Meta-proteomics for the discovery of protein biomarkers of beef tenderness: an overview of integrated studies

This meta-proteomics review focused on proteins identified as candidate biomarkers of beef tenderness by comparing extreme groups of tenderness using two-dimensional electrophoresis (2-DE) associated with mass spectrometry (MS). We reviewed in this integromics study the results of 12 experiments that identified protein biomarkers from two muscles, Longissimus thoracis (LT) and Semitendinosus (ST), of different types of cattle: young bulls, steers or cows from beef breeds (Charolais, Limousin, Blond d’Aquitaine), hardy breed (Salers) or mixed breed (PDO Maine-Anjou). Comparative proteomics of groups differing in their tenderness evaluated by instrumental Warner-Bratzler shear force (WBSF) or by sensory analysis using trained panelists, revealed 61 proteins differentially abundant (P < 0.05) between tender and tough groups. A higher number of discriminative proteins was observed for LT (50 proteins) compared to ST muscle (28 proteins). The Gene Ontology annotations showed that the proteins of structure and contraction, protection against oxidative stress and apoptosis, energy metabolism, 70 family HSPs and proteasome subunits are more involved in LT tenderness than in ST. Amongst the list of candidate biomarkers of tenderness some proteins such as HSPB1 are common between the 2 muscles whatever the evaluation method of tenderness, but some relationships with tenderness for others (MYH1, TNNT3, HSPB6) are inversed. Muscle specificities were revealed in this meta-proteomic study. For example, Parvalbumin (PVALB) appeared as a robust biomarker in ST muscle whatever the evaluation method of tenderness. HSPA1B seems to be a robust candidate for LT tenderness (with WBSF) regardless the animal type. Some gender specificities were further identified including similarities between cows and steers (MSRA and HSPA9) in contrast to bulls. The comparison of the 12 proteomic studies revealed strong dissimilarities to identify generic biomarkers of beef tenderness. This integrative analysis allowed better understanding of the biological processes involved in tenderness in two muscles and their variations according to the main factors underlying this quality. It allowed also proposing for the first time a comprehensive list of candidate biomarkers to be evaluated deeply to validate their relationships with tenderness on a large number of cattle and breeds.


Introduction
Beef tenderness is considered as the most important quality trait for consumer acceptability of fresh meat (Miller, Carr, Ramsey, Crockett, & Hoover, 2001). This quality trait strongly influences consumer satisfaction and thus the repeat of purchase, particularly of beef, at the point of sale (Grunert, Bredahl, & Brunsø, 2004;Troy & Kerry, 2010). It is well documented that the origin of tenderness is multifactorial Ouali et al., 2013). It depends on the structural and biochemical properties of the muscles that develop and evolve throughout the life of the animal under the influence of factors related to the animal itself (age, sex, breed) and its rearing practices Gagaoua, Picard, Soulat, & Monteils, 2018;Listrat et al., 2016). Furthermore, processes such as postmortem rigor development and ageing are dependent on these factors and could be modified by pre-slaughter (such as diet, handling, temperament, and stress), slaughtering and postmortem conditions (pH and temperature decline, chilling regime, hanging method and carcass handling). This diversity of factors from the farm-to-fork (Gagaoua, Monteils, Couvreur, & Picard, 2019; explains why this quality trait is very variable and why there is difficulty to control or to predict it accurately despite the many research works carried out since decades. Several technologies such as electrical stimulation, pelvis hung, tenderstretch, tendercut, pulsed electric field or super stretch (Bhat, Morton, Mason, & Bekhit, 2018;Y. H. B. Kim, Warner, & Rosenvold, 2014;Sørheim & Hildrum, 2002;Suwandy, Carne, van de Ven, Bekhit, & Hopkins, 2015;Warner et al., 2017) have been proposed to control the post-mortem conversion of muscle into meat and consequently increase tenderness. However, these technologies are not always used by industrials. Moreover, the available techniques to measure tenderness can only be applied after the death of the animal on aged meat, and they cannot be used online and in serial time on large number of samples. Today, tenderness is evaluated by sensory or mechanical analysis and the relationship between these two methods is variable (Gagaoua, Bonnet, Ellies-Oury, De Koning, & Picard, 2018). It is therefore necessary for both stakeholders and researchers to develop simple and fast tools applicable at slaughter or on the living animal, to evaluate tenderness. This type of tool could have applications to evaluate the potential tenderness of an animal or a cut of meat, to manage its marketing according to this potential and to develop genetic selection schemes based on this quality trait. For this purpose, numerous works, such as OMICs have been proposed during the last two decades.
Genomics has been used to discover and identify without a priori biomarkers of tenderness by applying approaches developed in medicine (Rifai, Gillette, & Carr, 2006). A biomarker can be defined as a measurable indicator of a specific biological state. Among OMICs tools, proteomics offers large interest to identify biomarkers because proteins and their enzymes entities determine the phenotypic diversity arising from a set of common genes. Furthermore, posttranslational modifications (PTMs) revealed by proteomics regulate structure, function, localization, maturation and turnover of proteins. Accordingly, Rifai and co-workers proposed a specific proteomic pipeline to identify biomarkers for clinical use, diagnosis or to monitor the activity of diseases and guide molecularly targeted therapy or assess therapeutic response (Rifai et al., 2006). This pipeline is composed of 6 steps « discovery, qualification, verification, research assay optimization, clinical validation and commercialization ». Therefore, the same pipeline was adapted by meat scientists for the discovery of beef tenderness biomarkers . The first phase (Figure 1) is the « discovery » considered by Rifai et al. (2006) as "an unbiased semi-quantitative process by which the differential expression of specific proteins between states is first defined". The products of the discovery phase are lists of twenty to hundred proteins found to be differential between the two compared situations or conditions. These putative differentially abundant proteins can be considered as candidate biomarkers of the studied trait. Then, the relationships between the abundances of the putative biomarkers and the trait(s) of interest (phenome: a set of phenotypes) have to be evaluated with specific tools (Gagaoua, Bonnet, Ellies-Oury, et al., 2018;Lopez-Pedrouso et al., 2019;Picard, Gagaoua, Al Jammas, & Bonnet, 2019;. At the end, the validated biomarkers may be selected for the development of diagnostic test for commercialization (Kraus et al., 2011;Pierzchala et al., 2014;Marinus FW Te Pas, Hoekman, & Smits, 2011;M. F. Te Pas et al., 2013;Woelders, Te Pas, Bannink, Veerkamp, & Smits, 2011).
The aim of the present meta-proteomics was to review for the first time the studies conducted under identical laboratory conditions during the last decade to identify candidate protein biomarkers of beef tenderness by comparing groups of samples of high versus low tenderness for a given muscle from cattle raised in similar conditions. These studies have been applied mainly in Longissimus thoracis (LT) muscle but also in Semitendinosus (ST) from young bulls, steers and cows grouping different breeds. Groups of meat samples differing in their tenderness have been constituted on the basis of sensory analysis or mechanical measurements (Warner-Bratzler Shear force) according to each proteomic study.
Thus, two-dimensional electrophoresis (2-DE) combined with mass spectrometry to separate and identify the proteins of the tender versus tough meat samples were used. We report in this meta-proteomic a comprehensive list of specific or common candidate biomarkers to several muscles and genders. These data are firstly a source of new knowledge to better understand the biological processes and pathways involved in beef tenderness and also a list of candidate biomarkers of tenderness that once validated can be proposed to develop routine evaluation tools according to the pipeline of biomarkers discovery. Furthermore, these data are the first to highlight the major protein biomarkers of beef tenderness that can be identified whatever the treatments and techniques cited above.
For each proteomic study, LT, a mixed fast oxido-glycolytic muscle sampled from 6 th and 7 th rib and ST, a mixed fast glycolytic muscle sampled from the center of the muscle, were excised from the carcasses 15 min p-m for 6 studies (1, 2, 5, 9, 10 and 12), 60 min p-m for 2 studies (3 and 11) and 24h p-m for 4 studies (4, 6, 7 and 8) as described in Table 1. The muscle samples were immediately frozen in liquid nitrogen and stored at -80°C until protein extractions for proteomic analysis.
For sensory analysis or mechanical measurements (WBSF) of tenderness, other muscle samples were taken at 24 h p-m, and aged for 14 (for 9 studies) or 21 days (for 3 studies) at 4°C (Table 1), then the chilled muscles were cut into steaks of about 1.5 cm thick, packed under vacuum and kept frozen at -20°C until analysis as previously described (Gagaoua, Bonnet, Ellies-Oury, et al., 2018).

