Functional analysis of inter-individual transcriptome differential expression in pig longissimus muscle
Summary
Selection of pigs for increased meat production or improved meat quality changes muscle mass and muscle composition. This will be related to transcriptome expression profile changes in muscle tissue, generating inter-individual differences. This study investigated the differentially expressed genes in the transcriptome profiles of the longissimus muscle of 75 Large White–Duroc cross sows and castrates. The use of a common reference design enabled to investigate the inter-individual transcriptome expression profile differences between the animals as compared with the means of all animals. The aim of the study was to identify the biological processes related to these inter-individual differences. It was expected that these processes underlie the selection effects. In total, 908 transcripts were differentially expressed. Among them, 762 were mainly downregulated and 146 were mainly upregulated. Gene Ontology and Pathways analyses indicated that the differentially expressed genes belong to three groups of processes involved in protein synthesis and amino acid–protein metabolism, energy metabolism and muscle-specific structure and activity processes. Comparing the functional biological analysis results with previously reported data suggested that the protein synthesis, energy metabolism and muscle-specific structure would contribute to meat production and the meat quality.
Introduction
Improving growth rate and muscularity has been the primary focus during the past decades in pig breeding (Merks 2000). During the last decade, consumer concerns changed the focus of breeding towards meat quality. Selection uses genetic differences between animals and improve the average level of traits. Changing the muscle transcriptome is one mechanism for the genome to change the level of traits during selection (Ponsuksili et al. 2008). Therefore, individual animals will show differential transcriptome expressions during selection related to the mechanism of the trait under selection. Both selection for intramuscular fat and muscularity were accompanied by transcriptome expression differences between individuals during prenatal life and at slaughter (Te Pas et al. 2005, 2010; Cagnazzo et al. 2006). This makes pigs a good animal model to study inter-individual transcriptome variation.
Numerous studies have demonstrated that there are several genes and chromosomal regions affecting meat production (Bidanel & Rothschild 2002; http://www.animalgenome.org/). At present, few genes are known to explain the variability of this complex trait (Ciobanu et al. 2004) including meat developmental rate and meat quality. Consequently, improvement of meat quality could benefit from the use of genomic approaches.
Analysis of these inter-individual differences may highlight the biological mechanisms underlying the selection effects. The objectives of this study were (i) to investigate inter-individual transcriptome variation in the pigs of a typical meat production cross-breed and (ii) to investigate the physiological mechanisms related to the inter-individual transcriptome expression differences using bioinformatics.
Material and methods
Animals and muscle tissue samples
Seventy-five Large White (LW) pigs bred with a terminal Duroc boar from two Duroc boar lines were slaughtered according to Spanish laws. The number of animals derived from the two Duroc lines was equal. In total 47 sampled animals were sows, and 34 were castrates. No entire males were sampled. Table 1 shows the variation in meat production–related traits. There was no difference between the two Duroc lines for average and variation in the data for the meat quality traits (data not shown). Meat quality traits were measured according to Honikel (1998). The longissimus muscle (LD) was sampled at the last rib. Tissue samples were taken within one hour after slaughter (bleeding) and immediately snap-frozen. Samples were stored at −80°C until use.
| Traits | Mean | SD |
|---|---|---|
| Carcass weight (kg) | 84.40 | 6.81 |
| Carcass lean (%) | 55.29 | 2.82 |
| Fat thickness (mm) | 15.22 | 4.55 |
| PH24 | 5.86 | 0.12 |
| NPPCa color (Score) | 2.68 | 0.52 |
| NPPCa marbling (Score) | 2.08 | 0.74 |
| DRIP (%) | 2.47 | 0.94 |
- aNational pork producers council.
Microarray analyses
Approximately 200 mg of tissue was used for RNA isolation. The RNA was isolated using the RNA Trizol (Sigma-Aldrich, St. Louis, MO, USA) procedure with an additional phenol/chloroform extraction. The RNA isolates were quantified using a Nanodrop apparatus (Nanodrop; Thermo Scientific, Wilmington, DE, USA) and quality checked on a 1% agarose gel.
