The GENOTEND chip: a new tool to analyse gene expression in muscles of beef cattle for beef quality prediction

Design of the GENOTEND chip

Whenever possible, several probes were used for each gene (e.g., 17 probes for DNAJA1). On average, four to five different probes were designed for each gene of interest.

Animals and samples

This study took place with two groups of 25 and 30 Charolais young bulls and two groups of 21 and 19 Charolais steers (castrated at 3 months of age) slaughtered in two different years (years 1 and 2) from the French MUGENE programme (Genanimal-APISGENE 2005-2009). All animals proceed from the same experimental farm (Bourges) and were progeny of the same seed stock sample. The young bulls were reared in the same experimental farm (Bourges) and slaughtered at 17 months on average. The steers were reared in the experimental farm “Le Pin au Haras” and slaughtered at 30 months on average. All animals were slaughtered in the same conditions and without any electrical stimulation at the INRA experimental slaughterhouse in compliance with current INRA ethical guidelines for animal welfare [22]. Muscle samples from the Longissimus thoracis were excised within 15 minutes after slaughter. Muscle samples were immediately frozen in liquid nitrogen and stored at -80°C until RNA extraction.

In addition, steaks for each muscle were aged at 4°C for 14 days, and then frozen at -20°C. They were thawed rapidly under flowing water. On the following day they were grilled to a core temperature of 55°C and immediately served to trained panellists who scored initial tenderness and overall tenderness on 10-point scales: from 1 (tough) to 10 (tender). A total of 10-12 trained panelists were used in each session. Panelist underwent 8-10 test sessions for training, and during these they evaluated meat from different muscles, type of animals, and cooking processes [12].

Hybridization studies

After extraction, total RNA was quantified with a Nanodrop ND.1000 spectrophotometer (Thermo Scientific, World Headquarters Location, Waltham, USA). RNA integrity was evaluated with the 2100 bioanalyzer (Agilent Technologies, Massy, France) and the RNA 600 Lab Chip kit. The total RNA was amplified and labeled with Cyanine 3 using Agilent’s Low RNA Input Linear Amplification Kit, PLUS, One-Color (Agilent Technologies) following the detailed protocol described by Agilent. Briefly, 500 ng of total RNA was reverse transcribed to double-strand cDNA using a poly dT-T7 promoter primer. cDNA products were then used as templates for in vitro transcription to generate fluorescent cRNA. Labeled cRNAs were finally purified using QIAGEN’s RNeasy mini spin columns and eluted in 30 μl of nuclease-free water. After amplification and labeling, cRNA quantity and cyanine incorporation were determined using a Nanodrop ND.1000 spectrophotometer (Thermo Scientific). For each hybridization, 600 ng of Cyanine 3 labeled cRNA were fragmented and hybridised at 65°C for 17 hours to an Agilent 8 x 15 K custom Oligo Microarray. After washing, microarrays were scanned using an Agilent DNA G2505B scanner. Feature extraction 9.1 software (Agilent Technologies) was used to assess fluorescent hybridization signals.

Statistical analyses

The total number of probes was 10257, of which 1614 were control probes used for normalization. After data extraction with Feature Extraction (Agilent Technologies), hybridization values were averaged for replicated probes. Generally, probes for the same gene gave similar results. For each array, normalization was applied with the median of the 1614 control probes whose average expression level did not significantly differ between animal groups.

Normalization was then calculated per probe from the median of the probe obtained from all arrays. This ensured to use the same scale for all probes. Then a log2 transformation was done.

A Principal Component Analysis was performed with 70 genes belonging to the heat shock protein family in the group of young bulls slaughtered year 1 (the reference group) in order to identify the most correlated genes to initial and overall tenderness. Correlation coefficients of Pearson and of Spearman (relevant in this case due to the relative small number of animals per group) between gene expression levels and scores of initial or overall tenderness were calculated using the Statistical software (Stat soft, World Headquarters Location, Tulsa, USA). Multiple regression analyses were performed using the Statistical software (Stat soft) with the heat shock protein genes the most correlated to tenderness scores from young bulls of the reference group to test if these genes may predict variability of tenderness in the other groups of animals. Multiple regression analyses were also performed by combining gene expression levels and muscle biochemical characteristics which were previously shown to be associated with shear force (a measure of beef tenderness), namely lipid content, insoluble and total collagen contents and muscle fiber mean area [38]. The pooled relationship between muscle characteristics and gene expression with beef tenderness was calculated across animal groups after removing the gender and the year effects, i.e. by working on the deviation of each observation to the animal group muscle mean divided by the standard deviation for the group.

A similar approach was conducted with 150 genes belonging to fat and energy metabolism.

