Metabolite Profile Differences between Beef Longissimus and Psoas Muscles during Display

The objective of this research was to compare metabolite profiles between beef longissimus and psoas muscles during display. Beef short loins were collected 3 d postmortem (n = 10). Steaks were cut from each longissimus lumborum (LL) and psoas major (PM) muscle and displayed under retail conditions for 7 d. Surface color, biochemical properties, and metabolites were analyzed during storage. PM decreased in redness (P < 0.05) by d 3 of display compared with LL. There were differences in metabolite concentrations (P < 0.05) between each muscle type at each time point. Sugars, amino acids, tricarboxylic acid cycle intermediates, and glycolytic substrates were detected in both muscles. Glycolytic metabolites such as pyruvic acid, glucose–6–phosphate, and fructose were greater (P < 0.05) in LL than PM at all display times. On d 0, the intensity of pyruvic acid in LL and PM were 142 and 42, respectively. Citric acid and succinic acid were lower on d 0, but were greater (P < 0.05) in LL compared with PM by d 7 of display. Carnitine was lower (P < 0.05) in LL than PM at all display times. On d 7, carnitine level in LL was 4.1 while in PM was 13,500. The results suggest that in addition to muscle-specific differences in mitochondrial and enzyme activities, inherent metabolite differences also may contribute to muscle color stability.


Introduction
Myoglobin is a sarcoplasmic protein that is primarily responsible for meat color.In fresh meat, myoglobin can exist in 3 different redox forms, namely deoxymyoglobin, oxymyoglobin, and metmyoglobin (AMSA, 2012).Deoxymyoglobin imparts purple color while predominant oxymyoglobin gives consumers the preferred bright-red color to beef.Oxidation of both oxy-and deoxymyoglobin result in the formation of metmyoglobin and discoloration of meat (Faustman and Cassens, 1990).Although various pre-and post-harvest factors can increase myoglo-bin oxidation, meat has an inherent capacity to delay metmyoglobin accumulation by a process called metmyoglobin reducing activity (MRA; Ledward, 1985).The concentration of reduced nicotinamide adenine dinucleotide (NADH), myoglobin chemistry, and mitochondrial activity play a significant role in metmyoglobin reduction (Tang et al., 2005;Ramanathan and Mancini, 2010;Nerimetla et al., 2017).In postmortem muscle, enzymes involved in the glycolytic and tricarboxylic acid cycle (TCA) retain activity.However, the metabolite concentration decreases with postmortem time.Various studies have shown that the addition of metabolites such as lactate, pyruvate, and succinate to meat or isolated mitochondria can regenerate NADH and can influence beef color (Tang et al., 2005;Kim et al., 2006;Ramanathan et al., 2011).Therefore, characterizing TCA and glycolytic substrate changes in postmortem muscle is critical to understand the fundamental basis for meat discoloration.
Muscles in a beef carcass can be classified as colorstable or color-labile depending on the proportion of red and white fibers.Longissimus lumborum (LL), which is commonly merchandised as New York strip steak, is a color-stable muscle while the tenderloin (Psoas major; PM) is a color-labile muscle (O'Keeffe and Hood, 1982;Seyfert et al., 2007).Various researchers have shown that biochemical properties such as MRA and oxygen consumption rate can vary between LL and PM (McKenna et al., 2005).A recent study suggests that the sarcoplasmic proteome differs significantly between LL and PM (Joseph et al., 2012).More specifically, an overabundance of enzymes involved with color stability such as aldose reductase, pyruvate dehydrogenase, β-enolase, and triose phosphate isomerase involved in the glycolytic pathway were reported in LL compared to PM. Muscles with more red muscle fibers (PM) will have more mitochondria and capillary density than muscles with more white fibers (LL).Hence TCA and glycolytic substrates will be utilized at different rates in red muscles than in white muscles (Kushmerick et al., 1992;Glancy and Balaban, 2011).Recently metabolomics was used to study color stability in ovine meat (Subbaraj et al., 2016).However, limited studies have utilized metabolomic techniques to characterize beef color.Therefore, the objective of this study was to determine the metabolomic profile differences between LL and PM using a gas chromatographymass spectrometry-based metabolomics approach to gain insights into muscle-specific differences in color stability.

