Gene drives in plants are moving from imagination to reality, with the first proof-of-concept achieved. What comes next?

1. What gene drives have achieved so far

A farmer walking through a maize field, pulling out weeds row after row, might not imagine that the same genetic tools designed to reshape insect populations could one day be applied to plants. That step has not yet been taken. To understand why it remains so distant, it is useful to look at where gene drives already show promise.

In insects, laboratory demonstrations have shown that CRISPR–Cas9-based gene drives can bias inheritance in fruit flies, malaria-transmitting mosquitoes, and even mice, either suppressing populations or modifying them to make control measures more effective (Bier 2022; Tek et al. 2022; Kuman et al. 2023). The type of drive employed influences the goal: suppression of a pest population or modification of its traits (Kuman et al. 2023). None of these systems has yet progressed to field trials, but the experimental results suggest that such outcomes could be achievable under controlled conditions.

These engineered systems echo processes that nature has long carried out, for example selfish genetic elements such as transposons, homing endonucleases, meiotic drivers and toxin–antidote systems all distort inheritance away from Mendelian ratios (Bier 2022; Oberhofer et al. 2024). Fungi and yeast have provided well-established models to study such dynamics, while in plants a striking case has been described in rice, where a hybrid sterility locus biases the transmission of gametes between subspecies (Wang et al. 2023).

Despite rapid advances in molecular tools and the diversity of natural examples across the tree of life, plants have so far remained on the margins of synthetic gene drive research. As Barrett et al. (2019) observed, their potential applications have rarely been examined in detail, and until very recently, no experimental demonstration existed.

This is the current point of departure: a powerful genetic concept, tested in several kingdoms of life, but only just beginning to find its place in plants.

2. Why plants are different

The attraction is obvious: if gene drives could be made to work in plants, they could transform breeding for example, to restore herbicide sensitivity, and perhaps even suppress weeds (Figure 1). Yet the step from promise to practice is not a small one. Plants present a set of obstacles that differ fundamentally from those seen in insects or fungi.

The first hurdle lies in biology itself. For a drive to succeed, processes at the molecular, cellular, organismal and population levels must all align. Plants complicate this picture with their remarkable diversity. They range from short-lived annuals to millennia-old trees, and their reproductive modes span from strict self-fertilisation to obligate outcrossing, with hermaphroditism, monoecy and dioecy in between (Barrett et al. 2019; Hay et al. 2025). Some species propagate clonally, others rely on cross-pollination by wind, water or animals. Because gene drives depend on sexual recombination, their spread is slowed, or even halted, by selfing or asexual reproduction. Modelling confirms that even relatively low levels of self-fertilisation can prevent a drive from spreading (Barrett et al. 2019; Croghan et al. 2023).

Seedbanks add another complication. Seeds can remain dormant in the soil for years or decades, creating a buffer against rapid genetic change. This means that in many species, drives would advance slowly, rather than sweep quickly through populations (Kim et al. 2025; Neve 2018). Large population sizes and limited pollen dispersal reduce the pace of spread even further, while long generation times in perennials could delay outcomes for centuries (Neve 2018; Croghan et al. 2023). Even in obligate outcrossers, there is the possibility that populations might evolve increased selfing as a form of resistance (Neve 2018). Genetic architecture poses an additional challenge. Many plants are polyploid, carrying multiple sets of chromosomes. This increases genome size, multiplies potential target sites and requires all copies of a gene to be modified before a trait is expressed (Barrett et al. 2019; Hay et al. 2025; Oberhofer et al. 2024). Identifying conserved guide RNA (gRNA) sites is therefore difficult and the dosage requirements for toxin–antidote systems are uncertain. Plants also show peculiarities in gametophyte biology, where haploid expression can create sex-specific biases in transmission (Oberhofer et al. 2024). Engineering drives to operate reliably within such complex systems is far from straightforward.

