A high deposition of muscle fatty acids, particularly UFAs, is correlated with high-quality meat
The predictive ability of the four machine learning models (RF, XGB, LGB, and SVM) was evaluated by training on a dataset of 291 samples from six pig studies (Supplementary Table 1). The prediction performance was ranked as RF > XGB > LGB > SVM, and the high regression coefficient (R2) values (≥0.83) indicated that the RF model was suitable for predicting the influence of muscle fatty acids on the meat quality (Fig. 1A, B). Therefore, the RF model was further used to predict the relative importance of muscle fatty acids in terms of meat quality (Fig. 1C, D). A set of the most informative features of muscle fatty acids, also referred to as ‘predictors’, was extracted from the RF model. The results revealed that muscle polyunsaturated fatty acids (PUFAs) and the monounsaturated fatty acids/total fatty acids (MUFAs/TFAs) ratio had strong positive predictive power for meat color, but the importance of PUFAs decreased and the MUFA/TFA ratio increased at 24 hours after sacrifice (Supplementary Fig. 1 A, B). The SFA/TFA ratio had a significant negative effect on the b (Supplementary Fig. 1 C, D) and L (Supplementary Fig. 1 E, F) values. The marbling score, which is an intramuscular fat correlation index, was positively associated with the muscle UFA/TFA ratio, especially the MUFA/TFA ratio (Supplementary Fig. 1 G). The muscle pH index reflects the oxidative reactivity of fatty acids, and the pH index at 45 min was positively related to PUFA content, while it was negatively correlated with TFA at 24 hours (Supplementary Fig. 1 H, I).
Specifically, 31 fatty acids in the muscle were evaluated for relative importance, and 6 features were found to be predictors of meat quality. The C22:0 content had a significant correlation with the meat color a value (Fig. 2A, B) and marbling score (Fig. 2 G). For the b value, C21:0 had a significant positive correlation at 45 min (Fig. 2 C), and C18:3n3 was positively correlated at 24 hours (Fig. 2 D). C12:0 and C21:0 were negatively correlated with the L value at 45 min (Fig. 2 E) and 24 hours (Fig. 2 F), respectively. Muscle pH was positively related to C18:2n6c content at 45 min (Fig. 2 H), while a negative correlation with C20:1 was observed at 24 hours (Fig. 2 I). In summary, high levels of muscle fatty acids produced high-quality pork with better meat color, and increasing UFA abundance was predicted to improve the marbling score of meat. However, muscle C21:0 and C20:1 were not desirable due to the negative effects on meat color and pH.
Microbial α diversity correlates with high-quality meat rich in PUFAs
The prediction models of the gut microbiota demonstrated that, in terms of microbial α diversity and microbial abundance at the phylum level, the RF model had a better fitting performance than did the XGB, LGB, and SVM models (Fig. 3A, B. R2≥0.87 for α diversity; R2≥0.85 for microbial phyla). Subsequent relative importance analysis was performed using the RF model for the prediction of microbial α diversity and phyla. At the genus level and for microbial functional annotations, the best prediction performance was observed for the LGB model (Fig. 3 C, D. R2≥0.92 for microbial genera; R2≥0.93 for microbial functional annotations). Thus, the roles of genre and microbial pathways in determining muscle fatty acid levels were predicted using the LGB model.
Community evenness and richness are generally evaluated by microbial α diversity using the ACE, Chao1, Simpson, and Shannon indices. According to the relative importance calculated by the RF model, ACE exhibited a negative correlation with muscle MUFAs (Fig. 4 A), PUFAs (Fig. 4 B), UFAs (Fig. 4 C), SFAs (Fig. 4 D), and TFAs (Fig. 4 E). Data mining revealed that a core subset, which contributed to microbial α diversity, exhibited strong predictive power for muscle-specific fatty acids. Primarily, microbial α diversity (Chao1 and ACE) was positively correlated with most UFAs (i.e., C14:1, C17:1, C18:3n6, C20:3n3, C20:4n6, C20:5n3, C22:1n9, and C22:6n3) and negatively correlated with muscle SFAs (i.e., C6:0, C8:0, C10:0, C12:0, C14:0, C15:0, C16:0, C17:0, C18:0, C20:0, and C21:0) (Fig. 4 F-O and Supplementary Fig. 2 A-U). In summary, increasing gut microbial evenness and richness was produced high-quality meat rich in GLA, DHA, and EPA.
