Meat quality traits in a commercial Duroc pig population are affected by genetics but not by Nuclear Magnetic Resonance-based metabolite profile measured in blood

Abstract

In the present study, single nucleotide polymorphism (SNP) markers and meat quality traits on animals from a commercial Duroc sire line were assessed, aiming to identify phenotypic and genetic variation within this line. In addition, it was assessed if a nuclear magnetic resonance (NMR) spectroscopy-based blood metabolite profile obtained from a blood sample taken at the end of the performance test could be used as an indicator for meat quality prior to slaughter. The results of the genetic analysis showed that the eight meat quality traits were lowly to moderately heritable: pH (0.41), meat color Minolta L* (0.24), Minolta a* (0.19), and Minolta b* (0.24), myofibril fragmentation index (0.10), intramuscular fat (0.16), Warner-Bratzler shear force (0.14) and cooking loss (0.10). A genome-wide association study (GWAS) showed that 2 SNP markers marked a quantitative trait locus (QTL) segregating on SSC2 for cooking loss (false discovery rate (FDR) < 0.10), while no QTL were discovered for the other meat quality traits. The NMR-based metabolome could explain part of the total variance for traits Minolta L* (0.11) and Log₁₀(MFI) (0.31), respectively. However, when accounting for the contemporary environment, the blood metabolome profiles explained almost no variance for any of the meat quality traits measured in this study, and therefore, the blood metabolite profile is not a good indicator trait for meat quality in this Duroc sire line.

This study explored how genetics and nuclear magnetic resonance (NMR) spectroscopy-based blood metabolite profiles impact meat quality in a Duroc sire line. While traits proved heritable, only one QTL on SSC2 for the meat quality trait cooking loss emerged. Additionally, NMR-based blood metabolite profiles were not reliable indicators for meat quality.

Graphical abstract

Graphical Abstract.

Introduction

In 2023, the EU produced 20.83 million metric tons of pork, accounting for 18% of global production, with Denmark among the top producers, generating around 1.3 million metric tons (https://www.fas.usda.gov/data/production/0113000). Modern pork production focuses on traits such as growth, lean muscle tissue, and feed efficiency (Tiezzi et al 2025). However, EU consumers are increasingly willing to pay for higher-quality cuts, which are more tender, juicier, and flavorful (Aaslyng and Meinert 2017). This shift emphasizes the importance of carcass composition and meat quality in enhancing pork’s economic value.

Genetic selection in commercial pig breeding programs typically targets economically significant traits. However, improving meat quality traits is challenging, as these traits—such as pH, tenderness, intramuscular fat (IMF), and meat color—are postmortem measurements. Thus, genetic markers or live animal indicator traits could aid genetic selection of parents for future generations. Scientific studies suggest that meaningful breed differences in meat quality exist and that meat quality is heritable, with varying heritability levels reported (Gjerlaug-Enger et al 2010; Miar et al 2014; Cabling et al 2015). Genome-wide association studies (GWAS) have identified over 30,000 quantitative trait loci (QTLs) linked to meat and carcass traits (Hu et al 2022). Additionally, candidate genes related to meat quality have been pinpointed in various pig breeds (Plastow et al 2005; Mohammadabadi et al 2021).

To identify potential biomarkers for meat quality, nuclear magnetic resonance (NMR) metabolomics could offer a promising approach, as it enables unbiased analysis of biological samples without altering them. NMR has shown its ability to assess meat quality in different pig breeds (Straadt et al 2011). In 2018, the Danish slaughter pig population exhibited low variability in meat quality traits, a factor both advantageous and limiting for genetic selection (Aaslyng and Hviid 2020). This study explores the DanBred Duroc sire line’s meat quality using genomics and NMR metabolomics, aiming to estimate genetic parameters, identify QTLs, and assess NMR-based blood profiles as potential indicator traits for meat quality.

Material and methods

The raising of the animals was conducted following Danish laws of management and welfare procedures for pig production. For taking the blood samples, an animal research permit was obtained from the Danish Veterinary and Food Administration. Meat quality phenotypes were recorded at slaughter, requiring no special permission. Humane slaughter practices were followed according to Danish legislation.

We examined eight different meat quality traits measured on DanBred Duroc males. For each meat quality trait, we estimated genomic, metabolomic, and environmental variances using REML and performed GWAS studies. For each pair-wise combination of traits, we estimated genetic and phenotypic correlations between the traits.