Pipeline of beef tenderness biomarkers discovery
The standard proteomic workflow described in  for the identification and discovery of protein biomarkers of beef tenderness was followed in this meta-proteomics (Figure 1). Briefly and according to the large literature, the main strategy that includes 6 steps used for the identification of biomarkers of meat quality including tenderness is based on comparative proteomics (Picard, Gagaoua, & Hollung, 2017). Therefore, two-dimensional electrophoresis (2-DE) (for review: (Görg, Weiss, & Dunn, 2004)) has been applied to analyze protein extracts from groups showing extremes (high vs. low) tenderness in order to reveal protein spots of interest with differential abundances.
Step 1 The first step of the pipeline of beef tenderness biomarkers discovery is the selection of extremes (contrasting) sample groups (Figure 1). According to the experimental design and facilities of each study, we used sensory panelists to assess meat cooked at end-point cooking temperature of 55°C as usual in France (Gagaoua, Terlouw, Richardson, Hocquette, & Picard, 2019) using an unstructured scale from 0 to 10 or Warner-Bratzler shear force on cooked meat as a routine instrumental measure for tenderness to select extreme tender (TE) and tough (TO) samples from each experiment population. The details of the evaluation methods of tenderness are briefly summarized in the supplementary data of Table S2. The two techniques were applied on the whole 120 meat cut samples as described in previous studies (Dransfield et al., 2003;Gagaoua, Bonnet, Ellies-Oury, et al., 2018;Wheeler, Shackelford, & Koohmaraie, 1996) to select a minimum of 2 samples per tenderness group for studies 9, 10 and 12; 3 for studies 1 and 11; 4 for studies 2 and 5; 5 for studies 3 and 4; and 10 for studies 6, 7 and 8 ( Table 1). The number of samples taken in each study varied from one project to another depending on the funding's and objectives of each research project. WBSF alone (mostly evaluated in N/cm² and in kg for study 3 only) was used to evaluate the meat samples of studies 1 -5, sensory panelists alone were used for studies 9 -12 and both sensory and instrumental methods were used for studies 6 -8.

Step 2
The second step is sample and protein extracts preparation (Figure 1). Briefly, muscle proteins were extracted following the same procedure using frozen samples (Bouley, Chambon, & Picard, 2004). The frozen muscle tissue was homogenized in a lysis buffer containing 8.3 M urea, 2 M thiourea, 1% DTT, 2% CHAPS and 2% IPG buffer pH 3 -10 and centrifuged at 10000 g for 30 min. The supernatant was harvested and the concentration of extracted protein was determined using the 2-DE Quant kit (Amersham Uppsala, Sweden).
All protein samples were handled at 4°C to minimize the proteolytic activity.
Step 3 The protein extract samples were then subjected in the third step to 2-DE analysis following the same protocol Picard et al., 2016). Briefly, 800 μg of proteins were first subjected to isoelectrofocalisation (IEF) on 18 cm length IPG strips pH range 4 -7 (Amersham-Biosciences, Uppsala, Sweden) in a Multiphor II gel apparatus at a temperature of 20.5°C. After a desalting step (50V, 7h), proteins were separated according to the following conditions: 200V for 1h, 200V increasing to 8000V over 5h, 8000V continuously until 73500Vh. Then, proteins were separated in the second dimension by SDS-PAGE on 12%T, 2.6%C separating polyacrylamide gels using Hoefer DALTsix system. 2-DE gels from the whole 12 studies were stained with G250 Colloidal Coomassie Blue for 72h.
After staining, gels were rinsed for 1h in fresh distilled water before digitization. Triplicate gels from each protein extract were analyzed. Spot detection and quantification were performed with ImageMaster 2D Elite software on scanned gels (scanner UMAX, Amersham Pharmacia Biotech) at 300 dots per inch (dpi). Image warping, spot detection and volume quantification were realized using the SameSpots v3 software (Nonlinear Dynamics, Newcastle, UK). For each experiment, a gel corresponding to tender meat was selected as reference. 2-D gel data for each study were normalized by dividing each spot volume by the total volume of all the matched spots in the 2-D gel image, to obtain a normalized spot volume value. The proteins differentially abundant between tenderness groups were identified by mass spectrometry using MALDI-TOF (Matrix-Assisted Laser Desorption/Ionizationtime-of-flight).
Step 4 The fourth step of the pipeline is the statistical comparisons of the protein spots between groups. Accordingly and for each proteomic study, the electrophoretic profiles of the two groups of tenderness (TE and TO) were all compared using the statistical approach described in Meunier et al. (2005). It is based on the use of the Significance Analysis of Microarray (SAM) method based on a modified t-test. This approach considers not only the individual variability but also the average variability of proteins with the same level of expression than the considered protein. Among the differentially abundant proteins between the 2 groups of tenderness, only the proteins with a fold change >1.5 were retained (ratio of abundance levels between the 2 groups).
Step 5 In the fifth step, significantly differential spots were carefully excised by hand from at least three different replicate gels and placed in clean Safe-Lock Tube (Eppendorf, Hamburg Germany). Each tube allowed spot distaining and acetonitrile (ACN) drying. First, the spots were washed in 25 mM ammonium bicarbonate 5% ACN for 30 min and twice in 25 mM ammonium bicarbonate 50% ACN for 30 min each. The spots were then dehydrated with 100% ACN.
Step 6 In the last and sixth step, and for protein identification by MALDI-TOF, the dried gels were re-swelled in 25 mM ammonium bicarbonate and digested at 37 °C for 5 h with 10 to 15 µL (depending on the gel volume to be treated) of trypsin solution (12 ng/µL; V5111, Promega, Madison, WI, USA). Generally, 8 to 12 µL of ACN (depending on gel volume) was added to extract the peptides. Trypsin digested peptides solution was diluted in equal amount of matrix (5 mg/mL alpha-cyano-4-hydroxycinnamic acid (CHCA) in 50% ACN/0.1% Trifluoroacetic acid (TFA)). Peptide mass fingerprints (PMF) of trypsin-digested spots were determined in a positive reflectron mode using a Voyager DE Pro MALDI-TOF (Perspective Biosystem, Farminghan, MA, USA). External calibration was performed with a standard peptide solution (Proteomix, LaserBio Labs, Sophia-Antipolis, France). Internal calibration was performed using peptides resulting from autodigestion of porcine trypsin with protonated masses of 842.509, 1045.564, and 2211.104. PMF were compared to Bos taurus Swiss Prot database or to other mammalian using MASCOT software. PMF of unidentified proteins (using bovine database) were compared to the nrMammalia database. The search criteria used were one missing trypsin cleavage site, partial methionine oxidation, partial carbamidomethylation of cysteine, and a mass deviation lower than 30 ppm. We required at least four matched peptides per protein for identification, using MASCOT probabilistic scores and accuracy of the experimental theoretical isoelectric point (pI) and molecular weight (MW). Protein spots with more than one protein identified in the mixture were considered for biological interpretation only if they passed both of the following criteria: i) the relative abundance based on the number of sequenced peptides per protein (EmPAI protein content (mol%) for the most abundant protein was larger than 50%; and, moreover; ii) the ratio of the EmPAI for this most abundant protein to EmPAI for the second most abundant protein in the spot mixture was at least 2.

