As carcasses were presented for grading (approximately 24 h postmortem), U.S. Select carcasses were evaluated using the tenderness prediction model in the VBG 2000 GigE beef grading system (Shackelford et al., 2003). Carcasses (n = 40) were selected from commercial production lots to have either tender (predicted slice shear force < 19 kg; n = 10) or tough (predicted slice shear force > 19.5 kg; n = 30) predicted slice shear force values. Our experience with the tenderness prediction model is that some carcasses predicted to be tough have low slice shear force values. Thus, we selected a greater number of carcasses predicted to be tough to ensure that a class of carcasses with adequately high slice hear force values were obtained. A total of 93 U.S. Select carcasses were screened to identify the 40 carcasses included in the experiment. As carcasses were fabricated, beef, loin, strip loin subprimals (similar to IMPS# 180; USDA, 2014) were obtained from the left side of each carcass, vacuum packaged, and transported under refrigeration to the U.S. Meat Animal Research Center. The next day, (2 d postmortem) subprimals were unpackaged and trimmed free of subcutaneous fat and accessory muscles. The longissimus lumborum muscle was cut into 28-mm-thick steaks perpendicular to the long axis of the muscle using a Graselli NSL400 slicer (Grasselli-SSI, Throop, PA). Identity of steak location was maintained. From each of the anterior and posterior halves of the longissimus lumborum, 4 pairs of adjacent steaks were blocked by location and assigned to one of 4 aging times (2, 7, 14, and 28 d postmortem) at 1°C. Thus, there were 2 pairs of adjacent steaks from each strip loin assigned to each aging time. Because vacuum packaging steaks can compress steaks, reducing the repeatability of belt grill cookery, and consequently, slice shear force determination, pairs of steaks were placed on plastic trays and packaged in oxygen impermeable bags (Clarity 3 mil standard barrier; Prime Source, St. Louis, MO; oxygen transmission rate = 0.01 mL/cm² per 24 h × 23°C and 1 atm) flushed with 75% N₂, 25% CO² (Linweld, Hastings, NE) to prevent spoilage during aging. One steak from each pair was designated for slice shear force determination and the other was designated for metabolite profiling.

After each aging time, packages of steaks assigned to that aging time were opened and steaks destined for slice shear force were transferred to a 5°C refrigerator and allowed to equilibrate before they were cooked on an electric belt grill as described by Wheeler et al. (1998). Slice shear force was measured using the protocol described by Shackelford et al. (1999). Steaks designated for metabolite profiling were trimmed free of epimysium and visible fat, diced, frozen in liquid nitrogen, and stored at –80°C until further analysis.

Slice shear force values from all aging times were utilized to identify the carcasses with the lowest (most tender) and highest (toughest) slice shear force values to create two tenderness classes (n = 10 for both the tough and tender class). Aliquots of diced, frozen samples from the steaks adjacent to the slice shear force steak at each aging time were transported to the Colorado State University Proteomic and Metabolomics laboratory. Additional diced and frozen aliquots were pulverized in liquid nitrogen into a homogenous powder and used for sarcomere length and desmin degradation determination.

Sarcomere length determination was made using the helium-neon laser diffraction method as described by Cross et al. (1981). Small aliquots of powdered tissue (approximately 0.5g; n = 6 per sample) were placed on microscope slides. Approximately 150 μL of 0.2 M sucrose in 0.1 M NaHPO4 buffer was added to each aliquot prior to determining sarcomere length. Six sarcomere laser diffraction patterns were recorded per aliquot on paper, for a total of 36 sarcomere lengths per sample. Diffraction patterns were scanned into JPEG (Joint Photographic Experts Group) images, and Image Pro (Media Cybernetics, Inc., Rockville, MD) software was used to measure the distance between primary diffraction bands and calculate sarcomere length using the equation reported by Cross et al. (1981).