Animals

In total, 971 DanBred Duroc boars from the DanBred breeding program were used in this study. Production traits were recorded for these boars at a test station, where high-index males were selected for breeding, and others were slaughtered for meat quality analysis. Born on 11 farms, the male piglets were transported to the test station (Bøgildgaard, Kjellerup, Denmark) at 7–10 kg and fed an adapted version of the Danish nutrient standard for purebred pigs (https://pigresearchcentre.dk/Nutrient-standards, accessed 15 April 2024). At the station, they were housed in separate sections, with each section sent to the same slaughterhouse on the same day. The animals were slaughtered across three slaughterhouses over 53 d from October 2020 to May 2022, with an average live weight of 106.64 ± 10.65 kg.

Phenotypes

Pigs are driven forward to the stunner in small groups at a time. The pigs are stunned with a high concentration of CO₂ (80%–90%), followed immediately by bleeding to ensure death while unconscious. The carcass is stored at 4 °C during hanging. At 3 d postmortem a 10 cm piece of the M. longissimus lumborum was taken from the 13–15th rib and stored at 5 °C for 1 d.

At 4 d postmortem, the following meat quality parameters were measured in the lab at the Department of Food Science, Aarhus University (Denmark):

The pH was measured using an electrode (Metrohm 826 mobile pH meter, Metrohm Nordic ApS, Glostrup, Denmark) calibrated at 20 °C using pH 4.005 and 7.000 IUPAC-buffers. The pH was determined in duplicate, and the average of the two pH measurements was used in the analysis. At the time of measurement, the meat temperature was 5 °C.

Meat color (CIE 1976 L*, a*, b*) was measured after 1 h of blooming at five sites per sample using a Minolta Chroma Meter CR-300 (Minolta, Osaka, Japan), calibrated against a white plate (L* = 92.30, a* = 0.32, b* = 0.33), and the average lightness (Minolta L*), redness (Minolta a*), and yellowness (Minolta b*) values were recorded. The color phenotype used in the analysis was based on an average of five measures.

IMF was determined based on a 10 g meat sample (1 cm cross-section of the M. longissimus lumborum) using the Weibull-Stoldt method (Stoldt 1949). The meat sample was exposed to an acid hydrolysis using 4M HCl in a Hydrotherm (C. Gerhardt GmbH & Co. KG, Königswinter, Germany), whereafter the fat extraction was performed using petroleumether on a Soxtherm (C. Gerhardt GmbH & Co. KG, Königswinter, Germany). IMF was expressed as a percentage of the fresh meat sample.

Myofibril fragmentation index (MFI) was determined as described by Lametsch et al (2007) with some modifications. In short, 2.5 g meat was taken and homogenized in 25 mL buffer (100 mM KCl, 20 mM potassium phosphate [pH 7], 1 mM EGTA, 1 mM MgCl₂, and 1 mM NaN₃) and centrifuged 2 times for 30 s at 20,500 rpm (1,000 g) (Ultra-Turrax T25, Ika Labortechnik, Staufen, Germany) on ice. Between homogenizing steps, the connective tissue was removed. The size of myofibril fragments was determined using a Malvern Mastersizer 2000S (Malvern instruments Ltd, Worcestershire, UK). The surface area mean diameter of the fragments was determined and used as an expression for MFI. The particle size calculation model is based on spherical particles; consequently, the measured sizes of the myofibrils should be regarded as apparent rather than actual dimensions (Lametsch et al 2007). The smaller the fragments, the more muscle protein is degraded in the sample, thereby increasing the tenderness of the meat (Lametsch et al 2007).

Cooking loss (CL) was determined as described by Oksbjerg et al (2019). At 4 d postmortem, meat samples were frozen until further use. Frozen meat was thawed at 5 °C overnight in a fridge. Thereafter, the meat sample was cut into pieces of 5 cm × 8 cm × 4 cm, weighted, and vacuum packed. These were then heated in a water bath at 70 °C for 60 min, unpacked, pat dried and weighted for the determination of the CL (%).

After the CL determination, the same samples were used to perform the Warner-Bratzler shear force (WBSF) test. Measurement of max WBSF was done as described in Hansen et al (2006) using a texture analyzer (Stable Micro System Ltd., Godalming, UK) equipped with a Warner-Bratzler shear blade with a rectangular hole, 11 mm wide and 15 mm high, and a blade thickness of 1.2 mm (Honikel 1998). The maximum shear force sheared across the fiber direction was recorded at a test speed of 50 mm/min according to the procedure described by Honikel (1998). Six repetitions were made, and the reported maximal shear force is an average of these six repetitions.

