NMR-based comparison of the metabolome of beef from Simmental and black-and-white young bulls during wet- and dry-aging

Taste as an eating quality factor of beef can be influenced by the selection of the raw material and aging method. The metabolic changes of different breeds during aging were analyzed in wet-aged and dry-aged beef up to 28 days of aging using samples from the M. longissimus thoracis et lumborum of the Simmental and Black-and-White breeds by 1H NMR spectroscopy. Breed affected the metabolome of beef samples before and during the aging of beef. The concentration of 24 of the 30 metabolites differed significantly (P < 0.05) in unaged samples of both breeds. In addition, aging time and aging type of beef samples showed partially different effects on the metabolome caused by the breed. Aging time significantly affected (P < 0.05) 30 metabolites in beef of both breeds. Aspartate, betaine, creatinine, glycine, and niacinamide correlated with aging time only in samples of Black-and-White breed. Five metabolites (carnitine, creatine, inosine 5’ monophosphate, lactic acid and O-acetyl-l-carnitine) varied significantly (P < 0.05) between dry-aged and wet-aged beef of both breeds.


Introduction
The metabolome of beef includes, among other, taste-active compounds and therefore affects the general taste of beef [1]. Taste in its basic description is characterized by the perceptibility of attributes such as sweet, sour, salty, bitter and umami by the tongue [2]. For example, the amino acids phenylalanine and tyrosine are associated with sour and bitter taste and aspartic acid, glutamic acid and the nucleotide inosine 5'-monophosphate (IMP) are associated with umami taste. Beef taste is a quality factor that can be improved or influenced by the specific selection of the raw material, specified by intrinsic and extrinsic factors, and aging of beef postmortem [1,3]. Intrinsic cues are described as those related to the physical and chemical properties of the meat such as breed, sex, age, muscle type postmortem changes in muscle tissues flavor and texture. Extrinsic cues are characterized by environmental, non-detectable parameters such as animal welfare, feeding/diet, storage conditions and packaging. Meat aging by itself is an intrinsic parameter due to its relationship with postmortem metabolic changes in muscle, while the aging method of beef as a postmortem treatment is an extrinsic influence due to the different aging types and conditions applied [1,4,5]. There are two conventional aging methods: traditional dry-aging and wet-aging under vacuum [3,6].
The applied aging methods influenced the metabolome of beef. It was found during the aging of beef, that the content of amino acids, such as leucine and isoleucine, increased and the content of inosine 5'-monophosphate (IMP) decreased [8][9][10][11][12]. Kim et al. [13] and Setyabrata et al. [14] revealed significant differences in the content of amino acids, such as alanine, valine and glutamine, between dry-aged and wet-aged beef, and Bischof et al. [10] observed marker metabolites such as anserine, O-acetyl-l-carnitine, and lactic acid for aging type and leucine, isoleucine, IMP, and hypoxanthine for the aging time in the beef of Simmental young bulls. The effect of aging method on the changes in metabolite contents and, thus, on the metabolome of beef varies in the literature [15]. In addition to variations in the aging method, this could be due to other influencing factors, such as breed.
Breed as an intrinsic factor also influences the metabolome of beef. Vopálenský et al. [16] analyzed beef from eight different breeds immediately after slaughter and found that the contents of amino acids, such as valine, methionine, and serine, varied significantly between breeds. In addition, Dashdorj et al. [17] reported significant differences in the content of amino acids, such as aspartic acid (0.57 vs. 0.41 µmol/g) and glutamic acid (1.99 vs. 1.27 µmol/g) in 28-days dry-aged beef from Hanwoo and Angus breeds. Fan et al. [18] compared three breeds Yunling, Simmental, and Wenshan, and showed significant differences in the IMP content of unaged beef from Yunling (896.11 µg/g) and the other breeds (805.98 and 822.17 µg/g, respectively), while these differences were not observed after seven days of wetaging. Koutsidis et al. [7] analyzed 10-days wet-aged beef from Aberdeen Angus and Holstein-Friesian and observed significant differences in the concentrations of ribose, hypoxanthine, creatinine, glycine, and arginine.
It is unknown so far whether the effect of the aging method is more pronounced than that of the breed or otherwise. The aim of this study was to analyze the polar fraction of the metabolome of unaged, dry-aged and wet-aged beef from two different breeds to determine the influence of breed and aging on the metabolome and thus on the aging outcome. It was postulated as a hypothesis, that the breed would have a dominant influence on the metabolome and superimpose the effects of the aging method.