Immuno-blotting for desmin was performed on a whole muscle extract of each sample to determine the extent of postmortem proteolysis (Wheeler et al., 2002). Each combination of tenderness class and aging time was represented on each gel. Potential lane effects were mitigated by blocking lane assignment of each treatment across gels. At-death standards from longissimus lumborum (collected within 40 min of exsanguination) were run in the center and two outermost lanes on each gel. The chemi-luminescent intensity of the entire desmin band of each sample was compared to the mean chemi-luminescent intensity of the entire desmin band of the at-death standards within each blot to calculate the percentage of desmin degraded in each sample.

Diced, frozen samples were lyophilized at –80°C and 0.008 mBar for 72 h before being homogenized with liquid nitrogen. Homogenized tissue (30 mg) was weighed into 2 mL glass vials and extracted in 1.5 mL 80% methanol 20% water (v/v) and vortexed at 4°C for 2 h. After centrifugation (3000 × g) for 20 min, the supernatant was transferred to a new 2 mL glass vial.

Extracts (300 μL) were dried under nitrogen gas and resuspended in 100 μL of 80% methanol 20% water (v/v). Samples (2 μL) were injected in duplicate in randomized order and separated using a Waters Acquity UPLC CSH Phenyl Hexyl column (1.7 μM, 1.0 × 100 mm), employing a gradient from solvent A (2 mM ammonium hydroxide, 0.1% formic acid) to solvent B (Acetonitrile, 0.1% formic acid). A pooled quality control (QC) sample comprised of equal amounts of all samples was injected after every 10 experimental samples. Injections were made in 100% A, held at 100% A for 1 min, ramped to 98% B over 12 min, held at 98% B for 3 min, and then returned to starting conditions over 0.05 min and allowed to re-equilibrate for 3.95 min, with a 200 μL/min constant flow rate. The column and samples were held at 65°C and 6°C, respectively. The column eluent was infused into a Waters Xevo G2 Q-TOF-MS with an electrospray source operating in positive mode, scanning 50 to 2000 m/z at 0.2 s per scan, alternating between MS (6V collision energy) and MSE mode (15 to 30V ramp). Calibration was performed using sodium iodide with 1 ppm mass accuracy. The capillary voltage was held at 2200 V, source temp at 150°C, and nitrogen desolvation temp a 350°C with a flow rate of 800 L/h.

Extracts (400 μL) were dried under nitrogen, resuspended in 50 μL of pyridine containing 25 mg/mL of methoxyaminehydrochloride, incubated at 60°C for 45 min, sonicated for 10 min, and incubated for an additional 45 min at 60°C. Next, 50 μL of N-methyl-N-trimethylsilyltrifluoroacetamide with 1% trimethylchlorosilane (MSTFA + 1% TMCS, Thermo Scientific) was added and samples were incubated at 60°C for 30 min, centrifuged at 3,000 × g for 5 min, cooled to room temperature, and 80 μL of the supernatant was transferred to a 150 μL glass insert in a GC–MS auto sampler vial. Metabolites were detected using a Trace 1310 GC coupled to a Thermo ISQ-MS. Samples were injected in duplicate using a 10:1 split ratio. The inlet was held at 285°C, and transfer line and ions source were held at 300 and 260°C, respectively. Separation was achieved on a 30m TG-5 M S column (Thermo Scientific, 0.25 mm i.d., 0.25 μm film thickness) with a 1.2 mL/min helium gas flow rate, and the program consisted of 80°C for 30 sec, a ramp of 15°C per min to 330°C, and an 8 min hold. Masses between 50 and 650 m/z were scanned at 5 scans/sec using electron impact ionization. Pooled QC samples were injected after every 12 experimental samples.

For each sample, raw data files were converted to.cdf format, and a matrix of molecular features as defined by retention time and mass (m/z) was generated using XCMS (Smith et al., 2006) software in R (R Development Core Team, 2008) for feature detection and alignment. Raw peak areas were normalized to total ion signal in R, outlier injections were detected based on total signal and PC1 of principle component analysis, and the mean area of the chromatographic peak was calculated among replicate injections (n = 2). Features were grouped based on a novel clustering tool, RAMClustR (Broeckling et al., 2014), which groups features into spectra based on co-elution and co-variance across the full data set. Compounds were annotated based on spectral matching to in-house, NISTv12, predicted MS1 and retention times spectral data bases using RAM search (Broeckling et al., 2016) and manual interrogation of the Metlin compound and spectral database. The peak areas for each feature in a spectrum were condensed as the weighted mean of all features in the spectrum to generate a single quantitative value for each metabolite.

