New insights into cheddar cheese microbiota-metabolome relationships revealed by integrative analysis of multi-omics data

Abstract

Cheese microbiota and metabolites and their inter-relationships that underpin specific cheese quality attributes remain poorly understood. Here we report that multi-omics and integrative data analysis (multiple co-inertia analysis, MCIA) can be used to gain deeper insights into these relationships and identify microbiota and metabolite fingerprints that could be used to monitor product quality and authenticity. Our study into different brands of artisanal and industrial cheddar cheeses showed that Streptococcus, Lactococcus and Lactobacillus were the dominant taxa with overall microbial community structures differing not only between industrial and artisanal cheeses but also among different cheese brands. Metabolome analysis also revealed qualitative and semi-quantitative differences in metabolites between different cheeses. This also included the presence of two compounds (3-hydroxy propanoic acid and O-methoxycatechol-O-sulphate) in artisanal cheese that have not been previously reported in any type of cheese. Integrative analysis of multi-omics datasets revealed that highly similar cheeses, identical in age and appearance, could be distinctively clustered according to cheese type and brand. Furthermore, the analysis detected strong relationships, some previously unknown, which existed between the cheese microbiota and metabolome, and uncovered specific taxa and metabolites that contributed to these relationships. These results highlight the potential of this approach for identifying product specific microbe/metabolite signatures that could be used to monitor and control cheese quality and product authenticity.

Figure 1

Principal component analysis (PCA) of industrial and artisanal cheeses for (a) microbiota, (b) GC-MS metabolites and (c) LC-MS metabolites. I: industrial, A: artisanal; C: core, S: Surface; number indicates different brand for each artisanal and industrial cheese.

Figure 2

Venn diagram representing the unique and shared OTUs between all cheeses (a), between industrial cheeses (b) and between artisanal cheeses (c). Microbiota composition of the cheddar cheeses at genus level (d). The ten most abundant genera are shown. I: industrial, A: artisanal; C: core, S: Surface; a-c: sample replicates, 1–2: different brands within each artisanal and industrial cheese.

Figure 3

MCIA analysis of core and surface samples of cheddar cheeses. (a,b) The first two axes of MCIA represents metabolomics (LC/MS and GC/MS) and microbiota composition of the industrial and artisanal core (a) and surface (b) cheeses. Different shapes (df1, diamond: GC/MS data set; df2, triangle: microbiota data set; df3, square: LC/MS data set) represent the different variables connected by lines, the length of these lines is proportional to the divergence between the data. Lines for each sample are joined at a common point, at which the covariance derived from the MCIA analysis is maximal. Color shows the 4 brands of cheeses. (c,e) Psedo-eigenvalue space representing the percentage of variance explained by each of the MCIA component for core (c) and surface (e) datasets. (d,f) Pseudo-eigenvalues space of all datasets for the core (c) and the surface (f), showing overall co-structure between three datasets and shows which dataset contributes more to the total variance.

Figure 4

The coordination of GC/MS metabolites (a), OTUs (b), and LC/MS metabolites (c) for the core samples and the coordination of GC/MS metabolites (d), OTUs (e) and LC/MS metabolites (f) for the surface samples are shown. The OTUs and metabolites at the positive end of PC1 are associated with the artisanal cheese and at the opposite side are associated with the industrial cheeses, also these features show strongest covariance. P120: tyramine, N40: citrulline, P242: unknown peptide, P11: Proline, P155: unknown (m/z = 628.55, RT = 19.14), P262: Unknown (m/z = 631.491 = , RT = 17.80), N19: 1-formyl pentanedioic acid, P80: dipeptide (asparagine-valine), N43: unknown (m/z = 268.15, RT = 6.76). (See putative formula and MS2 features in Supplementary Table S3). Abbreviations for metabolites are shown in Supplementary Table S1. Identity of OTUs is shown in Supplementary Dataset S5.

Figure 5

Inter-omic Spearman correlation networks. Pairwise Spearman correlation was performed for each OTU and GC/MS metabolites for the core (a) and the surface (b) samples; Any resulting correlations with q > 0.1 were removed. Red line shows positive and blue line shows negative correlation. Circles indicate taxa and diamonds indicate metabolites. Stronger correlations are shown as line thickness increases. Thicker line shows strangest correlations. Taxa keys: red nodes: Streptococcus, purple nodes: Lactococcus, cyan nodes: Lactobacillus, blue nodes: Macrococcus, green nodes: Leuconostoc, gray nodes: unassigned genus. Abbreviations for metabolites are shown in Supplementary Table S1.