Introduction

The Hanwoo (Bos taurus coreanae) is a member of the European cattle (Bos taurus) breeds that is native to the Korean Peninsula and Japanese Islands (Mannen et al. 1998, 2004; Yum et al. 2004; Seong et al. 2012, 2013). Hanwoo play an important role in the livestock industry, and to increase both the meat quality and quantity is major goal of research in Hanwoo (Lee et al. 2014). Meat quantity is determined by carcass weight (CW), longissimus muscle area (LMA), and back fat thickness (BF) (Hong et al. 2011; Kim et al. 2013). And marbling score (MS) which influences juiciness, tenderness, and taste is one of the factors affecting meat quality (Hovenier et al. 1993; Hausman et al. 2009; Bong et al. 2012). Marbling score is defined by the amount and distribution of intramuscular fat in the LMA (Kim et al. 2013).

To aid the early identification of beef cattle with superior genetic merits, researchers have identified qualitative trait loci (QTLs) for numerous economically important traits. Candidate gene analyses have been successfully conducted for economically important traits such as leptin (LEP), fatty acid binding protein 4 (FABP4), fatty acid synthase (FASN), and calpastatin genes (CAST) (Lee et al. 2011; Avilés et al. 2013). These studies have identified differentially expressed genes and investigated their role in signaling pathways in M. longissimus with a wide range of marbling phenotypes (Lee et al. 2010). These studies have also increased our understanding of the behavior of microRNAs (miRNAs) in the regulation of bovine adipogenesis and fat metabolism, and it was revealed that miRNA expression patterns and functions are associated with miRNA genomic location, organization and conservation (Romao et al. 2014). Clearly, genetic analyses and gene expression patterns are important for establishing the expression of economic traits. In addition, genes targeted by miRNAs may also be important in the expression of economically valuable traits. While a large number of candidate gene analyses and identified QTLs for economic traits have been previously reported in cattle, miRNAs and their expression patterns in cattle are not well studied.

miRNAs are a newly identified class of single-stranded endogenous non-coding small RNA molecules (~22 nucleotides), which regulate gene expression after transcription by binding to the 3′UTR of target mRNAs. In this way, miRNAs may regulate up to 30 % of all genes (Lewis et al. 2003; Bartel 2004; Sun et al. 2013). This phenomenon plays a significant role in the growth and developmental processes in plants and animals, and, depending on the level of miRNA–mRNA complementarity, the target mRNA can be degraded or its translation repressed (Bartel 2009; Vegh et al. 2013).

While our understanding of miRNAs is rapidly expanding, there has been no study of miRNA expression differences in cattle with respect to marbling score. Therefore, this study examines how the expression of miRNAs and mRNAs correlates with marbling score in Hanwoo M. longissimus tissue. These correlations were constructed using microarray data to screen for differentially expressed miRNAs and mRNAs in tissue samples with a wide range of marbling scores. Correlations were also constructed between miRNA expression and differentially expressed genes and pathways, as well.

Materials and methods

Animals and sampling

We performed microarray analysis on 10 samples from 46 Hanwoo aged 32 months (Table 1), validating our study with qRT-PCR. All animals in this study were maintained according to the Korea Institute for Animal Products Quality Evaluation (KAPE). M. longissimus tissue used in this study were obtained immediately after slaughter. CW was recorded at slaughter age. Backfat thickness and longissimus muscle dorsi area were measured at the 12th and 13th rib interface. Marbling was scored on a scale of 1–9, where 9 is associated with the most marbling. This animal study was approved by Institutional Animal Care and Use Committees, Hankyong National University (No. 2014-6).