Data were analyzed as a split-plot design with the PROC GLIMMIX procedure of SAS (SAS Inst. Inc., Cary, NC). Tenderness classification (tough or tender) was the whole-plot treatment with carcass being the experimental unit. Aging time (2, 7, 14, 28) was the sub-plot treatment and steak was the experimental unit. Orthogonal polynomial contrasts were used to determine trends in each metabolite regarding aging time. Least-squares means were generated for main-effects and the tenderness class × aging time interaction. When appropriate, means were separated using the DIFF and LINES options.

From a total of 2,562, metabolites (n = 651) with significant F-tests for at-least one main effect that exceeded Bonferroni correction of multiple testing (P ≤ 0.00002) and nominal (P < 0.05) for the remaining main-effect were included in further analysis. Pearson correlation coefficients and partial correlation coefficients (removing aging time effects) were generated between each metabolite and slice shear force, desmin degradation, and sarcomere length values across all aging times. Metabolites with the greatest partial correlation (significance levels < 0.00002) to slice shear force, desmin degradation, or sarcomere length were included in partial least-squares (PLS) regression, using the PROC PLS procedure of SAS, with slice shear force, desmin degradation, and sarcomere length as dependent variables and fixed effects of tenderness class and aging time. Bonferroni correction for multiple testing (P < 0.00002) was used for screening metabolites for inclusion in analyses. However, a pre-determined significance level of (0.05) was used for judgments of statistical significance for discussion.

Least-squares means for slice shear force values of longissimus lumborum steaks from carcasses classified as tender or tough after aging for 2, 7, 14, or 28 d are presented in Fig. 1. The carcasses were classified based on slice shear force values across aging times, and thus, the difference between tenderness classes was by design (i.e., carcasses classified as tender had to have consistently low slice shear force values at all aging times). At all aging times tested, steaks from carcasses classified as tough had higher (P < 0.001) slice shear force values than steaks from carcasses classified as tender. Each progressive increase in aging time resulted in lesser (P < 0.001) slice shear force values in steaks from carcasses classified as tough. In steaks from carcasses classified as tender, increasing aging time from 2 to 7 and from 14 to 28 d postmortem decreased (P < 0.001 and 0.02, respectively) slice shear force values. Steaks from carcasses classified as tender on d 2 postmortem had slice shear force values similar (P = 0.42) to those of steaks from carcasses classified as tough on d 28 postmortem. Increasing aging time resulted in substantial tenderization and was consistent with longissimus tenderization detected in previous reports (Gruber et al., 2006; King et al., 2009c).

Least-squares means for desmin degradation are presented in Fig. 2. Steaks classified as tender had a greater proportion of desmin degraded at 2 (P = 0.01), 7 (P < 0.001), 14 (P < 0.001), and 28 d (P = 0.03) than steaks from carcasses classified as tough. Each increase in aging time increased (P < 0.04) the amount of desmin degraded in steaks from both tenderness classes. However, the magnitude of the desmin degradation difference was not consistent with each incremental increase in aging time. Previous research from our laboratory has found increases in desmin degradation in various beef muscles through 42 d of aging (King et al., 2009a; King et al., 2009b; King et al., 2009c) which were associated with decreases in slice shear force, although statistical differences were not always detected with each incremental increase in aging time. By 28 d postmortem, desmin degradation in steaks classified as tough was similar (P = 0.38) to the desmin degradation at d 7 postmortem in steaks classified as tender. Sarcomere length did not differ across tenderness classes (P = 0.42) or aging times (P = 0.40) in the present experiment (Fig. 3).

