AI-driven protein design to understand effector function within the Magnaporthe oryzae rice pathosystem

Abstract

Magnaporthe oryzae is the causative agent of rice blast disease, threatening food security worldwide with economic losses estimated at $66 billion per annum. Like many phytopathogens, M. oryzae secretes a diverse repertoire of effectors that manipulate plant defences to promote infection. Using a structural phylogenetics approach, we describe the characterisation of a previously unstudied effector, Magnaporthe effector protein 26 (Mep26), and explore the utility of AI-based binder design to interrogate its function. To characterise the M. oryzae effector repertoire, we tooks a structural phylogenetics approach. AlphaFold2 was used to predict structures across the M. oryzae effector repetoire, and used to construct a structure-based phylogenetic tree using Foldtree. This analysis was complemented by identification of structural homologues within the Protein Data Bank (PDB) via Foldseek, as well as predicted subcellular localisation via EffectorP3. These combined analyses identified Mep26 to have a conserved fold suggestive of phytate-degrading enzymatic activity. Five additional structural orthologues were identified within the M. oryzae secretome, sharing low sequence identity but significant tertiary structure conservation, defining a putative Mep26 family. Using AI-guided de novo protein design, through software such as BindCraft, we generated candidate binders targeting the predicted active site of Mep26. In silico analysis was used to screen these binders for cross-reactivity across structural orthologues, aiming to identify designs capable of disrupting conserved catalytic mechanisms across the predictd Mep26-like family of effectors. De novo binders were designed to interact with Mep26, with some predicted to bind over the putatitve active site, possibly occluding ligand binding. By targeting the putative active site, the binders are expected to block substrate recognition and disrupt ligand turnover. The combined use of structural phylogenetics and de novo binder design provides a route to experimentally interrogate effectors that are challenging to study genetically. This proof-of-concept study demonstrates the potential of AI-designed protein binders as targeted molecular probes for effector functional dissection. More broadly, this strategy can be extended to redundant structural families and to other systems where genetic manipulation is limited. Our findings highlight how AI-driven protein design can be integrated into fungal molecular biology and deepen our understanding of effector-driven infection mechanisms. Planned future work focuses on enzymatic assays to assess whether Mep26 catalytic activity is present, and if so reduced in the presence of designed binders. Once binding has been demonstrated, tagged binder variants (e.g. localisation or degradation motifs) will allow in-cell functional validation and phenocopying of effector knockouts.