Genotypes

The animals were genotyped as part of the DanBred breeding program using a 50k GGP-Porcine Illumina Bead SNP array (https://www.neogen.com/categories/genotyping-arrays/ggp-porcine-50k/) at Neogen (Scotland, UK) according to laboratory practice. To ensure a 100% accurate link between the meat sample and animal ID, both a blood sample from the animal as well as the meat sample were genotyped. Quality control and selection of SNP was part of the DanBred system, and was as follows: SNP were selected when the GC-score > 60% and a call rate > 80%; filtering of the SNPs was performed using Plink version 1.07 (Purcell et al 2007) with the following settings for removal of SNP markers: minor allele frequency (MAF) < 0.05, Hardy-Weinberg equilibrium test (HWE) P <= 1e⁻⁵, and missingness test (GENO > 0.1). After frequency and genotyping pruning, 25,318 SNP markers for 921 animals remained for analysis.

The SNP markers were mapped based on the Sscrofa11.1 map locations (https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_000003025.6/). To check if the SNP genome location corresponded to an annotated gene location, the R package BiomaRt version 2.52.0 (Durink et al 2005) was used.

NMR spectroscopy

A detailed description of the NMR spectroscopy was presented in Guo et al (2025). In short, the blood samples for NMR analysis were typically taken 8 d before the animal was sent for slaughter but in some cases up to 29 d. The samples were sent in four batches to the Swedish NMR Centre at the University of Gothenburg, Sweden, in the period mid-2021 to early 2023.

Thawed blood plasma samples were mixed with a buffer (75 mM sodium phosphate pH 7.4, 0.1% sodium azide, 20% v/v D₂O, 0.08% TSP-d4) using a Bruker BioSpin robot (Ettlingen, Baden-Württemberg, Germany) at 2 °C. Bruker In Vitro Diagnostics for research (IVDr; Bruker BioSpin) standard operating procedures were used to acquire the ¹H-NMR data (Dona et al 2014) using a 600 MHz Bruker Avance III HD spectrometer (https://www.bruker.com/en/products-and-solutions/mr/nmr/avance-nmr-spectrometer.html, accessed 16-09-2024). NMR-tube racks were kept at 6 °C before data acquisition.

The 32,697 NMR intensity spectral data spanned a frequency range from 0 to 10 ppm and was transformed using a designed MATLAB script (Haggart 2019).

The NMR data was preprocessed by applying an exponential apodization function, which emulated a line broadening of 0.5 Hz, followed by Fourier transformation. Thereafter, the data were aligned using the glucose anomeric proton doublet (at ∼5.23 ppm) as a reference, phased, and a baseline correction was applied. Followed by removing the water peaks (4.4 to 4.9 ppm) and the region (−0.04 to 0.04 ppm) which is associated with an added standard. The final NMR dataset consisted of 28,119 intensities (ie, metabolic features [MF]).

These 28,119 intensities were subjected to the probabilistic quotient method (Dieterle et al 2006) followed by the icoshift algorithm for the final spectral alignment (Savorani et al 2010; Vu and Laukens 2013).

Statistical analysis

Explorative analysis of the data was performed using R version 4.2.0 (https://www.r-project.org/). The phenotypic correlations between meat quality traits are Pearson correlations using the cor.test() function in the stats package 4.2.0 in R (version 4.2.0; https://www.r-project.org/).

Metabolic similarity matrix

The relationship among the pigs based on their similarity in blood NMR profile (ie, including all MF) was estimated by constructing an NMR based similarity matrix (M). This matrix is computed as a variance-covariance matrix from the NMR profiles:

$$\mathbf{M}=\frac{X{X}^{\mathrm{\prime }}}{n}$$ (1)

Where $X$ is the centered and scaled NMR intensities and $n$ is the number of MF.

Variance estimation

To estimate the variance components and genetic parameters, the software DMU (Madsen and Jensen 2023) was used based on the following model:

$$Y_{\mathit{ijklm}}={\mathit{Slaugtherhouse}}_i+\mathit{Liv}{\mathit{eweight}}_j+{\mathit{Section}}_k+{\mathit{Animal}}_l+{\mathit{Metabolite}}_m+e_{\mathit{ijklmn}}$$ (2)

where $Y_{\mathit{ijklm}}$ is the meat quality phenotype, Slaughterhouse_i is the fixed effect (i = 1, …, 3), Liveweight_j is the weight (kg) of the live animal l at the time of slaughter fitted as a fixed regression effect, Section_k is a random contemporary group effect (k = 1, …,77) and represents which animals are in the same stable at the same time, Animal_l is a random effect of the animal l based on the pedigree relationships, Metabolite_m represents the NMR based metabolite effect for Animal_l following a distribution $N(0,\mathbf{M}{\sigma }_m^2)$, where ${\sigma }_m^2$ is the NMR variance and $\mathbf{M}$ is the NMR based similarity matrix between animals (model 1). Finally, $e_{\mathit{ijklmn}}$ is the random residual error.

