Meta-análise de redes metabólicas de plantas sob diferentes condições de estresse

Abstract

Due to the need to obtain genotypes with tolerance to abiotic stress, the metabolic response of plants under stress conditions has been widely studied. However, a holistic and integrative view of data to analyse proprieties at network scale and to distinguish components with greater robustness to survive adverse environment is scarce. In this context, the main objective of this study was to analyse the changes in topology and connectivity of metabolic networks of plants under stress conditions. Thereunto, we carried out a meta-analysis using eight previously published works that contain metabolomics data under different stress conditions such as water deficit, low temperature, salt excess, or nitrogen limitation. The data was obtained and analysed using the Cytoscape software, for the creation of correlation-based networks, and MetaboAnalyst, for multivariate analysis via PLS-DA (Partial Least Square Discrimante Analysis) and biomarker analysis. The data obtained of plants under stress were compared to plants in the absence of stress, in order to: (i) analyse if the appearance of hubs is a common response in metabolic networks of plants under stress; (ii) identify which metabolites have high degree of connection in metabolic networks of plants in absence and presence of stress conditions; (iii) analyse the capacity of PLS-DA in separating plants in absence and presence of stress; (iv) identify common metabolic responses to different stress conditions; (v) verify the similarities and differences among PLS-DA and metabolic network analyses. The results demonstrated that PLS-DA is capable to distinguish plants under stress from its respective control groups using metabolite profiling data. However, there was no clear pattern regarding the parameters obtained in the analysis of metabolic networks. Whilst some species presented a decrease in network density, others, under different stress condition, had an increase in this parameter. Interestingly, the calculation of the number of hubs in a network, recently established in our group, was positively and negatively correlated with network density (R2 = 0.90; P < 0.001) and heterogeneity (R2 = - 0.59; P = 0.014), respectively. This highlights the power of this parameters that, in addition to providing information about network density and topology, it also highlights which hubs (metabolites) have great influence on the network topology. Taking together the hubs, biomarker analyses and the metabolites that most contributed to the separation in PLS-DA, our work identified which metabolites are possible stress markers in plants, among which stands out asparagine, amino acids (Val, Ile, Leu, Gly, Phe and Trp), sugars (fructose and rafinose) and shikimate. Taken together, our meta-analysis provides a systemic data analysis approach for analyzing metabolite profiling data from plants under stress, in which different tools are used to identify metabolic markers, being possible its extension to areas unrelated to plant biology.