3.2. Impact of Metabolic Pathways and Involved Metabolites and Proteins for Drip Loss
In this study, metabolite profiling was based on an untargeted metabolomics approach to uncover the whole metabolome. Compared to that, proteins were profiled more specific by means of a targeted proteomics approach using the absolute quantification of 40 proteins that have been shown as important indicators for drip loss in previous investigations. For the final enrichment analysis 128 annotated metabolites and 35 proteins were used. Five proteins were rejected because of missing entrez gene identifier. The drastic reduction of the number of metabolites from 1865 to only 128 is a severe bottleneck, so that it is highly probable that even metabolites with strong influence on drip loss were excluded. This situation is caused by the fragmentary information of biochemical functions of metabolites that is stored in metabolome databases. According to Chagoyen and Pazos [
34], this lack of scientific fundamentals and principles of physiological and biochemical processes of higher life forms is a big challenge in systems biology studies. In a similar way, Chagoyen and Pazos [
34] argued that there is a need of more accurate profiling tools for omic phenotypes in order to get a more comprehensive insight into the metabolic processes.
Our enrichment analysis considered all available annotated metabolome and proteome information and revealed 10 functional KEGG pathways with significant (
p ≤ 0.05) enriched components. The applied test mean-rank gene-set enrichment (MR-GSE) statistic is based on Pearson’s correlation coefficients between metabotypes and drip loss and averages the ranks of the applied statistics instead of the statistics themselves. This procedure makes the results less influenced by individual components in the set of variables [
35] and is the main difference to other usually applied testing procedures, like the Tktest of Tian et al. [
36]. Further details are given by Ackermann and Strimmer [
37].
In summary, it can be expected that the underlying function of our applied enrichment test has enough power to detect overrepresented groups of variables (e.g., genes or metabotypes), even if the effects are very small or the amount of data is not sufficient to detect the important variables individually [
35]. This argument can be used to explain, why our enrichment analysis has resulted in functional sets of metabotypes although correlation coefficients between individual metabotypes and drip loss do not significantly deviate from zero (
Table 1).
In our study, we observed particularly pathways and corresponding key regulators which affect muscle metabolism related to meat quality traits. Glycolysis, pyruvate and methane metabolism are strongly connected and belong to the most important energetic processes that influence the muscle to meat conversion [
38,
39]. Because drip loss strongly depends on p.m. energetic processes in muscle, the meaning of glycolysis and pyruvate metabolism is obvious. After slaughtering, in muscle tissues, anaerobe metabolic processes predominate and, in glycolysis, glycogen is released via glucose to pyruvic acid. Under aerobic conditions, pyruvic acid is metabolized in citrate cycle and oxidative phosphorylation [
39]. In the case of stress before slaughtering, in hypoxic tissues the rate of oxidative processes like glycolysis is increased and pyruvic acid does not flow into glycolysis but is transferred to lactic acid. Accumulation of lactic acid goes along with pH decrease to 5.6 [
40]. The meaning of metabolic processes associated with energy metabolism for drip loss is confirmed by a multitude of studies. Among others, Binke [
41], Scheffler and Gerrard [
39] and D’Alessandro et al. [
26] allocated the relevance of glycolysis and pyruvate metabolism for meat quality. The coincidence of low early pH values and high temperature in muscle lead to partial denaturation of proteins and reduction of intercellular space. Thereby, lipids are dissolved from membranes, permeability of membranes is increased and drip loss is the result [
6]. In cell exudate dissolved lipids clarify the connection between drip loss and activity of sphingolipid metabolism that includes the metabolization of ceramides, phosphoethanolamine and serines. The relation between drip loss and associated lipids and acids has been already described by Lambert et al. [
42] and Poulsen et al. [
43].
As a result of our enrichment analysis, the metabolite glycine is associated with drip loss. In methane metabolism the enzyme glyoxylate transaminase catalyzes the metabolization of metabolite glyoxylate into glycine or hydroxypyruvic acid (
www.genome.jp). High glycine contents indicate a higher rate of glycolytic processes. A high glycolytic potential is known to be related with high drip loss. The link between drip loss and glycine was already described by Lim et al. [
44], who observed higher drip loss in the case of higher glycine level in porcine skeletal muscle cells.