Reduced models were applied to evaluate the impact of the metabolites and section on the meat quality traits:

$$Y_{\mathit{ijk}}={\mathit{Slaugtherhouse}}_i+\mathit{Liv}{\mathit{eweight}}_j+{\mathit{Metabolite}}_k+e_{\mathit{ijk}}$$ (2a)

$$Y_{\mathit{ijkl}}={\mathit{Slaugtherhouse}}_i+\mathit{Liv}{\mathit{eweight}}_j+{\mathit{Section}}_k+{\mathit{Metabolite}}_l+e_{\mathit{ijkl}}$$ (2b)

Where the effects are defined as in Model 2. Model 2a excludes the section and animal effects, whereas Model 2b includes the section effect but excludes the animal effect. The pedigree contains 6,674 animals traced back 10 generations.

Heritability is computed as

$$h^2=\frac{{\sigma }_A^2}{({\sigma }_{\mathit{Section}}^2+{\sigma }_A^2+{\sigma }_m^2+{\sigma }_e^2)},$$ (3)

where ${\sigma }_A^2$ is the estimated additive genetic variance, ${\sigma }_{\mathit{section}}^2$ is the estimated section variance, ${\sigma }_m^2$ is the estimated NMR variance, and ${\sigma }_e^2$ is the estimated residual error variance.

The proportion of phenotypic variance explained by the metabolites ($m^2$) was calculated for the three models as:

$$m_{\left(2\right)}^2=\frac{{\sigma }_m^2}{({\sigma }_m^2+{\sigma }_A^2+{\sigma }_{\mathit{section}}^2+{\sigma }_e^2)},$$ (4)

$$m_{\left(2a\right)}^2=\frac{{\sigma }_m^2}{({\sigma }_m^2+{\sigma }_e^2)},$$ (4a)

$$m_{\left(2b\right)}^2=\frac{{\sigma }_m^2}{({\sigma }_m^2+{\sigma }_{\mathit{section}}^2+{\sigma }_e^2)},$$ (4b)

Similarly, the proportion of phenotypic variance explained by the section effect (${\mathit{section}}^2$) was calculated as:

$${\mathit{section}}_{\left(2\right)}^2=\frac{{\sigma }_{\mathit{section}}^2}{({\sigma }_m^2+{\sigma }_A^2+{\sigma }_{\mathit{section}}^2+{\sigma }_e^2)},$$ (5)

$${\mathit{section}}_{\left(2b\right)}^2=\frac{{\sigma }_{\mathit{section}}^2}{({\sigma }_m^2+{\sigma }_{\mathit{section}}^2+{\sigma }_e^2)},$$ (5a)

where the variance components are defined as in Model (2).

The genetic correlations are estimated in two-trait models, where the sub-models for each of the traits are as in model 2, but with the random metabolite effect omitted from the model.

Genome-wide association mapping

The GWAS analysis was performed based on the following model:

$$Y_{\mathit{ijklm}}={\mathit{Slaugtherhouse}}_i+{\mathit{SNP}}_j+\mathit{Liv}{\mathit{eweight}}_k+{\mathit{Section}}_l+{\mathit{Animal}}_m+e_{\mathit{ijklmn}},$$ (6)

where the variables are as described in model 2, SNP_j is a covariate (0,1,2), ie, an additive allelic effect. The GWAS analysis was performed using Restricted Maximum Likelihood (REML) implemented in software DMU (Madsen and Jensen 2023). The test for each SNP was a Wald test:

$$W^2=\frac{{(\widehat{\beta }-{\beta }_0)}^2}{\mathit{VAR}\left(\widehat{\beta }\right)}\sim {\chi }_1^2,$$ (7)

where $\widehat{\beta }$ is the REML estimate of the allele effect, and ${\beta }_0$ = 0 is our null hypothesis, $\mathit{VAR}\left(\widehat{\beta }\right)$ is the asymptotic variance of the REML estimate, and ${\chi }_1^2$ is the Chi-Square distribution with 1 degree of freedom. A SNP association was considered significant at the false discovery (FDR) level of q < 0.10 (Storey et al 2022). Proportion of variance explained by the SNP marker (PVE) was calculated as:

$$PE=\frac{2{\widehat{\beta }}^{2}MAF(1-\mathit{MAF})}{2{\widehat{\beta }}^{2}MAF\left(1-\mathit{MAF}\right)+{SE\left(\widehat{\beta }\right)}^{2}2N\mathit{MAF}(1-\mathit{MAF})},$$ (8)

where $\widehat{\beta }$is the estimated SNP effect size, MAF is the minor allele frequency of the SNP, $SE\left(\widehat{\beta }\right)$ is the standard error of the estimated SNP effect size, and N is the sample size.