Muscle characteristics

Relationships between beef tenderness scores and gene expression levels in the reference group

Cross-validation of the identified tenderness markers

Five groups of probes (corresponding to four genes) belonging to the HSP family (Hsp27 encoded by the HspB1 gene, identified from proteomic studies with the accession number Gene Bank ID: GI:71037405, HspB1 Gene Bank ID: NM_013560.2, DNAJA1 Gene Bank ID: NM_001015637.1, HspH1 Gene Bank ID: NM_001075302.1, and HspA8 Gene Bank ID: NM_174345.3) were found to be significantly expressed in all animals across genders and years of slaughter (which means expression levels differ from zero). In contrast, only two genes related to fat metabolism (FASN and HADHB) were found to be significantly expressed in all animals and were not validated for all animal groups. Therefore, we attempted to validate only the five groups of probes belonging to the HSP family in all animal groups.

With the same approach, we observed that the four muscle biochemical traits (intramuscular fat content, total and insoluble collagen contents and muscle fiber area), which were previously identified to differ between shear force classes in the Longissimus thoracis of the same animals [38], explained only 8% of the variability in tenderness in young bulls slaughtered year 1 (not significant). As well, they explained only 4-8% and 19-21% in young bulls and steers respectively slaughter year 2 (not significant). However, they could explain 42% of the variability in tenderness in steers slaughtered year 1 (P = 0.06, Table 5).

Relationships between beef tenderness scores and expression levels of genes encoding heat shock proteins

After slaughter of the animals, there is generalized cell death in all organs and tissues due to the lack of oxygen and hence anoxia of all cells. This stage precedes rigor-mortis and subsequently meat ageing during which protein degradation occurs thanks to the activity of different proteolytic systems (for a review, see [42]). More and more evidence is now described in the literature indicating that the muscle death process at the beginning of conversion of muscle into meat may play an important role in regulating meat quality [41]. This cell death process, often referred to as apoptosis, is regulated by many proteolytic enzymes including caspases, the activity of these enzymes being itself regulated by many factors.

Cell death induces considerable cell stress and, in response to stress, cells rapidly produce heat shock proteins (HSP) that play a universal role in maintaining cellular homeostasis. HSP help in maintaining the integrity of the cell and have an anti-apoptotic activity [43]. For instance, proteomic studies showed that Hsp27 in fresh Blonde d’Aquitaine muscle and levels of Hsp27 fragments in aged meat explained up to 91% of variation in sensory scores [23]. It was hypothesized that higher levels of Hsp27 associated with limited aggregation of muscle proteins could facilitate the action of proteolytic enzymes during meat ageing. Therefore, the positive correlation between tenderness scores and levels of specific Hsp27 protein isoforms or fragments in Blonde d’Aquitaine [23] contrasts with the negative correlation we observed in this work between Hsp27 mRNA levels and tenderness scores. This may be explained by the different sources of variability between the two studies: ageing time [23] or animal variability in this study. However, our results fit well with the positive correlation of Hsp27 protein level and shear force value in Korean cattle [44]. Indeed, a recent study with French breeds confirms that correlation of Hsp27 level may be positive or negative depending on the breed [45]. Other authors have described the complex effects of meat pH upon the sub cellular distribution of muscle HSP during ageing of beef [46]. In other experiments, HspB1 (encoding Hsp27) and its regulator genes were shown to be negatively correlated with intramuscular fat content while being associated with a high shear force [47] and hence low tenderness as in our work.

Our results with the GENOTEND chip also confirm the negative correlation between DNAJA1 expression and the tenderness score [12], but for young bulls and steers of year 1 only, which confirms our hypothesis that some markers of beef quality are highly dependent on rearing practices and environmental conditions. DNAJA1 encodes a member of the large 40 kDa heat shock protein family (Hsp40). This protein is a co-chaperone of the 70 kDa heat shock protein (Hsp70) and is believed to play a role in protein folding and mitochondrial protein import. The DNAJA1/Hsp70 complex also directly inhibits programmed cell death (or apoptosis), which supports the hypothesis of Ouali et al. [41] that apoptosis is important for the tenderization of beef during ageing. Depending on the animal group (young bulls or steers slaughtered year 1), DNAJB9 DNAJC3 or DNJAC10, which are also members of the Hsp40 family, were negatively correlated to tenderness scores (data not shown) just like DNAJA1, which supports the role of members of the Hsp40 family in tenderness. In humans, the DNAJ family has over 40 DNAJ members [48] and can be subdivided into three subfamilies: DNAJA proteins, DNAJB proteins and DNAJC proteins with different structural features. Our results also support the role of some HspA (Hsp70), and some HspB (small HSPs) including Hsp27, in addition to that of DNAJA1 (Hsp40). For instance, CRYAB, which encodes αB-crystallin and shares homology with Hsp27, was also negatively correlated with tenderness in steers. However, the precise genes of this family which are individually negatively correlated to tenderness scores depends on the animal group and hence on its characteristics (gender, breeding management, etc).

