Experimental procedures and animal holding facilities had respected French animal protection legislation, including licensing of experimenters. They were controlled and approved by the French Veterinary Services (slaughterhouse and cattle experimental facilities license numbers are #63 345 01 and #63 345.17, respectively). The experiment was conducted on 15 Normand cull cows, of which meat is frequently consumed in France. Animals were selected on live weight (649 ± 41 kg), age (48 - 60 month-old) and body fat score (2.0 ± 0.3) estimated by visual and manual procedures. Cows were randomly assigned to three isoenergetic and isonitrogenous diets (n = 5 per diet) during a 101 ± 3 d. of finishing period. The basal diet consisted in straw (30%) and concentrate (70%); the concentrate was supplemented with 40 g of lipids/kg DM obtained from extruded linseeds (60%) and rapeseeds (40%). Animals were given either 1) the basal diet (C group, n = 5), or 2) the basal diet supplemented with vit. E as dl-α-tocopheryl acetate (INZO Company, Château-Thierry, France), at the maximal dose used in practice (155 IU/kg diet DM; 2.8 g/animal/d) (E group, n = 5) or 3) the basal diet supplemented with vit. E and the PERP mixture (Phytosynthèse Company, Riom, France) (7 g/kg diet DM; 126 g/animal/d) (EP group, n = 5). Cows were kept in groups of 4 in straw bedded pens in a single naturally ventilated shed. Pens were equipped with electronically controlled feeding gates randomly attributed to a feeding condition. Cows were equipped with an individual electronic transponder mounted on a collar opening only the defined feed gate. They received twice daily the concentrate-based diet and straw, adjusted for a target growth rate of 1150 g/day.

At the end of the finishing period, cows were slaughtered at the INRA experimental slaughterhouse (Saint-Genès-Champanelle, 63122, France). Carcasses were refrigerated at 4˚C for 24 h. Longissimus thoracis (LT) and Semitendinosus (ST) muscles from the right half carcass were removed, vacuum packed and stored for ageing at 4˚C for 10 d. (vacuum ageing). The left half carcass was whole-aged at 4˚C for 10 d. (ageing in carcass). Both muscles were cut into 10 - 15 mm (LT) and 8 - 10 mm (ST) thick steaks corresponding to the type commonly found on the French market realized by the Association for Meat Institute Development (ADIV, 63000 Clermont-Ferrand, France). Meat samples were stored at 4˚C under a standard supermarket fluorescent light. A first sample was placed in an expanded polystyrene (PSE) tray type 049405 (Boulegon-Parry, France) overwrapped in an aerobic vinyl stretchable film 9 µm thick (Soussana, France) for 4 d. (A). A second sample was placed in a polystyrene (EVOH) tray type 2450 (Form’plast, France) and packed under 70% O₂/30% CO₂ with a Multivac T200 using OPP-T504 AF/20/30 film 52 µm thick (Soussana, France) for 7 d. (MA). The packaging gas was provided by Linde-gas (France). Finally, the last sample was placed under vacuum with a Multivac C400 in a bag type 102353 (Boulegon-parry, France) for 14 d. (V). Just after the storage period, samples were used to determine color attributes.

Surface color measurements were performed after a 4 d. (A), a 7 d. (MA) and a 14 d. (V) storage period. Color measurements were realized immediately after opening of each pack in order to maintain the effects of the specific gas composition (A or MA). Visible reflectance spectra (from 360 to 760 nm) were determined with an UVIKON 933 spectrophotometer. Color coordinates were calculated in the CIE-L*a*b* system (1976) according to the method of [19] .

The relative metmyoglobin at the surface of meat after 4 d. (A), 7 d. (MA) and 14 d. (V) storage was measured by spectrophotometry as described by [19] , which allows the proportions of met-, oxy- and myoglobin fractions forms to be calculated. Reflectance at 473, 525, 572, 580, 630 and 730 nm wavelengths were measured using UV-V is spectrophotometer. According to the mathematical method of [19] , metmyoglobin% of total pigment in meat was calculated by using 1395-((R572-R730)/(R525-R730)) equation and oxygenation index by R630-R580 calculation.