Results

Summary statistics

Transformation of the meat quality phenotypes were explored and based on QQ-plots, it was decided to use a log₁₀ transformation for MFI and WBSF (data not shown). The meat quality traits with the lowest coefficient of variation (CV) were pH with 2.3%, Minolta L* with 4.6%, and CL with 6.2%. The traits with the highest CV were IMF, Minolta b*, and log₁₀(MFI) with 27.9%, 19.1%, and 18.27%, respectively (Table 1). When applying a log₁₀ transformation on MFI and WBSF the CV was reduced from 61.7% to 18.27% for MFI and from 40.4% to 9.96% for WBSF (data not shown).

Table 1.

Mean, standard deviation, coefficient of variation, minimum, maximum, number of animals, heritability (h²), variance explained by the blood metabolites (m²), and variance explained by section (section²) for the meat quality traits measured in the Danbred Duroc pigs¹.

Trait	Mean	SD	CV, %	Minimum	Maximum	number of animals	h2 (SE)	${\mathit{m}}_{\left(2\right)}^2$  (SE)	${\mathbf{section}}_{\left(2\right)}^2$  (SE)
pH	5.59	0.13	2.3	5.29	6.52	957	0.41 (0.14)	0.00 (0.00)	0.16 (0.04)
Minolta L*	56.83	2.60	4.6	47.99	65.46	955	0.24 (0.09)	0.00 (0.00)	0.32 (0.08)
Minolta a*	6.25	1.11	17.8	1.71	10.20	955	0.19 (0.09)	0.02 (0.02)	0.25 (0.06)
Minolta b*	5.82	1.11	19.1	2.67	9.89	955	0.24 (0.09)	0.01 (0.01)	0.25 (0.07)
Log10(MFI, %)	1.27	0.23	18.27	0.80	2.06	945	0.10 (0.04)	0.00 (0.00)	0.57 (0.15)
IMF, %	1.65	0.46	27.9	0.25	5.75	926	0.16 (0.09)	0.02 (0.02)	0.02 (0.02)
Log10(WBSF, N)	1.51	0.15	9.96	1.17	2.05	707	0.14 (0.09)	0.01 (0.01)	0.32 (0.10)
CL, %	32.45	2.02	6.2	17.20	38.07	710	0.10 (0.09)	0.00 (0.01)	0.05 (0.03)

¹SE between brackets.

Heritabilities and variances explained by NMR and section

Table 1 shows the heritabilities and variances explained by NMR and section for the meat quality traits. For all analyses, the REML procedure converged within 200 steps. The highest heritability was found for pH (0.41) and the color measures Minolta L* and Minolta b* (both with a heritability of 0.24). Lowest heritability was found for log₁₀(MFI) (0.10) and CL (0.10). Heritability for IMF, Minolta a* and WBSF were 0.16, 0.19, and 0.14, respectively. Variance explained by section was highest for log₁₀(MFI) (0.57) and lowest for IMF (0.02) and CL (0.05). Variance explained by the metabolome was either zero or very close to zero for all the meat quality traits using model 2 (Table 1).

To test the influence of the blood metabolites and section on the meat quality, we analyzed the data using model 2a and model 2b with either including the metabolite relationship matrix or a combination of the relationship matrix and section, respectively (Table 2). Results of model 2a showed that the metabolome could explain part of the total variance for traits Minolta L* (0.11) and Log₁₀(MFI) (0.31), while for the other meat quality trait, the m² was not significant (Table 2). When including a section in the model (Model 2b) NMR-based metabolome variance components went to zero. The section explained part of the total variance for traits pH (0.18), Minolta L* (0.32), Minolta a* (0.25), Minolta b* (0.26), log₁₀(MFI) (0.58), and log₁₀(WBSF) (0.32), respectively.

Table 2.

Variance explained¹ by the blood metabolites (m²), variance explained by section (section²) for the meat quality traits measured in the Danbred Duroc pigs with either including the metabolite relationship matrix or a combination of the relationship matrix and section, respectively.

