Flavor Development of Individually Vacuum-Packaged Beef Steaks During Extended Wet Aging

Product selection and subprimal fabrication

Product collection and fabrication were conducted following the methods of Barker et al. (2022). Modifications are as follows: strip loins (IMPS 180, NAMP 2014), top sirloin butts (IMPS 184, NAMP 2014), and paired tenderloins (IMPS 189A, NAMP 2014) were selected from USDA Low Choice (Small⁰⁰ to Small¹⁰⁰ marbling score) beef steer carcasses (n = 16) at a commercial processing facility. Carcass characteristics, including back fat thickness, ribeye area, marbling scores, lean maturity, and skeletal maturity, were collected by trained personnel for all carcasses and displayed in Table 1. Vacuum-packaged subprimals were wet-aged in darkness and under refrigeration (0°C to 4°C) for 28 d postmortem to mirror industry practice. Following 28 d of aging, all subprimals were fabricated to produce 2.54-cm-thick steaks (n = 480) representing the longissimus lumborum (LL), psoas major (PM), and gluteus medius (GM) per respective subprimal. From each subprimal, 10 steaks were produced. Fabricated steaks were individually allotted to VRP (Multivac Inc., Kansas City, MO) and randomly assigned to 1 of 5 aging treatments: 28, 35, 42, 49, and 56 d. Packaged steaks were stored under darkness and refrigerated conditions (0°C to 4°C; mean: 1.98°C ± 0.87°C) for their respective aging intervals and frozen at −20°C following aging.

Cooking method

Steaks were tempered for 24 h at 2°C to 4°C prior to cooking. Following methods similar to Barker et al. (2022), steaks were cooked in electric rational ovens (Model SCC WE 62G, Rational USA, Rolling Meadows, IL). Raw and cooked weights and temperatures (Thermapen, Classic Super-Fast, ThermoWorks, American Fork, UT) were recorded for all samples.

Cooked homogenate

Samples assigned to volatile analysis were immediately processed following procedures described in Barker et al. (2022). Processed cubes were flash frozen by submersion in liquid nitrogen, pulverized to a free-flowing powder (Blixer 3 Food Processor, Robot Coupe, Ridgeland, MS), and stored at −80°C.

Descriptive trained sensory analysis

Descriptive trained sensory analyses were performed following the methods of Barker et al. (2022) and the American Meat Science Association sensory guidelines (American Meat Science Association, 2015). Panelists were trained for approximately 40 h to objectively identify descriptors described by Adhikari et al. (2011) (Table 2): beef flavor identity, brown/roasted, bloody/serumy, fat-like, liver-like oxidized, buttery, fishy, umami, earthy/musty, salty, bitter, and sour. Steaks from all carcasses, aging interval, and muscle types (n = 240) were randomly evaluated over 30 panel sessions. Panels were completed over a 15-d period.

Steaks were thawed and cooked following methods previously described and following postcooking procedures listed by Barker et al. (2022). Cooked samples were held (Cambro Ultra Heated Holding Pan Carrier, 214UPCH400, Cambro Manufacturing, Huntington Beach, CA) at 50°C to 55°C for no more than 5 min prior to serving. Prior to serving, cooked steaks were cube pieces measuring the steak thickness × 1.27 × 1.27 cm (1/2” sensory box, Tallgrass Solutions Inc., Manhattan, KS).

Volatile compound analysis

Volatile compound analyses (n = 240) were performed on cooked steaks following methods similar to Gardner and Legako (2018) and Barker et al. (2022). In brief, 5 g of homogenate was placed in 20-mL glass vials (Gerstel, Linthicum, MD), and 10 μL of an internal standard (1,2-dichlorobenzene, 25 ng/μL) was added before sealing. Vials were loaded onto a Gerstel autosampler for volatile extraction, separation, and measurement by head space solid phase microextraction with an 85-μm film thickness carboxen polydimethylsiloxane fiber (Supelco, Bellefonte, PA) and gas chromatography mass spectrometry (7890B series, Agilent, Santa Clara, CA; 5977A, Agilent). Volatile compound identities were confirmed by authentic standard (Sigma-Aldrich, St. Louis, MO), retention times, and comparing unique ion fragmentation patterns. Internal standard calibration was used to quantify compounds of interest, and data were expressed as quantity extracted in nanogram per sample weight in grams.