The meaning of PKM that is involved in pathways glycolysis/gluconeogenesis, pyruvate metabolism and type II diabetes mellitus (
Table 2) was already clarified by several studies. For example, D’Alessandro et al. [
26] confirmed that the PKM level appeared to be highly related to many meat quality criteria (WHC, meat color). Beneath PKM, PGAM2 and DG3P are also involved in glycolysis/gluconeogenesis and pyruvate metabolism. Under anaerobic conditions PGAM2 catalyzed the degradation of DG3P to 2-phosphoglycerates (
Figure S2). Because high levels of glycolytic enzymes like phosphoglycerates are associated with increased drip loss [
45], PGAM2 might be considered as an appropriate indicator for drip loss [
46]. In addition, Davoli et al. [
47] appreciated that the corresponding gene PGAM2, is a potential candidate gene for drip loss. The non-essential α-amino acid glycine is also product of catabolism of DG3P and is thus part of the same metabolic process as PGAM2.
Another section of glycolysis/gluconeogenesis illustrates the interactions of the enzymes FBPase and TPI1 and the metabolite dihydroxyacetone phosphate (glycerone-p). In gluconeogenesis FBPase converts fructose-1,6-biphosphate to F6P and in glycolysis phosphofructokinase catalyzes the metabolisation of F6P to fructose-1,6-bisphophate. In the following process of glycolysis, the enzyme fructose-bisphosphate aldolase converts fructose-1,6-bisphophate to glycerone-p. In the next step, glycerone-p is metabolized to glyceraldehyde-3-phosphate catalyzed by TPI1 (see
Figure S2). Laville et al. [
48] revealed a significant correlation between high TPI1 and tender meat with low drip loss. The meaning of FBPase for meat quality in pigs was described by Nam et al. [
49]. They detected a lower FBPase expression in pigs with high drip loss and weak pH decrease p.m. [
49].
Beneath metabolic processes whose activity directly depends on the individual energy resources, also sphingolipid metabolism is significantly associated with drip loss. With a
p-value of 0.014, metabolic compounds in sphingolipid metabolism are the most strongly enriched metabolites and proteins in our study and thereby have an obvious effect on drip loss. According to Heidt et al. [
24] there is a negative correlation between drip loss in DuPi pigs and transcripts associated with sphingolipid metabolism. According to our analysis, the metabolites ceramide, glucosylceramide, phosphoethanolamine and serine are involved in sphingolipid metabolism. Ceramides are lipid signaling molecules that activate proliferative or apoptotic pathways. They are products of the metabolism of free fatty acids to long-chain fatty acyl-CoAs (LCACoAs). LCACoAs can either be used for energy production through β-oxidation or undergo conversion to various signaling molecules, such as ceramide and diacylglycerol [
50]. In the analysis of differentially expressed transcripts in DuPi pigs, Ponsuksili et al. [
10] concluded that low drip loss is associated with ceramide pathways. Especially, drip loss is associated with ceramides as lipid signaling molecules that can activate proliferative or apoptotic pathways. The ceramide biosynthesis is part of the sphingolipid metabolism and ceramides arise from the conversion of complex sphingolipids such as glucosylceramides. According to Dobrowsky and Kolesnick [
51], the levels of ceramides and glucosylceramides and the enzymes regulating their metabolism are associated with the cells response to stress. The degradation of membranes accompanies with cell stress and as a consequence drip loss has a relation to metabolites that indicate cell stress. This connection explains the relationship between drip loss and transformation products of sphingolipid metabolism.
The metabolic processes and their involved components and overlapping are presented in
Figures S2–S5. Several metabolic components, such as glucose and pyruvic acid are involved in five of ten pathways relevant for drip loss. The connective position of these metabolites confirms their specific role as metabolic key players in the regulation of meat quality. The meaning of the disease related pathways (e.g., type II diabetes mellitus) and other processes (meiosis in yeast) for drip loss (
Table 2) are based on the strong influence of specific involved metabolic components like glucose and pyruvic acid. It is not to be expected that there is in fact a physiological connection between meiosis in yeast and meat quality in pigs.