Raw material
Seven young bulls of the 'Black-and-White' breed (German code: 01-SBT) and seven young bulls of the 'Simmental' breed ('Fleckvieh' in German, code: 11-FL) were used, randomly selected from German high-input production systems. The twenty-months old young bulls were fed on maize silage and fattening feed. Until delivery on the 5th day postmortem, the carcasses were hung in a cold storage according to a standardized industrial process. The carcasses were classified with a 2-3 (slight to average fat cover) and U, R or O (very good, good or fair meat grader) according to the EU classification system EUROP [19]. The right and left bone-in strip loin (M. longissimus thoracis et lumborum) was taken five days postmortem from the 5th rib to 6th lumbar vertebra.

Aging method
The 28 strip loins were cut into ten 8 cm-thin pieces. Samples for different aging times (two pieces per time point) were taken cranial to caudal. The beef was aged in four aging runs (D1 to D4) with samples of three or four animals each. The right bone-in strip loins were used for dryaging in an aging chamber (AT 1-6/E, Autotherm Ludwig Brümmendorf GmbH & Co. KG, Waxweiler, Germany) at D1, 2.6 ± 1.1 °C and 73.6 ± 7.9% relative humidity; D2, 2.5 ± 1.0 °C and 77.7 ± 11.5% relative humidity; D3, 2.2 ± 1.0 °C and 73.8 ± 12.4% relative humidity; and D4, 2.0 ± 1.0 °C and 81.7 ± 12.0% relative humidity. After dryaging, the surface of the beef (approx. 1 cm) was trimmed. The left deboned strip loins were vacuum-packed in polyamide/polyethylene bags (BST-090; Rolf Bayer Vacuumverpackung GmbH, Veitsbronn, Germany; oxygen permeability < 60 mL/m 2 per 24 h per bar at 0% relative humidity and 23 °C and water permeability < 4 g/m 2 per 24 h at 85% relative humidity and 23 °C) and utilized for wet-aging in vacuum (< 15 mbar) in the same conditions as for dry-aging. Both aging methods were applied for 0, 7, 14, 21 and 28 days, with time point 0 corresponding to the 5th day postmortem. All aging times refer to the time point 0 i.e. an aging time of e.g. 21 days plus five days postmortem.

H NMR spectroscopy
A Bruker 400 MHz Ascend NMR spectrometer (Bruker Biospin GmbH, Rheinstetten, Germany) was utilized for the NMR analysis of the beef extract samples. The measurement and evaluation of 1 H NMR spectra were performed according to the method of Bischof et al. [9]. A noesygppr1d pulse sequence was used to acquire the spectra with the following parameters: Temperature, 300 K; number of points, 65 k; number of scans, 128; spectral width, 8.403 kHz; acquisition time, 3.9 s; and presaturation field strength for water suppression, 25 Hz. Automatic processing of NMR spectra was performed using the Software Topspin 3.6 (Bruker Biospin GmbH, Rheinstetten, Germany). The metabolite qualification was based on the comparison with the NMR database (Bruker Biospin GmbH), literature [9,[20][21][22] and our own measured standard components. Validation of the method, metabolite qualification and quantification were performed in our previous study [9].