Materials and Methods
This research did not include animal or human subjects and therefore does not require Animal Care and Use Committee approval.

Raw materials and processing
Ten USDA Choice short loins were purchased from a major packing facility 3-d postmortem.Vacuum packaged loins were transported on ice to the Robert M. Kerr Food and Agricultural Products Center at Oklahoma State University.Ten beef strip loins (longissimus lumborum; IMPS #180) and 10 tenderloins (psoas major; IMPS #190A) were separated from each short loin, and each was cut into five 2.5-cm-thick steaks using a meat slicer (Bizerba USA Inc., Piscataway, NJ).The steaks were placed onto foam trays with absorbent pads, overwrapped with PVC film (oxygen-permeable polyvinyl chloride fresh meat film; 15,500-16,275 cm 3 O 2 /m 2 per 24 h at 23°C, E-Z Wrap Crystal Clear Polyvinyl Chloride Wrapping Film, Koch Supplies, Kansas City, MO) and stored in a simulated retail display under fluorescent lighting for 7 d.The surface color was measured daily using a HunterLab Miniscan spectrophotometer (HunterLab Associates, Reston, VA).Following surface color measurements, steaks were cut in half.The first half was used to measure metmyoglobin reducing activity (MRA) and oxygen consumption (OC), and the second half was used to measure metabolite profile and pH.The steak half assigned to MRA and OC was then bisected parallel to the oxygenated surface to expose the interior of steak (resulting in 2 interior pieces).The first interior section was used to measure MRA and the second interior piece was used to measure OC.From the second half, representative samples that contained both oxygenated and interior sections were taken for both metabolomics and pH measurements.

Muscle pH
Ten grams of LL and PM steak samples were blended with 100 mL of deionized water at 25°C, and homogenized for 30 s in a Sorvall Omni tabletop mixer (Newton, CT).The pH of the muscle homogenates was measured using an Accumet combination glass electrode connected to an Accumet 50 pH meter (Fisher Scientific, Fairlawn, NJ).

Retail display
After packaging, steaks were placed in a coffin-style open display case maintained at 2°C ± 1 under continuous lighting (1,612 to 2,152 lux, Philips Deluxe Warm White Fluorescent lamps; Andover, MA; color rendering index = 86; color temperature = 3,000° K).All packages were rotated daily to minimize the effects of variation in light intensity or temperature due to location.

Surface color measurement
At each display time, the surface color was measured on the steaks assigned to that display time at 2 random locations using a HunterLab MiniScan XE Plus spectrophotometer (HunterLab Associates) with a 2.5cm diameter aperture, illuminant A, and 10° standard observer.Reflectance (R) at isobestic wavelengths from 400 to 700 nm was used to quantify myoglobin redox forms on the surface of steaks.Reflectance at 474, 525, 572, and 610 nm was converted to K/S values using the following equation: K/S = (1-R) 2 /2R.These values were then substituted into the appropriate equations (AMSA, 2012) to calculate the percentage of deoxymyoglobin (DeoxyMb), oxymyoglobin (OxyMb), or metmyoglobin (MetMb).Percentage myoglobin form values also were used to calculate MRA and OC.

Metmyoglobin Reducing Activity (MRA)
Metmyoglobin reducing activity was determined according to the procedure described by Sammel et al. (2002).Samples from the interior of steak halves were submerged in a 0.3% solution of sodium nitrite for 20 min (Sigma Aldrich Corp., St. Louis, MO) to facilitate MetMb formation, and then removed, blotted dry, vacuum packaged (Prime Source Vacuum Pouches, 4 mil, Koch Supplies Inc., Kansas City, MO), and scanned with a HunterLab MiniScan XE Plus spectrophotometer to determine pre-incubation MetMb values (AMSA, 2012).Each sample was incubated at 30°C for 2 h to induce MetMb reduction.Upon removal from the incubator, samples were rescanned to determine the percentage of remaining surface MetMb.The following equation was used to calculate MRA: (% surface MetMb pre-incubation -% surface MetMb post-incubation).Increased MRA is associated with improved color stability.

