organism and resulting populations. This may also have to be considered for elimination/suppression drives, though cause and effect chains are less clear and may possibly be related to chasing dynamics, mating behaviour, resistance build-up, genetic variation, etc. Potential effects include: More on this... displacement of related species onset of speciation 4) Uncontrolled & unintended spread Both the engineered gene drive element (at the DNA level) and the gene drive organism could potentially spread beyond an intended target population or target species. The following points are examples of what may need to be considered: More on this... The gene drive could spread beyond the target population, affecting populations of the target species in geographic areas it was never intended to reach. Eradication or modification of target organisms in their native range and habitat - for example when used on alien invasive species, such as rabbits, mice or agricultural insect or plant pests. The gene drive could spread into non-target species via vertical (sexual) or horizontal gene transfer (asexual). The resulting potential impacts in these other species include the potential for unintended elimination or suppression, or for altered characteristics. This in turn may have multiple knock-on effects, such as impact on other species due to effect caused by the integration of novel genetic sequences and element into their genomes. The unintended spread of gene drive elements into non-target species can also be seen as a conservation issue and concern in itself – as it may unintentionally turn wild species into GM organisms. 5) Increased toxicity Whilst increased toxicity may be present from the onset, it may also arise under certain environmental conditions or within specific genetic backgrounds within wild populations. This may potentially impact predators, decomposers, and food-webs. In case of vector organisms, such as mosquitoes, higher levels of allergen or toxin in saliva may impact humans and animals. III. Risks - viewed at ‘systems levels’ In complex biological systems the hypothesis-driven approach to risks will likely fail to identify the full breadth of risks and consequences. So if one does not assume one has all the required knowledge, where does that leave one? This problem has been recognised for example by the concept of general monitoring or general surveillance. In this approach the consequences and harms of any technological intervention are being observed/monitored at the systems level without any predetermined hypothesis. In the case of gene drives - were gene drive organisms to be released into the open environment - consequences could for example manifest at many levels such as: 1) Molecular level effects. More on this... This includes for example the DNA and RNA levels, such as altered gene regulation, altered DNA sequences, altered proteins or protein composition, altered inter- or intracellular communication/messaging. 2) Population genetics and evolutionary level effects: More on this... This could include for example Decreased genetic diversity in target population Build-up of resistance (evolution) to engineered gene drive in target species/population Change of target site specificity of the CRISPR/Cas element of engineered gene drives over time and space. Presence of active CRISPR/Cas molecules resulting in low level background modifications, making outcomes unpredictable 3) Ecological level effects, More on this... e.g. due to both intended and unintended outcomes and knock-on effects across impact levels. Impacts may include reduction of ecosystem functions, loss of resilience, destabilisation, emergence of new pests or pathogens. 4) Physiology, whole organisms, and life cycle of organism levels. 5) Human & animal health level effects. More on this... This could for example be due to increased or altered disease transmission; or increased toxicity or allergenicity. 6) Socio-economic level impacts. More on this... There could for example be opportunity costs of neglecting safer and proven alternatives, such as rebuilding functional food webs, adjusting agricultural practices, or – in the case of vector transmitted diseases - addressing access to medicines, housing improvements, environmental sanitation. 7) Cultural and ethical level impacts. 8) Biosphere and climate level impacts. In the case of gene drives, however, general surveillance does not provide the answer, because the risks must be fully understood before any release as to avoid to possibility of serious and irreversible harm. This brings us full circle back to this fundamental tension underlying risk identification and risk assessment of gene drives. – namely the impossibility to comprehensively predict risk prior to any release.