Meta-proteomics database description and bioinformatics
Across the 12 proteomic experiments, a total of 61 putative protein biomarkers (unique gene names) of beef tenderness from young bulls, cows and steers grouping different continental breeds evaluated by sensory panelists or WBSF on Longissimus thoracis (LT) and Semitendinosus (ST) muscles were differentially abundant (P < 0.05) between tender and tough groups ( Table 2). Considering the 12 experiments, the number of differential proteins between extreme tenderness groups (Table 2) varies from 3 (study 12, ST muscle from young Salers bulls categorized by sensory panelists) to 22 proteins (study 2, LT muscle from young Charolais bulls categorized by WBSF). A higher number of putative protein biomarkers was observed for LT muscle (n = 50) compared to ST with 28 differential proteins (Table S3 and   Table S4). Panther screening of the whole proteins allowed their clustering into 6 biological pathways ( Table 2) Gene ontology annotations of proteomic data was realized using the ProteINSIDE workflow tool (Kaspric, Picard, Reichstadt, Tournayre, & Bonnet, 2015) to analyze the 61 proteins and gather biological information provided by functional annotations. The proteinprotein interactions (interactomics) of biological function of the proteins by muscle or for both LT and ST muscles were realized using STRING webservice database (http://stringdb.org/). Default settings of confidence of 0.5 and 4 criteria for linkage: Co-occurrence, experimental evidences, existing databases and text mining were used. Furthermore, the ProteQTL tool included in ProteINSIDE (http://www.proteinside.org) was used to search for tenderness Quantitative trait loci (QTL) among the list of putative protein biomarkers of tenderness. This tool interrogates a public library of published QTL in animal science: CattleQTLdb (Hu, Park, Wu, & Reecy, 2013) that contains cattle QTL and association data curated from published scientific papers.
The degree of similarity between the protein biomarkers among the 12 proteomic studies to explain tenderness was estimated as the percentage of proteins shared based on the F coefficient (López-Pedrouso, Bernal, Franco, & Zapata, 2014): F = 2nxy/(nx + ny). Where nxy is the number of tenderness biomarkers shared by study x and y; and nx and ny are the total numbers of biomarkers retained in studies x and y, respectively (Gagaoua, Terlouw, et al., 2019). Then, a pairwise of biomarker distance (D) expressed as a percentage based on the abundances of the biomarkers under the different factors was built as estimated by D = 1 -F.

Gene Ontology annotations results of the protein biomarkers of beef tenderness
The corresponding Gene Ontology annotations: cellular components, biological pathways and molecular functions are illustrated in Figure 2A,B for LT and ST muscles, respectively.
The distributions of the molecular functions are deeply described in Figure 3. Considering the two muscles together, the most represented functions are: catalytic, metabolism and ATP metabolic process (26%) and contractile and associated proteins (25%) followed by chaperones and heat shock proteins (18%) (Table 2 and Figure 3). Oxidative stress and cell redox homeostasis represents 12% of the biological functions in LT and only 4% in ST muscle (Figure 3). Proteasome and associated proteins were found only in LT muscle (4%).
The interactomes constructed with the differential proteins for each muscle illustrated the main biological functions (Figure 4A,B). The contractile proteins MYL1, MYL2, MYH1, TnnT1, TnnT3, WDR1 and MYBPH are common to both muscles. On the contrary, structural proteins such as DES, CAPZB, CAPZA3 and ACTA1 were differential proteins in only LT muscle. For chaperones and Heat shock proteins pathways, 3 small HSP (HSPB1, HSPB6 and CRYAB) as well as STIP1 and YWHAE were differential in both 2 muscles. However, proteins of the HSP 70 family: HSPA1B, HSPA1A, HSPA2, HSPA8 and HSPA9 were specific to LT muscle. HSPA5 was found differential in only ST muscle. Concerning the catabolic and metabolic pathways, 3 proteins (TPI1, PGM1 and NDUFS1) were differential in both LT and ST muscles. A large number of glycolytic and oxidative proteins were found differential and specific to LT muscle ( Figure 4A), and two proteins: COX4I1 and SUCLA2 were differential only in ST muscle. The interactome in LT muscle illustrates numerous interactions among the different biological pathways and putative biomarkers. The oxidative stress and cell redox homeostasis in ST muscle is represented by only PDIA3 protein which was differential also in LT muscle with 5 other proteins: MSRA, GSTP1, PRDX6, PARK7 and SOD1. Among the proteins of the binding, cofactor and transport proteins, only ALB was common to the two muscles with connections with all other biological pathways whatever the muscle. The other proteins of this cluster are specific to the considered muscle ( Figure 4A,B).

Description of the results for each factor of the meta-proteomics
As stated above, the number of proteins is variable from one study to another and was also found to be variable from one biological pathway to another (Table 2 and Figure 3). The comparison of the common proteins among the 12 proteomic studies in terms of their redundancy revealed a strong dissimilarity based on the pairwise biomarker distance ( Figure   5). The dissimilarity was particularly strong for muscles as there was an unbalance between the number of proteins identified for LT and ST muscle. The pairwise distance given in Figure 5 highlights the strong dissimilarities to identify generic biomarkers of beef tenderness if we refer to the factors included in this study.
The direction of the putative protein biomarkers with tenderness is also variable ( Table 2,   Table S3 and Table S4). Taken the 16 proteins belonging to catalytic, metabolism and ATP metabolic process together, 18 times positive and 14 times negative directions were observed between the proteins and tenderness for both LT and ST muscles and whatever the other factors ( Table 2). For LT (13+, 11-) and ST (5+, 3-) muscles separately, the same trend was observed. From the second pathway of 15 contractile and associated proteins, a high number of associations was found for both muscles taken together with 23 times positive, 16 times negative and 2 times with both positive and negative directions. For LT muscle alone, there were 12 times positive, 13 times negative and 2 times both directions. For ST muscle, more positive (11 times) than negative (3 times) directions were found. In the third biological pathway of chaperones and heat shock proteins grouping 11 proteins, an equivalent 15 times associations of positive and negative directions was found with 5 times both signs depending of the isoform mostly of HSPB1 (4 times) and HSPB6 (1 time). For LT muscle alone, more negative association (12 times) than positive (9 times) were found with the same 5 in both directions. For ST muscle, few associations were found, with 6 times positive and 3 times negative associations with tenderness. In the fourth biological pathway grouping 11 binding, cofactor and transport proteins for the two muscles, 10 times positive and 5 times negative associations with tenderness were found ( Table 2). For LT muscle, less association compared to ST were reported: 2 positive vs. 8 and 4 negative vs. 1, respectively. In the fifth pathway grouping 6 proteins belonging to oxidative stress and cell redox homeostasis process, 6 times positive and 5 times negative associations were reported for both muscles. Most of the associations were for LT muscle 5 times positive and 5 negative, compared to only 1 positive association for ST. finally, the 2 proteasome and associated proteins reported in this metaproteomic were specific to LT muscle and were both positive.