Oxygen consumption (OC)
Oxygen consumption was determined according to a modified procedure of Madhavi and Carpenter (1993), on the fresh-cut surface of the bottom half portion of the cube removed prior to MRA analysis.The samples were allowed to oxygenate for 30 min at 1°C, vacuum packaged, and then scanned twice (as described in the MRA procedure) on the bloomed surface (representing the previously unexposed interior of the original cube) to measure OxyMb.The surface color was measured through the vacuum packaging.The instrument was also standardized using the same packing material.Oxygen consumption (measured by conversion of OxyMb to DeoxyMb) was induced by incubating samples at 30°C for 30 min.Samples were rescanned immediately on removal to determine remaining surface OxyMb as a percentage by using K/S ratios and equations (AMSA, 2012).To calculate OC, changes in OxyMb values preand post-incubation were used.

Metabolomic analysis
Sample preparation.The metabolites were extracted from the muscle samples, following a modified procedure of Brown et al. (2012).Briefly, 0.5 g of intact muscle was kept in 1.5 mL of methanol (GC-MS grade, J.T Baker, Pittsburgh, PA) in borosilicate glass vials with PTFE-lined caps.The vials were then vor-texed for 30 s and incubated for 20 h at room temperature (22 to 26°C).Following incubation, the samples were vortexed for 10 s and centrifuged for 5 min at 2,000 rpm.From the supernatant, 200 μL was transferred to amber colored vials, and 2 µg ribitol was added as an internal standard.The samples were then dried under a gentle stream of nitrogen gas.
One microliter of the extract was injected using an Agilent 7683B auto sampler injector in splitless mode into an Agilent 6890 GC coupled with a 5973N MSD (Agilent Technologies).The temperature of the inlet was set at 250°C to vaporize the sample, and a splitless glass liner with a tapered bottom was used to focus the vapors to a DB-5MS GC capillary column (Agilent Technologies; 30 m × 250 µm × 0.25 µm).Ultra-pure helium (Stillwater Steel Supplies, Stillwater, OK) was used as a carrier gas at constant flow mode (1 mL/min).The oven was set to an initial temperature of 50°C for 5 min followed by a ramp of 5°C/min to a final temperature of 315°C, which was held for 3 min.The MSD was operated in electron ionization mode with transfer line and source temperatures maintained at 230°C and quadrupole temperature maintained at 106°C.Mass spectra ranging from m/z 50 to 650 were recorded in scan mode.Data were then processed by the MSD Chemstation (Agilent Technologies).

Statistical analysis
The experimental design was a split-plot with randomized block design in the whole plot (n = 10 replications).Each short loin from an animal served as a block.Within the whole plot, 10 LL and 10 PM muscles were considered as experimental units.Within the subplot, each LL and PM muscle was divided into five 1-inch steaks and randomly assigned to 0, 1, 3, 5, and 7 d (splitplot experimental unit).The fixed effects for the whole plot included muscle type (LL or PM) and random effects included error A (muscle × unit).The fixed effects for the sub-plot included display time, muscle type × display time interaction, and random effects were residual error (Error B).Surface color, pH, MRA, and OC were analyzed as a split-plot design.Type-3 tests of fixed effects for muscle, display time, and their interactions were performed using the MIXED PROC of SAS (Version 9.3.SAS Inst.Inc., Cary, NC).Least square means for the protected F-tests (P < 0.05) were separated using the PDIFF option and were considered significant at P < 0.05.
Metabolomics data from the Chemstation were deconvoluted using an Automated Mass Spectral Deconvolution and Identification System (AMDIS) program.To extract good quality data, the signal to noise ratio was set at 20, with medium settings for resolution, sensitivity, and shape requirements.A new mass spectral library was created by combining mass spectral data of compounds included in the Fiehn Metabolomics RTL Library (Agilent Technologies) and the NIST 05 library (National Institute of Standards and Technology, Gaithersburg, MD).The compounds in the samples were identified by matching the mass spectra with those in the newly created library with the minimum match factor set at 80, to rule out any possible false positives.Peak alignment, normalization, and statistical analysis of the identified compounds were performed using an Agilent Mass Profiler Professional (MPP) software (Agilent Technologies).
The intensity of the detected masses was normalized using the internal standard (ribitol), and log transformed followed by baseline transformation to the median of all samples.The compounds identified in only one sample were then omitted, and to further increase the quality of the data, only those which were present in at least 50% of samples in one condition were selected for statistical analysis.The compounds identified by AMDIS using the mass spectra of the libraries (Fiehn Metabolomics RTL Library and NIST 05) were analyzed to determine significantly different compounds based on the normalized intensity values in the samples using the MPP.Twoway ANOVA (P < 0.05) was used to find the statistical significance of differences, and to further increase the data quality, Benjamini-Hochberg multiple test corrections (P < 0.05) were applied.An unsupervised principal component analysis (PCA) and hierarchical clustering analysis (HCA) were performed to differentiate the muscle type depending on metabolites.The normalized metabolites were clustered based on muscle type and intensities using Euclidean distance and Ward's linkage rule.Our initial analysis indicated that there were no significant differences in metabolite levels between d 0 and 1.Similarly, between d 3 and 5. To avoid the complexity of data analysis and the result presentation, metabolite levels in d 0, 3, and 7 were considered.