Descriptive trained sensory analysis

There were no interactions between aging time and muscle type for any flavor attributes detected by trained panelists (P ≥ 0.062). However, there was a main effect for days of age on flavor development for 11 out of 16 attributes evaluated (P ≤ 0.033) (Table 3). As days of age increased, positively associated attributes, including beef flavor, buttery, and brown/roasted, decreased (P < 0.05) in score at day 49. Meanwhile, these attributes did not differ (P > 0.05) among samples between 28 and 42 d. Similar studies have found a decrease in signature beef flavors following extended aging beyond 35 d (Lepper-Blilie et al., 2016; Foraker et al., 2020). When considering extended aging of beef steaks, vacuum packaging may be detrimental to consumer perception as signature beef flavors decline with age (Garmyn et al., 2020). Contrastingly, those flavors that are often considered negative (i.e., liver-like, oxidized, fishy, sour, etc.) were more prevalent to panelists by day 42 (P < 0.05). Overall juiciness and tenderness did not differ over the aging period (P > 0.05). Furthermore, umami flavors did not change during aging durations, disagreeing with others suggesting umami notes tend to increase during refrigerated aging (Dashdorj et al., 2015). The increased presence of off-flavors, specifically sour notes, may be due to the accumulation of lactic acid bacteria, commonly developed in extended vacuum-packaged aging (Pierson et al., 1970; Egan, 1983). Additionally, the increased presence of off-flavors may lead to a masking effect, recognized in prior studies in which positive notes, such as umami, are often hidden by more offensive notes (Stutz et al., 1991; Jackson et al., 1992). Although this study did not evaluate the development of flavors prior to 28 d, a companion paper evaluated similar attributes following 14 d of subprimal aging, in addition of up to 10 d of retail display under vacuum as individual steaks (Barker et al., 2022). The results in said study also recognized a decrease in umami notes at 6 d in retail display (20 d postmortem and under vacuum), and an increase in off-flavors − liver-like and sour − as soon as 2 d in retail display (16 d postmortem and under vacuum).

Muscle type influenced 11 out of 16 attributes evaluated by trained panelists (P ≤ 0.040) (Table 4). Two attributes most often associated with positive eating experiences in beef − beef flavor ID and brown/roasted − were highest in LL steaks (P < 0.05), whereas GM and PM steaks did not differ for beef flavor ID (P > 0.05) and were lower than LL steaks (P < 0.05). For other positive attributes, including umami, salty, and overall juiciness, values across muscles did not differ (P > 0.05). Undesirable flavors like liver-like, oxidized, earthy/musty, and sour were highest in PM steaks (P < 0.05), intermediate in GM steaks, and lowest in the LL (P < 0.05). The PM is often considered a bland muscle, lacking strong flavors under normal conditions. However, as a muscle made up of predominantly red muscle fibers, the PM is more receptive to oxidation, proving that extended aging may accelerate the development of off-flavors as oxidation of those fibers occurs (O’Keeffe and Hood, 1982). Furthermore, the lack of stability in the PM is often noticed during retail display, when the PM continues to prove itself as a muscle lacking color stability (O’Keefe and Hood, 1982; Lanari and Cassens, 1991). Although this study did not evaluate color of the muscle described, it may be worth noting that discoloration or color stability may be related to flavor development because discoloration and lipid oxidation may be products of myoglobin oxidation (Mancini and Hunt, 2005; Suman and Joseph, 2013). It is still unclear if lipid or myoglobin oxidation occurs first, but the two often go together, and where there is oxidation, there are typically off-flavors. Furthermore, studies have noted an increase in undesirable liver-like flavors in cuts like the PM and GM as iron content and myoglobin concentrations increase (Yancey et al., 2006). Interestingly, there were no interactive effects between muscle type and aging. Thus, this implies that each factor strongly influences flavor but that the impacts of aging were similar in nature for each muscle.

Volatile compound analysis

An interaction occurred between days of age and muscle type (P = 0.021, standard error of the mean =0.34) (Figure 1) for one compound, 2-pentyl furan. The compound 2-pentyl-furan has been described as an oxidation product of linoleic acid (Ho and Chen, 1994; Mallick et al., 2021). At 28 d of aging, 2-pentyl furan was at its highest concentration (P < 0.05) in all 3 muscles but decreased at 35 d, with similar concentrations (P > 0.05) for the remainder of the aging duration. Mallick et al. (2021) further suggests that the oxidation of lipids prior to cooking may dictate 2-pentyl-furan products. It may be possible that degradation of lipids has occurred to a point that the production of this compound is no longer possible to the same extent with prolonged aging methods.

Figure 1.

Two-way interaction (P = 0.021) between muscle type (gluteus medius (GM); longissimus dorsi (LL); and psoas major (PM)) and aging duration (28, 35, 42, 49, and 56 d) for the compound 2-pentyl-furan.