Microarrays were hybridized according to a common reference design. A reference sample was created by mixing equal amounts of each RNA sample. Dobbin and co-workers showed that for dual colour reference design, at least 10% of the samples should have a dye swap to correct for colour bias (Dobbin & Simon 2002; Dobbin et al. 2003). Therefore, we decided that half of the 75 samples were hybridized in dye swap (Te Pas et al. 2010).
One microgram of RNA was Cy3 or Cy5-labelled using labelled Cytidine triphosphate’s QuickAmp Labeling kit (Agilent, Santa Clara, CA, USA) according to the manufacturers’ protocol. The labelled cRNA was purified with Qiagen’s RNeasy mini spin columns (QIAGEN Inc, Hilden, Germany), and then cRNA was fragmented according to the manufacturers’ protocol. The cRNA was quantified according to the manufacturers’ protocol, and 1.65 μg of each colour was hybridized per microarray. The dual colour hybridization was performed according to the protocol of the manufacturer of the hybridization kit (Agilent) using the 70-mer 20 400 probes containing whole genome swine protein-annotated oligonucleotide microarray (http://www.pigoligoarray.org, gal file version 01.02). The hybridization was carried out in a humidified slide chamber. After hybridization, the arrays were washed twice for 4 min in 1 × SSC/0.2% Sodium dodecyl sulphate (SDS), 0.1 × SSC/0.2% SDS and 0.1 × SSC, respectively, at room temperature. The slides were dried by centrifugation and were scanned at 532 nm (for Cy3) and 635 nm (for Cy5) with the GenePix 4000 B microarray scanner (Molecular Devices; MDS Analytical Technologies, Sunnyvale, CA, USA) using auto PMT setting. First, analyses were performed using the GenePix Pro (version 6.1) (Molecular Devices) software including quality control and flagging of bad spots (irregular shaped, outside the spot area). Background subtraction used local background measurements measured immediately next to the spot.
The microarray normalization was carried out using functions from the limma package [version 3.2.1, Bioconductor (http://bioconductor.org/)] (Smyth 2004). The quality of the arrays was evaluated through several diagnostic plots. The ‘normexp’ method (Ritchie et al. 2007) was used for background correction, followed by normalization within individual microarrays using the default ‘print tip loess’ method and normalization between arrays using the ‘quantile’ method. The M-values (i.e. the M-value of a specific gene expression is the log2 ratio of the expression in the investigated animal divided by the expression of the same gene in the common reference sample) for all expressed genes were calculated. Arbitrarily, a gene was considered differently expressed when |M-value| > 1 in at least 11 carcasses. Differential expression was evaluated with a moderated t-statistic using the lmFit and eBayes functions of the limma package. The p-values were adjusted for multiple testing using the Benjamini and Hochberg method (Benjamini & Hochberg 1995) within limma.
Bioinformatics analysis
Gene function analysis of the differentially expressed genes was carried out with the The Database for Annotation, Visualization and Integrated Discovery (DAVID; http://david.abcc.ncifcrf.gov/ home.jsp; Dennis et al. 2003; Huang et al. 2009) software (sixth version). The list of genes was uploaded using the human ENSEMBL gene ID equivalent of the pig genes enabling functional analysis using the human gene annotations. Analysis was carried out using the total human gene content as suggested by DAVID (Dennis et al. 2003; Huang et al. 2009; Te Pas et al. 2010). The human ENSEMBL ID’s were available in the annotation file of the microarray. Only results significant after correction using the Benjamini method (Benjamini & Hochberg 1995) were reported.
Results
Transcriptome profile–differential expression
Comparative transcriptome profiling of the carcasses with the common reference revealed 908 differentially expressed transcripts, 762 showed mainly downregulation and 146 showed mainly upregulation compared to the reference sample. These 908 genes were used for further analysis in DAVID. Table S1 shows the raw data of all differently expressed genes.
