The risks of genome-edited crop varieties should be considered alongside their benefits, and in the context of plant breeding. Traditional conventional breeding is not free of risks, such as unintended increased levels of toxic alkaloids in fava bean and potato, introduction of disease susceptibility, or the reduction of protein content when breeding for increased grain yield17. Secondly, mutations occur spontaneously with every generational advance, giving rise both to favorable and unfavorable (sometimes lethal to the variety, such as chlorophyll deficiency) alleles that drive natural selection for fitness and enable farmer- and consumer-guided selection for preferred traits. These risks provide a baseline and context to assess the risks of genome-edited plants.
Non-target edits
One frequently cited risk of genome editing is whether it could lead to additional mutations that compromise the safety or agronomic performance of a variety. From the initial deployment of clustered regularly interspaced short palindromic repeats (CRISPR) technology, and as for previous double-strand break technologies (such as transcription activator-like effector nucleases (TALENs), zinc-finger nucleases or meganucleases), it was apparent that sites other than the target site could be inadvertently edited following introduction of a double-stranded break18. An appropriate context when considering the implications of this potential non-target mutagenesis is the frequency of spontaneous natural mutagenesis (10–8–10–9 per base pair (bp), or perhaps 5–140 per plant genome)19,20. For perspective, the genomic data of 3,010 accessions of Asian cultivated rice identified diversity as high as 1 single-nucleotide polymorphism for every 22 bp in the 370-Mbp rice genome21. Another consideration is whether non-target edits in plants present new concerns relative to the risks that are inherent to other breeding approaches. The frequency of chemical- or radiation-induced mutations introduced during the development of more than 3,200 horticultural and crop varieties is around 1,000 times greater than natural frequencies. Current genome-editing technology options generate non-target mutations at a similar frequency to natural mutations that result from spontaneous mutagenesis15,16,17, and at a frequency much lower than that of induced mutation methods.
Continuous improvement in the bioinformatic tools and approaches used to design genome-editing targets works to mitigate and reduce the likelihood of occurrence of non-target edits in crop plants. Much effort has been invested to develop robust assays for mutations, and to understand the frequency and nature of non-target site mutations resulting from CRISPR technology22. These efforts have resulted in CRISPR systems with increased fidelity and fewer non-target edits. With technological advances that help limit non-target mutations, we believe that there are no significant safety concerns that are unique to the deployment of genome-edited crops when assessed in the context of the long history of safe use of conventionally bred varieties, including those derived from mutagenesis.
Breaking of natural reproductive barriers
Another concern could be that genome editing, similar to chemical- or radiation-induced mutations17, breaks reproductive barriers that would prevent some mutations from occurring in nature. For example, some DNA segments are tightly linked, effectively preventing recombination during sexual reproduction. Although this feature can be seen equally as an advantage or an opportunity of mutation and genome-editing technologies, ultimately, as in conventional breeding, genome editing is followed by extensive field evaluations in target environments to select and deliver to farmers only those crops that are superior to current varieties, considering all agronomic and consumer criteria.
Inadequate stewardship
Today, most genome-edited plants are produced using an intermediate step that involves insertion of foreign DNA sequences that are removed in subsequent steps, so that the final genome-edited plants are not transgenic. However, use of an intermediate transgenic step presents technical risks that require appropriate stewardship, both in the lab and greenhouse. Molecular tools can then be used to demonstrate that the transgenic intermediate has been resolved before field trials, with country-appropriate stewardship for edited crops. As genome-editing technologies evolve, the use of an intermediate transgenic step may become unnecessary, further simplifying genome-edited plant development and greatly enhancing its utility for editing clonally propagated crops — for which removal of the intermediate transgenic elements is challenging.