Table 1 Sample carcass traits used in microarray analysis
Full size table

RNA extraction

Total RNA was extracted by homogenizing the M. longissimus tissue samples with Trizol (TaKaRa Biotechnology, Japan). The concentration of total RNA was measured using a NanoDrop ND-1000 spectrophotometer (Nanodrop Technologies, Inc., USA) and RNA integrity was measured using the Agilent Bioanalyzer 2100 (Agilent Technology, USA). RNA with integrity number (RIN) > 8.0 was used for microarray and qRT PCR analysis.

miRNA microarray and data analysis

miRNA profiling of M. longissimus tissue samples from 10 cattle (grade 1, one cattle; grade 2, three cattle; grade 5, three cattle; grade 9, three cattle) was performed. Total RNA from M. longissimus tissue samples from 10 cattle were hybridized to the GeneChip® miRNA2.0 Array (Affymetrix Inc.). This array has miRBase v15 coverage (www.mirbase.org) with 15,644 probe sets of 131 organisms, including 679 probe sets for Bos taurus miRNAs. In brief, total RNA (1000 ng) was labeled with the Affymetrix FlashTag Biotin HSR kit (Affymetrix Inc.) and added to the polyA tailing master mix (Affymetrix Inc.). The labeled RNA was hybridized with GeneChip® miRNA2.0 Array (Affymetrix Inc.) and GeneChip® Hybridization Oven 645(Affymetrix Inc.), and incubated at 48 °C for 16 h. Finally, the arrays were washed on the GeneChip Fluidics Station 450 (Affymetrix Inc.) and scanned on a GeneChip® Scanner 3000 (Affymetrix Inc.).

The quality of the arrays was assessed through standard quality control measures for Affymetrix Expression console (Affymetrix Inc.). The normalized signal intensity was log2 transformed and data analysis was performed with RMA, and array performance was measured as the percentage of probesets flagged as “present” with a conservative cutoff (% detection above background (%DABG) fold change > 1.3, P < 0.05).

Gene expression microarray and data analysis

Gene expression profiling was performed on samples of M. longissimus tissue from 10 cattle (grade 1, one cattle; grade 2, three cattle; grade 5, three cattle; grade 9, three cattle). cDNA was synthesized from 500 ng total RNA using a GeneChip® 3′IVT Expression kit (Affymetrix Inc.). Biotin-labeled antisense RNA (aRNA) was synthesized in vitro using the cDNA and IVT biotin master mix (Affymetrix Inc.). Biotin-labeled antisense cRNA was purified and the cDNA was fragmented in 5× fragmentation buffer (Affymetrix Inc.). The labeled RNA was hybridized with GeneChip® Bovine Genome Array (Affymetrix Inc.) and GeneChip® Hybridization Oven 645 (Affymetrix Inc.), and incubated at 45 °C for 16 h. Finally, the arrays were washed on a GeneChip Fluidics Station 450 (Affymetrix Inc.) and scanned on a GeneChip® Scanner 3000 (Affymetrix Inc.).

The quality of the arrays was assessed through standard quality control measures for Affymetrix Expression console (Affymetrix Inc.): pseudoimages of the arrays, MA scatter plots of the arrays versus a pseudomedian reference chip, and other summary statistics, including histograms and ox plots of raw log intensities, box plots of relative log expressions, box plots of normalized unscaled standard errors and RNA degradation plots (Bolstad et al. 2005; Lee et al. 2010). Transcription intensities in log2 scale were estimated from the probe-level data using two summary methods: MAS 5.0 and RMA (Irizarry et al. 2003; Lee et al. 2010). All arrays were scaled to the same mean for normalization and were summarized by a log2 scale average using 1-step Tukey biweight. In MAS 5.0, each probe was adjusted using a weighted average. For RMA, the background was corrected by convolution. The data were quantile normalized and summarized as median expression.

Bioinformatics analysis (miRNA target prediction, KEGG pathway analysis, and GO term enrichment)

To investigate the molecular function of the 11 differentially expressed miRNAs in M. longissimus tissue, we used the GeneSpring GX v12.5 (Agilent Technologies, CA)-based Target-Scan database (www.targetscan.org) to predict their target mRNAs. This approach was based on the assumption that miRNA target sites in the 3′UTR of the genes are more highly conserved among closely related animals than in more distant species (Chen and Rajewsky 2006; Rajewsky 2006; Wang et al. 2013). With the use of DAVID bioinformatics resources (DAVID v6.7: http://david.abcc.ncifcrf.gov/) (Dennis et al. 2003; Wang et al. 2013). The genes were classified according to KEGG functional annotations to identify pathways that were actively regulated by miRNA in M. longissimus tissue. In addition, GO term enrichment of the target genes was calculated using the GO: Term Finder Pearl module in the Term Enrichment tool.