Data analysis
Data analysis was performed using Matlab R2018a software (The Mathworks, Natick, Massachusetts, USA). The 1 H NMR spectra were bucketed in 250 buckets of equal size from 0.5 to 4.7 ppm. The buckets were scaled by subtraction of their mean values and division by their standard deviation. An orthogonal projection to latent structures discriminant analysis (OPLS-DA) was calculated with k-fold cross validation (k = 25) and a test set of 25 (all samples) and 10 (0 d samples) randomly chosen samples, respectively, for the analysis of differences between the breeds. The partial least squares regression (PLS-R) with k-fold cross validation (k = 10) was performed for the analysis of the aging time. Regarding the univariant analysis, the 1 H NMR spectra were bucketed in 500 buckets from 0.5 to 9 ppm for statistical analysis of the aging type using univariant scaling and linear discriminant analysis (LDA) with k-fold cross validation (k = 10). The concentrations of metabolites were analyzed by a Kruskal-Wallis test with a post-hoc Bonferroni test (α = 0.05) to analyze significant differences between samples. A Spearman correlation coefficient r SP was calculated for the analysis of correlation between the metabolite concentration and aging time. MetaboAnalyst 5.0 (https:// www. metab oanal yst. ca) was used for the pathway analysis using the Bos taurus library.

Analysis of the metabolome of beef from Simmental and black-and-white breed
The left and right M. longissimus thoracis et lumborum of fourteen animals (seven young bulls each) of the two breeds Simmental and Black-and-White were analyzed during dryaging and wet-aging up to 28 days. The 1 H NMR spectra of unaged beef from both breeds are shown in Fig. 1 and demonstrated that the most pronounced differences were obtained between 3.0 and 4.2 ppm. Several overlapping signals from different metabolites, for example, glucose, lactic acid, creatine, glycine, betaine and carnitine, are in this region of the 1 H NMR spectrum. The comparison of 1 H NMR spectra indicates that the samples from the two breeds already show differences. The OPLS-DA of 1 H NMR spectra of unaged beef showed that the polar fraction of metabolome differed between the two breeds with a model accuracy of 96.8%, coefficient of determination of R 2 = 0.980 and crossvalidated R 2 of Q 2 = 0.924 (Fig. 2b). In addition, the OPLS-DA of all samples (unaged or aged) resulted in an accuracy of 99.6%, R 2 = 0.977 and Q 2 = 0.945 for the classification of the breed (Fig. 2a), which showed better goodness of prediction Q 2 than unaged samples only. This could be due to the number of samples used for the analysis. All 1 H NMR spectra of samples from aging time points 0, 7, 14, 21 and 28 days were used to calculate the OPLS-DA to represent that the differences between breeds were reflected throughout the aging. The 1 H NMR spectra and the OPLS-DA imply that the metabolome of the two breeds is significantly different from each other. The OPLS-DA for aging type analysis with an accuracy of 57.2% show, that the separation of aging type with all aged samples independent of breed is not possible. Furthermore, the statistical analysis of aging time (0 and 28 days) using the OPLS-DA model shows an accuracy of 99.5%, R 2 = 0.983 and Q 2 = 0.973 that indicates a good discrimination of unaged and 28-day aged samples by their 1 H NMR spectra. Based on the OPLS-DA of all samples, the effect of breed on the metabolome of beef is more pronounced than the effect of aging type. The differences in the metabolome of the unaged beef already showed differences between breeds that were not compensated during aging. The results of the Kruskal-Wallis test showed significant differences in metabolite concentrations for 24 (0 d) and 21 (0-28 d) of 30 metabolites between the two breeds (Tables 1  and 2). The concentration of 13 (alanine, anserine, betaine, carnitine, α-and β-glucose, glucose-6-phosphate, glycine, IMP, Inosine, niacinamide, O-acetyl-l-carnitine and succinic acid) of the 24 metabolites that differed at the onset of aging were higher in samples of the Black-and-White breed, and 11 metabolites (aspartate, carnosine, creatine, glutamate, hypoxanthine, isoleucine, leucine, methionine, phenylalanine, tyrosine and valine) were higher in samples of the Simmental breed. The concentration of glucose, for example, is significantly higher (P < 0.05) in samples of the Black-and-White breed (4.16 ± 1.48 µmol/g) than those of the Simmental breed (1.83 ± 0.91 µmol/g). Lee et al. [23] compared beef from Hanwoo, Jeju Black and Wagyu breeds and found significant differences in the content of amino acids such as alanine, aspartic acid, glutamic acid, leucine, isoleucine, methionine and phenylalanine. Furthermore, Antonelo et al. [24] reported differences in the metabolome of 7-days-wet-aged beef of Nellore and crossbreed Angus x Nellore. The authors identified 10 metabolites (acetate, alanine, carnosine, creatinine, glucose, glutamate, glycine, IMP, methionine and succinate) that differed significantly between the two breeds, showed that carnosine, glutamate and IMP correlated with sensory attributes such as overall liking, flavor, juiciness and tenderness of beef, and concluded that the major metabolic pathways affected by breed were glutathione metabolism, alanine, aspartate and glutamate metabolism, valine, leucine and isoleucine metabolism and primary bile acid biosynthesis. By comparison, the pathway analysis for the two breeds Simmental and Blackand-White carried out in this study showed that phenylalanine, tyrosine and tryptophan biosynthesis, D-glutamine and D-glutamate metabolism, Alanine, aspartate and glutamate metabolism, phenylalanine metabolism, glycine, serine and threonine metabolism and histidine metabolism were influenced by the breeds (Fig. 3). This demonstrated that the metabolome of the both breed Black-and-White and Simmental differed, which may also have an influence on the eating quality of beef.