The gut microbiota is involved in fatty acid deposition in muscle
According to the relative importance results calculated by the RF model at the phylum level, two phyla were positively correlated with muscle fatty acids. Notably, one of the subgraphs containing muscle UFAs (Fig. 5 C) and one containing TFAs (Fig. 5 E) was found in the gut, including Actinobacteria. The subgraphs correlated with Spirochaetes contained MUFAs (Fig. 5 A), PUFAs (Fig. 5 B), and SFAs (Fig. 5 D).
For each specific fatty acid, Actinobacteria was negatively correlated with muscle C6:0 (Supplementary Fig. 3 A), C8:0 (Supplementary Fig. 3 B), C10:0 (Supplementary Fig. 3 C), and C12:0 (Fig. 5 F), and positively correlated with C22:2 content (Supplementary Fig. 3 R). Spirochaetes was positively correlated with C14:0 (Supplementary Fig. 3 D), C15:0 (Supplementary Fig. 3 F), C16:0 (Supplementary Fig. 3 G), C16:1 (Supplementary Fig. 3 H), C17:0 (Supplementary Fig. 3 I), C17.1 (Supplementary Fig. 3 J), C18:2n6c (Fig. 5 G), C18:1n9c (Fig. 5 L), C18:3n6 (Fig. 5 M), C20:0 (Supplementary Fig. 3 L), C20:1 (Fig. 5 I), C20:2 (Supplementary Fig. 3 M), and C20:4n6 (Supplementary Fig. 3 P) depositions in the muscle. Muscle C18:0 (Supplementary Fig. 3 K), C20:3n6 (Supplementary Fig. 3 O), and C21:0 (Fig. 5 J) concentrations were negatively correlated with Proteobacteria. C18:3n3 (Fig. 5 H) and C20:3n3 (Supplementary Fig. 3 N) were principally, positively correlated with Cyanobacteria. Firmicutes (negative for C22:0, C23:0, and C24:0) (Fig. 5K; Supplementary Fig. 3 S, T) and Bacteroidetes (positive for C20:5n3 and C22:6n3) (Fig. 5 N, O) were the two most abundant phyla in most pigs, and the F/B ratio, a microbial marker for lipid metabolic disorder [26], was positively correlated with C14:1 (Supplementary Fig. 3 E) and C24:1 (Supplementary Fig. 3 U) but negatively related to C22:1n9 (Supplementary Fig. 3 Q) and C22:1 (Supplementary Fig. 3 Q).
According to the predictions of the LGB model, the relative importance of gut microbiota at the genus level in predicting muscle fatty acids was further evaluated. Only Escherichia, a common pathogen in the gut, was identified to be negatively correlated with muscle fatty acids, including MUFAs (Fig. 6 A), PUFAs (Fig. 6 B), UFAs (Fig. 6 C), SFA (Fig. 6 D), and TFAs (Fig. 6 E), indicating that pathogenic elimination promote muscle fatty acid deposition in pigs.
The relationships between gut genera and specific fatty acids were further analyzed. Eleven of 50 genera were associated with muscle fatty acids, including Psychrobacter (C6:0, C8:0, and C10:0) (Supplementary Fig. 4 A-C), Bacteroides (C12:0) (Fig. 6 F), Escherichia (C14:0, C14:1, C15:0, C16:0, C16:1, C18:0, C18:1n9c, C18:2n6c, C20:0, C21:0, and C22:1n9) (Fig. 6 G, J, L and Supplementary Fig. D, F, G, H, K, L), Clostridium (C17:0, C20:1, C20:3n6, and C22:2) (Fig. 6 I and Supplementary Fig. 4 I, O, R), Campylobacter (C17:1 and C20:3n3, C22:0, and C24:0) (Fig. 6 K and Supplementary Fig. 4 J, N, T), Bradyrhizobium (C18:3n6) (Fig. 6 M), Phascolarctobacterium (C18:3n3, C20:5n3, and C22:6n3) (Fig. 6 H, N, O), Terrisporobacter (C20:2) (Supplementary Fig. 4 M), Ruminococcus (C20:4n6) (Supplementary Fig. 4 P), Corynebacterium (C23:0) (Supplementary Fig. 4 S), and Dehalobacter (C24:1) (Supplementary Fig. 4 U). Overall, Bradyrhizobium, Phascolarctobacterium, Terrisporobacter, Ruminococcus, and Dehalobacter were the main species associated with the abundance of very long-chain UFAs, such as C18:3n3, C18:3n6, C20:2, C20:4n6, C20:5n3, C22:6n3, and C24:1.