Results and Discussion
The enzymes involved in glycolysis and TCA cycle remain active in postmortem muscles.However, substrates required to regenerate reducing equivalents are depleted continuously.Hence, characterizing the metabolite changes using the traditional wetlaboratory methods can be challenging.We utilized a GC-MS based non-targeted metabolomics approach to study metabolome profile changes in LL and PM muscles.Metabolomic techniques allow simultaneous detection of hundreds of low molecular weight metabolites such as sugars, amino acids, nucleosides, fatty acids and other compounds such as nucleotides in a biological system (Kanani et al., 2008).The application of metabolomics has been applied in diverse research areas such as human medicine, drug discovery, plant science, human nutrition, and food science (Kaddurah-Daouk and Krishnan, 2008;Wishart, 2008;Cevallos-Cevallos et al., 2009).Few studies have utilized metabolomics techniques to characterize the role of metabolites in meat tenderness and water-holding capacity (Bertram et al., 2010;Graham et al., 2010;D´Alessandro et al., 2011;Warner et al., 2015).
There was a significant muscle type main effect for pH.Psoas major had greater (P < 0.05) pH than LL (PM = 5.72, LL = 5.61, SE = 0.02).As expected, PM discolored quickly compared with LL.By d 3 of display, PM had a-e Least squares means within a row with a different superscript letter differ (P < 0.05).
x,y Least squares means within a column and trait with a different superscript letter differ (P < 0.05).
1 SE = standard error for muscle type × display time interaction.
lower redness compared with LL (Table 1).The changes in redness during display were greater (P < 0.05) for PM (d 7 -d 0 = 16.4 a* units) than LL (d 7 -d 0 = 6.2 a* units).Interestingly, both OC and MRA were greater (P < 0.05) for PM on d 0 of display than LL (Table 1).However, OC and MRA decreased more rapidly for PM by d 3 of display than LL.This suggests that oxidative changes were greater for PM than LL.Previous studies have reported color-labile nature of PM when compared with LL (McKenna et al., 2005;Joseph et al., 2012).
The concentration of muscle fiber types can influence the capillary network and metabolite use for cellular metabolism.Further, metabolism of carbohydrates, amino acids, and fatty acids are inter-related.Nevertheless, metabolism of carbohydrates can be attributed to the generation of reducing equivalents such as NADH and FADH (flavin adenine dinucleotide reduced).Mitochondrial content is greater in PM than LL (Mohan et al., 2010;Ramanathan et al., 2015) which can significantly influence the utilization of metabolites postmortem.One hundred and forty-one metabolites were identified in LL and PM.In the current study, an intensity difference of 2-log fold was considered significant, and 29 compounds were different (P < 0.05) after Benjamini-Hochberg multiple test correction.Of these, 19 compounds were found to have a fold change difference greater than 2 on a logarithmic scale.The significantly different metabolites between muscles across all display times are presented in Table 2, with P-value (after Benjamini-Hochberg correction), mass and retention time.
The principal component analysis scores plot (Fig. 1) clearly displayed the separation of metabolites between  LL and PM muscles.Both LL and PM were clustered separately.Positive loadings on the PC1 axis were obtained for PM while negative loadings were obtained for LL.The component 1 explains 55.75% variation in metabolite changes between samples (across muscles and display times) while the component 2 explains 22.12% of the variation.Further, the PCA plot indicates LL metabolite separation among d 0, 3, and 7 for the component 2. However, for PM, d 3 and 7 had little separation compared to d 0 across component 2. From the loadings plot the metabolites responsible for maximum variation were identified and the absolute loading values for component 1 are provided in Table 3. Uracil, hypoxanthine, malic acid, carnitine, and dihydroxy acetone had positive loadings indicating their effect on PM; whereas fructose, glucose-6-phosphate, methionine, and succinic acid had negative loadings explaining their influence on LL.