Results model 2a
Trait	${\mathit{m}}_{\left(2\mathit{a}\right)}^2$ (SE)	Section2 (SE)	Variancemetabolite (SE)	Variancesection (SE)	Varianceerror (SE)
pH	0.01 (0.01)		1.48e-4 (0.01)		0.02 (0.13)
Minolta L*	0.11 (0.05)		0.71 (0.84)		5.82 (2.41)
Minolta a*	0.08 (0.04)		0.09 (0.29)		1.01 (1.00)
Minolta b*	0.08 (0.04)		0.09 (0.30)		1.02 (1.01)
Log10(MFI, %)	0.31 (0.08)		0.01 (0.11)		0.03 (0.17)
IMF, %	0.03 (0.02)		0.01 (0.09)		0.21 (0.46)
Log10(WBSF, N)	0.04 (0.03)		6.74e-4 (0.23)		0.01 (0.12)
CL, %	0.01 (0.02)		0.06 (0.25)		3.99 (1.99)
Results model 2b
Trait	${\mathit{m}}_{\left(2\mathit{b}\right)}^2$  (SE)	${\mathbf{section}}_{\left(2\mathit{b}\right)}^2$ (SE)	Variancemetabolite (SE)	Variancesection (SE)	Varianceerror (SE)
pH	0.00 (0.01)	0.18 (0.05)	3.91e-6 (8.36e-5)	2.93e-3 (7.22e-4)	0.01 (9.69e-4)
Minolta L*	0.00 (0.01)	0.32 (0.08)	2.97e-6 (0.04)	2.15 (0.45)	4.51 (0.23)
Minolta a*	0.03 (0.02)	0.25 (0.07)	0.04 (0.03)	0.28 (0.06)	0.80 (0.04)
Minolta b*	0.02 (0.02)	0.26 (0.07)	0.02 (0.02)	0.29 (0.07)	0.84 (0.04)
Log10(MFI, %)	0.00 (0.00)	0.58 (0.15)	1.24e-4 (1.89e-4)	0.03 (4.86e-3)	0.02 (9.56e-4)
IMF, %	0.03 (0.02)	0.02 (0.02)	0.01 (0.01)	4.39e-3 (3.66e-3)	0.21 (0.11)
Log10(WBSF, N)	0.01 (0.01)	0.32 (0.10)	9.71e-5 (1.93e-4)	0.01 (1.31e-3)	0.01 (6.45e-4)
CL, %	0.01 (0.02)	0.05 (0.03)	0.05 (0.09)	0.21 (0.11)	3.81 (0.23)

¹SE between brackets.

Phenotypic and genetic correlation

Table 3 shows the genetic and phenotypic correlations between the meat quality traits. Genetic correlations were almost zero between Minolta L* and Minolta a* (−0.01), Minolta L* and log₁₀(WBSF) (0.02), Minolta a* and log10(MFI) (0.02), Minolta a* and CL (−0.01), Minolta b* and log₁₀(WBSF) (0.002), as well as between log₁₀(MFI) and CL (−0.01). In addition, there were negative genetic correlations between pH and Minolta L* (−0.77), Minolta b* (−0 to 66), and CL (−0 to 72), respectively, and a positive genetic correlation between pH and log₁₀(MFI) (0.43). In general, when both the genetic and phenotypic correlations were statistically significant from zero, they had the same sign and similar magnitude. An exception to this was the correlation between log₁₀(MFI) and log₁₀(WBSF), where the genetic correlation (0.41) and the phenotypic correlation (0.14) had the same sign but differed in magnitude.

Table 3.

Genetic correlation¹ (above the diagonal), and phenotypic correlation (below the diagonal) between the meat quality traits measured in the Danish Duroc pigs.

Trait	pH	Minolta L*	Minolta a*	Minolta b*	Log10(MFI)	IMF	Log10(WBSF)	CL
pH		−0.77 (0.11)	−0.23 (0.23)	−0.66 (0.14)	0.43 (0.19)	−0.27 (0.27)	−0.34 (0.27)	−0.72 (0.22)
Minolta L*	−0.42*		−0.01 (0.24)	0.65 (0.13)	−0.21 (0.22)	0.23 (0.27)	0.02 (0.31)	0.51 (0.33)
Minolta a*	−0.25*	−0.06		0.15 (0.21)	0.02 (0.26)	0.49 (0.27)	0.12 (0.39)	−0.01 (0.54)
Minolta b*	−0.47*	0.63*	0.48*		−0.16 (0.23)	0.59 (0.24)	0.002 (0.32)	0.41 (0.23)
Log10(MFI, %)	0.13*	−0.16*	0.05	−0.16*		0.06 (0.29)	0.38 (0.25)	−0.01 (0.39)
IMF, %	−0.04	0.05	0.10*	0.12*	−0.06		−0.16 (0.34)	0.65 (0.45)
Log10(WBSF, N)	0.04	−0.24*	−0.18*	−0.35*	0.14*	−0.09		0.76 (0.24)
CL, %	−0.35*	0.28*	−0.02	0.14*	−0.01	0.01	0.27*

¹SE between brackets; Genetic correlation > 2 $\times$ SE is significant.

^*Significant phenotypic correlation (P < 0.05).