Gender effect
In Charolais LT muscle and among 3 proteomic studies, 28 proteins were differentially abundant between groups of tenderness evaluated by WBSF: 22 proteins in young bulls, 9 in steers and 10 in cows (Figure 6). Only two proteins HSPA1B and TPI1 were found to be differentially abundant for the 3 genders. The abundance of HSPA1B was always lower in tender meat for the three genders. The abundance of TPI1 was lower in tender meat of young bulls and higher in tender meat of steers and cows. These differences could be explained mainly by differences in sex hormones between the genders. Indeed, the effect of estrogens on skeletal muscle properties has been largely studied (Enns & Tiidus, 2010;Sauerwein & Meyer, 1989) with important regulation of energy metabolism pathways, including glucose transport, glycolysis and tricarboxylic acid cycle (Y. Xu & López, 2018).
On another hand, two proteins: TnnT3 and YWHAE were found differentially abundant only in males but inversely in steers and young bulls. In young bulls, TnnT3 was lower abundant in tender meat, while in steers it was higher. Two proteins MSRA and HSPA9 (GRP75) were differentially abundant in groups of divergent tenderness in steers and cows but not in young bulls. MSRA was higher abundant and GRP75 was lower abundant in groups of high tenderness. Five proteins were common to cows and young bulls: HSPB1 and MYBPH in the same direction, and ENO1, MYH1 and PGM1 in opposite direction. In cows, PGM1 and ENO1 were higher abundant in tender group and MYH1 (MyHC-IIx) was lower abundant, the opposite was observed in young bulls. ENO3 was higher abundant in tender LT only of cows.

Breed effect
Considering LT tenderness groups only evaluated by sensory panelists in young bulls from 3 breeds, a total of 26 proteins were found differentially abundant: 8 in Charolais (CH), 14 in Limousin (LI) and 15 in Blond d'Aquitaine (BA) (Figure 7). Two biomarkers, ACTA1 and HSPB1 were common to the three breeds. PGM1 and TnnT3 were lower abundant in tender LT in BA and CH but not different in LI. The oxidative protein PARK7 was higher abundant in tender LT in CH but lower abundant in LI and not different in BA breed. Four proteins were common to Li and BA: HSPB6, CRYAB, CAPZB and ENO3. This later protein was always lower in abundance for both LI and BA tender group and, the three other proteins showed inverse relations with tenderness.
In ST muscle, 17 proteins were differentially abundant between the groups of tenderness evaluated by sensory analysis in young bulls and 9 proteins were identified for Charolais and Blond d'Aquitaine, 4 for Limousin and 3 for Salers (Figure 8). This number of divergent proteins is lower than observed in LT muscle. Among the putative biomarkers of ST muscle, PVALB was found for CH, BA and LI breeds and never for Salers. Three proteins were common between CH and BA young bulls: HSPB1, MYL1 and TnnT3 were higher abundant in tender ST from these 2 breeds. This is coherent with earlier studies indicating that ST muscle is tenderer when it is likely fast glycolytic oriented . HSPB1 was differently abundant between groups of tenderness but inversely between the two breeds.

MYL2 was lower abundant and MYBPH higher abundant in tender ST in LI and CH breeds.
If we consider only the population of cows, from CH and PDO MA breeds, 21 putative protein biomarkers were found differently abundant between groups of tenderness evaluated by WBSF, 17 in PDO MA and 10 in CH cows (Figure 9). Six proteins, ENO1, ENO3, MYH1, HSPB1, HSPA1B and MSRA, were common to the two cow breeds. ENO1, ENO3 and MSRA were higher abundant and MYH1 and HSPA1B were lower abundant in the tender LT. HSPB1 was inversely related with tenderness in the two cow breeds. Trim72 was found differentially abundant between groups of tenderness in only PDO MA breed in which it was lower abundant in the tender LT steaks (Figure 9).

Muscle type effect
The comparisons of ST and LT muscles from young Charolais bulls allowed identifying a total of 34 differentially abundant proteins between groups of tenderness evaluated by WBSF ( Figure 10). A total of 22 proteins were identified in LT versus 16 in ST. Among them, 4 biomarkers, TPI1, HSPB6, HSPB1 and YWHAE, were common to the two muscles. HSPB6 was higher abundant in the tender group whatever the muscle. The 3 others were inversely related with tenderness in the two muscles. TPI1 and YWHAE were higher abundant in tender ST and lower abundant in tender LT. The inverse was observed for HSPB1. An inverse relationship between some proteins such as HSPB1 and tenderness was already reported in the large literature. For example, the results illustrated in Figure 10 highlighted that in LT muscle the proteins related with tenderness correspond mainly to contractile, structural and Heat shock proteins whereas in ST the proteins related with tenderness are mainly involved in metabolism in coherence with the Gene Ontology of Figure 2.

Effect of evaluation method of tenderness
For LT muscles of young Charolais bulls, 22 proteins were differentially abundant between groups of tenderness evaluated by WBSF and 8 between groups of tenderness evaluated by sensory panelists (Figure 11). Among a total of 25 proteins, 5 biomarkers, PGM1, ACTA1, TnnT3, HSPB1 and PARK7, were common to the two evaluation methods of tenderness.
PGM1 and TnnT3 were lower abundant in tender LT whatever the methods used.
For ST muscles of young Charolais bulls, 16 proteins were differentially abundant between groups of tenderness evaluated by WBSF and 9 between groups of tenderness evaluated by sensory panelists (Figure 12). Three biomarkers, MYL2, HSPB1 and PVALB, were common to the two evaluation methods of tenderness. Among them, HSPB1 and MYL2 were lower abundant and PVALB higher abundant in the tender groups whatever the evaluation method.

Overall discussion
This integrative meta-proteomics study which gathered 61 proteins identified following the same procedure from one laboratory, provides new insights and robust results on the putative protein biomarkers of beef tenderness. It allowed us to identify the main proteins that are common to several muscles, animal types (gender) and evaluation method of tenderness, as well as those that are specific to each of these factors. This integrative review allows a better understanding of the biological pathways involved in beef tenderness that would explain its variability according to muscle, breed, gender or evaluation method of tenderness. For the two muscles considered in this review that are Longissimus thoracis (LT) and Semitendinosus (ST), a higher number of differential proteins discriminating tender versus tough meat groups were found for WBSF compared to sensory panelists. In the following discussion sections, we first compared the list of the candidate biomarkers (by biological family) identified in the 12 proteomic studies to those of the literature by linking each putative biomarker to some earlier proteomic studies and then, we focused on the major biomarkers candidates of great interest (reported more than 4 times in the 12 studies) deserving further evaluation and validation studies following the pipeline of tenderness biomarkers discovery (Figure 1).