To visualize the metabolite changes during display, heat maps (Fig. 2) were plotted for metabolites that were different (P < 0.05).The heat map helps to view the changes in intensity levels of metabolites over the display time and between samples.The hierarchical clustering analysis indicated that there were differences between LL and PM as they sorted into different clusters (Fig. 3 and  4).The clustering analysis classifies the samples based on their similarities in metabolite intensity levels.In the  current study, LL and PM were grouped into 2 different clusters indicating a difference in metabolite intensity levels.For both LL and PM, metabolite clusters were similar on d 0 and 3, but were different on d 7.
Glycolytic compounds (fructose, glucose-6phosphate, pyruvic acid) were abundant in LL (Tables 4 and 5) than PM.Previous research also indicated that color stable muscles have predominantly glycolytic metabolism (O' Keeffe and Hood, 1982;Joseph et al., 2012).Citric acid was greater in PM on d 0 and 3 compared with LL.However, on d 7, citrate levels were greater in LL than PM.This can be attributed to the utilization of citrate in PM for mitochondrial activity.A previous study indicated that pyruvate dehydrogenase was overabundant in color stable muscle (Joseph et al., 2012).This suggests that in LL, pyruvate may be entering the Krebs cycle, and thereby, increasing citrate levels.Further, PM has a higher mitochondrial aconitase (an enzyme that converts citrate to isocitrate; Joseph et al., 2012), which may be another reason for the lower level of citrate in PM.
Succinate is a complex II substrate and is an intermediate metabolite in the TCA cycle.The addition of succinate can limit MetMb formation by electron transport mediated metmyoglobin reduction (Tang et al., 2005) and also by reverse electron transport (Belskie et  al., 2015).On d 3 and 7, succinate levels in PM were lower than LL (Tables 4 and 5).Hence, rapid discoloration in PM can be attributed to lower levels of succinate.Greater levels of succinate in LL can be due to (1) less utilization by mitochondria and (2) formation from glutamic acid or valine.In the current study, valine content in LL was greater than PM.However, further research is necessary to validate the conversion of valine to succinate in postmortem muscles.
The amino acid derivative carnitine is involved in transporting fatty acids into mitochondria for oxidation.Psoas major had greater carnitine levels compared to LL.This can be due to greater mitochondrial content in PM than LL.In contrast to the findings of Subbaraj et al. (2016), methionine levels were higher in LL.Hypotaurine, a metabolic breakdown product of cysteine and methionine metabolism, was lower in LL.Therefore, it is possible that PM and LL differ in cysteine-methionine metabolism.Joseph et al. (2012) reported that greater levels of methionine sulfoxide reductase prevented oxidation of methionine in longissimus than psoas muscles.In the current study, a greater concentration of methionine sulfoxide reductase may have increased methionine level in LL than PM.
Nucleotide degradation rate varied between muscle type.For example, there were differences in hypoxanthine and ribose-5-phosphate levels between LL and PM.These compounds are formed from the hydrolysis of inosine-5-monophosphate by nucleosidases (Koutsidis et al., 2008).Hypoxanthine can 2 Down represents on a specific day (either 0, 3, or 7) metabolite level in LL was lower compared with PM.Similarly, up represents on a specific day (either 0, 3, or 7) metabolite level in LL is greater compared with PM.