Effect of aging time on the metabolome of beef
The PLS regression of the samples for aging time analysis revealed that the metabolome changed over the time during aging (Fig. 4). The PLS-R of all samples showed a larger scatter a lower R 2 (R 2 = 0.929) and a higher root mean square error (RMSE) of calibration (RMSEC; 2.56) than that of the single breeds due to the differences between the breeds. The PLS-R of NMR spectra from Black-and-White samples have the best predictability with Q 2 = 0.938 and a prediction error RMSE of  (Table S1) of the buckets between 0.93 and 1.06 ppm, 3.21 and 3.28 ppm, 3.51 and 4.13 are the highest (> 1) and are associated with the metabolites leucine, isoleucine, and valine, among other. The Kruskal-Wallis test of all samples for analysis of aging time showed that aging time significantly (P < 0.05) affected all of the 30 metabolites observed, while the concentration of the metabolite glucose-6-phosphate in Simmental samples and that of creatine, carnitine and O-acetyl-L-carnitine in Blackand-White samples had no aging time effect (Tables 1  and 2). The amino acids isoleucine, leucine, methionine, phenylalanine, tyrosine, and valine correlated positively with the aging time (r sp ≥ 0.80), which also agreed with the calculated VIP scores in PLS-R. Furthermore, hypoxanthine showed a positive correlation with aging time (r sp ≥ 0.81), while IMP correlated negatively with aging time (r sp ≤ − 0.86) ( Table 1). Metabolites such as aspartate, betaine, creatinine, glycine, and niacinamide correlated with aging time only in samples of the Black-and-White breed. The concentration of the dipeptides anserine and carnosine increased with aging time in samples of Simmental young bulls, while their concentrations showed a maximum at 14 days of aging in samples of Black-and-White young bulls. The increase of amino acids is consistent with the literature [8,11,13,21] as well as our previous study [10] and is the result of proteolysis during aging. Amino acids are taste-active compounds and support taste attributes such as sweet, sour, bitter, and umami. The observed amino acids correlated with aging time, Fig. 2 Orthogonal projections to latent structures discriminant analysis (OPLS-DA) of 1 H NMR spectra from samples of Simmental and Black-and-White breeds. a 1 H NMR spectra used from all samples (unaged, dry-aged, wet-aged, 0 to 28 days aging) of Simmental (light grey) and Black-and-White (dark grey) breeds; b 1 H NMR spectra used from unaged samples (d = 0 days) of Simmental (light grey) and Black-and-White (dark grey) breeds. c 1 H NMR spectra used from all dry-aged (dark grey) and wet-aged (light grey) samples. d 1 H NMR spectra used from 0-(light grey) and 28-days aged (dark grey) samples. The models were performed with a k-fold cross validation with k = 25 (a) or k = 10 (b, c, d), respectively. The 95% confidence interval is demonstrated by the ellipses. The accuracy is 99.6% (a), 96.1% (b), 57.2% (c) and 99.5% (d), respectively. The coefficient of determination (R 2 ) and cross-validated R 2 (Q 2 ) are R 2 = 0.977 and Q 2 = 0.945 (a), R 2 = 0.980 and Q 2 = 0.924 (b), R 2 = 0.719 and Q 2 = 0.561 (c) and R 2 = 0.983 and Q 2 = 0.973 (d) isoleucine, leucine, phenylalanine, tyrosine, and valine are all associated with a bitter taste. This suggests that the taste of beef changes during aging. Further, peptides such as carnosine and anserine are taste-active compounds for the attribute 'umami' [1]. The pathway analysis for both breeds with significant metabolites over the aging time shows that the aging time affects phenylalanine, tyrosine and tryptophan biosynthesis, D-glutamine and D-glutamate metabolism, alanine, aspartate and glutamate metabolism, phenylalanine metabolism and glycine, serine and threonine metabolism (Fig. 3).