Quantitative real-time PCR (qRT-PCR) and statistical analysis

The candidate differentially expressed miRNAs were selected based on their pattern of expression as seen in the microarray results (fold change > 1.3, P < 0.05). We synthesized single-stranded cDNA from 2 µg of total RNA using the TaqMan MicroRNA Reverse Transcription Kit. Analysis of miRNA expression was carried out with TAQMAN miRNA assays according to the manufacturer’s instructions (Applied Biosystems, USA). Fluorescence signal was detected with an ABI STEPONEPLUS Real-time PCR System (Applied Biosystems, USA) using the following conditions: 10 min at 95 °C, 40 cycles of 15 s at 95 °C, and 60 s at 60 °C.

In order to assess target gene expression, cDNA synthesis was performed using 2 µg of total RNA. Each PCR reaction consisted of 2 µg of template cDNA, Fast universal master mix (Roche) and Bos taurus TaqMan Gene Expression Assays (Applied Biosystems, USA) were used. Fluorescence was detected with an ABI STEPONEPLUS Real-time PCR System (Applied Biosystems, USA) using the following conditions: 2 min at 50 °C, 10 min at 94 °C, 40 cycles of 15 s at 95 °C, and 1 min at 60 °C. The primer sets used in the real-time PCR were KLF11 TaqMan Gene Expression Assays (Bt04301987_m1, Applied Biosystems) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) TaqMan Gene Expression Assays (Bt03210909_g1, Applied Biosystems). The the ΔΔCt method was used to determine the relative fold changes and data was normalized with GAPDH.

After qRT-PCR, The association between expression level of miRNAs and MS were evaluated using the least square method (generalized linear model procedure of the SAS software package, SAS Institute, USA). The following adjusted linear model as follow was applied: Yij = u + Expressioni + eij, where, Yij is the observation of the marbling score grade; u is the overall mean for marbling score grade; Expressioni is the expression level; eij is the random residual effect.

Results

Correlation between differentially expressed miRNAs and marbling score grade

To investigate the relationship between differentially expressed miRNAs and marbling score, samples from M. longissimus were taken from 10 unrelated animals with marbling score grades of 9 (highest), 5, 2, and 1 (lowest). Microarray data indicated that 11 miRNA transcripts were differentially expressed (P < 0.05) out of the 679 bovine miRNAs detected on the microarray based on the marbling score grade 2. Six miRNAs were upregulated in M. longissimus tissue with high marbling scores: bta-miR-16a, bta-miR-27a-3p, bta-miR-145, bta-miR-2343, bta-miR-2360, and bta-miR-2392 showed >1.3-fold increases in expression. Five miRNAs were downregulated in M. longissimus tissue with high marbling scores: bta-miR-487b, bta-miR-494, bta-miR-660, bta-miR-671, and bta-miR-2477 showed >1.3-fold reduction in expression (Table 2). qRT-PCR was conducted to corroborate this microarray data, and the expression levels of five of the selected upregulated miRNAs (bta-miR-16a, bta-miR-27a-3p, bta-miR-145, bta-miR-2343 and bta-miR-2392), and the four downregulated miRNAs (bta-miR-487b, bta-miR-660 and bta-miR-671) were determined: The expression levels of these 8 differentially expressed miRNAs were in concordance with the normalized microarray data and bta-miR-660, bta-miR-16a, bta-miR-27a-3p, bta-miR-2392 expression levels were associated with a significant effect on marbling score grade (P < 0.05, Fig. 1).