Mean values of meat were calculated by using the STATISTICA software Version 12 (DELL, Texas, USA, 2015). Data were analyzed by an ANOVA fitting diet, muscle, ageing and packaging conditions and their interactions as fixed effects. Whatever the diet or the muscle type, no significant differences were observed on tested values between under vacuum or in half carcass meat ageing. Color deterioration of stored beef was not influenced by the two ageing methods tested, which agreed with our previous evaluation of lipid oxidation intensity showing no differences between the two ageing conditions [7] . Consequently, ageing effect was not considered in the model. Color coordinate b* was not analyzed, because the dataset did not follow a normal distribution. The fixed effects (diet, muscle and packaging) were calculated based on 10 samples repetitions.

Meat visual appreciation was carried out by evaluation of different parameters. In our study, average of L* value from processed LT in C group (Table 1(a)) was 37.7, which was slightly higher than L* value of 30.6 determined for LT muscle just after slaughter in Normand cows by [20] corresponding to a dark red meat. In this study the processing period could have increased meat lightness, since full bloom, based upon low lightness L*, occurs immediately after the muscle went into rigor mortis [21] .

In the CIE Lab system, positive a* value is associated to meat redness. Here, a* values from retailed meats stayed under 21 (Table 1(b)). According to [21] , the maximum redness of muscles from Normand cows amounted to 25.7 just after slaughter. The redness slight decrease observed in our samples resulted probably from the meat ageing and packaging exposing meat to air and light [6] . Moreover, cows of this study were relatively fat with a higher body score condition at

Significant differences (P < 0.05) between LT and ST muscles were reported in the results and discussion section. Means values reported at the bottom of the table are the means of the corresponding column (all the packaging and muscle conditions in a same diet). ^a-bMeans bearing different superscripts are significantly different at P < 0.05 and highlight the diet effect. NS = not significant.

slaughter than standard Normand cows (on average 3.5 vs. 3.0). By this way, fatty meat could exhibit slightly different color values than standard meats. The two colors represented by b* (blue and yellow) are not typical or intuitively related to beef color. Nevertheless, b* coordinate was more correlated to the brown color as described by sensory panelists than to the sensory evaluation of blue and yellow descriptors [22] . Unfortunately, b* data set did not follow a normal distribution and was not analyzed by ANOVA procedure.

The oxygenation index (R630-R580) represents the bright red meat impression and the jury’s preference for a meat [20] . As previously reported by [12] , the mean value of the oxygenation index obtained in our study was 15.2 (Table 2).

For meat metmyoglobin percentage, the mean value from the LT and ST muscles in the C group (30.6%) (Table 3) was higher than 22.3% reported by [12] , but close to 29.2% reported by [23] . Such differences can be explained by the different muscle type or ageing duration twice lower in the [12] study. In our cows given PUFA rich diets, meat color differed from that of steers used in the two other reported studies. This could be in relation with the fatty composition of muscle or the effect of terminal lipid oxidation products on the myoglobin stability. Yin and Faustman [24] showed an increase of heme and lipid oxidation with the increase of FA unsaturation. In the same way, [25] reported that 4-hydroxy-2-nonenal (a by-product of 18:2 n-6 peroxidation) induces a redox instability of oxymyoglobin present in cattle skeletal muscle via its adduction to

Significant differences (P < 0.05) between LT and ST muscles were reported in the results and discussion section. Means values reported at the bottom of the table are the means of the corresponding column (all the packaging and muscle conditions in a same diet). ^a-bMeans bearing different superscripts are significantly different at P < 0.05 and highlight the diet effect. NS = not significant.