Of the 56 compounds evaluated in samples, a main effect of days of aging (P ≤ 0.046) occurred for 22 compounds (Table 5). Of the 22 compounds affected, 16 were lipid-derived compounds, whereas the remaining 6 were Maillard reaction derived. Compounds including ethanol, nonanoic acid, and decane showed increased (P < 0.05) concentrations as aging duration increased. Ethanol is a common product of vacuum packaging because of the fermentation of lactic acid bacteria (Argyri et al., 2015). In a similar study, Barker et al. (2022) also recognized substantial increases in ethanol concentrations with vacuum-packaged display up to 10 d. Furthermore, Shahidi and Pegg (1994) further suggested that aldehydes, including decane, and carboxylic acids like nonanoic acid are sensitive to oxidation and are excellent indicators of flavor deterioration as they may contribute to rancid or sour notes (Burdock, 2005). Interestingly, in the current study, concentrations of both decane and nonanoic acid remained steady from day 28 to 49 (P > 0.05) but increased almost 3 times the original concentrations by day 56 (P < 0.05). Acetic acid also showed substantial increases in concentration over the aging period (P < 0.05) while also being the most prevalent compound when comparing the main effect. Like ethanol, acetic acid is often associated with vacuum-packaged products and the production of sour notes (O’Quinn et al., 2012). However, whereas sour flavors only appeared to increase by day 56 for trained panelists, acetic acid showed a steady numerical increase, although statistical significance was not shown until day 56 (P < 0.05). Contrastingly, compounds that are most often associated with positive eating experiences and beef flavor, like 3-methylbutanal and 3-hydroxy-2-butanone, decreased as aging increased (P < 0.05). Although the masking effect is theorized extensively in terms of distinguishable flavor, this provides evidence of a clear decrease in these Maillard reaction–derived compounds, aiding in positive attributes. However, many studies have also suggested that 3-hydroxy-2-butanone is also correlated with the oxidation unsaturated fatty acids (El-Magoli et al., 1996; Stetzer et al., 2008). Because the final stage of lipid oxidation is termination, it may be possible that the production of 3-hydroxy-2-butanone had increased to its full potential and could no longer increase in concentration by 56 d.

Only one compound—3-hydroxy-2-butanone—was affected by muscle type (P = 0.009) (Table 6). The LL showed the highest concentration of the compound (P < 0.05), whereas the GM and PM were similar (P > 0.05). Previous studies have recognized 3-hydroxy-2-butanone’s contribution to buttery flavors in samples with higher fat content (El-Magoli et al., 1996). Although all cuts analyzed in the current study came from the same carcasses and were all low Choice quality, there may be slight fat differences among muscles contributing to differences in concentrations. Nonetheless, intramuscular fat content in prior studies has shown no differences in GM and LL steaks (Hunt et al., 2016) or LL steaks versus other round and chuck cuts (Nyquist et al., 2018) of the same quality grade.

Discriminant function analysis

A DFA was utilized to determine maximum variation between groups while minimizing variation within groups. Variables that were highly correlated were removed from the functions, and those attributes that were less correlated remained. Loadings and standardized coefficients were determined for each treatment DFA, and models were developed according to each fixed effect (muscle type and aging duration) to further interpret relationships among variables. Both fixed effects, muscle type and aging duration, varied among treatments (P < 0.001) within their respective model.

Muscle type

The DFA corresponding to muscle type showed the canonical correlations on DF1 explained 75.7% of variation, whereas DF2 explained 24.3% of the variation (P < 0.001) (Figure 2A and 2B). The DFA explained 100% of the variation between treatments overall. The DF1 was most effective at explaining variation and showed maximal separation between the LL and GM from the PM. On the DF2 function, the LL and GM were maximally separated from each other, with the PM lying between the 2 factors.

Figure 2.

(A) Plot of muscle types (gluteus medius (GM), longissimus lumborum (LL), and psoas major (PM)) on the discriminant functions derived from the combination of trained sensory panels and volatile flavor compounds (P < 0.001). (B) Loadings of sensory attributes and volatile flavor compounds corresponding to the plot of muscle type.

The DF1 function explained much of the variation between the 2 functions. However, all the variables in the output loaded heavily on the function around the LL and GM, apart from overall tenderness, which loaded heavily on the function corresponding with the PM. Of the flavor attributes selected by the model, many could be considered negative and include sour, liver-like, oxidized, and bitter. Acetic acid was a notable volatile compound present on the DF1 function. Acetic acid is a contributor to sour flavors in beef, especially when vacuum packaged because of lactic acid bacteria (Seideman et al., 1976; O’Quinn et al., 2012). In a similar study, results from Barker et al. (2022) were similar for muscle loadings, in which acetic acid also corresponded to the GM and LL but not the PM.