Genome-wide association

Figure 1A and B present the Manhattan plot and a QQ-plot of the genome-wide association scan for CL. The GWAS showed 2 significant SNP markers at the FDR <0.10 level on SSC2 (rs326506642 and rs341197410). The location of these SNP markers (131429753 bp and 131995508 bp, respectively) did not correspond to any annotated gene location. Both markers explained only a small part of the total variation of CL (2.94% and 2.83%, respectively). No significant SNP markers were detected for pH, Minolta L*, Minolta a*, Minolta b*, log₁₀(MFI), IMF, or log₁₀(WBSF), respectively.

Figure 1.

A) Manhattan plot of the genome-wide association analysis for cooking loss (CL). The X-axis represents the position on the chromosome. On the Y-axis the −log₁₀(P-value) is presented. Each blue or orange dot represents a SNP marker ordered based on the location on the chromosome. The red dots represent the significant SNP markers at FDR <0.10. B) A quantile-quantile (QQ) plot of the −log₁₀(P-values) for CL. On the Y-axis the observed −log₁₀(P-value) is presented. On the X-axis, the expected −log(P-value) is presented.

Discussion

In this study, genomic and metabolomic techniques were applied to a commercial Duroc sire line to investigate genetic variation for meat quality traits and the potential of using blood NMR profiling as an indicator trait for meat quality. This is particularly important as meat quality traits must be heritable to be included in breeding programs.

Heritability of meat quality traits

Our study demonstrated that meat quality traits are heritable in the DanBred Duroc population, which suggests potential for breeding for improved meat quality. The heritability for pH in our study was 0.41, which is higher than the values reported in previous studies ranging from 0.07 (Suzuki et al 2005) to 0.39 (Willson et al 2020). For meat color traits (Minolta L*, Minolta a*, Minolta b*), the heritabilities were 0.24, 0.19, and 0.24, respectively, which are lower than those found in other studies (ranging from 0.30 to 0.72) (Gjerlaug-Enger et al 2010; Willson et al 2020; Lopez et al 2020). For IMF, the heritability estimated in our study (0.16) is lower compared to Gjerlaug-Enger et al (2010) (0.50). Lopez et al (2020) presented heritabilities of CL (0.28) of shear force (0.19) based on female Duroc finishing pigs. These heritabilities are higher than the heritabilities for CL (0.10) and WBSF (0.14) detected in our study. Differences in heritability for the meat quality traits estimated in our study compared to heritability estimates in the literature could be due to the fact that in our study we measure meat quality based on intact males only. In addition, the average live weight at slaughter of the animals used in this study is 106.64 kg, which is light compared to the average slaughter weight in literature. It has been shown that differences in sex and weight have an impact on meat quality (Zomeño et al 2023).

Genetic correlations between meat quality traits

In general, the standard errors on the estimates of the genetic correlation were large in this study (Table 3). For the discussion, we focused on the correlations, where the estimates were at least twice the size of the associated standard errors.

Fresh meat typically has a pH ranging from 5.4 to 5.8, making it slightly acidic. In contrast, living muscle has a slightly basic pH of just over 7. During the slaughter and chilling processes, the muscle’s pH decreases as it depletes its stored energy. The meat quality trait pH is an important factor influencing meat color, water retention during storage and cooking, and tenderness. For example, lower ultimate pH results in lighter-colored meat (Huff-Lonergan et al 2002; Smith et al 2011). When the pH drops fast while the carcass is hanging, some water-soluble proteins like myoglobin or, under more extreme circumstances, myofibrillar proteins in the meat partially denature, resulting in a lighter color of the meat. However, these myofibrillar proteins, like actin and myosin, are essential for water retention in pork and thereby influence the CL (Zielbauer et al 2016). CL refers to the liquid and soluble substances that are lost from the meat during cooking. When CL is high, eating quality of the meat is expected to be less optimal (Aaslyng et al 2003). The results in our study showed that the meat quality trait pH had a strong negative genetic correlation to meat color traits L (−0.77) and b (−0.66) and CL (−0.72) but showed a positive correlation to log₁₀(MFI) (0.43) (Table 3).

IMF plays a role in the sensory quality of pork. A positive correlation was estimated between IMF and Minolta b* (0.59), which was much higher compared to the study of Gjerlaug-Enger et al (2010). Genetic correlations to the other meat quality traits were not significant, mainly due to the high standard error of the estimates. The high standard errors could be circumvented by increasing the sample size in a follow-up study.