Technical barriers are equally significant. Homing drives, so efficient in animals, encounter strong resistance in plants because DNA repair favours end-joining over homology-directed repair, producing resistant alleles instead of spreading the drive (Liu et al. 2024a). In vitro transformation is another bottleneck: many species, particularly weeds, lack efficient tissue culture or regeneration protocols (Kuman et al. 2023). Even when transformation succeeds, transgenes may be silenced before the drive can act (Liu et al. 2024b).

Ecological dynamics complete the picture. Traits designed for human benefit (such as restored herbicide sensitivity or reduced fertility) often impose fitness costs on the plants themselves, making them unlikely to persist under natural selection (Liu et al. 2024b; Neve and Barrett 2024). At the same time, the scale of many plant populations makes it logistically demanding to release enough engineered individuals. Dormant seeds and long life cycles further extend timelines, while hybridisation across related species creates additional risks of spread beyond the intended target (Hay et al. 2025).

3. What we know so far in plants

In the last couple of years, plant gene drive research has shifted from theoretical discussions to tangible advances. The field now spans natural systems discovered in rice, synthetic proof-of-concept drives in Arabidopsis and modelling studies that test how such systems might behave in real populations.

3.1. When nature does the driving

Rice provides one of the clearest demonstrations that drive-like systems can exist in plants. Wang et al. (2023) identified the locus RHS12 which carries two interacting genes: iORF3 (DUYAO) and iORF4 (JIEYAO). DUYAO encodes a mitochondrion-targeted toxin that kills developing pollen, while JIEYAO encodes an antidote that redirects DUYAO for degradation, rescuing the gametes that carry it. This interaction gives RHS12 a transmission advantage (Figure 2).

The impact is not just theoretical. The frequency of functional DUYAO–JIEYAO alleles rose sharply in japonica rice: absent in landraces before the 1970s, they were found in 86% of registered varieties by 2019. This suggests that the system has acted as a natural drive during modern rice breeding. The toxin–antidote pair also functions across kingdoms, as JIEYAO was shown to neutralise DUYAO toxicity in both yeast and Drosophila cells (Wang et al. 2023).

Unlike other reported gamete-killers, DUYAO–JIEYAO can also affect somatic cells, making it a distinctive type of drive. Beyond its biased inheritance, it plays a role in reproductive isolation between indica and japonica rice subspecies, helping to maintain species boundaries (Awan et al. 2024; Wang et al. 2023).

3.2. First experiments in Arabidopsis

Arabidopsis thaliana has recently become the first plant where engineered gene drives have been demonstrated. Two independent groups reported their results in the same issue of Nature Plants, marking a major milestone for the field.

Liu et al. (2024a) developed CAIN (CRISPR-Assisted Inheritance utilising NPG1), a toxin–antidote system. In CAIN, Cas9 and gRNAs disrupt the essential No Pollen Germination 1 (NPG1) gene, which normally allows pollen to germinate. Pollen carrying the drive is rescued by a recoded, drive-resistant version of NPG1, while wild-type pollen fails. This biased inheritance produced transmission rates of 88–99% over two generations, with minimal formation of resistance alleles. Modelling suggested that CAIN could, in principle, spread from just 1% of individuals to fixation in about 17 outcrossed generations. The system was tested using a fluorescent marker as cargo and demonstrated its effectiveness even in the self-pollinating background of Arabidopsis (Liu 2024a; Liu 2024b).

A second group engineered a Cleave-and-Rescue (ClvR) drive in Arabidopsis. Oberhofer et al. (2024) targeted the essential gene YKT61. Cas9 and gRNAs disrupted the native gene, while a recoded version provided rescue. Gametes that lacked ClvR were nonviable, resulting in clear non-Mendelian inheritance. This strategy, adapted from earlier Drosophila work, demonstrates that toxin–antidote gene drives can also be built in plants.

Commentaries by Neve and Barrett (2024) and Hay et al. (2025) compared the two approaches. Both CAIN and ClvR are notable for avoiding reliance on homology-directed repair, a mechanism inefficient in plants and a frequent obstacle for homing drives (Figure 3). While the proof-of-concept drives only carried fluorescent markers, modelling suggested that, depending on mating system and fitness costs, such systems could theoretically spread fitness-reducing traits through wild populations in 10–30 generations.