3.3. Significant Markers and Candidate Genes for Drip Loss and Associated Metabolic Traits
Drip loss is a complex trait that is genetically controlled by a variety of different genes [
10] and is influenced by interaction of metabolic processes and participants like genes, transcripts, proteins and metabolites [
6]. Against this background, it is problematic to identify genes with a strong influence on drip loss using classical GWAS approaches. Moreover, statistical problems like stratification within the investigated population increase the risk of false positive results. In order to adjust for population stratification we included PCs as fixed effects into the model of the GWAS procedures as suggested by Aulchenko et al. [
52] and applied among others by Becker et al. [
53] and Utsunomiya et al. [
54]. Depending on the investigated trait (drip loss, protein, metabolite) the models contain 2 to 10 PCs, which lead to λ-values close to one. From these results we conclude a sufficient elimination of population stratification without unacceptable reduction of the genetic variation.
Instead of a Bonferroni correction, that favors the occurrence of false negative associations [
55], we used the q-value which based on the FDR to correct for multiple testing. Storey and Tibshirani [
56] suggested including the FDR in GWAS to provide a better balance between statistical significance and power to detect true effects. As it has been recommended by Benjamini and Hochberg [
57], we set a relaxed significant threshold of
q ≤ 0.10.
The performed GWAS procedures resulted in a varying number of significant SNPs for drip, 11 metabolites and three proteins. The total of 871 significant SNPs are spread across the entire porcine genome, but concentrated on SSC 14, 17 and 18. For drip loss itself, promising candidate genes are located on SSC 18. This region has been earlier described by Jennen et al. [
58] and Liu et al. [
11]. In the region around 12 Mb, the meaning of “
Sus scrofa pleiotropic factor beta” (
PTN) (
q ≤ 6.26 × 10
−2) is highlighted by the direct neighborhood of gene “
cAMP responsive element binding protein” (
CREB3L2)
. CREB3L2 was identified by the GWAS of the protein PGAM2, which revealed an intronic SNP (ALGA0107449) as one of the most significant marker (
Table 5). The family of cAMP response element binding proteins is crucial for a variety of cellular processes including cell proliferation, differentiation, apoptosis, extra-stimuli and stress response [
59]. Although the meaning of
CREB3L2 so far was not precisely described for meat quality, our results suggest that this gene seems to have a relevant influence in energy metabolism in skeletal muscle that is indicated by its interacting effect on PGAM2, glycine and drip loss (
Figure 3).
In the second interesting region on SSC 18 from 15.9 to 16.1 Mb, two intronic SNPs located in the gene “
Leucine-rich repeats and guanylate kinase domain containing” (LRGUK) were found. These SNPs are ranked in the Top 10 list for drip loss as well as for glycine. The nearby gene “
Exocyst complex component 4” (
EXOC4) is also significantly associated with drip loss.
EXOC4 is part of the exocyst complex (Exo70), which is involved in insulin-stimulated glucose transport. Due to Laramie et al. [
60], in humans polymorphisms near
EXOC4 and
LRGUK on chromosome 7 are associated with type 2 diabetes and fasting glucose. The metabolic pathway that is regulated by the polymorphisms near
EXOC4 and
LRGUK potentially is also relevant for drip loss in pork, because fasting glucose also effectsthe pH decrease in muscle p.m. and drip loss. The investigations of Leheska et al. [
61] demonstrated that fasting before slaughtering yielded in a significant lower glucose level and weaker pH decrease in muscle p.m. and in less drip loss. In the third interesting region on SSC 18 around 20 Mb, directly next to each other genes “
Adenosylhomocysteinase-like 2” (
AHCYL2) and “
Smoothened, frizzled class receptor” (
SMO) are located and significantly associated with drip loss. Just like the polymorphism between
EXOC4 and
LRGUK,
AHCYL2 is associated with type 2 diabetes [
62]. Until now, there is no further evidence that this chromosomal region has an influence on meat quality. The effect of gene “
Nuclear factor, erythroid 2-like 3” (
NFE2L3) at 51 Mb, associated with protein PGAM2, fits into the same metabolic background like the previously described genes [
63]. In summary, the multitude of significant SNPs detected for drip loss and associated metabotypes gives an ambiguous indication that in the described regions on SSC 18 promising candidate genes for drip loss can be expected.