Functional physiological analyses
DAVID analysis first revealed that the differentially expressed genes were mainly specific for skeletal muscle tissue (Table 2). This further indicates that the differentially expressed genes were specifically expressed in muscle tissue and may relate to skeletal muscle tissue-specific processes.
| Tissues | Fold enrichment | Benjamini | FDR |
|---|---|---|---|
| Skeletal muscle | 2.8 | 5.20 × 10−09 | 2.73 × 10−08 |
| Epithelium | 1.5 | 8.76 × 10−06 | 9.20 × 10−05 |
| Placenta | 1.4 | 1.26 × 10−04 | 0.002 |
Functional analysis of the differentially expressed genes using the GO term annotations showed that the differentially expressed genes can be functionally grouped: (i) muscle-specific structure genes (muscle structural genes, contractile fibres, muscle contraction), (ii) protein synthesis (including amino acid metabolism, ribosome structure proteins) and (iii) energy metabolism (cellular respiration, oxidative phosphorylation, mitochondrion). Table S2 shows the details for all significant GO terms. Similar results were obtained analysing the protein functional databases (data not shown).
Mining the KEGG database retrieved seven pathways (Table 3). The number of genes with microarray information varied per pathway (Figure 1). Pathways with at least two differentially expressed genes localized on one biochemical path (Te Pas et al. 2007) were further analysed. The number of differentially expressed genes varied from 5 to 29 per pathway. Table 4 shows the genes participating in the pathways. Remarkably, the pathways showed only downregulated genes. These results specify the same three biological functional categories mentioned earlier, but here directly indicating physiological relevant mechanisms: (i) Muscle-specific activity: cardiac muscle contraction pathway and the calcium signalling pathway. In this pathway, Ca2+ ions released to bind to troponin C, resulting in the release of inhibition induced by troponin I. The Ca2+ binding to troponin C thereby triggers the sliding of thin and thick filaments. (ii) Protein synthesis metabolism and the ribosome pathways including a number of ribosome protein (RP) genes. (iii) Energy metabolism pathway: the group consists of five pathways. Among these pathways, the genes participated also in Alzheimer’s disease pathway, Parkinson’s disease pathway and Huntington’s disease pathway. Pathways found in other physiologic pathway databases confirmed and extended the results of the KEGG pathways database analysis. (Table 5).
| Terms | Fold enrichment | Benjamini | FDRa |
|---|---|---|---|
| Ribosome | 7.24 | 4.23 × 10−15 | 3.30 × 10−14 |
| Alzheimer’s disease | 3.47 | 4.93 × 10−6 | 7.69 × 10−5 |
| Parkinson’s disease | 3.73 | 1.24 × 10−5 | 2.89 × 10−4 |
| Oxidative phosphorylation | 3.34 | 2.14 × 10−4 | 6.67 × 10−3 |
| Huntington’s disease | 2.78 | 6.02 × 10−4 | 0.02 |
| Cardiac muscle contraction | 3.62 | 4.86 × 10−3 | 0.23 |
| Glycolysis/Gluconeogenesis | 3.62 | 0.03 | 1.73 |
- aFalse discovery rate.
Number of genes per KEGG pathway.
| Pathways | Genes |
|---|---|
| Ribosome | RPL10, RPL11, RPL12, RPL14, RPL21, RPL24, RPL34, RPL6, RPL27a, RPL35a, RPL36, RPL37a, RPL38, RPS10, RPS13, RPS6, RPS4X, RPS3, RPS21, RPS2, RPS17, RPS19, RPS7, RPS8, RPLP0, RPLP1, RPLP2 |
| Alzheimer’s disease | ATP5G3, ATP5J, ATP5C1, CYTB, COX1, NDUFA1, NDUFA10, NDUFA2, NDUFA3, NDUFB2, NDUFB1, APH1a, CALM1, COX4I1, COX7b, COX7c, COX6a2, COX6b1, CYCS, CYC1, GRIN1, PLCB3, PPP3CB, RYR3, UQCRB, UQCR11 |
| Parkinson’s disease | ATP5G3, ATP5J, ATP5C1, CYTB, COX1, NDUFA1, NDUFA10, NDUFA2, NDUFA3, NDUFB2, NDUFAB1, PARK7, COX4I1, COX7b, COX7c, COX6a2, COX6b1, CYCS, CYC1, GRIN1, PLCB3, UQCRB, UQCR11 |
| Oxidative phosphorylation | ATP5G3, ATP5J, ATP5C1, ATP6V1H, CYTB, COX1, NDUFA1, NDUFA10, NDUFA2, NDUFA3, NDUFB2, NDUFAB1, COX4I1, COX7b, COX7c, COX6a2, COX6b1, CYC1, UQCRB, UQCR11 |