3.3. What models reveal when seedbanks are involved

Gene drive modelling has a long tradition across insects, fungi and other organisms. What is new in plants is the attempt to adapt these frameworks to their distinctive ecology. The Arabidopsis proof-of-concept studies by Liu et al. (2024a) and Oberhofer et al. (2024) included simulations showing that CAIN and ClvR could spread efficiently under a simplified assumption: fertilisation is panmictic, where outcrossing populations have no ecological constraints. These models provided an initial view of how engineered drives might behave in plants but did not capture the demographic realities of wild species.

Kim et al. (2025) advanced this work by explicitly incorporating seedbanks into the same modelling frameworks. This was a crucial addition, as dormant seeds are a defining feature of many weeds and wild plants but were not considered in the earlier studies. Their models also accounted for germination probabilities and fecundity, producing results that differed from the earlier projections. Long-lived seedbanks were shown to slow the spread of modification drives and increase the threshold required for suppression drives to establish. By contrast, reduced fecundity and lower germination rates could favour suppression, since fewer individuals are needed to reduce population growth.

Another key insight was that self-fertilisation can evolve as a resistance mechanism, impeding drive spread, whereas dioecious species, where male and female reproductive organs are not present together in the same individual, are more promising targets. Among the different strategies tested, ClvR male-suppression consistently outperformed both ClvR female-suppression and CAIN male-suppression. This contrasted with the more optimistic predictions of Liu (2024a) and Oberhofer (2024), underlining how seedbanks and other plant-specific traits reshape expectations.

For a broader discussion of modelling in gene drive research, see the article “From lab to landscape: How models predict gene drive spread”.

3.4. Testing ideas on paper: The case of common tansy

Beyond proof-of-concept experiments and modelling, Croghan et al. (2023) offered a different type of contribution: a theoretical risk assessment centred on common tansy (Tanacetum vulgare), an invasive species in North America. Tansy is a perennial, self-incompatible plant that can produce up to 50,000 seeds per individual. It spreads mainly through seed rather than rhizomes and it is toxic to livestock as well as a host for plant viruses. These traits make it both a problematic invader and a potentially interesting candidate for a drive-based suppression approach.

Croghan and colleagues examined how a drive that reduced or eliminated seed production might work in tansy, since this would directly target the plant’s main mode of reproduction. They noted that transformation could be feasible, given that related species can be regenerated in vitro and transformed using Agrobacterium methods. Yet several unknowns remain critical for risk assessment: the persistence of tansy seedbanks, the possibility of resistance evolution and the long generation times associated with perennials.

One of the main contributions of the study lies in its framework for assessing candidate species. The authors compiled a set of biological characteristics, such as mating system, generation time, dispersal mode and seedbank longevity, that determine whether a plant species is a strong or weak candidate for gene drive development. They then evaluated common tansy against these traits (Figure 4). This exercise not only highlights the potential of tansy as a model but also illustrates how regulators and risk assessors might identify endpoints for evaluating gene drive plants more broadly.

For a complementary discussion of risk assessment for gene drive organisms, see the article “Before the first release, how is problem formulation shaping gene drive risk assessments?”.

4. How plant drives might be used

The studies reviewed so far suggest that gene drives in plants are technically feasible, at least under controlled conditions and in a model species. The natural next question is where such tools might find application if they were ever to progress beyond proof-of-concept. Researchers have started to outline potential niches or applications.

4.1. Weed management

Weed control is the most frequently discussed application of plant gene drives. Two main strategies have been proposed: population suppression, which directly reduces weed fertility or seed production, and population sensitisation, which restores susceptibility to herbicides or other control measures (Barrett et al. 2019; Neve, 2018; Kuman et al. 2023).

The appeal is clear. Herbicide resistance is spreading rapidly: more than 200 resistant species have been documented worldwide, posing a growing challenge for agriculture and increasing production costs (Kuman et al. 2023; Liu et al. 2024b). Sensitisation drives could be designed to target specific mutations that confer resistance, effectively resetting weeds to their herbicide-sensitive form. For target-site resistance, this might be as simple as correcting a single nucleotide change. More complex non-target-site resistance could require multiplexed approaches with several gRNAs (Kuman et al. 2023).

