Rumen microbiome and fat deposition in sheep: insights from a bidirectional mendelian randomization study

Abstract

Rumen microbiotas are known to influence the fat deposition (FD) in sheep, but controversy over causality remains unresolved. Here, we performed microbiome-wide association studies (MWAS), microbiome genome-wide association analysis (mbGWAS) and bidirectional mendelian randomization (MR) analyses on 1,150 sheep with genotype data from whole-genome resequencing, 16S rRNA sequencing and multilevel FD-traits data. We quantified the proportion of individual variation in FD-traits explained by host genetics, rumen microbiota, and their interaction effects. We identified 32 rumen microbiota biomarkers including Bifidobacterium that were associated with FD-traits (P_adj <0.05). Further, utilizing five MR methods, we identified eight causal associations between marker genera and FD-traits (P_adj <0.05), including Butyrivibrio, Olsenella, p-2534-18B5 gut group, Prevotellaceae UCG-003, and Pseudobutyrivibrio causing forward causal effects on FD, and changes in Flexilinea and Suttonella induced by FD. To our knowledge, this is the inaugural attempt to employ MR in sheep to investigate the causal relationships between gastrointestinal microbiota and complex phenotypes, underscoring the potential for developing interventions related to adipose deposition in sheep from the perspective of the rumen microbiome.

Fig. 1. Study design.

Flowchart for the study of the casual association between rumen microbiota and fat deposition traits in sheep. LD linkage disequilibrium, SNP single nucleotide polymorphism, MWAS microbiome-wide association studies, mbGWAS microbiome genome wide association study.

Fig. 2. Characteristics of fat deposition phenotypes.

a The correlation between the phenotype of sheep fat deposition. All sheep fat deposition traits were classified into two types: overall level and local level, based on their sources. The Spearman and Pearson correlation methods were used above and below the diagonal, respectively. The size and color of the squares, as well as the area and color of the pie chart, represent the magnitude of the correlation coefficient. “*” represents P value < 0.05, “**” represents P value < 0.01, and “***” represents P value < 0.001. b–e The principal component analysis (PCA) analysis was performed on all adjusted data of sheep fat deposition phenotypes to identify representative indicators. The horizontal axis represents the first principal component (PC), the vertical axis represents the second PC, and the percentage represents the contribution of the PC to the sample differences. Each point represents a sample. Figure (b) show the PCA results visualization based on the Total-FW grouping (categorical data). Samples in the same group are represented by the same color. H_Total-FW represents individuals with Total-FW values greater than “mean + 3sd” (n = 178); M_Total-FW represents individuals with Total-FW values between “mean + 3 SD” and “mean –3 SD” (n = 770); L_Total-FW represents individuals with Total-FW values less than “mean –3 SD” (n = 190). Figure (c) shows the PCA visualization based on the Total-RFW grouping, using the same grouping method as Total-FW (H_Total-RFW: n = 174, M_Total-RFW: n = 775, L_Total-RFW: n = 187). The figure (d, e) shows the PCA visualization based on the continuous values of Total-FW and Total-RFW, where the color changes represent the magnitude of Total-FW and Total-RFW values. f Above and below the diagonal lines are the “microbial correlation” and “genetic correlation” of sheep fat deposition phenotypes, respectively. All correlations were positively correlated. The size and color variation of the circles represent the high and low values. BMI body mass index, BF backfat thickness, GR Rib thickness, FW Absolute weight of fat (Measured using an electronic scale), RFW relative weight of fat (FW/body weight).

Fig. 3. Contributions of host genetics, rumen microbial features, and their interaction effects to sheep fat deposition traits.

a The individual differences in sheep fat deposition phenotypes explained by host genetics, rumen microbiome, and their interaction were estimated using the variance component method. Ho² holobiability, BMI body mass index, BF backfat thickness, GR Rib fat thickness, FW absolute weight of fat (Measured using an electronic scale), RFW relative weight of fat (FW/body weight). b Mantel-test (based on Spearman’s rank correlation) was performed between the rumen microbial relationship/similarity matrices (based on ASV data) and the genome relationship matrix (GRM). Each square represents the correlation coefficient between two matrices, and the darker color indicates a higher correlation. GRM genomic relationships matrix, Euclidean: Euclidean distance, Bray-Curtis: Bray-Curtis dissimilarity, Jaccard: Jaccard index, CCA constrained correspondence analysis, DCA Detrended Correspondence Analysis, MRM Microbial Relationship Matrix.

Fig. 4. Association of rumen microbiota with sheep fat deposition traits.

a–d Microbiome-wide association studies (MWAS). Identification of rumen microbial features associated with fat deposition phenotypes using MWAS. The results were visualized using Manhattan plots. Each circle represented a genus, and the X-axis indicated that all the genera were sequenced at the phylum level, and the Y-axis indicated the relevant statistical significance. Solid red lines and dashed red lines indicate Bonferroni corrected P value = 0.01 and 0.05, respectively. e Venn diagram depicting unique and shared marker rumen microbial features among four MWAS outcome datasets (2 Traits × 2 models). The box plots on the left show details of the distribution of marker microbial features among the four MWAS outcome datasets. f The bar graphs illustrate the detection rate and average relative abundance of marker microbiota. g Pearson (lower diagonal) and Spearman (upper diagonal) correlations between the rumen microbial genera associated with fat deposition in sheep. Blue and red indicated positive and negative correlations, respectively; The gray background labels represent the modules divided by clustering. h The Sperman correlation between marker genera related to fat deposition and rumen VFA profile. The size and color of the circle indicated the size of the correlation coefficient, while red and blue indicated positive and negative correlation, respectively. The square color block in the background indicated the P value of correlation test.

Fig. 5. Mendelian Randomization analysis of the effect of marker rumen microbiota on fat deposition in sheep.

Forest plot comparing results from inverse variance weighting (IVW), weighted mode (WM), MR-Egger regression, weighted median estimator (WME), simple mode method, and multiplicative random-effects IVW model [IVW (muti−random)].

Fig. 6. MR scatter plots.

a–k Scatterplot of SNP potential effects on marker rumen microbial features vs sheep fat deposition traits, with the slope of each line corresponding to estimated MR effect per method.