A transcriptome multi-tissue analysis identifies biological pathways and genes associated with variations in feed efficiency of growing pigs

Abstract

Background: Animal's efficiency in converting feed into lean gain is a critical issue for the profitability of meat industries. This study aimed to describe shared and specific molecular responses in different tissues of pigs divergently selected over eight generations for residual feed intake (RFI).

Results: Pigs from the low RFI line had an improved gain-to-feed ratio during the test period and displayed higher leanness but similar adiposity when compared with pigs from the high RFI line at 132 days of age. Transcriptomics data were generated from longissimus muscle, liver and two adipose tissues using a porcine microarray and analyzed for the line effect (n = 24 pigs per line). The most apparent effect of the line was seen in muscle, whereas subcutaneous adipose tissue was the less affected tissue. Molecular data were analyzed by bioinformatics and subjected to multidimensional statistics to identify common biological processes across tissues and key genes participating to differences in the genetics of feed efficiency. Immune response, response to oxidative stress and protein metabolism were the main biological pathways shared by the four tissues that distinguished pigs from the low or high RFI lines. Many immune genes were under-expressed in the four tissues of the most efficient pigs. The main genes contributing to difference between pigs from the low vs high RFI lines were CD40, CTSC and NTN1. Different genes associated with energy use were modulated in a tissue-specific manner between the two lines. The gene expression program related to glycogen utilization was specifically up-regulated in muscle of pigs from the low RFI line (more efficient). Genes involved in fatty acid oxidation were down-regulated in muscle but were promoted in adipose tissues of the same pigs when compared with pigs from the high RFI line (less efficient). This underlined opposite line-associated strategies for energy use in skeletal muscle and adipose tissue. Genes related to cholesterol synthesis and efflux in liver and perirenal fat were also differentially regulated in pigs from the low vs high RFI lines.

Conclusions: Non-productive functions such as immunity, defense against pathogens and oxidative stress contribute likely to inter-individual variations in feed efficiency.

Keywords: Feed efficiency; Multi-tissues; Multiple factor analysis; Pig; Residual feed intake; Transcriptome.

Fig 1

Molecular changes in tissues due to RFI line. Microarray data obtained in longissimus muscle, liver, perirenal (PRAT) and subcutaneous (SCAT) adipose tissues were separately analyzed for the main effects of line (low RFI: residual feed intake below the average; high RFI: residual feed intake above the average). Genes were declared as differentially expressed (DEG) between pigs from the low or high RFI lines according to the same cutoffs in the four tissues (P < 0.01 and FC between conditions > |1.1|). FC: fold-change between mean values calculated in pigs from the low or high RFI lines. Values are inversed and preceeded by a minus sign for FC < 1 (e.g., FC = 0.5 was indicated as FC = −2)

Fig 2

Reliability of microarray data. A subset of genes was analyzed by qPCR in longissimus muscle, liver, perirenal adipose tissue and subcutaneous adipose tissue (n = 48 in each tissue). The fold-changes (FC) calculated between expression levels of target genes in pigs from the low to high RFI lines in microarrays on one hand and qPCR on the other hand, were plot together. Correlation coefficient was calculated between the two measures. Forward and reverse primers used for qPCR are listed in Additional file 3

Fig 3

VENN diagram representing DEG due to RFI line. The lists of differentially expressed genes (DEG) due to line effect in longissimus muscle, liver, perirenal (PRAT) and subcutaneous (SCAT) adipose tissues were considered to edit the VENN diagram, using the VENNY tool [http://bioinfogp.cnb.csic.es/tools/venny/index.html]. A total of 147 DEG were commonly found in the four tissues (black circle). Black arrows point to the DEG exclusively found in one tissue. RFI: residual feed intake

Fig 4

Examples of functional networks of regulated genes shared by different tissues. The 147 differentially expressed genes commonly listed in longissimus muscle, liver, perirenal and subcutaneous adipose tissues were submitted to Ingenuity® Pathway Analysis to visualize small networks of co-expressed genes. Highly-interconnected non-redundant networks were released from specified genes (represented as nodes) and relationships (represented as edges) between them or with neighboring genes were established based on literature records in the Ingenuity® Pathway Knowledge Base. This allows identifying functional networks related to response to oxidative stress, apoptosis, immunity and(or) protein metabolism. Two small functional networks are shown. More networks can be viewed in Additional files 5 and 6. a Relationships through low density lipoproteins (LDL), between SOD2 and PON1 (response to oxidative stress), IL1 and CTSC (immunity) and NTN1 (apoptosis). b Relationship between different molecules acting in protein translation (RPL6, RPL14), protein catabolism (PSMA3, 26S proteasome, and heat-shock proteins HSP70) and protein transport (YWHAZ)

Fig 5

Multi-way mathematical datasets analysis: consensus between distinct tissues transcriptomes underlying separation of molecular data due to RFI line. a Pigs were represented on the scatter plot created with the first two dimensions (Dim) of the multiple factor analysis (MFA) which aggregated the whole transcriptomic variability across four tissues. Only the first dimension of the MFA (Dim1; 24% of the whole transcriptomic variability) discriminated pigs from the low RFI line and pigs from the high RFI line. b Synthetic latent variables were calculated from correlated molecular probes in the longissimus muscle (Dim 1_LL), liver (Dim 1_liver), perirenal adipose tissue (Dim 1_PRAT) and dorsal subcutaneous adipose tissue (Dim_1 SCAT) and projected in the correlation circle of MFA. The first latent variables in muscle (Dim1_muscle), liver (Dim1_Liver), perirenal adipose tissue (Dim1_PRAT) and subcutaneous adipose tissue (Dim1_SCAT) had a contribution near |1| to Dim1 in the diagnostic plot. This allows considering Dim1 of the MFA as a relevant summary of the main common molecular responses across the four tissues