WHC is influenced by many factors including genetic and environmental effects. We addressed the problem to dissect the genetic background of this complex trait by using the strategy of combining (1) the correlation of expression of many thousands of genes measured simultaneously by microarray technology with quantitative phenotypic data of drip loss, (2) mapping of QTL for the trait drip loss, and (3) mapping of QTL for the expression levels of genes with trait associated expression (Figure 1). QTL analyses provide information suitable to address positional candidate genes whereas trait associated expression studies reveal functional candidate genes. Taking both together, i.e. taking into account the localisation of functional candidate genes in QTL regions, enables to define functional positional candidate genes. Additional insight from eQTL analysis derives from three cases. (1) eQTL are detected within the pQTL but the functional candidate genes itself are located elsewhere, i.e. they are under trans control. These are genes that are likely to be trait dependent expressed due to hierarchically superior genes located in the pQTL that actually represent candidate genes (positional candidate genes). Here the eQTL analysis provides a link between functional and positional candidate genes. In this study 90 functional candidate genes were found with their corresponding 96 eQTL being situated within the previously detected pQTL. These genes point to biological pathways, which are relevant to the trait, and perhaps to causal genes underlying the QTL. However these genes themselves may either be not polymorphic, or the power of the pQTL analysis was not sufficient to detect them, or trait associated expression of these genes is rather an effect than a cause of variation. (2) For functional positional candidate genes being under trans control it can be speculated that the nature of variation affecting the phenotype is differential expression due to polymorphisms in hierarchically superior genes and different responsiveness of the candidate genes to regulatory mechanisms. Here the eQTL combined with pQTL and trait associated expression directs to biological pathways and genes relevant for the trait of interest. In total 66 positional functional candidate genes which corresponded to 119 eQTL with trans mode of expression regulation were found. (3) For genes categorized positional functional candidate genes, mapping of their corresponding eQTL in cis highlights them as genes showing variation with impact on the trait of interest and the expression level, indicating that the nature of the variation is likely a polymorphism in regulatory regions of the gene. Eight genes of this category were identified in this study. These genes are regarded as primary targets for further analysis.
However, it is important to mention that phenotypic variation may be due to genetic variation causing differential expression and/or structural variation of the gene products. The later are not addressed by eQTL analyses. Further, there are functional candidate genes that are under cis mode of transcriptional regulation, where there is no link between eQTL and pQTL analyses.
Trait dependent expression analysis
The association between a quantitative phenotype and gene expression can be examined pair wise using a Pearson correlation coefficient between the expression of a single gene and a continuous phenotype. The approach of trait correlated expression analysis already proofs to be useful by many studies [32–34]. Kraft  used the within-family correlation analysis to remove the effect of family stratification. Here we used general linear models to account for systematic effects of family and environment on both drip phenotypes and expression levels in the correlation analysis. The pre-adjustment of individual phenotypes and expression levels increased the power to identify genetic effects compared to analyses conducted with raw data and revealed biologically meaningful relationship among the traits. Blalock et al. , considered correlation significant at p ≤ 0.05 corresponding to false discovery rates of 20%. In this study, genes were considered for further analysis showing correlation coefficients between gene expression and drip loss of r ≥ 0.37, with p ≤ 0.001 and corresponding q ≤ 0.004.
Biological categories and/or pathways of positively correlated genes
Currently, we do not completely understand the specific biochemical and/or biophysical mechanisms underlying differences in meat water holding capaCity. The processes of muscle conversion to meat occurred in post mortem stage. One possible explanation for some of the variation that exists resides in the structure of the muscle cell itself. Most studies concentrated on post mortem process of drip [35, 36]. In this study, transcript levels of muscle at slaughter were correlated with drip loss at post mortem meat stage in order to reveal insight into the biological processes that are initiated during life and thus are under genetic control and finally determine the liability to develop elevated drip loss. The mechanism underlying this liability trait may also be valid for the (patho-) physiological processes that take place during muscle damage due to biochemical and physical burden at prolonged exercises. Functional annotation analysis is essentially based on the extrapolation of pathway information and gene ontology data of human to the pig. Thus general cellular physiological processes are taken into account, whereas any pig-specific functional annotation data and in particular information on the physiology of the trait drip loss are not addressed during the automated bioinformatics analyses. However, the relevant knowledge has been taken into account in the biological interpretation of the results. The study revealed changes in genomic regulation of multiple cellular pathways that correlate with drip loss. The genes with positive correlation of transcript abundance and drip loss were genes of the group of receptor activity, non-membrane-bound organelle, cytoskeleton, plasma