Suzuki et al (2005) studied the correlations between meat quality traits in an experiment where Duroc pigs were selected for average daily gain, longissimus muscle area, backfat thickness, and IMF content. These genetic correlations are very different from the correlations estimated in our study. In particular, estimated correlations between CL and Minolta L*, pH and IMF, respectively, had opposite signs from the estimates found in our study. Furthermore, in cases where the sign of the correlation was the same, ie, pH and L (−0.10) and pH and IMF (−0.51), the magnitude of the genetic correlations estimated in our study was very different, −0.66 and −0.27, respectively. A possible explanation could be that Suzuki et al (2005) used barrows and gilts, whereas in our study we used intact males.

NMR profiling and its potential as an indicator trait

Instead of focusing on individual metabolites, we used the full NMR spectrum from blood samples to assess whether it could explain variance in meat quality traits. Although previous studies have shown that NMR profiles from meat samples can indicate meat quality (Straadt et al 2011), we were interested in using blood NMR profiling to predict meat quality traits non-invasively before slaughter. Metabolites and their expression can be variable due to environmental influences, time of the day, and age, but also across body tissues within the animal. In our study, we have used NMR profiling based on a blood sample with the idea to use this as an indicator trait for meat quality without having to slaughter the animal.

NMR-based blood metabolite profile could explain part of the variation for the traits Minolta L* and log₁₀(MFI) when the random factor section was excluded from the model (Table 2). To account for the environment a pig is assigned to, the random factor section was added to the model. This section represents the environment the pigs are kept in, defined as pigs kept in the same room at the same time, up to the date of slaughter. By including section as a random factor to the model, the blood NMR profile no longer explained much variance associated with the meat quality traits, ie, reduced to almost zero. Therefore, this observation suggests that the blood NMR profile before slaughter is not a good indicator trait for meat quality in this Duroc sire line population.

Genome-wide association study

In general, the fact that only few SNP markers are detected can be caused by the relatively low number of animals that were used in this study, which could make detection of SNP markers with a small effect on the trait more difficult. Another issue may be that the animals were intact males only, and these were a selected group not used for further breeding. This means that the sampled animals were a subsample of the total Duroc population, which could potentially influence the QTL detection. The GWAS conducted in this study identified one significant QTL for CL on chromosome 2 (SSC2). This finding is consistent with previous reports that have detected QTLs for CL on different chromosomes. For instance, studies have identified QTLs for CL on SSC2 ranging from 2.7 to 8.1 Mbp (Rohrer et al 2006; Liu et al 2008) and 41.4 to 157.8 Mbp (Harmegnies et al 2006; Wimmers et al 2007). While the QTL detected in our study is located within these regions, its precise overlap with earlier findings is unclear due to differences in detection methods (interval mapping vs. candidate gene analysis) and large confidence intervals in previous studies. Despite this, the PVE of the detected SNP markers for CL in our study was small, indicating that further research and independent validation are necessary before using these markers for selection purposes.

Conclusion

This study provides important insights into the genetic basis of meat quality traits in the DanBred Duroc population. The heritability of several traits, including pH, meat color, IMF, and tenderness, suggests that these traits could be targeted in breeding programs. However, the potential of using blood NMR profiling as an indicator trait for meat quality was limited by environmental effects, which reduced its predictive value. This indicated that the blood NMR profile 8–29 d before slaughter is not a good indicator trait for meat quality. Additionally, the GWAS identified a QTL for CL on SSC2, but further validation is needed before these markers can be used in selection programs. Future studies with larger sample sizes and additional validation are needed to confirm these findings and improve breeding strategies for enhanced meat quality.

Glossary

Abbreviations

CL

cooking loss

CV

coefficient of variation

FDR

false discovery rate

GWAS

genome-wide association study

HWE

Hardy-Weinberg equilibrium

IMF

intra-muscular fat

MAF

minor allele frequency

MF

metabolic features

MFI

myofibril fragmentation index

NMR

nuclear magnetic resonance

PVE

proportion of variance explained

QTL

quantitative trait loci

REML

restricted maximum likelihood

SNP

single nucleotide polymorphism

SSC

Sus scrofa chromosome

WBSF

Warner-Bratzler shear force

Author contributions

Albert J. Buitenhuis (Data curation, Formal analysis, Methodology, Visualization, Writing—original draft), Xiangyu Guo (Data curation, Writing—review & editing), Anders C. Sørensen (Conceptualization, Data curation, Writing—review & editing), Tage Ostersen (Conceptualization, Funding acquisition, Project administration, Writing—review & editing), and Ole F. Christensen (Conceptualization, Funding acquisition, Methodology, Writing—review & editing)

Conflicts of interest. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. XG, ACS, TO, and OFC are currently employed by Danish Agriculture & Food Council. The funders had no influence on the study design or choice of analysis methods.

Data availability

Data supporting this study are not publicly available due to data protection policy by a private company. Please contact the corresponding author for possibility.