Brief comparison of the list of the putative biomarkers to bovine proteomic studies from the large literature
During the last two decades, meat scientists used OMICs tools more precisely proteomics, to identify accurate biological markers of beef tenderness that would make possible the categorization of meat cuts soon after slaughter on the basis of their potential final tenderness and/or to propose those biomarkers to optimize the genetic selection of beef animals on the basis of this quality. As highlighted above and following the guidelines of several studies in the large literature D'Alessandro & Zolla, 2013;Gagaoua, Bonnet, Ellies-Oury, et al., 2018;Gagaoua, Terlouw, et al., 2019;Longo, Lana, Bottero, & Zolla, 2015;Mullen, Stapleton, Corcoran, Hamill, & White, 2006;Ouali et al., 2013;Picard et al., 2015;Picard, Lefevre, & Lebret, 2012), 6 distinct but interconnected biological pathways were identified Guillemin, Bonnet, et al., 2011;Picard et al., 2018).
The 61 biomarkers identified in this meta-proteomics and their directions (positive, negative or both) with mechanical (WBSF) or sensory beef tenderness are given in Table 2.
The discussion of the most interesting candidates according to the corresponding biological functions in which they are involved allow producing new insights about tenderness determinism. For instance, in the following sections, a brief description for each pathway and a comparison to the large literature are given. The directions will not be discussed in the following sub-sections and only non-exhaustive cattle references from the large literature for reported putative protein biomarkers by biological family are given.

Biomarkers belonging to oxidative stress & cell redox homeostasis
Among

Biomarkers belonging to proteasome & associated proteins
The 2 putative protein biomarkers of this biological pathway (PSMB2 and PSMC2), were both earlier reported to be related to beef tenderness (Chaze et al., 2013;Ouali et al., 2013).

Quest for quantitative trait loci (QTL) from the list of the putative protein biomarkers
Among the list of the 61 proteins candidate biomarkers of tenderness, the ProteQTL tool (http://www.proteinside.org/) revealed 15 of them (shown by blue ovals for both muscles in the networks) for which a QTL for WBSF or sensory tenderness has been reported (Figure 4).
The 15 QTL proteins belong to the following biological functions:

Important role of HSPB1 and HSPB6 in beef tenderness
Earlier meat proteomic studies whatever the quality trait and the species, have confirmed the implication of heat shock proteins (HSPs) including small, large and co-chaperones (For From the most interesting HSPs proteins identified in this meta-proteomics and whatever the muscle, the small HSPB1 (HSP27) and HSPB6 (HSP20) proteins were reported 9 and 5 times respectively from the 12 studies. The former was identified in different breeds and irrespective of the evaluation method of tenderness and the later was specific to young bulls but found irrespective of the same factors of HSPB1 (Table 2). However, the relations with tenderness for the two sHSP were not always in the same direction, a phenomenon that we already reported in previous studies (Gagaoua, Terlouw, et al.,  interaction between all these proteins during meat tenderization. Moreover, in silico analysis identified the main interactors of HSPB1, among them YWHAE, also found differential between groups of tenderness (Guillemin, Bonnet, et al., 2011) and discussed below (subsection 7.3.4). It is worthwhile to note, that differences in proteins interacting with HSPB1 were observed between Gastrocnemius (fast glycolytic) and Soleus (slow oxidative) muscles, suggesting muscle-specific effects and thus the role that this protein would play in tenderness might differ from muscle cuts with different metabolic and contractile properties . Overall, these data suggest that HSPB1 seems to play a central role in beef tenderness determinism by interacting with five main protein families (HSPs, pro/antiapoptotic factors, translation factors, cytoskeletal proteins, and oxidative proteins) in accordance with the interactomes and correlation networks of tenderness proposed by (Guillemin, Bonnet, et al., 2011) and .
HSPB6, or HSP20, is another sHSP retained 5 times whatever the muscle for young bulls.
Accordingly, it can be proposed as for HSPB1, as a potential biomarker of tenderness due to the role it can play in muscle to meat conversion ( Furthermore, in skeletal muscle, this protein protects against atrophy, ischemia, hypertensive stress, and metabolic dysfunction (Dreiza et al., 2010). As HSPB6 has a binding domain to troponin 1 (Rembold, Foster, Strauss, Wingard, & Eyk, 2000), it could affect skeletal muscle contraction through the troponin complex. Moreover, in obese subjects with decreased insulin sensitivity, it promotes blood flow and increases glucose utilization during exercise (Y. Wang, Xu, & Cooper, 1999). So, HSPB6 is involved both in muscle contraction and metabolism. Phosphorylation of HSPB6 also leads to the disruption of the actin cytoskeleton. Therefore, its role in tenderness could be through these biological functions in different types of muscles.

Troponin T Fast, a biomarker of tenderness irrespective of muscle and gender
TnnT3 was found differential between groups of tenderness in 7 of the 12 proteomic studies ( Table 2). After HSPB1, it is the most frequently identified differential biomarker in this meta-proteomic review. In LT muscle of young bulls, TnnT3 was less abundant in tender meat and the inverse was observed for steers. Similar to LT muscle of steers, a higher abundance in tender ST meat of young BA and CH bulls evaluated by sensory panelists was observed. A difference between genders was further observed for sensory tenderness as Troponin T has been used as a marker of ongoing proteolysis (Iwanowska et al., 2010;Muroya et al., 2004) due to the appearance of polypeptides giving a product of 30 kDa and whose quantity progresses during the tenderization process and was found correlated to meat tenderness (Ho, Stromer, & Robson, 1994 Picard et al., 2018;Picard et al., 2014). Differences in line to this meta-proteomics among breeds or genders were further reported (Chaze et al., 2013;Rodrigues et al., 2017;. For example, the recent study by Silva et al. (2019) found greater abundance of fast troponin T isoform (TNNT3) in steers LT muscle, while slow troponin T isoform (TNNT1) tended to be lower, compared to bulls.

PGM1, a biomarker of toughness in bulls and of tenderness in cows
Phosphoglucomutase 1 (PGM1) plays a central role in glucose homeostasis (Anderson, Lonergan, & Huff-Lonergan, 2012;Anderson et al., 2014) and its identification as a biomarker of beef tenderness is in agreement with the idea that in the early post-mortem period, glycolysis will continue, thereby affecting different quality traits of meat. PGM1 catalyzes the interconversion of glucose-1-phosphate and glucose-6-phosphate through the intermediate glucose-1,6-bisphosphate and thus plays an important role in the regulation of glycogen metabolism. Several posttranslational modifications (PTMs) through phosphorylation, acetylation and methylation of PGM1 were reported (Anderson et al., 2014).
For example, 9 sites on PGM1 have been identified as phosphorylation sites and these include 2 serine, 3 tyrosine, and 4 threonine residues. PTMs were proposed to influence metabolism and post-mortem muscle to meat conversion (Huang et al., 2011;Huang, Larsen, & Lametsch, 2012;Huang, Larsen, Palmisano, Dai, & Lametsch, 2014;Jiang, Liu, Shen, Zhou, & Shen, 2019;Li et al., 2015;Zhang et al., 2016). In fact, previous proteomic studies by  and  suggested that phosphorylation of muscle proteins, including PGM1, can be responsible for the differences in the final beef tenderness of steaks. Moreover, PGM1 is among the QTLs that were identified in this review whatever the muscle (Figure 4). We found this enzyme to be the lowest in young bulls (5 times) whatever the muscle and the highest in LT Charolais cows (1 time). Accordingly and in line to an earlier work, we reported that protein extracts from tender beef samples has less PGM1 compared to proteins extracted from tough beef steaks (Picard et al., 2010). It is worthwhile to note, that most proteomic studies identified multiple isoforms of PGM1 and their relationships with the targeted meat quality differ among isoforms according to the studied factors that are muscle, animal types, stress statue of the animal or response to postmortem electrical stimulation (Anderson et al., 2014;Bjarnadottir et al., 2010;Laville et al., 2009;Li et al., 2015).