membrane and cell signal. Recently, many studies have shown that degradation of cytoskeleton and other structural proteins plays an important role in drip loss at post mortem [35, 37–39]. As shown in this study the transcript abundance of genes of the cytoskeleton and other structural proteins increased with increasing drip loss. Extra cellular matrix proteins binding integrins and interacting with the cell cytoskeleton are important in controlling many steps in cell membrane-cytoskeleton attachments and in signalling pathways . The degradation of integrin has been suggested to increase the drip channel formation between the cell and cell membrane and thus to be associated with drip loss during post mortem storage on pork [37, 41]. The degradation of integrin may be due to the activity of the calpain system which requires high concentration of calcium for activation . In this study, the enrichment of transcripts of extra cellular matrix receptor pathways among the positively drip correlated genes suggested that WHC may be involved with a breakdown of this extracellular matrix that activate the proteolytic system and thereby promote enzymatic degradation . Calcium signalling pathways are very peculiar in nature. When there is an extracellular change, cells get the message either by introduction of calcium ions into cytoplasm or by evacuation to outside through ion channels. Increase in intranuclear Ca2+ initiates gene expression and cell cycle procession, but also can activate degradative processes in programmed cell death or apoptosis . Gene sets associated with calcium signalling pathways were enriched with decreasing water holding capaCity. For example, epidermal growth factor receptor (EGFR) showed highest positive correlation with drip loss (r = 0.67, p < 0.0001). An early signal generated by the activation of EGFR upon ligand binding is a transient increase in the cytosolic concentration of free calcium ion ([Ca2+]cyt) . Entry of extracellular Ca2+, and Ca2+ release from intracellular stores, both appear to contribute to the generation of the EGF-mediated [Ca2+]cyt spike [45–47]. Early post mortem higher Ca2+ concentration causes rapid contraction, an increase in the rate of muscle metabolism, and accelerated pH decline with resulting higher drip . Another hypothesis is that higher Ca2+ concentration present in muscle fibres early post mortem is a source for the activation of Ca2+ dependent protease, phosphatases and phospholipases like the calpain system which influences drip production. Increased cytoplasmic Ca2+ levels are also observed due to excessive exercises. This may initiate vicious cycles of cell degradation because of the Ca2+ dependent activation of proteolytic enzymes such as calpain that by themselves digest structural elements of the muscle fibres leading to membrane damage, leakage of intracellular water and proteins and further accumulation of Ca2+ . Together, increase in transcript levels of genes involved in cytoskeleton, and extracellular matrix receptor pathways as well as calcium signalling pathways in muscle play an important role in final meat quality.
Biological categories and/or pathways of negatively correlated gene
Though the energy metabolism is crucial for muscles, the biochemical processes involved in the change from aerobic metabolism ante mortem to anaerobic metabolism post mortem, which associates to drip loss, is not much investigated. The negatively correlated transcripts were enriched in mitochondrion, transporter activity and protein metabolism GO categories as well as oxidative phosphorylation pathway. A dominant role of mitochondria is the production of ATP by several different biochemical routes, i.e. via aerobe glycolysis and via oxidative phosphorylation. At the pre-slaughter stage in living animals with the presence of oxygen, aerobic processes take place. When oxygen is limited (post mortem) the glycolytic products will be metabolised by anaerobic respiration, a process that is independent of the mitochondria. A shift from aerobic to anaerobic metabolism - favouring the production of lactic acid - results in a pH decline post mortem and thereby influence the water holding capaCity in muscle . In our study, 63 transcripts belong to mitochondrion GO category and 20 transcripts belong to the oxidative phosphorylation pathway. The negative correlations with drip loss may indicate reduced activity of biochemical processes of ATP production via oxidative pathways in mitochondria of animals with high drip loss, reduced number of mitochondria in their muscle, i.e. higher content of glycolytic fibers, or reduced ATP reserves in the muscle.
Together, analysis of trait correlated expression revealed that the complex relationships between biological processes taking place in live skeletal muscle and meat quality are driven on the one hand by the energy reserves in the muscle and their metabolisation as well as on the other hand by the muscle structure itself.
cis and trans mode of regulation of gene expression in QTL regions for WHC
Expression-QTL for genes showing high correlation with the phenotype may provide the necessary information required for identifying genes that control quantitative phenotypes. Categorizing eQTL has the potential to enable reverse genetics approaches for the identification of genes controlling quantitative traits, and may also help to enhance the rate of QTL cloning . In particular, if the pQTL for drip loss were caused by interstrain differences of gene expression, the genetic determinants responsible for the pQTL would be restricted to the genes that were encoded inside the pQTL region and provide variations of gene expression under cis acting transcriptional fashion in the F2 population. In this case, their eQTL were found to reside at the same chromosomal positions at which they were encoded and the lod score curves with the peak of eQTL should coincide with those of the pQTL. Local eQTL where expression phenotypes map to the genes themselves, are of great interest, because they are direct candidates for previously mapped pQTL.
