Microbiota and Metabolite Profiling Combined With Integrative Analysis for Differentiating Cheeses of Varying Ripening Ages

Abstract

Cheese maturation and flavor development results from complex interactions between milk substrates, cheese microbiota and their metabolites. In this study, bacterial 16S rRNA-gene sequencing, untargeted metabolomics (gas chromatography-mass spectrometry) and data integration analyses were used to characterize and differentiate commercial Cheddar cheeses of varying maturity made by the same and different manufacturers. Microbiota and metabolite compositions varied between cheeses of different ages and brands, and could be used to distinguish the cheeses. Individual amino acids and carboxylic acids were positively correlated with the ripening age for some brands. Integration and Random Forest analyses revealed numerous associations between specific bacteria and metabolites including a previously undescribed positive correlation between Thermus and phenylalanine and a negative correlation between Streptococcus and cholesterol. Together these results suggest that multi-omics analyses has the potential to be used for better understanding the relationships between cheese microbiota and metabolites during ripening and for discovering biomarkers for validating cheese age and brand authenticity.

Keywords: 16S rRNA-based microbiota analysis; GC-MS untargeted metabolomics; cheese; cheese maturity; integrative analysis.

FIGURE 1

Variation in the relative abundance of bacterial genera in cheeses between brands and between cheeses of different maturity. Brands (A–C) are as indicated. Only genera detected at >0.1% relative abundance are shown.

FIGURE 2

Variation in bacterial community composition in cheese between brands and between cheeses of different maturity. Bray-Curtis non-metric multi-dimensional scaling (nMDS) plots and PERMANOVA results are shown comparing (A) variation between cheeses from different manufacturers and (B–D) variation between cheeses of different maturity for each manufacturer: brand A (B), brand B (C), and brand C (D). Different colors and shapes show different cheese manufactures and different ages, as indicated. PERMANOVA results are shown; Pseudo-t = the Pseudo-t statistic, whereby higher values indicate greater differences between the community structure of cheeses of the two groups compared; P (MC) = Monte Carlo P-value, with significant values (P < 0.05) indicated by an asterisk.

FIGURE 3

Principal Component Analysis (PCA) of untargeted GC/MS metabolomics of cheeses made by different manufacturers. (A) PCA of untargeted GC/MS metabolomics of cheeses made by different manufacturers. PC1 and PC2 account for 32.2% and 27.5% of the variance, respectively. (B–D) The biplot superimposed on the scores and loadings of PCA analysis based on a correlation scaling method for cheeses for brand A (B), brand B (C), and brand C (D) Brands and maturity of cheeses are as indicated. p(corr), t(corr) is a combined vector, p(corr) represents loading p scaled as correlation coefficient between X and t; t(corr) represents score t scaled as correlation coefficient resulting in all points falling inside the circle with radius 1. Different colors represent different brands and different classes of metabolites: black; brand A, cyan; brand B, red; brand C, green; amino acids and amines, blue; carboxylic acids, pink; fatty acids and sterols, orange; sugar and sugar phosphates. Orn, ornithine; Tyr, tyrosine; GABA, gamma amino butyric acid; Lys, lysine; Gly, glycine; Val, valine; Ser: serine; Leu, leucine; Noreleu, noreleucine; Thr, threonine; Pro, proline; Pip, piperedine; Asn, asparagine; AspA, aspartic acid; Glu, glutamic acid; Met, methionine; Arg, arginine; PDA, pentadecanoic acid; HepA, heptadecanoic acid; LA, lauric acid (dodecanoic acid); PA, palmitic acid (hexadecanoic acid); STA, stearic acid (octadecanoic acid); MA, myristic acid (tetradecanoic acid); Oxal, oxalic acid; Succ, succinic acid; GlyA, glyceric acid; Glt, glutaric acid; MalA, malonic acid; HGlt, hydroxy-glutaric acid; Citric, citric acid; GalA, galactonic acid, Pglu, pyroglutamic acid; Oro, orotic acid; UA, uric acid; PhA, phosphoric acid; Mann, mannose; Gal, galactose; MI, inositol myo; Lac, lactose; Gly3p, glycerol-3-phosphate.

FIGURE 4

Multifactorial analysis (MFA) of cheese microbiota and cheese metabolite profiles for cheeses of different maturities from different manufacturers (A, B, and C). (A–C) MFA scatter plots for cheeses from brands A, B, and C, respectively. Ellipses representing the barycentre of the sample groups with 95% confidence. Maturity (age) of chesses is as indicated. (D–F) MFA group representations to illustrate the relationships between variables (bacterial genus composition, metabolite profiles, maturity of cheeses) and Dim 1and Dim 2 for brands A, B, and C, respectively.

FIGURE 5

Multifactorial analysis (MFA) showing correlations between bacterial genera and metabolites in cheese. The Correlation circle depicts correlations in normalized abundance between cheese, bacterial genera (blue) and cheese metabolites (red) along the MFA axes for (A) brand A, (B) brand B, and (C) brand C cheeses. Those metabolites and genera which are depicted together in the same direction along an axis indicate positive correlations; those which are depicted together in the opposite direction indicate negative correlations; metabolites and genera depicted in different directions indicate no correlations. The strength of the correlation is shown by the increasing distance from the center of the plot. To improve the readability of the plots, only the greatest correlations for metabolites in each dimension are shown.