Continued
be formed from purine metabolism.An increased concentration of uracil in PM than LL is suggestive of muscle-specific differences in rates of nucleic acid hydrolysis postmortem.In the current study, polar metabolites were extracted in methanol; however, some non-polar fatty acids were also detected.This can, in part, be attributed to the ability of methanol to extract some non-polar metabolites (Lisec et al., 2006).Fatty acids such as stearic acid, palmitoleic acid, and linoleic acid were lower in LL compared with PM.Further research is required to determine the role of fatty acids in postmortem muscle metabolism and meat color.

Conclusion
The gas chromatography-mass spectrometry based non-targeted metabolomic approach was utilized to understand the differences in color stability between LL and PM during simulated retail display.The results indicate that the GC-MS based metabolomic technique is a valuable tool to study the metabolite changes in muscles during storage.The differences in metabolite utilization between LL and PM can partially explain shorter color stability of PM.The metabolomics approach will improve our understanding of post-harvest processes such as effects of aging, temperature, and packaging on meat quality.Further, this method can be utilized to validate how the metabolites change with the addition of ingredients such as lactate or succinate in postmortem muscle.Future research using targeted-metabolomics will aid in quantification of metabolite levels in muscles.The application of other omics techniques such as proteomics and genomics, in combination with metabolomics, will help to characterize the role of biomolecules in postmortem meat biochemistry.*Proceeded by the number 1, indicates the intensity value is less than 10,000 counts.The cutoff level for the signal intensity was 10,000; a lower value was considered as noise.
2 P-value represents muscles × display time interaction.

Figure 1 .
Figure 1.The PCA score plots of differentially abundant metabolites 1 in longissimus and psoas muscle during display time with Hotelling's T2 ellipse (95% confidence interval). 1Total quantitative variances of 29 annotated peaks were clustered to reveal the difference and relative similarities of metabolite profiles of muscle type during display time.X-axis represents principal component 1 and Y-axis represents principal component 2.

Figure 2 .
Figure 2. Heat map showing the changing patterns in metabolite concentrations based on normalized intensity values across muscle type and display time.The metabolites were normalized against the abundance of internal standard ribitol.The red color indicates lesser abundance and green color indicates a higher abundance.LL0 = longissimus lumborum on d 0, LL3 = longissimus lumborum on d 3, LL7 = longissimus lumborum on d 7, PM0 = psoas major on d 0, PM3 = psoas major on d 3, PM7 = psoas major on d 7.

Figure 3 .
Figure 3. Hierarchical clustering of all identified metabolites from longissimus and psoas muscles during display time.Rows indicate muscle type and day of display, and columns indicate metabolites.Clustered based on metabolites normalized against the abundance of internal standard ribitol.The red color indicates lesser abundance and green color indicates a higher abundance.LL0 = longissimus lumborum on d 0, LL3 = longissimus lumborum on d 3, LL7 = longissimus lumborum on d 7, PM0 = psoas major on d 0, PM3 = psoas major on d 3, PM7 = psoas major on d 7.

Figure 4 .
Figure 4. Hierarchical clustering of significantly different metabolites from longissimus and psoas muscles during display time.Rows indicate muscle type and day of display, and columns indicate metabolites.The metabolites were normalized against the abundance of internal standard ribitol.The red color indicates lesser abundance and green color indicates a higher abundance.LL0 = longissimus lumborum on d 0, LL3 = longissimus lumborum on d 3, LL7 = longissimus lumborum on d 7, PM0 = psoas major on d 0, PM3 = psoas major on d 3, PM7 = psoas major on d 7.

Table 3 .
Metabolites with their weighted loadings and absolute loadings for the principle component 1 (PC1) 1 Metabolites (P value < 0.05) with a fold change of 2 on a logarithmic scale are presented.