TPI1 and YWHAE as biomarkers of tenderness whatever the muscle but with inverse relationships
Triosephosphate isomerase (TPI1) was identified only in Charolais breed, with lower abundance in tender LT of young bulls but not in ST and higher abundance in tender LT of steers and cows evaluated by WBSF. TPI1 is a glycolytic enzyme playing an important role in energy generation. It is the last enzyme of the first preparatory phase of glycolysis where it catalyzes the reversible conversion of D-glyceraldehyde 3-phosphate from phosphate dihydroxyacetone. Proteomic studies cited in sub-section 7.1.1 reported mainly positive and also negative associations with beef tenderness Gagaoua, Bonnet, Ellies-Oury, et al., 2018;Grabez et al., 2015;Rosa et al., 2018). For example, it has been reported to be up regulated in LT tender meat of bulls by D'Alessandro,  and Grabez et al. (2015). TPI1 gene was found to be significantly higher expressed in marbled cattle and proposed by several studies as a biomarker of intramuscular fat (Bazile, Picard, Chambon, Valais, & Bonnet, 2019;Ceciliani, Lecchi, Bazile, & Bonnet, 2018;N. K. Kim et al., 2009;Mao et al., 2016). From these, one can speculate that greater abundances in steers and cows known to be fatter than young bulls  could explain the inverse relationships with tenderness observed in the animal groups of this meta-proteomics. TPI1 including its phosphorylation were further related with post-mortem muscle pH decline rate, comforting its role in meat tenderization Huang et al., 2011).

YWHAE was reported 4 times in this review and mostly with samples of LT and ST
muscles of Charolais bulls and Steers evaluated by WBSF and also by sensory analysis in BA bulls ( Table 2). The gene of this protein encodes tyrosine 3-monooxygenase/tryptophan 5monooxygenase activation protein, 14-3-3 epsilon. The 14-3-3 proteins (phospho-serine/threonine binding proteins) belong to a large, highly conserved family functioning as a dimer in diverse biological processes such as signal transduction, metabolism, protein trafficking, signal transduction, apoptosis, cell cycle regulation and potassium channel regulation (Jin, Lyu, Kozak, & Jeang, 1996). This protein was reported as an interesting biomarker of beef tenderness in very recent studies (Chaze et al., 2013;Mahmood et al., 2018;Rodrigues et al., 2017). It was further reported by E. B. Carvalho et al.
(2019) as a differential biomarker related to energy metabolism in skeletal muscle of beef cattle identified for low and high residual feed intake. This would explain as for TPI1, its link with tenderness of meat. Indeed, an earlier phosphoproteomic study reported that 14-3-3 proteins are able to regulate glucose homeostasis in response to insulin or to energetic stress (Ogihara et al., 1997). Furthermore, YWHAE is known to negatively regulate apoptosis (J. Kim et al., 2012), an important phase of muscle to meat conversion (Ouali et al., 2013;Ouali et al., 2006) with major consequences on final tenderness (Gagaoua, Hafid, et al., 2015). It was also reported to interact with HSPs such as HSPB1 as described in Picard et al. (2016).
The results allowed to propose this protein as a candidate biomarker of mechanical tenderness in both LT and ST muscle mainly in males as it was higher abundant in tender ST and LT steers and lower abundant for LT young bulls, but not differential in cows.

Importance of structural MYH1, MYL1, MYL2 and MYBPH in beef tenderness
Myosin heavy, light and binding protein chains were identified as expected (Lana & Zolla, 2016;Mato et al., 2019) to play important role in the determinism of meat tenderness of young bulls and also of cows in both LT and ST muscle. Most of these proteins are grouped in the M-band which consists of a series of parallel electron dense lines in the central region of the A-band that comprises a filament system cross-linking the myosin filaments (For review: (Ertbjerg & Puolanne, 2017)). MYL1 was identified to be related positively only (3 times) in LT muscle of cows and ST of CH and BA of young bulls. Inversely, MYL2 that is also identified with a tenderness QTL, and negatively was less abundant in tender meat whatever the evaluation method of tenderness and specific to young CH bulls (3 times) in LT and ST muscles and one time for ST muscle of LI bulls. MYBPH was less abundant in LT muscle of CH young bulls and cows, and highly abundant in ST muscle of LI and CH young bulls. Finally, MYH1 was negatively linked with tenderness in cows (LT muscle of CH and PDO Maine-Anjou cows) and positively in young bulls (ST and LT muscles of CH and BA young bulls). Myosins are the most abundant proteins of thick filaments and interaction of these structural proteins occur through the actomyosin complex that associates myosin light chains with myosin heavy chains and actin (Anderson et al., 2012). The identification of the 4 structural proteins as good biomarkers of beef tenderness agrees with several studies that we cited in sub-section 7.1.2 and further summarized in some recent comprehensive reviews (Ouali et al., 2013;. Among some studies, recent cattle proteomics on LT muscle showed similar findings with MYL1 that was reported in high abundance in less tender meat (Franco et al., 2015;Rodrigues et al., 2017;Rosa et al., 2018).
In accordance to the findings of this meta-proteomics, Bjarnadottir et al. (2012) reported that MYL1 was more abundant in muscle biopsies from tender meat, while MYL2 was present in less abundance. The identification of MYL2 only in young bulls agrees with the findings by Rodrigues et al. (2017) showing increased muscle abundance in LT muscle of leaner breeds.
Another work by Franco and co-workers identified that several muscle myosin light chains (MYL3 and MYL6B) and regulatory light chain 2 isoforms (MYL2 and MYLPF) participate in the conversion of cattle muscle to DFD meat in Rubia Gallega animals affected by preslaughter stress (Franco et al., 2015). In line to our findings namely the protein-protein networks of both muscles, the authors found also that all these structural proteins appeared to be involved in the main network of functionally associated proteins. Indeed, this is not a surprising finding when the myosin is a major structural protein of the muscle sarcomere in association with actin and other contractile proteins. Accordingly, all these structural proteins, well discussed in the comprehensive review by Lana and Zolla (2016), deserve further evaluation following the pipeline of biomarkers discovery and under other conditions to validate their robustness as biomarkers of beef tenderness.