Significant differences (P < 0.05) between LT and ST muscles were reported in the results and discussion section. Means values reported at the bottom of the table are the means of the corresponding column (all the packaging and muscle conditions in a same diet). ^a-bMeans bearing different superscripts are significantly different at P < 0.05 and highlight the diet effect. NS = not significant.

histidine residues. Consequently, meat samples rich in n-3 PUFA were more susceptible to lipid oxidation [7] and could be more susceptible to myoglobin oxidation and consequently had less stable color than meats from no PUFA supplemented cattle. Metmyoglobin values from our experimentation stayed relatively close to that reported in the literature. These values are in acceptable limits, since meats containing 40% metmyoglobin or more are downgraded or rejected for purchase [6] .

Mean L* value from ST was 12% higher (P < 0.05) than LT muscle in C group (Table 1(a)). Moreover, oxygenated index from ST was 11% higher (P < 0.05) than that of LT muscle (Table 2). In contrast, no significant muscle effect was observed for metmyoglobin percentage (Table 3). Each muscle has an inherent biochemistry and hence varied in their biochemical and color attributes [26] [27] . Differences could be attributed to differences in composition (collagen content or lipid composition) or in fiber types between these muscles.

In the C group, mean L* values from retailed meats (Table 1(a)) were 17% higher than value obtained at slaughter by [21] , but largely under threshold of tolerance for red meat which is of about 50 to 55 [20] . So, our beef can be considered as very acceptable for consumers. The significant packaging effect on L* value was not clear regarding data evolution. The a* values showed no significant differences between the different packaging systems. High levels of O₂ were known to support oxymyoglobin formation [22] and therefore to promote lipid oxidation, which can eventually adversely affect color especially in animals given PUFA rich diet such as in the present study. However, no significant color differences between V, A and MA packaging were observed (Table 1(b) and Table 2). Metmyoglobin percentages in beef under V was 7% and 9% lower than that in A and MA packaging, respectively (P < 0.05) (Table 3), probably in relation with the absence of O₂ [28] . Meat under V, having purple color, seemed to be better protected against discoloration than that in A and MA packs. These observations were in accordance with results obtained on lipid oxidation measured on the same samples [7] showing a higher antioxidant status of meat after retail storage in V vs. A and MA packaging systems. This provides evidence that under vacuum meats were better protected against lipid oxidation and color deterioration than meats in contact with oxygen. A and MA packed meats were more susceptible to lipid oxidation [7] associated to a slight deterioration of color, according to metmyoglobin percentages (Table 3). A meat protection must be provided against these two deteriorative reactions in these packaging systems.

In the present study, the studied variables indicated no improvement of color stability of PUFA rich meat by vit. E. In cows from E group (given 2.3 g α-tocopherol/animal/d for the finishing period) α-tocopherol content in processed beef amounted 4.2 µg/g fresh tissue, which was 2.6 times higher than beef from the control group (means between ST and LT muscles) [7] . This data and those of the present study were in agreement with data from [12] study on steers given 1 g vit E/animal/d with a mean α-tocopherol content in processed beef (from longissimus lumborum and triceps brachii) of 5.3 µg/g (1.4-fold higher than in control animals) but without any significant improvement of color parameters (a*, R630-R580, MetMb percentage). Nevertheless, several studies reported that vit. E is able to delay beef color deterioration [8] [9] [10] . [11] reported that vit. E content must be about 3.5 µg/g to assure a good retail of beef from animals without lipid supplement. In the present study, beef vit. E content was higher than this threshold, but did not provide any effective protection against meat color deterioration. This could be explained by the higher beef n-3 PUFA content increasing risk of lipid oxidation and color degradation, these two reactions being interrelated [29] [30] [31] . [11] reported that dietary vit. E (2 g/kg diet) can protect stored beef against lipid oxidation and color deterioration simultaneously from no lipid supplemented animals. Currently, vit. E is able to protect meat from lipid oxidation, with its color-stabilizing effect indirectly by delaying oxymyoglobin oxidation through the inhibition of lipid oxidation [29] . However, vit. E alone was not enough effective to limit lipid oxidation intensity in all stored meats from n-3 PUFA supplemented animals [7] . These results might explain the lack of protective effect of vit. E towards color deterioration in the present study.