Among the most prominent candidates are dioecious weeds such as Amaranthus palmeri and A. tuberculatus. Their reproductive biology (separate male and female individuals, short annual life cycles and identified sex-determining regions) makes them more amenable to drive-based strategies (Oberhofer et al. 2024; Kuman et al. 2023). Potential applications include biasing sex ratios, disrupting reproductive genes, or reversing herbicide resistance (Tek et al. 2022; Croghan et al. 2023).

Yet the challenges remain considerable. Amaranthus species are prolific seed producers and readily hybridise with related species, raising risks of unintended spread. They also hold cultural significance in their native range, which complicates the social acceptability of interventions (Oberhofer et al. 2024). Seedbanks add another layer of difficulty. Large reserves of dormant seed can slow the spread of a drive, but they may also serve as a natural safeguard by increasing the threshold needed for establishment after an accidental release (Kim et al. 2025).

4.2. Faster breeding

Beyond weed management, plant gene drives are being considered as tools to accelerate breeding and enhance crop resilience. Natural systems such as DUYAO–JIEYAO in rice show how hybrid sterility loci could be harnessed to overcome barriers between subspecies, unlocking new opportunities for heterosis and for the introgression of useful genes from wild relatives into elite cultivars (Wang 2023; Awan 2024). This could broaden the genetic base of crops and improve access to traits that are otherwise difficult to incorporate.

Engineered drives may also shorten the time required to develop resistant parental lines. For instance, targeting susceptibility genes can create resistant plants, but conventional breeding requires years of repeated backcrossing to introduce these traits into hybrid lines. A gene drive could speed this process by copying the resistant allele automatically into new parental lines, removing the need for repetitive transformations or long breeding cycles (Tek et al. 2022) (Figure 5).

The CAIN system in Arabidopsis illustrates the principle that gene drives can be used to modify populations by embedding resistance-restoring cassettes. In agriculture, such an approach could, in theory, re-establish herbicide sensitivity while also carrying additional traits. More broadly, gene drives could be envisioned for adding stress tolerance traits such as drought or disease resistance, though the persistence of these traits in wild populations would be limited by fitness costs (Liu et al. 2024a; Liu et al. 2024b; Hay et al. 2025).

Agricultural systems may also provide practical advantages for early testing. Managed fields make it easier to control population densities, apply selective pressures and monitor outcomes, at least compared to natural ecosystems (Legros et al. 2021). This controlled setting could make agriculture a more feasible arena for piloting gene drive strategies, though it does not resolve the regulatory or societal challenges that accompany them.

4.3. Conservation

Potential conservation applications extend the scope of plant drives beyond agriculture. Barrett et al. (2019) highlighted their possible use in genetic rescue, where drives could spread beneficial alleles through small or declining populations. In theory, this might help mitigate inbreeding depression or accelerate adaptation to rapid environmental change. The main barrier lies in identifying single genes of major effect that could serve as reliable targets.

Gene drives are also discussed as potential tools for managing invasive species. Suppressing invasive alien plants, or restoring herbicide sensitivity, could reduce their ecological footprint and help native communities recover (Neve and Barret 2024; Oberhofer et al. 2024; Croghan et al. 2023). Here too, ecological risks are central: traits that benefit human management often reduce plant fitness, meaning that they may be quickly eliminated by natural selection once released (Liu et al. 2024b).

The CAIN system provides an example of how a single framework could, at least conceptually, address different ecological and conservation challenges. In endangered species, it could carry alleles that improve environmental resilience, and in invasive plants, it could target fertility genes to suppress populations (Liu et al. 2024b) (Figure 6).