Although the DF1 function explained more variation, differences between loadings on the DF2 function were clearer. The GM and LL on the function were maximally separated, with the PM between the two. Negatively associated attributes, such as sour, oxidized, liver, and fishy, weighed heaviest on the side of the function corresponding with the GM. On the opposite side of the function, positive attributes like beef flavor ID, umami, buttery, and fat-like corresponded with the canonical correlations, LL and PM. These flavors may be associated with the volatile compounds that also loaded heavily with the standardized coefficients. Ethyl benzene is a lipid oxidation product responsible for roast beef flavors (MacLeod and Ames, 1986) and may be contributing the umami and beef flavors in the PM and LL samples. Other compounds loading on the function most heavily with the PM, including 2,3-butandiol and acetic acid, suggest anaerobic fermentation occurring in those samples. As aging under vacuum increases, spoilage organisms tend to increase, eventually contributing to a loss of product quality and consumer acceptance. A study examining the shelf life of tenderloins under vacuum using total viable microorganism counts, pH, and mass spectrometry recognized a decrease in pH for tenderloins under vacuum from 0 to 21 d of aging (Mansur et al., 2019). Although the study did not hold steaks under vacuum as long as the current study, it can be speculated that lactic acid bacteria, common in vacuum-packaged beef, are responsible for the production of fermentation products and organic acids (Jalilsood et al., 2015). Hernandez et al. (2022) further confirms this during long aging methods, in which lactic acid bacteria increased over a 56-d aging period under vacuum. Furthermore, Hernandez et al. (2022) showed a substantial increase in acetic acid concentrations with extended wet aging of subprimals beyond 14 d. These data suggest extended aging in GM steaks produces fewer desirable flavors and associated compounds compared with LL and PM samples.

Aging duration

The DFA corresponding to aging duration showed the canonical correlations on DF1 explained 75.7% of variation, whereas DF2 explained 24.3% of the variation (P < 0.001) (Figure 3A and 3B). The DFA explained 100% of the variation between treatments overall. The DF1 was most effective at explaining variation and showed maximal separation between days 56 and 28, with days of age following sequential order across the function. On the DF2 function, days 56 and 49 were maximally separated from each other, with days 42, 35, and 28 falling between the 2 factors.

Figure 3.

(A) Plot of aging duration (28, 35, 42, 49, and 56 d) on the discriminant functions derived from the combination of trained sensory panels and volatile flavor compounds (P < 0.001). (B) Loadings of sensory attributes and volatile flavor compounds corresponding to the plot of display duration.

The DF1 function explained most of the variation for days of aging. Although days 56 and 28 were maximally separated across the function, the loadings corresponding to these 2 d also drastically differed from each other. The loadings corresponding with the canonical correlations, 28 and 35 d, are primarily composed of positively associated attributes and volatile compounds. Positive palatability traits such as buttery, beef flavor ID, fat-like, and overall juiciness were most associated with the shortest aging treatments, suggesting a desirable eating experience is still maintainable after 28 and 35 d of vacuum packaging. Furthermore, numerous Maillard reaction–derived volatile compounds, such as 2-ethyl-3,5/6-dimethylpyrazine and 3-hyrdroxy-2-butanone, loaded heavily corresponding with days 28 and 35 on the DF1 function. In general, pyrazines like 2-ethyl-3,5/6-dimethylpyrazine have been extensively associated with roasted foods resulting from the condensation of sugars and amino acids (Hodge, 1953; Mussinan et al., 1973). With increased aging, free amino acid availability also increases (Koutsidis et al., 2008; Foraker at al., 2020; Hernandez et al., 2022), therefore suggesting increased Maillard reaction products and thus a beefier, roasted flavor with increased aging. However, loadings on the function suggest that as aging increases beyond 42 d, off-flavors such as liver-like, earthy/musty, bitter, fishy, and oxidized become more present. This could be due to a masking effect theory, suggesting that positive notes are hidden as negative ones become more prevalent (Stutz et al., 1991; Jackson et al., 1992).

The DF2 function explains less variation in the function. However, it may provide a better timeline for individual vacuum-packaged aging by showing maximal separation between days 56 and 49. Many positive flavor attributes and volatile compounds are centered at the intersection of the 2 functions, loading heavily in correspondence with days 42, 35, and 28. On the bottom of the function, fewer flavor attributes correspond with day 49. However, those that do—beef flavor ID and brown/roasted—are considerably more positive than those loading heavily on the function with 56 d. As aging increases beyond 49 d and up to 56 d, the function shows more attributes associated with negative eating experiences—sour, oxidized, and liver-like—suggesting that individual vacuum-packaged steaks begin to noticeably decline in flavor following 49 d of age.