Metabolites (n = 2562) were profiled in the beef samples using both GC–MS and UPLC-MS analytical platforms. Initial statistical screening of the data was accomplished by ANOVA F-tests for the fixed effects of tenderness class and aging time. A total of 651 metabolites had F-tests significant at the Bonferroni correction level (all for the aging time effect; Supplementary Table 1). For the tenderness class main effect, 3 metabolites met the Bonferroni correction criterion, although 254 were significant at a nominal significance level (P < 0.05). Additionally, 173 metabolites displayed an interaction (P < 0.05) between tenderness class and aging time.

Correlation analysis was used to examine the relationship of the remaining 254 metabolites to slice shear force, desmin degradation, and sarcomere length (Supplementary Table 2). Because these metabolites were selected, in part, based on postmortem changes, partial correlations were generated to remove the postmortem aging time effect. In some cases, partial correlations were much lower than those including the aging effect, suggesting that changes in the abundance of the metabolite were not related to changes in tenderness. Metabolites were selected for discussion and further analysis if the level of significance of the partial correlation to slice shear force, desmin degradation, or sarcomere length exceeded 0.00002. No metabolite had this level of correlation to sarcomere length. A total of 102 metabolites had highly significant (P < 0.00002) partial correlation to either slice shear force or desmin degradation. Of these, 61 met this criterion for slice shear force and 61 met this criterion for desmin degradation (19 were correlated to both slice shear force and desmin degradation). At a nominal significance level (P < 0.05), all 102 metabolites were correlated (partial) to slice shear force, 79 were correlated (partial) to desmin degradation, and 19 were correlated (partial) to sarcomere length.

The process of assigning a molecular identity to a detected mass spectrometry signal (metabolite annotation) is challenging and represents a current constraint in the field of metabolomics (Broeckling et al., 2016). Thus, in this study of the 102 metabolites included in the final analysis, 28 were annotated. As the primary objective of the present experiment was to identify molecular changes reflective of beef longissimus tenderness and the aging response in the longissimus, unknown metabolites were retained in the final analyses and the complete results are presented in Supplementary Tables 1 and 2. However, the discussion below is focused on the 28 annotated metabolites.

Least-squares means for the tenderness class and aging time effects and the tenderness class × aging time interaction for annotated metabolites are presented in Table 1. Some of the annotated metabolites could be loosely grouped into classes of similar compounds.

Pearson correlation and partial correlation coefficients between metabolites and slice shear force, desmin degradation, and sarcomere length are presented in Table 2. Partial correlation coefficients represent relationships to variation in slice shear force, desmin degradation, and sarcomere length within aging times and removes correlation simply due to changes during aging that may be coincidental.

The loadings plot of the metabolites included in the PLS analysis are presented in Fig. 4. All 102 metabolites meeting the previously described criteria were included in the PLS model. However, only those that could be annotated are shown in Fig. 4. The loadings for all 102 included metabolites are available in Supplementary Table 3. The first 2 factors derived in the PLS analysis explained 41 and 9% of the variance in the dependent variables, respectively. The first factor was highly related to both increased slice shear force (positive values) and increased desmin degradation (negative values). Aging time was strongly associated with Factor 1 with d 2 postmortem associated with positive values and increased aging times being associated with lower and negative values of Factor 1. The tender class of steaks was associated with negative values for Factor 1, while the tough class of steaks was associated with positive values. The second factor was highly associated with the tenderness classes with the tough class and increased slice shear force associated with positive values and the tender class being associated with negative values. Amino acids were primarily related to negative values for factor 1, and clustered around the loadings for desmin degradation. The exception to this was the peptide L-ɑ-aspartyly-L-valylglycyl-L-proline, which was strongly associated with Factor 2, and had loadings very close to the tender class of samples.

The results presented here provide new insights on factors regulating initial beef tenderness and tenderization during postmortem aging. Moreover, candidate biomarkers, which may be useful in future investigations, were identified. Further study is needed to refine the current understanding of the mechanisms by which these metabolites influence postmortem changes in the protein structure of meat. Taken together, the results provide new insights into the complexity of the postmortem tenderization process and lay the groundwork for future studies.