| Huntington’s disease | ATP5G3, ATP5J, ATP5C1, CYTB, COX1, NDUFA1, NDUFA10, NDUFA2, NDUFA3, NDUFB2, NDUFAB1, COX4I1, COX7b, COX7c, COX6a2, COX6b1, CYCS, CYC1, GRIN1, PLCB3, POLR2I, UQCRB, UQCR11 |
| Cardiac muscle contraction | CYTB, COX1, ACTC1, COX4I1, Cox7b, Cox7c, Cox6a2,COX6B1, CYC1,UQCRB, TPM3,TNNC1 |
| Glycolysis/Gluconeogenesis | ALDH3A2, AKR1A1, ALDOA, ENO1, ENO2, FBP2, LDHA, PGAM2, PKM2 |
| Category | Terms | Fold enrichment | Benjamini | FDRa |
|---|---|---|---|---|
| PANTHER_PATHWAY | Glycolysis | 6.63 | 0.04 | 0.49 |
| REACTOME_PATHWAY | 3′ -UTR-mediated translational regulation | 5.22 | 9.66 × 10−14 | 1.65 × 10−12 |
| REACTOME_PATHWAY | Influenza Infection | 3.70 | 7.06 × 10−1 | 2.42 × 10−8 |
| REACTOME_PATHWAY | Metabolism of proteins | 3.15 | 9.66 × 10−1 | 4.97 × 10−8 |
| REACTOME_PATHWAY | Muscle contraction | 7.71 | 7.15 × 10−8 | 4.91 × 10−6 |
| REACTOME_PATHWAY | Gene Expression | 2.20 | 3.48 × 10−6 | 2.99 × 10−4 |
| REACTOME_PATHWAY | Integration of energy metabolism | 2.26 | 4.30 × 10−4 | 0.04 |
| REACTOME_PATHWAY | Diabetes pathways | 1.98 | 1.35 × 10−3 | 0.16 |
- aFalse discovery rate.
Finally, analysis of protein domains and functions confirmed that the above-mentioned groups could also be found in specific protein functions such as acetylation, ribosylation, methylation and phosphorylation of proteins (data not shown).
Discussion
Selection exploits inter-individual genetic variation in the traits. It is expected that the inter-individual variation in the transcriptome profiles is at least partly genetically regulated, although environmental influences such as food and animal handling also affect the transcriptome. Different from variation at the genomic DNA level, variation at the transcriptome level is continuous. This type of variation may be induced by the demand of selection and this type of variation shows a high plasticity. For traits with a low heritability, especially, where the environmental component and the genotype–environment interaction makes up a large part of the variation in the trait, the variation in gene expression may be important for the regulation of the traits. Meat quality is typically such traits. However, because all animals in a batch receive the same treatment, the environmental effects that remain are either similar or different because of the interaction with different genotypes. For future improvements in meat production, it is interesting to know the genetic variation and the related molecular effects on inter-individual transcriptome profiles. Here, we focus on the inter-individual variation in the longissimus muscle of pigs in a final cross-breed and analyse the biological mechanisms related to the differentially expressed genes. It was expected that this will relate to the selection of the pigs. Here, we also realize that all biological activities of most genes are still unknown (even in the best annotated genomes) and that we cannot exclude that selection also can act on other–presently unknown–mechanisms. Therefore, unrelated inter-individual variation in transcriptome profiles may be found. However, by investigating the biological mechanisms of the genes underlying the transcriptome profile variation and relating these biological mechanisms to muscle growth and composition, we select the relevant biological variation.
The differentially expressed genes were classified into three biological processes including muscle-specific structural proteins, protein synthesis and energy metabolism. The expression levels of muscle-specific structural proteins can be expected to represent the muscle mass. More general protein synthesis can also attribute to this, but may also relate to muscle tissue composition. Changes in muscle composition may affect the selection for meat quality improvement. Also, energy metabolism may be involved to this because energy metabolism may be associated with muscle fibre type composition (Simoneau & Bouchard 1989; Stecchini et al. 1990; Klont et al. 1998; Karlsson et al. 1999).
