Modeling of Glucosinolate Biosynthesis During Biotic Stress as a Function of mRNA

Abstract

Glucosinolates are an important group of specialized metabolites in the Brassicaceae family, playing a role as defensive compounds against biotic attackers. In response to biotic stress, plants upregulate glucosinolate biosynthesis in part by increasing the abundance of enzymes in the glucosinolate biosynthetic pathway. As an increase in enzyme abundance is often preceded by an increase in the corresponding mRNA levels, the dynamic changes in mRNA levels should capture the information required to infer how metabolite levels change over time. In order to test this hypothesis, a time series of experimental glucosinolate content data collected from Arabidopsis thaliana, exposed to either a mock or methyl jasmonate (MeJA) treatment, as a proxy for biotic stress, was combined with existing mRNA abundance data over time at the same developmental stage and treatment. We propose the GEEM model, a multilevel mechanistic ordinary differential equation (ODE) model, which goes from Gene expression to an enzyme level model, followed by a Michaelis Menten kinetics metabolite model, to simulate the dynamics of a segment of the indolic glucosinolate pathway. In order to constrain the GEEM model, three models were fit to experimental de novo specialized metabolite data, using different degrees of freedom by utilizing both a Gradient Boosted Tree model with a tested architecture to predict the kinetic constants, and augmenting these predictions with a literature review of the known Michaelis Menten kinetic constants from the glucosinolate pathway. Using Sequential Monte Carlo - Approximate Bayesian Computing to fit the GEEM model to the experimental data, we showed that given the mRNA levels and initial concentrations of metabolites, the changes in specialized metabolites over time and treatment can be modeled. Author SummaryWe study how plants adjust their natural chemical defenses over time when they are under attack from living organisms. In the mustard family, including the subject of our experiment Arabidopsis, one important group of defense chemicals is called glucosinolates. When Arabidopsis is under attack, certain gene pathways can be activated or deactivated, allowing the plant to modulate the amount of enzymes they produce, which in turn modulates the levels of these defensive chemicals. In this work, we combine measurements of gene activity and glucosinolate levels from Arabidopsis treated with a compound used in stress signal that mimics insect or pathogen attack. We then constructed a mathematical model that goes from gene activity, to amount of enzyme present, and ends with the amounts of specific glucosinolates over time. By fitting this model to experimental data, we show that it is possible to predict how glucosinolate levels change over time from the gene activity and initial glucosinolate levels. Our approach offers a way to connect gene expression datasets to real changes in plant defense chemistry, with potential applications in plant breeding and insight into how these pathways change due to stress.