5. Rules being discussed for animals, not plants

So far, most governance discussions of gene drives have been shaped by animal research (Neve, 2018; Neve and Barrett, 2024). Traits such as mating system, dispersal and demographic structure strongly influence drive dynamics, and in plants, these vary across a wider spectrum than in animals. Seed dormancy, polyploidy and selfing rates all affect how quickly or reliably a drive might spread, and such parameters remain poorly understood when applied to gene drives (Barrett et al. 2019; Kim et al. 2025; Croghan et al. 2023). Rather than obstacles, these features highlight the importance of integrating ecological and demographic traits into assessments for all drive organisms, not just plants.

Existing biosafety frameworks, such as those that regulate living modified organisms, already provide a foundation for evaluating gene drives. The key will be adapting these case by case, ensuring that assessments capture both molecular mechanisms and ecological context, rather than creating entirely new systems (Neve, 2018; Neve and Barrett, 2024).

Safeguards and countermeasures are another dimension of governance. Proposals include the design of resistant alleles, reversal drives or anti-Cas9 transgenes that could neutralise an unwanted release (Oberhofer et al. 2024; Liu et al. 2024b). These are part of a shared toolbox across all gene drive organisms and can strengthen stewardship. A detailed review of these strategies is available in the article “How to stop a gene drive".

Dual-use considerations also remain relevant. Oberhofer et al. (2024) note that the clandestine introduction of drives into crops is improbable, given the controls around seed production and the detectability of engineered alleles. Still, Liu et al. (2024b) caution that vulnerabilities could arise where seeds are farmer-saved or uncertified, particularly in regions without strong monitoring systems. Continuous oversight will therefore be an important part of responsible governance.

As Hay et al. (2025) suggest, governance should evolve in tandem with science, combining precaution with the flexibility to allow innovation. In this sense, experiences from plants can help illuminate how governance of gene drives as a whole might proceed: not through separate rules, but through frameworks that are broad, adaptive and informed by ecological realism.

6. What future for plant drives

Plant gene drives remain more of a prospect than a practice. Despite decades of investment in genome engineering for crop improvement, little explicit effort has gone into developing genetic tools for wild plant populations. Only recently, with proof-of-concept studies in Arabidopsis and the rediscovery of natural elements in rice, has the idea shifted from speculation to demonstration.

These early steps show that the concept is more than imagination. CAIN proves that a toxin–antidote drive can function in a model plant and the conservation of its target gene across crops suggests it could be adapted more widely. DUYAO–JIEYAO demonstrates that natural drive elements have already influenced modern rice breeding, reminding us that nature itself sometimes provides the proof before technology follows. The challenge is to take these examples beyond the laboratory and into the diverse, complex realities of polyploid genomes, long-lived perennials and heterogeneous wild populations.

The possible applications are broad: weed suppression, breeding acceleration, resilience to climate stress and even conservation of threatened species. Could plant gene drives one day be designed with the same precision and reliability that breeders now expect from conventional tools? If so, they might help manage resistance, extend genetic diversity and support sustainable agriculture in ways that no other method can.

Any potential future use, however, would not be decided by scientists alone. Farmers, seed producers and rural communities would be directly affected, whether through new management options or through changes in local ecosystems. Their voices will shape not only whether plant gene drives are acceptable, but also how they are designed and deployed. Early engagement with those who stand to benefit or bear the risks will be essential if the technology is to move forward responsibly. For a broader discussion on community engagement in gene drive research, see “How community voices shape the future of gene drive research".

For now, plant gene drives remain a field in waiting, but a field where the outlines of a future are just starting to take shape (Figure 7). Imagine that same farmer walking through the maize rows: today still pulling weeds by hand, but tomorrow perhaps working alongside tools that harness the very rules of inheritance. Whether that vision comes to fruition will depend not only on how science, ecology and governance advance, but also on how farmers and communities choose to be part of that journey.

7. References

The studies cited in this version of the living literature review are visually summarised in Figure 8, which maps their citation links, publication dates and thematic focus.