Nevertheless, a major finding of our study is that a subset of genes belonging to the heat shock protein family may be suitable to better explain than muscle biochemical characteristics the variability of tenderness scores for both two groups of Charolais animals of the same gender or reared in similar environmental conditions. However, when the animals differ from the reference population by more than one factor (gender, or environmental conditions, etc), the expression levels of these genes cannot explain anymore the variability of tenderness scores. Further work is thus needed to extent the list of gene markers to have a more generic prediction of tenderness in other conditions of rearing and/or slaughtering and, if possible, in other breeds.

Relationships between beef tenderness scores and expression levels of genes related to muscle metabolism

As previously demonstrated recently by Zhao et al. [49], we observed that some genes known to be expressed only in adipocytes (FABP4), or involved in lipid deposition (PPARG, DGAT2, FASN) were characterized by expression levels positively associated with tenderness in young bulls slaughtered year 1. This strongly suggests that intramuscular fat level contributes to tenderness in agreement with previous research. Indeed, whereas it is well-known that intramuscular fat level affects mainly juiciness and flavor of beef, it also affects indirectly tenderness [50]. In addition, with the same samples of this study, Guillemin et al [38] observed that intramuscular fat content was 45% higher in the most tender meat samples (determined from the shearforce measurement) compared to the least tender meat samples. Intramuscular fat is mainly stored within intramuscular adipocytes embedded in a connective tissue matrix in close proximity to a blood capillary. So, in fat-rich beef, intramuscular fat deposited between fibers fascicules may disrupt the structure of muscle connective tissue, thus contributing to increase meat tenderness [51]. This is important in beef meat, in which the abundance and solubility of collagen (the main component of connective tissue) are a source of variation in tenderness especially in young bulls. Generally, intramuscular fat has a diluting effect on connective tissue (known to increase toughness) and consequently a tenderizing effect on meat. Meat from animals with a high intramuscular fat level often has a low shear force [47], and hence high tenderness. Therefore, it is not surprising to observe that FABP4 expression level, known to be a marker of intramuscular adipocyte number [26], was positively correlated with tenderness. Similarly, the more fatty acids are stored in these adipocytes, the bigger the adipocytes are, which favors tenderness by diluting connective tissue. Storage of fatty acids is controlled by two mechanisms. The first is due to the action of LPL which hydrolyses circulating triglycerides and hence favors uptake of resulting fatty acids. The second is associated with i) the activity of FASN which is a key enzyme involved in the de novo fatty acid synthesis and more precisely the elongation of fatty acids and (ii) the activity of DGAT which catalyses the formation of triglycerides from diacylglycerol and Acyl-CoA.

It is also well-known that muscles composed of fast fibers which low intramuscular fat content are more susceptible to early post-mortem proteolytic degradation than muscles mainly composed of slow fibers. Fast fibers tend to have higher levels of stored glycogen (which favors pH decline during ageing) and higher amounts of Ca2 + -activated myosin ATPase (which favors the speed of ageing). Therefore, after a short period of ageing, we expect fast muscles to generate tender beef compared to slow muscles due to a faster rate of ageing (for a review, see [52]). However, this may be not true after 14 days of ageing when a significant part of differences in ageing speed have disappeared. This may explain why the proportion of slow fibers has been reported to improve tenderness in cattle, especially for Longissimus muscle although this relationship may be not true for other muscle types or not always valid with data across muscle types (for a review, see [53]). However, the observation that oxidative fibers are associated with tender beef has been confirmed in our study by a higher expression of the HADHB known to be mainly expressed in oxidative fibers and also in tender beef in our case.

In contrast to the situation with HSP genes, we found no subset of genes related to fat metabolism or muscle fiber metabolism which were significant across environmental conditions and genders to predict beef tenderness. In some ways, this observation confirms results recently obtained by de Jager et al. [54] who observed differences in expression of the lipid storage genes between tough and tender genotypes, but also observed large differences in the expression of genes involved in fat and energy metabolism across two different experimental sites in Australia independently of genotype. A likely explanation is that these gene expression changes are reflecting differences in metabolic status of the animals such as turnover rates of nutrients (fatty-acid precursors such as glucose, triglycerides and fatty acids) or mitochondrial activity, especially considering nutritional regimens, feed density, ambient temperature, slaughtering conditions, etc between the different experimental sites or the different years of slaughtering.