Effect of aging type on the metabolome of beef
Statistical analysis of all 28-days aged samples of both breeds by aging type resulted in an LDA model with an accuracy of 88% (Fig. 5a). By comparison, the LDA models separated for the breeds had an accuracy of 94% x-z Different letters in the same column indicate a significant difference (P < 0.05) between aging type and breed r sp Spearman correlation coefficient (Black-and-White; c) and 94% (Simmental; d). The LDA model distinguished for aging type and breed showed that the breeds were clearly separated, while the aging type ellipses overlapped for the same breed (Fig. 5b). These results suggest that differences in samples by aging type have less impact on the beef metabolome than differences by breed.
The results of the Kruskal-Wallis test showed five metabolites (carnitine, creatine, IMP, lactic acid and O-acetyll-Carnitine) that differed significantly (P < 0.05) between the aging types of samples of both breeds across aging. The metabolites aspartate, betaine, fumaric acid and inosine varied significantly (P < 0.05) between dry-aged and wetaged beef of Black-and-White samples over the entire aging period ( Table 2). Metabolites such as acetic acid, carnitine, and creatine differed significantly (P < 0.05) between dryaged and wet-aged beef at single days of aging of Simmental young bulls, while anserine, betaine, creatine, inosine and lactic acid had significantly different concentrations in dryaged and wet-aged samples of Black-and-White young bulls of single days of aging ( Table 1). The concentration of lactic acid increased significantly (P < 0.05) over time in dry-aged beef of both breeds, while consistent concentrations were observed in wet-aged beef (Table 1). Setyabrata et al. [14] compared various aging types and reported that most amino acids differed significantly between dry-aged and wet-aged samples of the Holstein breed. Kim, et al. [13] showed that the aging type significantly affected the concentration of anserine, carnitine, carnosine and creatine in samples of Holstein steers up to 28 days of aging. Kim et al. [20] compared dry-aged and wet-aged beef using NMR spectroscopy and sensory analysis. The authors found that dry-aged beef resulted in enhanced flavor and reported metabolites such  as tryptophan, phenylalanine, valine, tyrosine, glutamate, isoleucine and leucine were higher in dry-aged than wetaged beef. In addition, compounds such as lactic acid and acetic acid are associated with a sour taste, while peptides associated with bitter or umami taste [1]. The aging type applied affected the purine metabolism, histidine metabolism, glycine, serine and threonine metabolism, and citrate cycle in samples of the Black-and-White breed, while the purine metabolism and arginine and proline metabolism were affected in samples of the Simmental breed.

Conclusions
The metabolome of the Simmental and Black-and-White breeds varied significantly. The breed-related differences in the metabolome during aging remain and other metabolic pathways were partially affected depending on the breed. These differences in the contents of metabolites may influence the taste of beef due to changes in tasteactive compounds. In the future, sensory evaluation will be used to assess these differences of the metabolome and to what extent the samples of different breeds, aging types and aging times differ from each other in sensory terms. The influence of breed is stronger than of aging type, which is why investigations with other breeds would be interesting.
Funding Open Access funding enabled and organized by Projekt DEAL.
Data availability Data available upon request.