The receptors in question belong to the nucleotide-binding leucine-rich repeat family, the workhorse intracellular immune sensors of plants. NLR proteins carry a modular architecture: a variable N-terminal signalling domain, a central nucleotide-binding domain that acts as a molecular switch, and a C-terminal leucine-rich repeat domain that serves as the recognition module. In the resting state, the receptor is autoinhibited, folded upon itself. When the leucine-rich repeat domain perceives, directly or indirectly, the presence or activity of a pathogen effector protein, the switch flips, the receptor oligomerises into a resistosome, and a cascade of defensive responses follows, often culminating in a localised cell death that walls off the invader.

The trouble with engineering these receptors has always been the recognition surface. Leucine-rich repeats are exquisitely tuned by evolution to particular effector proteins, and their binding interfaces are diffuse, conformationally dynamic and poorly tolerant of casual mutation. Attempts to retune an existing NLR to recognise a new effector have met with limited success, because the coupling between recognition and activation is idiosyncratic to each receptor. Natural diversity, meanwhile, is finite. When a pathogen delivers an effector for which no plant in the gene pool carries a matching sensor, breeders have no card to play. This is the gap that synthetic receptors are designed to close.

The new approach sidesteps the problem of retuning natural recognition surfaces by building the sensing module afresh. Advances in de novo protein design, accelerated by deep-learning structure prediction tools such as AlphaFold and by generative design methods demonstrated over the past several years, allow researchers to specify a protein binder with a chosen shape and interaction profile and then compute an amino acid sequence that should fold into it. Rather than asking an existing NLR to learn a new partner, the strategy designs a synthetic binder against a defined pathogen effector and wires that binder into an immune receptor scaffold, so that effector binding triggers the conserved activation machinery of the NLR.

Wiring is the crux. A binder that merely sticks to an effector accomplishes nothing unless binding is mechanically transduced into receptor activation. The engineering challenge is therefore one of allostery: the designed recognition domain must be coupled to the nucleotide-binding switch such that the receptor sits stably in its off state in the absence of pathogen, yet flips decisively when the effector is present. Getting this balance wrong in either direction is costly. A receptor that fires without its ligand would impose autoimmune damage and yield penalties, while one that fails to respond leaves the plant defenceless. The design-build-test cycle must therefore evaluate both specificity and the absence of spontaneous activation.

This is where the second half of the strategy, evolution in vivo, earns its place. Computational design produces starting points, not finished products; predicted interfaces routinely require empirical optimisation. Directed evolution, in which libraries of receptor variants are generated and screened for the desired activity, has already proven its worth in plant immunity. Previous work showed that natural NLRs can be turned into robust sensors of specific effectors through targeted mutagenesis of their recognition domains, and that such engineered receptors can be deployed in crops. The synthetic pipeline extends this logic: rather than evolving a natural receptor toward a new specificity, researchers evolve a designed receptor toward reliable function, screening variants in planta for strong responses to the target effector and no response to plant proteins.

The significance of combining these steps is that each compensates for the other’s weakness. De novo design offers unlimited starting diversity, unconstrained by what nature has provided, but its predictions are imperfect. Directed evolution offers empirical optimisation with exquisite sensitivity to function, but it needs a plausible starting scaffold to improve upon. Designed receptors give evolution a head start; evolution gives design its final polish. The result, as the Nature Plants commentary describes it, is on-demand design and refinement of resistance genes: a workflow in which a pathogen effector of agricultural concern can be identified, a synthetic receptor raised against it in silico, and functional variants selected in plants within a timeframe that conventional breeding cannot approach.

The broader context makes the timing notable. Structure prediction has matured to the point that protein complexes and designed binders can be modelled with remarkable fidelity, and generative protein design has produced functional enzymes, sensors and therapeutics in other domains. Plant immunology, meanwhile, has accumulated a detailed mechanistic understanding of how NLRs activate, including structural snapshots of resistosomes that reveal how effector perception is coupled to signalling. The convergence of these threads is what makes synthetic immune receptors plausible now, when the same attempt even five years ago would have lacked both the design tools and the mechanistic template.

There are, of course, substantial hurdles between a working prototype and a deployed resistance gene. Pathogens evolve; an effector targeted by a synthetic receptor can mutate its surface, and durable resistance may require stacking multiple sensors or targeting effectors whose function the pathogen cannot easily relinquish. Regulatory frameworks for crops carrying designed genes, and public acceptance of them, remain open questions. There is also the matter of scale: each new effector target demands its own design and evolution campaign, and the reliability of the pipeline across diverse effector families has yet to be demonstrated broadly. The commentary’s authors, writing from a laboratory with deep roots in plant immunity research, frame the advance as the opening of a route rather than its completion.

Even so, the conceptual shift is hard to overstate. Plant immunity has until now been a scavenging enterprise, drawing on variation that pathogens and plants co-authored over millions of years. If synthetic receptors can be designed against arbitrary pathogen proteins and matured by directed evolution in planta, the defensive repertoire of crops becomes a matter of engineering choice rather than evolutionary inheritance. For a world in which plant diseases claim a substantial share of harvests and climate change keeps redistributing pathogen pressure, the ability to write a resistance gene to order, and to have it refined in the very organisms it will protect, marks a genuine inflection point in how humanity arms its crops.

Subject of Research: Engineering synthetic plant NLR immune receptors through de novo protein design and directed evolution in planta

Article Title: Plant immune receptors built in silico and evolved in vivo

Article References: Jones, J. D. G., & Zhao, H. (2026). Plant immune receptors built in silico and evolved in vivo. Nature Plants. https://doi.org/10.1038/s41477-026-02398-2

Image Credits: AI Generated

DOI: 10.1038/s41477-026-02398-2

Keywords: plant immunity, NLR receptors, de novo protein design, directed evolution, synthetic biology, crop disease resistance, pathogen effectors, protein engineering, plant biotechnology, resistance genes, AlphaFold, resistosome