Table 2 Differentially expressed miRNAs associated with marbling
Full size table
Fig. 1
figure 1

The expression patterns of miRNAs by marbling score grade in M. longissimus tissue. The error bars represent the standard error of the mean expression level for a marbling score grade

Full size image

Predicted targets of differentially expressed miRNAs

To understand the molecular function of the 11 differentially expressed miRNAs, we analyzed the correlations between miRNA expression and differentially expressed genes. A total of 339 putative targets were identified (Table 3). To better understand miRNA functions, we subjected the putative target genes to GO analysis using the GO-TermFinder Peal module (http://amigo.geneontology.org/cgi-bin/amigo/term_enrichment). A total of 11 unique gene targets (sulfotransferase family 1E estrogen-preferring, member 1, SULT1E1; oxysterol binding protein-like 7, OSBPL7; synuclein alpha, SNCA; ATP-binding cassette, sub-family A member 1, ABCA1; glycerophosphodiester phosphodiesterase domain containing 2, GDPD2; cystic fibrosis transmembrane conductance regulator, CFTR; phospholipase C-like 2, PLCL2; dehydrogenase/reductase (SDR family) member 9, DHRS9; acyl-CoA dehydrogenase, long chain, ACADL; phosphatidylinositol-5-phosphate 4-kinase, type II, gamma, PIP4K2C; podoplanin, PDPN) were selected from the enrichment results as having biological functions related to lipid metabolism (Table 3).

Table 3 Predicted target genes regulated by miRNAs
Full size table

Differentially expressed genes by marbling score grade in M. longissimus tissue

To investigate the relationship between differentially expressed genes and marbling score, muscle samples were taken from 10 unrelated animals with the marbling score grades 9, 5, 2, and 1. Following hybridization with the Affymetrix bovine genome array, gene expression intensities were measured using two pre-processing methods: MAS 5.0 and RMA.

We found 763 differentially expressed genes in marbling score grades 9, 5, and 2 using MAS5.0 (639 transcripts) and RMA (228 transcripts). A total of 104 genes showed significant differential expression in at least two normalization methods (fold change > 1.5, P < 0.05, additional file 1). Of the 104 differentially expressed genes (DEGs; listed in additional file 1), 39 were identified as upregulated and 65 were downregulated in M. longissimus tissue with high marbling score.

The analysis of correlation between miRNA and differentially expressed genes and pathway analysis

By performing miRNA microarray analysis, we identified 11 miRNAs that were significantly differentially expressed relative to marbling score. We confirmed this finding with RMA and MAS 5.0. Therefore, we found that 763 genes gradually changed their patterns of expression. Importantly, we analyzed the interactions between specific miRNAs and target genes: Downregulated bta-miR-494 was associated with the upregulation of kruppel-like factor 11 (KLF11) and NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 5 (NDUFA5) genes, and upregulated bta-miR-16a was associated with the downregulation of mitogen-activated protein kinase kinase 4 (MAP2K4) and protein phosphatase, Mg2+/Mn2+ dependent, 1D (PPM1D) genes by RMA. In the case of MAS 5.0 analysis, the downregulated bta-miR-494 was associated with the upregulation of KLF11 and glycoprotein nmb (GPMNB) genes, bta-miR-660 was associated with zinc finger protein 503 (ZNF503) gene, and bta-miR-2477 was associated with insulin-like growth factor 1 (IGF-1) and erythrocyte membrane protein band 4.1-like 2 (EPB41L2) genes. The upregulated bta-miR-145 was associated with the downregulation of rhotekin (RTKN) gene, bta-miR-27a-3p was associated with golgi SNAP receptor complex member 2 (GOSR2), LOC539475 and gamma-aminobutyric acid A receptor, pi (GABRP) genes, and bta-mir-2360 paired with nuclear receptor subfamily 2, group E, member 3 (NR2E3), histone deacetylase 6 (HDAC6), and poly(rC) binding protein 4 (PCBP4) genes (Table 4). Fifteen genes were found to be targeted by miRNAs overall.

Table 4 Differentially expressed miRNAs and targeted genes analyzed by RMA and MAS 5.0 methods
Full size table

We accessed the GO biological processes associated with these 15 genes. Due to the incomplete annotation of the bovine genome, only 8 out of 15 differentially expressed probe sets were annotated (Table 5).