Increasing vit. E level in the diet would be a recommended strategy to limit both meat lipid oxidation and color deterioration. However, an excess of vit. E will be catabolized or excreted [32] . In human, high dietary vit. E doses did not improve protection and moreover could have an adverse effect by increasing mortality [33] . Indeed, it had been reported that high α-tocopherol concentration would increase in vitro mitochondrial O₂ consumption [34] . This effect would increase production of reactive oxygen species (ROS) from mitochondria in aqueous phase [35] involving higher risks of lipid oxidation and consequently, a possible decrease of myoglobin stability. In case of n-3 PUFA rich diets, vit. E was not sufficient to effectively limit lipid oxidation and color deterioration in stored beef and a high α-tocopherol supplementation of cattle could have deleterious effects. Thus, other antioxidants such as hydrosoluble polyphenols would be necessary to reinforce the vit. E protective action.

L* coordinates (Table 1(a)) were 6% higher (P < 0.01) in meats from the EP group when compared to E group. Values stayed in common standards for retail beef. Redness (a*) (Table 1(b)) was not altered by dietary PERP. The oxygenation index (Table 2) for all tested packaging systems was 14% higher (P < 0.05) in beef from the EP group as compared to E group indicating a better meat oxygenation (redness recovery) in these steaks, that could be attractive for consumers. In all packaging systems (V, A and MA), PERP with vit. E were able to significantly protect meats towards discoloration as regarding metmyoglobin percentage (−7%, P < 0.05) (Table 3) when compared to values from C group, whereas vit. E alone was not (+7.5%, NS). We previously demonstrated that similar doses of vit. E + PERP effectively protected PUFA rich beef against lipid oxidation as shown by meat malondialdehyde (MDA) concentrations systematically lower than 0.8 µg/g tissue in all stored meats [7] . Consequently, this antioxidant mixture was able to stabilize beef color from PUFA rich beef by limiting lipid oxidation, as already proposed for vit. E in standard meats [29] . We calculated the correlation between MDA content from PUFA rich meat from ST muscle stored under modified atmosphere packaging, which represent the most deleterious conditions regarding lipid oxidation [7] (Figure 1). MDA concentration explained 26.6% of metmyoglobin percentage variability. Reactions of myoglobin with other lipid oxidation products, such as hydroxyalkenals (4-hydroxy-2-nonenal (4-HNE) and 4-hydroxy-2-hexenal (4-HHE)), could explained a resting part of variability in the formation of metmyoglobin via oxidation [25] . The later in vitro study suggested that 4-HNE accelerates oxidation of oxymyoglobin in bovine skeletal muscle by covalent connection with histidine residues. It was hypothesized that histidine 64 residue by 4-HNE could alter the oxymyoglobin structure around the heme cleft and subsequently impact the redox stability. Recently, [36] reported that hydroxyalkenals, final products of n-3 PUFA

(4-HHE) and n-6 PUFA (4-HNE) peroxidation, are strongly reactive. Generated in the hydrophobic part of cells, hydroxyalkenals would disperse within the cells and attack distant targets in other compartments, such as hydrophilic part of cell. Oxymyoglobin, located in the hydrophilic phase of muscle cells, could be attacked by 4-HHE and 4-HNE, which react readily with nucleophilic molecules such as thiols and amines. We hypothesized that the antioxidant association of PERP with vit. E would limit meat lipid oxidation intensity by limiting the formation of these aldehydes and consequently could have a protective action on red oxymyoglobin towards effect induced by covalent linkages. PERP activity could also limit the myoglobin oxidation by direct ROS uptake in the aqueous phases. Indeed, it is well known that polyphenols such as those of the PERP mix could be present at the aqueous/lipid interface [37] , near to the hydrophilic myoglobin. By this way, they would limit lipid oxidation and oxidation of myoglobin by the capture of ROS produced in aqueous phases.