Ethan Bier (2022). Gene drives gaining speed. https://doi.org/10.1038/s41576-021-00386-0

Muhammad Jawad Akbar Awan, Rubab Zahra Naqvi, Imran Amin, Shahid Mansoor (2024). Highlight natural selfish genetic elements in rice (the one in Wang 2023) as inspiration for plant-adapted gene drive systems. https://doi.org/10.1016/j.tplants.2023.10.009

Luke G. Barrett, Mathieu Legros, Nagalingam Kumaran, Donna Glassop, S. Raghu, Donald M. Gardiner (2019). Gene drives in plants: opportunities and challenges for weed control and engineered resilience. https://doi.org/10.1098/rspb.2019.1515

Lori Croghan, Alan G. Smith, Matthew A. Tancos, Neil O. Anderson, Roger L. Becker (2023). Benefits and risks of gene drives for invasive plant management - the case for common tansy. https://doi.org/10.3389/fagro.2023.1290781

Bruce A. Hay (2025). Applications and status of gene drive in plants. https://doi.org/10.1371/journal.pbio.3003148

Isabel K. Kim, Leqi Tian, Ryan Chaffee, Benjamin C. Haller, Jackson Champer, Philipp W. Messer, Jaehee Kim (2025). Gene drive dynamics in plants: the role of seedbanks. https://doi.org/10.1101/2025.04.24.649389

Yaiphabi Kuman, Harold N Trick, P.V. Vara Prasad, Mithila Jugulam (2023). Transformative approaches for sustainable weed management: The power of gene drive and CRISPR-Cas9. https://doi.org/10.3390/genes14122176

Mathieu Legros, John M. Marshall, Sarina Macfadyen, Keith R. Hayes, Andy Sheppard, Luke G. Barrett (2021). Gene drive strategies of pest control in agricultural systems: Challenges and opportunities. https://doi.org/10.1111/eva.13285

Yang Liu, Bingke Jiao, Jackson Champer, Wenfeng Qian (2024a). Overriding Mendelian inheritance in Arabidopsis with a CRISPR toxin–antidote gene drive that impairs pollen germination. https://doi.org/10.1038/s41477-024-01692-1

Yawen Liu, Bingke Jiao, Wenfeng Qian (2024b). Synthetic gene drives in plants: Development strategies, potential applications, and ethical considerations. https://doi.org/10.1002/moda.30

Paul Neve (2018). Gene drive systems: do they have a place in agricultural weed management? https://doi.org/10.1002/ps.5137

Paul Neve, Luke Barrett (2024) Driving evolution in wild plants. https://doi.org/10.1038/s41477-024-01723-x

Georg Oberhofer, Michelle L. Johnson , Tobin Ivy , Igor Antoshechkin, Bruce A. Hay (2024). Cleave and Rescue gamete killers create conditions for gene drive in plants. https://doi.org/10.1038/s41477-024-01701-3

Mumin Ibrahim Tek, Kubra Budak (2022). A new approach to develop resistant cultivars against the plant pathogens: CRISPR drives. https://doi.org/10.3389/fpls.2022.889497

Chaolong Wang, Jian Wang, Jiayu Lu, Yehui Xiong, Zhigang Zhao, Xiaowen Yu, Xiaoming Zheng, Jing Li, Qibing Lin, Yulong Ren, Yang Hu, Xiaodong He, Chao Li, Yonglun Zeng, Rong Miao, Mali Guo, Bosen Zhang, Ying Zhu, Yunhui Zhang, Weijie Tang, Yunlong Wang, Benyuan Hao, Qiming Wang, Siqi Cheng, Xiaojuan He, Bowen Yao, Junwen Gao, Xufei Zhu, Hao Yu, Yong Wang, Yan Sun, Chunlei Zhou, Hui Dong, Xiaoding Ma, Xiuping Guo, Xi Liu, Yunlu Tian, Shijia Liu, Chunming Wang, Zhijun Cheng, Ling Jiang, Jiawu Zhou, Huishan Guo, Liwen Jiang, Dayun Tao, Jijie Chai, Wei Zhang, Haiyang Wang, Chuanyin Wu, Jianmin Wan (2023). A natural gene drive system confers reproductive isolation in rice. https://doi.org/10.1016/j.cell.2023.06.023