Table 5 Gene Ontology for differentially expressed genes associated with marbling score
Full size table

The pathway Studio v9.0 program (Ariadne Genomics, Inc.) was used to identify molecular associations between the proteins encoded by these 15 differentially expressed genes. The program searches through the ResNet database for all known interactions between genes and proteins such as physical interactions and regulation of expression. Out of the 15 genes, two main pathway “hubs” (KLF11 and IGF-1) were detected in the pathway analysis. The KLF11 gene, which is upregulated in high marbling score tissue, appears to be involved in lipid metabolism. The KLF11 and IGF-1 genes appears to be regulated by insulin and by FASN, LEP and corticotrophin releasing hormone (CRH) genes (Fig. 2).

Fig. 2
figure 2

Pathway analysis of highly expressed genes

Full size image

Discussion

While role of miRNAs in the regulation and expression genes has been studied in many contexts, even in other cattle breeds, the influence of miRNAs on Hanwoo marbling score has not been reported or presently studied. To date, microarray based gene expression analyses of adipogenesis has focused on miRNA detection in other breeds of cattle, such as British-continental crossbred beef steers (Romao et al. 2014). And previous study identified that the beef quality of Angus Cattle sharply diversifies after acute stress: 13 miRNAs were shown to exhibit significant differential expression in the stressed group relative to the control group (Zhao et al. 2012). Other researchers identified 417 known miRNAs and 104 novel miRNAs in the longissimus thoracis of fetal and adult Qinchuan bovine using deep sequencing technologies (Sun et al. 2013). Our study on miRNA expression in Hanwoo, indicated that 6 miRNAs (bta-miR-16a, bta-miR-27a-3p, bta-miR-145, bta-miR-2343, bta-miR-2360, and bta-miR-2392) were upregulated and that 5 miRNAs (bta-miR-487b, bta-miR-494, bta-miR-660, bta-miR-671, and bta-miR-2477) were downregulated in M. longissimus tissue with high marbling score phenotypes.

Furthermore, we investigated the relationship between differentially expressed genes and the marbling score, finding 763 differentially expressed genes. To date, microarray base gene expression analysis for beef meat quality have focussed on detection of differentially expressed genes in cattle such as Hanwoo, Japanese Black (Wagyu) and Holstein–Friesian cattle for a wide range of marbling phenotypes (Wang et al. 2005a, b; Lee et al. 2010). Earlier results showed that ADAM metallopeptidase with thrombospondin type 1 motif 4 (ADAMTS4) gene was overexpressed in high marbling score Hanwoo (Lee et al. 2010) and our study concurs with this finding.

We also analyzed the interaction between specific miRNAs and target genes. Fifteen genes were found as target genes of specific miRNAs. Among those differentially expressed genes found to be associated with marbling score. These result implied that the downregulated bta-miR-2477 was associated with the upregulation of IGF-1 gene. A previous study reported that serum IGF-1 concentration was associated with growth traits in Angus cattle (Huang et al. 2011) and IGF-1 gene expression level was significantly associated with marbling score in Hanwoo (Yoon et al. 2014).

Furthermore, the pathway Studio v9.0 program has been used to identify the molecular association between the proteins encoded by these 15 differentially expressed genes. The results showed that out of the 15 genes, 2 main pathway “hubs” (KLF11 and IGF-1) were also detected in the pathway analysis. The IGF-1 and KLF11 genes were regulated by insulin, FASN, LEP, and CRH genes. It was reported that FASN and CRH genes were associated with marbling score in cattle (Wibowo et al. 2007; Oh et al. 2012). So, RT-qPCR corroborated the microarray data, and the expression levels of KLF11 and bta-miR-494 were correctly determined. The results show that downregulated bta-miR-494 was associated with the upregulation of KLF11 gene (P < 0.05, Fig. 3). The specific association among these miRNA and genes show a significant connection to the marbling score phenotype. The results provide evidence that bta-miR-494 and bta-miR-494 targeted KLF11 genes might be important genetic factors influencing marbling traits in Hanwoo.

Fig. 3
figure 3

The expression patterns of KLF11 and miR-494 by marbling score grade in M. longissimus tissue. The error bars represent the standard error of the mean expression level for a marbling score grade

Full size image