In this study, the most significant SNPs were detected on SSC 1. Two SNPs (
p ≤ 2.23 × 10
−5 and
p ≤ 1.59 × 10
−5) associated with glycerone-p and glucosylceramide, are located within the genes “
Ectonucleotidepyrophosphatase/phosphodiesterase 3” (
ENPP3) and “
Sterile alpha motif domain containing 4a” (
SAMD4A).
ENPP3 is associated with lipid and fatty acid metabolism and it has been reported by to Liu et al. [
64] that this gene affects fat deposition and skeletal muscle growth in pigs.
SAMD4A is also associated with lipid metabolism [
65] and influences the metabolisation of glucosylceramides that is part of sphingolipid metabolism. Combining biological knowledge found in literature and the highly significant results of our enrichment analysis leads to the conclusion that the sphingolipid metabolism is one of the most important metabolic pathways associated with drip loss.
Beneath glucosylceramides, phosphoethanolamines are also key players in sphingolipid metabolism. Two genes significantly associated with this metabolite were detected on SSC 6 (
Table 5). “
Phosphatidylinositol 3-kinase, catalytic subunit type 3” (
PIK3C3) is involved in the regulation of hepatic glucose output, glycogen synthase, and antilipolysis in typical insulin target cells such as those in the liver, muscle and fat tissue [
66]. Among others,
PIK3C3 influences the cellular response to glucose starvation (GO term: 0042149). This biological process describes the change in state or activity of a cell (in terms of movement, secretion, enzyme production, gene expression, etc.) as a result of deprivation of glucose. According to Kim et al. [
66] and Hirose et al. [
67] a polymorphism in
PIK3C3 is associated with body weight and carcass fat in Landrace and Duroc pigs.
Moreover, we identified potential candidate genes for several metabolic components involved in glycolysis/gluconeogenesis. The protein PKM is one of the most prominent members of these pathways. The activity of PKM is decreased in the case of low glucose availability in muscle that is positive correlated with anabolic cellular processes. During the conversion of muscle to meat, the metabolic processes change into the catabolic range and if glucose is used up very early, the PKM level is significantly associated with the aberrant glycolysis leading to PSE development [
39]. In our analysis, it was shown that PKM is influenced by six candidate genes on SSC 4. In the chromosomal region of 139 Mb, genes “
Guanylate-binding protein 4” (
GBP4) and “
Protein kinase N2” (
PKN2) are located. Zhao et al. [
68] have identified GBP4 as a significant QTL for lean meat content of pigs by comparing two divergent pig breeds with respect to carcass composition traits. Fontanesi et al. [
69] have reported markers close to
PKN2 that were associated with back fat thickness. SSC 4 harbors two genes
(ZNHIT6, DDAH1) within a region of 142–143 Mb which were significantly associated with average daily gain in Large White pigs [
69]. These polymorphisms seem to have a strong impact on the metabolic rate and the deposition of skeletal muscle mass.
Two SNPs on SSC 17 give evidence that “
Protein tyrosine phosphatase, receptor type” (
PTPRT) and “
VAMP (vesicle-associated membrane protein)—associated protein B and C” (
VAPB) are candidate genes that affect the metabolite DG3P. The protein encoded by
PTPRT is a signaling molecule that regulates a variety of cellular processes including cell growth, differentiation, mitotic cycle, and oncogenic transformation. In humans,
PTPRT is strongly associated with high-fat diet-induced obesity and insulin resistance [
70,
71]. Moreover, in beef cattle, Tiziato et al. [
72] identified
PTPRT as candidate gene for shear force. With respect to the negative correlation between intramuscular fat content and shear force both studies came to homogeneous results. The importance of
PTPRT is additionally indicated by the fact that the most important intronic SNP of
PTPRT is an overlapping SNP that is also significantly associated with protein FBPase (
Figure 1). DG3P and FBPase are strongly connected in glycolysis/gluconeogenesis and PTPRT might be a key player in regulation of glycolysis and thus a promising candidate gene for several meat quality traits.