Microbial association networks in cheese: a meta-analysis

Abstract

Interactions among starter and non-starter microorganisms (starter bacteria, naturally occurring or intentionally added non-starter bacteria, yeasts and filamentous fungi, spoilage and pathogenic microorganisms and, finally bacteriophages and even arthropods) deeply affect the dynamics of cheese microbial communities and, as a consequence, multiple aspects of cheese quality, from metabolites affecting the taste, aroma and flavour, to body, texture and colour. Understanding and exploiting microbial interactions is therefore key to managing cheese quality. This is true for the simplest systems (fresh cheeses produced from pasteurized milk using defined starters composed solely of Lactic Acid Bacteria) and the more so for complex, dynamic systems, like surface ripened cheese produced from raw milk, in which a dynamic succession of diverse microorganisms is essential for obtained the desired combination of sensory properties while guaranteeing safety. Positive (commensalism, protocooperation) and negative (competition, amensalism, predation and parasitism) among members of the cheese biota have been reviewed multiple times. Although the complex, multidimensional datasets generated by multi-omic approaches to cheese microbiology and biochemistry are ideally suited for the representation of biotic and metabolic interactions as networks, network science concepts and approaches are rarely applied to cheese microbiology. In this review we first illustrate concepts relevant to the description of microbial interaction networks using network science concepts. Then, we briefly review methods used for the inference and analysis of microbial association networks and their potential use in the interpretation of the cheese interactome. Since these methods can only be used for mining microbial associations, a review of the experimental methods used to confirm the nature of microbial interactions among cheese microbes. Finally, we demonstrate the potential of microbial association network inference by mining metataxonomic data stored in the public database DairyFMBN, a specialized version of FoodMicrobionet which collates data on 74 metataxonomic studies on dairy products. Microbial association networks were inferred from 34 studies on cheese with up to 4 different methods and the results discussed to evaluate several aspects (choice of method, level of taxonomic resolution for the analysis, network, node and edge properties) which provide insight on the usefulness of this approach as explorative tool in the detection of microbial interactions in cheese. HighlightsO_LIApproaches for inference of association networks from metataxonomic data were reviewed C_LIO_LIA metastudy on association networks in cheese was carried out using 34 recent studies C_LIO_LIInference method and taxonomic resolution should be chosen carefully C_LIO_LISPIEC-EASI may be used as a conservative method for microbial association inference C_LIO_LIEdge and node properties support the formulation of testable hypotheses for microbial interactions C_LI