Many investigations have reported the successful mapping of quantitative trait loci for gene expression phenotypes (eQTL) in rat or mice [51–53]. Such genetical genomics analyses in livestock are still scarce. Among livestock species, poultry is well placed to embrace this technology. De Koning et al.  identified the cis and trans effects for a functional body weight QTL on chicken chromosome 4 in breast tissue samples from chickens with contrasting QTL genotypes. Kadarmideen and Janss  presented a comparative systems genetic analysis on the physiology of cortisol levels in mice and pigs with the aim to show the potential of a comprehensive computational approach to quickly identify candidate genes. Here, the first expression QTL study is presented performed in a segregating pig population with focus on the trait drip loss. In a first step we analysed the correlation between trait dependent gene expression and the phenotype drip loss, which revealed biologically meaningful relationship. In the second step, eQTL were identified for transcripts that showed trait correlated expression, which supplies us with information about the genomic location of putative regulatory loci. This strategy reduces the number of several thousand eQTLs which were not associated with drip loss. The trans acting eQTL represent transcripts whose abundance is regulated by loci remote from the genomic locus of each of these genes. In our study the proportion of trans eQTL is higher (92%) than in other studies (60%–65%) [51, 56]. Here eQTL analysis was focussed on functional positional candidate genes for a trait that varies in degree, i.e. the study was driven by transcriptional and positional restrictions on the genes analysed. A network of genes relevant to the traits was addressed representing additive and pleiotropic as well as non-additive epistatic effects on the trait. This may lead to higher proportion of trans regulated genes compared to studies were eQTL were identified independent from any positional restrictions on the corresponding genes. Cis acting eQTL serve as an important new resource for the identification of positional candidates in QTL studies. We detected 8 out of 104 genes acting in cis, whereas Yashimita et al.,  and Dumas et al.,  reported 9 out of 13 genes and 1 out of 5 genes, respectively, acting in cis.
Candidate genes for WHC
The candidacy of cis regulated functional positional candidate genes has three-fold experimental evidence. In particular for AHNAK a number of reasons for its candidacy for drip loss have been put forward: (i) This gene is located in the SSC2 QTL region for drip loss as confirmed by RH-mapping. The pQTL for drip in this region was also found in other studies [14, 16, 17]. (ii) The correlation between drip loss and AHNAK is high (r = 0.53; p < 0.0001). (iii) The eQTL for AHNAK indicates a cis acting mode of regulation with genome wide significance (lod score = 6.4; F = 18.2). Real time RT-PCR performed for AHNAK support the microarray data in terms of trait correlated expression. Also significant correlation was observed of expression values obtained from microarrays and real time RT-PCR, respectively. Further, eQTL analysis of real time RT-PCR data matches those of microarray data. AHNAK is a functional candidate gene due to its role in muscle contraction, cell adhesion and proliferation as well as its interaction with calcium. AHNAK, a nuclear phosphoprotein with the estimated molecular mass of 700 Da, is expressed in all muscular cells [58, 59]. AHNAK is implicated in calcium flux regulation. At low calcium concentrations, AHNAK proteins are mainly localized in the nucleus, but the increase in intracellular calcium levels leads the protein to translocate to the plasma membrane . AHNAK relocates from the cytosol to the cytosolic surface of the plasma membrane during the formation of cell-cell contacts . The main localization of AHNAK is at the plasma membrane in adult muscle cells . AHNAK contains three distinct structural regions: the NH2-terminal 251-amino acid region, a large central region of about 4300 amino acids with 36 repeated units, and the COOH-terminal 1002 amino acids region. The carboxyl-terminal region of AHNAK proteins mediates cellular localization and interaction with L-type Ca2+ channels, calcium-binding S100B protein, as well as actin of thin filaments for muscle contraction [62–64].
MAP4K4 a member of the serine/threonine protein kinase family is involved in MAPK signalling for cell proliferation and differentiation as response to stressors and in cell adhesion via integrin beta 1 [65, 66]. Here MAP4K4 appeared as a prominent candidate for drip loss. MAP4K4 expression is induced by TNF-alpha and promotes insulin resistance , whereas silencing of MAP4K4 prevents insulin resistance in human skeletal muscle and enhances glucose uptake . This evidence promotes our finding of a positive correlation of MAP4K4 with drip loss. Reduced MAP4K4 expression, promotes glucose uptake, therefore increasing glucose content in muscle cells. By increasing energy depots in the muscle prior to slaughter, the anaerobic production of lactate post mortem may be delayed, thereby delaying of decline in pH and reducing drip loss.
Candidacy of SLC3A2 was confirmed by real time PCR. SLC3A2 is a member of the solute carrier family and encodes a cell surface, transmembrane protein. It associates with integrins and mediates integrin-dependent signalling related to normal cell growth. Information about function of BBC2, PRCC, USP39, LOC162073 and COQ9 are too limited to allow deducing functional links to the trait drip loss or other candidate genes for this trait.