Enhanced inequity between rich and poor
Another risk can arise if advanced technologies disproportionately benefit wealthy players, including multinational corporations and large-scale farmers, or disadvantage smallholders or farmers engaged in alternative agricultural systems such as organic agriculture4,23. One response to mitigate this risk may be misuse of regulatory processes, such as differential labeling, that work to stigmatize and inhibit adoption by food companies and discourage consumption. Some organizations have already sought to define products developed using genome-editing technologies as transgenic, which could lead to unwarranted avoidance of genome-edited crops by food and ingredient companies, smallholders and trade-dependent developing countries. Labeling rules should be framed in a harmonized global system that is based on transparent science-based consideration of risks, in which new traits in food would be included as a label if they introduce new allergens or toxins or fundamentally change the composition of the food; production method should not be part of mandatory labeling requirements. We propose that the most effective approach to mitigate this risk is to ensure that genome-editing technology remains accessible to those who will use it to democratize its benefits, particularly for resource-poor farmers and consumers in LMICs.
Lack of transparency
Lack of transparency regarding the products of genome-editing technologies would create a ‘social license risk’ by fuelling a lack of trust in product developers, regulators, producers and ultimately in the resulting genome-edited products4. By ‘social license for a new technology’, we refer to the willingness of potential users and consumers, and society at large, to accept products developed using that technology. Although social license is influenced by governmental policies, including local regulatory frameworks, global regulatory harmonization, trade and product-labeling requirements, and by public perceptions of risks and benefits, it is ultimately granted by the public locally and globally. One mechanism for transparency is an easily accessible registry through which developers of genome-edited crops can disclose the use of genome-editing technologies and meet public interest in knowledge about how specific foods are produced. Such registries should remain separate from the patent and regulatory risk-assessment systems. One such registry, developed by The Center for Food Integrity through their Coalition for Responsible Gene Editing in Agriculture (https://foodintegrity.org/programs/gene-editing-agriculture/), uses a consumer-focused approach that is designed to address transparency concerns, and incorporates needs of the concerned public and civil society through consumer and related engagements.
Unclear intellectual property landscape
Another risk for genome-editing technologies is the intellectual property landscape, which has evolved around the foundational patent dispute between the Broad Institute (Harvard–Massachusetts Institute of Technology) and Berkeley groups24,25. Although this dispute is not fully resolved, those who control the foundational intellectual property for agricultural applications have signaled their willingness to license their technologies to public institutions and companies to develop and commercialize genome-edited products26. This has been demonstrated through the granting of licenses to multiple CGIAR centres and others27 that are working on crops for smallholders. The availability of CRISPR alternatives to CRISPR-associated protein 9 (Cas9), some of which may have independent patent estates, could facilitate the development of genome-edited crops. However, lack of resolution about the ultimate ownership generates long-term uncertainty for the products developed using the technology. Clear legal rulings are needed to guide plant breeders, especially those in the resource-constrained public sector, who otherwise may avoid or delay using genome-editing technologies until these intellectual property uncertainties are resolved28.
Although the business model for commercialization of crop varieties developed using genome editing within CGIAR is not yet fully developed, CGIAR implements various models to advance new varieties through national agricultural research programs, and local and global seed companies that serve smallholder farmers. For example, maize hybrids are licensed to seed companies, which then compete naturally in the marketplace. Various models would be explored to maximize the value of edited products to smallholder farmers and their communities, ensuring access.
Inadequate public sector institutional infrastructures to support use of genome-editing technologies
While the innovation infrastructure for genome-edited crops and other biotechnologies within institutional frameworks in LMICs varies significantly, there are projects, institutes, strategic alliances and explicit biotechnology and bioeconomy policies that can help overcome some of the environmental factors that limit widespread use of genome-edited crops. Examples from Africa include the African Orphan Crops Consortium (http://africanorphancrops.org) and the African Agricultural Technology Foundation (https://www.aatf-africa.org). In Latin America, the Inter-American Institute for Cooperation on Agriculture (IICA) proposed the Hemispheric Program for the Bioeconomy and Productive Development29, and existing networks such as BIOTECSUR and MAIZALL30 provide models to help develop effective partnerships that support deployment of genome-edited crops. Public sector institutions are already engaged in genome-editing research to develop improved crop varieties5 (Fig. 1). Pertinent global science, policy and trade communities, including G20, should seize this opportunity to support the development of the necessary institutional capabilities.
