A Computation Method Based on the Combination of Chlorophyll Fluorescence Parameters to Improve the Discrimination of Visually Similar Phenotypes Induced by Bacterial Virulence Factors.

Phenotyping biotic stresses in plant-pathogen interactions studies is often hindered by phenotypes that can hardly be discriminated by visual assessment. Particularly, single gene mutants in virulence factors could lack visible phenotypes. Chlorophyll fluorescence (CF) imaging is a valuable tool to monitor plant-pathogen interactions. However, while numerous CF parameters can be measured, studies on plant-pathogen interactions often focus on a restricted number of parameters. It could result in limited abilities to discriminate visually similar phenotypes. In this study, we assess the ability of the combination of multiple CF parameters to improve the discrimination of such phenotypes. Such an approach could be of interest for screening and discriminating the impact of bacterial virulence factors without prior knowledge. A computation method was developed, based on the combination of multiple CF parameters, without any parameter selection. It involves histogram Bhattacharyya distance calculations and hierarchical clustering, with a normalization approach to take into account the inter-leaves and intra-phenotypes heterogeneities. To assess the efficiency of the method, two datasets were analyzed the same way. The first dataset featured single gene mutants of a strain which differed only by their abilities to secrete bacterial virulence proteins. This dataset displayed expected phenotypes at 6 days post-inoculation and was used as ground truth dataset to setup the method. The efficiency of the computation method was demonstrated by the relevant discrimination of phenotypes at 3 days post-inoculation. A second dataset was composed of transient expression (agrotransformation) of Type 3 Effectors. This second dataset displayed phenotypes that cannot be discriminated by visual assessment and no prior knowledge can be made on the respective impact of each Type 3 Effectors on leaf tissues. Using the computation method resulted in clustering the leaf samples according to the Type 3 Effectors, thereby demonstrating an improvement of the discrimination of the visually similar phenotypes. The relevant discrimination of visually similar phenotypes induced by bacterial strains differing only by one virulence factor illustrated the importance of using a combination of CF parameters to monitor plant-pathogen interactions. It opens a perspective for the identification of specific signatures of biotic stresses.