Inferred microbial networks lack replicability: consequences for next-generation biomonitoring

Abstract

Plant-associated microbial interaction networks protect plants against disease. There is, therefore, a need to monitor in real time their responses to environmental changes to predict disease risk and adjust crop protection strategies. Next-Generation Biomonitoring (NGB) proposes to reconstruct automatically these networks from metabarcoding data, to complement ecological community properties commonly used for ecosystem health assessment. This study aimed to evaluate the benefits and shortcomings of community-level and network-level properties for biomonitoring. We specifically investigated whether microbial networks inferred from metabarcoding data show robust responses to agricultural practices, using the grapevine microbiota as a study system. Our results demonstrate a strong footprint of the agricultural practice on the metabarcoding data, when analyzed at the community level. The richness, diversity and evenness of fungal communities were significantly higher in organic than conventional plots. The cropping system also affected the composition of grapevine foliar fungal communities significantly. Contrary to our expectations, microbial networks were less sensitive to changes in agricultural practices than microbial communities, confirming that NGB should not only consider network-level properties but also community-level properties. Moreover, we found that microbial networks lacked replicability within a cropping system but that consensus networks, built from several network replicates, could generate relevant hypotheses of microbial interactions. As things stand, community-level properties appear to be a more reliable and statistically powerful monitoring option than network-level properties. Future developments, especially in network inference methods, are likely to challenge our findings and help to improve the monitoring of the ecosystem services provided by the plant microbiota.