Gut microbiota-targeted fatty acid degradation is negatively correlated with muscle fatty acid deposition
According to the LGB model, the relative importance of level 3 KEGG microbial functional annotations for muscle fatty acids is summarized in Supplementary Figs. 5 and 6. Notably, UFAs (Supplementary Fig. 5 C), especially PUFAs (Supplementary Fig. 5 B), were negatively correlated with microbial metabolism, likely related to fatty acid degradation. Indeed, the fatty acid degradation pathway was negatively correlated with muscle C8:0 (Supplementary Fig. 6 B), C17:1 (Supplementary Fig. 6 J), C18:2n6c (Supplementary Fig. 5 G), C18:3n3 (Supplementary Fig. 5 H), C18:3n6 (Supplementary Fig. 5 M), C20:2 (Supplementary Fig. 6 M), C20:3n3 (Supplementary Fig. 6 N), C20:4n6 (Supplementary Fig. 6 P), C20:5n3 (Supplementary Fig. 5 N), C22:0 (Supplementary Fig. 5 K), C23:0 (Supplementary Fig. 6 S), and C24:0 (Supplementary Fig. 6 T) content. Microbial metabolism in diverse environments and muscle MUFAs (Supplementary Fig. 5 A), SFAs (Supplementary Fig. 5 D), and TFAs (Supplementary Fig. 5 E) were also correlated, indicating the potential role of environment-related microbial changes in the muscle fatty acid phenotypes of pigs. Typically, muscle fatty acids can also be influenced by propanoate metabolism (C14:1) (Supplementary Fig. 6 E); glycine, serine and threonine metabolism (C6:0, C10:0, C12:0, and C22:6n3) (Supplementary Fig. 5 F, O and Supplementary Fig. 6 A, C); alanine, aspartate and glutamate metabolism (C22:2) (Supplementary Fig. 6 R); microbial metabolism in diverse environments (C14:0, C15:0, C16:0, C16:1, C17:0, C18:0, C18:1n9c, C20:0, C20:3n6, C21:0, C22:1n9, and C24:1) (Supplementary Fig. 5 L, J and Supplementary Fig. 6 D, F, G, H, K, L, O, Q, U); and aminoacyl-tRNA biosynthesis (C20:1) (Supplementary Fig. 5I).
Microbial manipulation impacts the production of functional fatty acid-rich meat
A higher dietary level of UFAs, especially ALA, oleic acid, GLA, DHA, and EPA, is recommended for a healthy lifespan. Therefore, we further predicted how to produce high-quality pork rich in ALA, oleic acid, GLA, DHA, and EPA through microbial manipulation. According to the LGB model, 4 genera had the most significant impact on muscle C18:1n9c (oleic acid), with a positive relation to Dorea and negative relation to Escherichia, Terrisporobacter, and Actinobacillus (Fig. 6 L). In addition, positive associations between the abundances of Bradyrhizobium and Lachnoclostridium and C18:3n6 (GLA) (Fig. 6 M); Bradyrhizobium, Lachnoclostridium, Phascolarctobacterium and C20:5n3 (EPA) (Fig. 6 N); Bradyrhizobium, Escherichia, Lachnoclostridium and C22:6n3 (DHA) (Fig. 6 O); Enterococcus and Phascolarctobacterium and C18:3n3 (ALA), a precursor of DHA and EPA (Fig. 6H) were observed. Two microbial species from the genera (Bradyrhizobium and Lachnoclostridium) shared with the correlations, which were expected to provide an opportunity to promote GLA, DHA, and EPA deposition in the muscle. Although positively related to C22:6n3, Escherichia was also negatively correlated with C18:1n9c, C18:3n3, C18:3n6, and C20:5n3; thus, Escherichia inhibition was desirable for the production of high-quality meat rich in functional fatty acids.
The muscle PUFA content was negatively correlated with microbial metabolism related to fatty acid degradation (Supplementary Fig. 5 B). Here, we further analyzed the relationships between the key predictors of functional fatty acids (i.e., Lachnoclostridium, Bradyrhizobium, and Escherichia) and the fatty acid degradation pathway. The important microbes that influenced the degradation of fatty acids were ranked in the following order: Lachnoclostridium > Escherichia > Bradyrhizobium. Correlation analysis revealed a positive correlation between Escherichia and fatty acid degradation, while Lachnoclostridium and Bradyrhizobium exhibited a negative correlation (Supplementary Table 4). These findings further underscored the fact that promoting the growth of Lachnoclostridium and Bradyrhizobium or inhibiting the growth of Escherichia in the gut might facilitate the deposition of muscle functional fatty acids.