HSPA1B, a robust candidate biomarker of WBSF whatever the gender
HSPA1B, a large HSP protein, for the three genders (young bulls, steers and cows) of different breeds (mostly CH and PDO Maine-Anjou) was less abundant in meat samples of high tenderness evaluated by WBSF only,. This protein belongs to the most highly conserved 70 kDa HSPs family playing key roles as molecular chaperones and involved in promoting cellular proteostasis and survival during periods of stress (Mayer, 2013). They further regulate various cell signaling pathways involved in cell growth and inflammation (Daugaard, Rohde, & Jaattela, 2007). By interacting with other HSPs, HSPA1B protein stabilizes existing proteins against aggregation and mediates the folding of newly translated proteins in the cytosol and in organelles (M. E. Carvalho et al., 2014). It is particularly involved in the ubiquitin-proteasome pathway in accordance with the 2 proteasome subunits PSMB2, PSMC2 found differential between groups of tenderness only in LT muscle in the present review.
These data suggest a role of ubiquitin-proteasome pathway in tenderness mainly in muscles with mixed oxido-glycolytic properties as LT. As cited above in sub-section 7.1.3 and not only for HSPA1B but also for the other 70 kDa HSPs proteins (HSPA1A, HSPA8 and HSPA5), dozen studies reported their involvement in meat tenderness under several factors such as breed (Chaze et al., 2013;Keady et al., 2013;Picard et al., 2014;Rodrigues et al., 2017), gender/animal type Picard et al., 2019;Silva et al., 2019), rearing practices , muscle type Picard et al., 2014), evaluation method of tenderness  and country origin of the panelists and endpoint cooking temperature of meat (Gagaoua, Terlouw, et al., 2019). Recently, we proposed three main actions by which 70 kDa HSPs (HSPA1B) would intervene in post-mortem muscle and cooked meat, namely by i) responding to cellular stress since these proteins has been reported to be induced in skeletal muscle, ii) binding to the structural proteins to maintain homeostasis, or iii) protection roles of post-mortem muscle cells (Gagaoua, Terlouw, et al., 2019;.

ACTA1, a candidate biomarker specific to LT tenderness of young bulls and cows
ACTA1 or α-actin, a structural protein was related 5 times with tenderness in LT muscle only of young bulls whatever the breed (CH, LI and BA) and PDO Maine-Anjou cows. It was more abundant in the tender group samples and its direction varies in LI and CH breeds evaluated by WBSF and sensory LT of BA breed ( Table 2). We previously stated that structural proteins, in general, are more related to tenderness in LT than in ST muscle, in coherence with the data of this review (Lana & Zolla, 2016;. According to Ouali et al. (2013), actin release from muscle cytoskeleton suggests apoptosis onset where caspases initiate with other interacting proteases the breakdown of actin into fragments of different molecular weights (Goll, Neti, Mares, & Thompson, 2008;Ouali et al., 2013). ACTA1 is among the structural protein biomarkers of beef tenderness that were mostly identified in proteomic studies (Beldarrain et al., 2018;Bjarnadottir et al., 2012;Franco et al., 2015;Gagaoua, Bonnet, Ellies-Oury, et al., 2018;Lana & Zolla, 2016;Laville et al., 2009;Morzel et al., 2008;Polati et al., 2012;Zapata et al., 2009). From these, ACTA1 deserves special attention as a robust biomarker of beef tenderness as it is further considered as a hallmark of apoptosis (Ouali et al., 2013). Indeed, during apoptosis, ACTA1 is the first protein targeted by effector caspases and some actin fragments have been considered to be accurate markers of apoptosis (Mashima, Naito, & Tsuruo, 1999;Yang et al., 1998). According to experiments by Du et al. (2004) and during apoptosis, the initial step in myofibrillar proteolysis would be the breakdown of the actomyosin complex by direct action of caspase 3. The authors pointed out that both intact monomeric actin and fragments of actin appear since early post-mortem including bovine muscle (Ouali et al., 2013). In a second step, these products will be then further degraded by the ATP-dependent ubiquitin-proteasome system, in line with the identification of the two proteasome subunits PSMB2 and PSMC2 found differential between groups of tenderness only in LT muscle of the same breeds where greater abundance of ACTA1 was found.

ENO1 and ENO3 highlight similarities between young bulls and cows
Enolase 1 (ENO1) and 3 (ENO3) are two isoforms catalyzing the conversion of 2phosphoglycerate to phosphoenolpyruvate during glycolysis. ENO1 is ubiquitous in cell cytoplasm, whereas ENO3 is mainly present in skeletal muscle. These two enzymes are predominately reported in proteomic studies to be related to beef tenderness and other qualities. From those studies and as highlighted in the sub-section 7.1.1, we report for ENO1 the studies by Gagaoua, Bonnet, Ellies-Oury, et al., 2018;Polati et al., 2012) and for ENO3 those by (Bjarnadottir et al., 2012;Gagaoua, Bonnet, Ellies-Oury, et al., 2018;Marino et al., 2014;Polati et al., 2012;Zhao et al., 2014). In this meta-proteomics and in line with the findings of most of the above references, the two enolases were found 4 times. ENO1 was positively related with tenderness in LT muscle of cows (CH and PDO MA) and LI young bulls but negatively for CH young bulls. ENO3 was as for the former positively related with tenderness for cows whatever the breed (CH and PDO MA) but negatively for young bulls whatever the breed (BA and LI). The link of these two enzymes with meat tenderness was reported to be isoform specific . The associations of tenderness with enolases may partly reflect a cellular stress response to the deprivation of oxygen supply and to glucose metabolism (low glucose levels) (Sedoris, Thomas, & Miller, 2010). Indeed, these last authors reported that ENO3 expression is influenced by differences in glucose levels and deprivation of oxygen supply. It is worthwhile to note that ENO3 participates in multi-enzyme complexes present on the sarcomere (Keller et al., 2000) and their role in the tenderization process may further depend on the presence of other proteins identified in this meta-proteomics such as HSPs (Wulff, Jokumsen, Højrup, & Jessen, 2012).

