AI helps plants track down mutant bacteria

A group of Californian researchers has used artificial intelligence to design improved versions of a receptor, expanding its ability to recognize threatening bacteria. This has updated the plants’ defense system.

by Matteo Cavallito

Like animals, including humans, plants also have an immune system. This includes specific receptors that detect the presence of hostile bacteria. The problem, however, is that these microorganisms have developed the ability to change their appearance, thus evading detection. So how can we help plants defend themselves against an ever-changing threat? According to some American researchers, the answer may lie in artificial intelligence.

Bacteria change shape to deceive plants

This hypothesis was put forward in a study by the University of California Davis, recently published in the journal Nature Plants. The authors used AI to help plants recognize a wider range of bacterial threats. They explain that this result could pave the way for the development of new protection strategies for staple crops that are often affected by pathogens.

“Bacteria are in an arms race with their plant hosts, and they can change the underlying amino acids in flagellin to evade detection,” explained Gitta Coaker, lead author of the study and professor in the Department of Plant Pathology at UC Davis.

Flagellin is a protein found in the “tails” that bacteria use to move. The researchers’ attention therefore focused on a specific receptor, called FLS2, which helps plants recognize it. Unless, as mentioned, it changes its appearance to confuse the plants. And this is exactly where algorithms come into play.

AI provides an upgrade to defense systems

Specifically, the authors used a tool called AlphaFold, which can predict the three-dimensional shape of proteins. This allowed them to engineer the FLS2 receptor and enhance it to identify a greater number of intruders. In other words, they designed improved versions of the receptor, thereby upgrading the plants’ defense system. The group focused on receptors that were already known but not present in the main cultivated species.

By comparing them with more specific receptors, the researchers were able to identify the amino acids that needed to be modified in order to recognize a wide variety of bacteria.

“The surface-localized receptor kinase FLS2 detects the flg22 epitope (or the part of an antigen that is recognized by the immune system, ed.) from bacterial flagella,” the study explains. “Using diversity analyses, AlphaFold modelling and amino acid properties, key residues enabling expanded recognition were mapped to FLS2’s concave surface, interacting with the co-receptor and polymorphic flg22 residues.” The data collected then made it possible to modify the receptor, expanding its ability to recognize bacteria.

New prospects against plant pathogens

According to the authors, the results demonstrate the theoretical ability of prediction-based design to improve broad-spectrum resistance to pathogens. Among these, they explain, is Ralstonia solanacearum, the bacterium responsible for plant wilting. Some strains, they explain, can infect over 200 plant species, including key crops such as tomatoes and potatoes.

In the future, the researchers want to develop machine learning tools to predict which immune receptors need to be modified. They also aim to reduce the number of amino acids that need to be altered to achieve maximum results. “Narrowing specificity to a few amino acid residues is compatible with genome editing technologies (allowing accurate modification of DNA in organisms, ed.) that will expedite the deployment of the engineered resistance genes,” the research concludes.