Total protein synthesis in muscle is influenced by both the ribosome content of the tissue and the activity per ribosome (Liu & Hsu 2005). The ribosome pathway may be related to genetic information processing and translation processing, which could affect protein synthesis rate. A high ribosome concentration may be associated with a rapid rate of protein synthesis in LD (Zhao et al. 2003). The increase in protein synthesis may not be attributable solely to an increase in ribosomal and mRNA in muscle. The skeletal muscle protein synthesis varies with fibre type and may be regulated by ribosome number and a specific accumulation of myofibrillar proteins (Liu & Hsu 2005). This suggests that the protein synthesis would be the basic process for muscle-specific structure protein synthesis.
The intensive selection for lean muscle growth in modern high muscled pigs has, over time, induced a shift in muscle metabolism towards a more glycolytic and less oxidative fibre type (Lefaucheur et al. 2004). The muscle fibre type composition is considered an important factor influencing muscle growth and meat quality (Klont et al. 1998; Fiedler et al. 2004). Our data showed differentially expressed genes participating in the glycolysis and oxidative phosphorylation pathway in the mitochondria, indicative for oxidative-type muscle fibre type. Our data suggested that the glycolysis and oxidative phosphorylation would be indicative of levels of fast-oxidative fibre types, which would contain higher intramuscular fat.
Within the energy metabolism group, several neurological disease pathways were also found. This may be due to the involvement of a neurological mechanism, e.g. through the participation of neuronal tissue in muscle tissue composition. Alternatively, it may simply indicate that within these neurological disease pathways, the energy metabolism takes an important position. However, this result may also reflect the use of human annotation of genes (because it is more advanced than the porcine annotation) and be therefore meaningless in pigs.
It is interesting to note that the genes in these pathways all were downregulated as compared to the reference pool. The reason for this is not sure. We observed no animals with massive upregulated gene expression, so outliers influencing the observation were not found. It may indicate that reduced expression relates to improved traits, but the biological explanation for such a conclusion is unknown. In this case, the upregulated genes are just random, but any biological reason for this is lacking. An alternative explanation may be that pathway information for most genes is still lacking. Thus, this could just indicate that this observation just happened by chance.
Finally, similar biological mechanisms were previously reported in a study by Te Pas et al. (2010). This study used a highly different breed (Pietrain versus Duroc cross) differing in muscle mass and muscle composition (Jones 1998; Sellier 1998). The biological mechanisms reported by Te Pas et al. (2010) were highly similar to the mechanisms reported in the present study. These similarities may suggest that these mechanisms may be general mechanisms for the response of the longissimus muscle transcriptome to pig selection for improved meat production. Te Pas et al. (2010) also reported genetic association studies between meat quality traits and expression patterns next to DAVID analysis. It is tempting to speculate on the meaning of the present results for meat quality–especially because both studies showed independently that the differently regulated genes relate to muscle-specific processes, but associations were outside the objectives of the present study. Future studies may explore these aspects in more detail.
In conclusion, we report inter-individual transcriptome variation in pig muscle tissue that could reflect the selection for improved meat production in these animals. Our results suggest that this inter-individual variation may relate to the genetic capacity of the animals and may therefore be used for selection purposes.
Acknowledgements
The authors gratefully acknowledge financial participation from the European Community under the Sixth Framework Programme for Research, Technological Development and Demonstration Activities, for the Integrated Project Q-PORKCHAINS FOOD-CT-2007- 036245. The views expressed in this publication and the experimental design of the study are the sole responsibility of the author(s) and do not necessarily reflect the views of the European Commission. Neither the European Commission nor any person acting on behalf of the Commission is responsible for the use, which might be made of the information. The information in this document is provided as is and no guarantee or warranty is given that the information is fit for any particular purpose. The user thereof uses the information at its sole risk and liability. Furthermore, additional finances were from the ‘Kennisbasis’ (Knowledge Base) grant no KB05-003-02 of the Dutch Ministry of Agriculture, Nature and Food security. SZ was the recipient of a local P.R. China bursary. We acknowledge the supply of microarrays by the contribution of the US Pig Genome Coordinator, Prof. Rothschild, Iowa State University, and the contributions of the swine genome array coordination committee, Dr. Fahrenkrug, University of Minnesota, Chair.