HSPA9 and MSRA highlight similarities between steers and cows
HSPA9 (Stress-70 protein, mitochondrial GRP75) and MSRA (Mitochondrial peptide methionine sulfoxide reductase) were identified only in LT muscle and were specific to steers and cows ( Table 2). The former was negatively related with tenderness and further reported one time and positively with sensory LT of young Limousin bulls. The later was positively related to tenderness of CH steers and cows and PDO Maine-Anjou cows.
HSPA9 is a member of Hsp70 family of chaperone proteins that is not heat-inducible.
HSPA9 identified also as a tenderness QTL is known as mitochondrial HSP70, GRP75 or mortalin, that is predominantly localized in mitochondria but also in other cellular compartments such as endoplasmic reticulum, plasma membrane and cytoplasm (Kaul, Deocaris, & Wadhwa, 2007;Liu, Liu, Song, & Zuo, 2005). It is described as a multifunctional DnaK-type chaperone involved in many biological processes such as transportation of nucleus-encoded proteins to the mitochondrial matrix, cell protection against oxidative stress and apoptosis, import and translocation of cytosolic proteins, control of cell proliferation, muscle activity and proteasomal degradation of proteins (Kaul et al., 2007). This protein is the only chaperone described to be regulated by glucose privation, Ca 2+ homeostasis and perturbation of glycolysis (Mayer, 2013). Accordingly, we previously found HSPA9 to be related to pH decline with strong correlation with pH3h and ultimate pH (Gagaoua, Terlouw, Micol, et al., 2015). Numerous proteomic studies identified this protein as a good biomarker of beef tenderness (Grabez et al., 2015;Jia et al., 2009;Rodrigues et al., 2017). HSPA9 is able to enhance Ca 2+ trafficking from endoplasmic reticulum toward mitochondria and cytosol (Glancy & Balaban, 2012).
In previous studies in cattle where this protein was identified as an interesting biomarker of beef qualities Gagaoua, Terlouw, Micol, et al., 2015;, we suggested that increased Ca 2+ levels may exert regulatory effects on many proteolytic systems, namely, μcalpain and proteins involved in metabolic and apoptotic processes. Accordingly, a biological mechanism involving HSPA9 in muscle to meat conversion was described . The differential abundance of HSPA9 between groups of tenderness only in cows and steers and in LT muscle could mean that the protection against apoptosis and oxidative stress is higher in oxidative than in glycolytic muscles. The same hypothesis could be proposed for PARK7, or DJ-1, identified as a candidate biomarker of tenderness ( Table 2) only in LT muscle in coherence with an interaction between these two proteins .
MSRA is one of two mammalian methionine sulfoxide reductase isoforms. It is ubiquitous and highly conserved with mitochondrial and cytosolic forms translated from a single gene with localization dependent on alternative translation initiation sites. It is involved in the repair of oxidatively damaged proteins to restore their biological activity. Its main function is to convert methionine sulfoxide formed after methionine oxidation due to exposure to reactive oxygen species (ROS), to reduced methionine (Bin, Huang, & Zhou, 2017). Therefore, the action of MSRA has many cellular and physiological consequences as protection of cells against oxidative damage (Kantorow, Lee, & Chauss, 2012). The normal distribution of MSRA in the mammalian cell is approximately three folds greater in the cytosol than in mitochondria, hence suggesting potential role as antioxidant repair enzyme but with important role in apoptosis (Hunnicut, Liu, Richardson, & Salmon, 2015).
This meta-proteomics is the first to identify MSRA as a tenderness biomarker, but it was recently reported for beef color (Joseph, Suman, Rentfrow, Li, & Beach, 2012;Wu et al., 2016). The role of MSRA in tenderness and other meat quality traits could be through its cellular response to oxidative stress induced by post-mortem physiological conditions. It is well documented that oxidative stress induces damage to cellular proteins compromising their functions and leading to cellular dysfunction and finally cell death Ouali et al., 2013;L. L. Wang et al., 2018;Xing, Gao, Tume, Zhou, & Xu, 2019). Some authors described associations between MSRA with lipids and lipid-binding proteins (Lim, Kim, & Levine, 2017;Y. Y. Xu et al., 2015). This could partly explain why MSRA was found to be positively related to tenderness only in steers and cows which are known to be fatter than young bulls . We could speculate that in muscles with high intramuscular fat content that are likely with greater oxidative metabolism, oxidative stress is high and proteins such as MSRA would play pivotal role.

Major beef tenderness biomarkers specific to ST muscle
PVALB or Parvalbumin, was found in this review (4 times) specifically related to tenderness in ST muscle and only in young bulls whatever the breed and evaluation method ( Table 2). To our knowledge, this meta-proteomic study is the first to report PVALB as a biomarker of beef tenderness. It was further identified by the ProteQTL tool of ProteINSIDE (http://www.proteinside.org/) as tenderness QTL. In Charolais (in 2 experiments) and Limousin young bulls, PVALB was higher abundant in tender groups a PVALB is a low molecular mass (9 -11 kDa) calcium binding protein that regulates muscle contraction and relaxation (Cai et al., 2001). Several studies showed that PVALB is located exclusively in the sarcoplasm of type II (fast-twitch) mammalian skeletal muscle fibers (Muntener, Kaser, Weber, & Berchtold, 1995). The basic mechanism underlying the contraction-relaxation cycle of vertebrate muscles is based on a Ca 2+ exchange between the sarcoplasmic reticulum and the myofibrils. Relaxation is achieved by retrieval of Ca 2+ from the myofibrils and transfer to the sarcoplasmic reticulum. For example, mice with knockout of PVALB cannot properly regulate their intracellular Ca 2+ concentration and require longer time to attain peaktwitch tension than their littermates (Cai et al., 2001). Its affinity to Ca 2+ is higher than troponin C, but less than the sarcoplasmic reticulum Ca 2+ ATPase pumps. From these, the role of PVALB in tenderness could be explained by its role in muscle relaxation through calcium binding and reduction of intracellular Ca 2+ . The fact that PVALB is expressed in fast fibers could explain why it discriminated groups of tenderness in ST muscle and not those of LT. It is well documented in cattle that ST contains a higher proportion of fast fibers than LT . Moreover, muscles of the hardy Salers breed are characterized by higher proportions of slow oxidative fibers than the other CH, LI or BA breeds (Jurie et al., 2005), thereby, explaining why PVALB was not identified as a discriminating biomarker of Salers tenderness groups. On another hand, it has been reported that ST muscle from continental and French beef breeds are tenderer when they are likely fast glycolytic . This is totally coherent with a higher abundance of PVALB in this type of muscle.

Conclusion and future perspectives
This integromics work of comparative proteomic studies from one laboratory aimed to identify robust candidate biomarkers of beef tenderness. Combined with data about published tenderness QTL and proteomics studies from the large literature, the results of this review allow selecting among the list of 61 potential candidates those, which are more frequently differential between groups of tenderness according to the evaluation method of tenderness, muscles differing in their contractile and metabolic properties and breeds for three genders (young bulls, steers and cows). This meta-proteomics highlighted the importance of myriad biological pathways, namely the involvement of heat shock proteins with both small and large HSPs such as HSPB1 and HSPB6 identified whatever the factors described above. On another hand, HSPA1B, was found as a robust candidate biomarker of WBSF whatever the gender.
Structural proteins were as expected identified in several studies and the most interesting for validation seemed to be Troponin T Fast (TnnT3), MYH1, MYL1, MYL2 MYBPH and ACTA1. Among the metabolic enzymes, PGM1, was found as a biomarker of toughness in bulls and of tenderness in cows and ENO1 and ENO3 highlight similarities between young bulls and cows. Among the list of the biomaekers, some of them highlighted muscle specificities such as Parvalbumin (PVALB) foud a major beef tenderness biomarker specific to ST muscle. Taken all together, the selected candidates are proposed for a deep analysis in the step "qualification and verification" of the pipeline of biomarkers discovery (Rifai et al., 2006) in order to validate their relationships with tenderness on a large number of samples and using several proteomics based approaches.

Highlights
x First meta-proteomics gathering a comprehensive list of candidate biomarkers of beef tenderness.
x Integration of 61 putative biomarkers of beef tenderness from 12 proteomic studies from the same laboratory.
x Biomarkers of Longissimus thoracis and Semitendinosus muscles were grouped into 6 biological pathways.
x Biological mechanisms underpinning muscle to meat conversion.
x Robust biomarkers under several factors of genders, breeds, muscles and evaluation method of tenderness.
x Hsp27 (HSPB1) is an interesting biomarker of beef tenderness.

Declaration of Competing Interest
Authors declared no conflict of interest.