We examined the number and diameter of myofibers at each developmental stage in Texel and Ujumqin sheep via histological analysis (hematoxylin and eosin [H and E] staining) (Figures 1 and 2). Significant difference was observed in myofiber diameter between Texel and Ujumqin sheep at each development stage, except at 70 d. However, the myofiber diameter patterns of the two breeds were similar throughout the five developmental stages.

Unexpectedly, we found that the myofiber hyperplasia profiles differed between Texel and Ujumqin fetuses (Figure 2). The proliferation of myofibers in Texel sheep was fastest at 70 and 85 d, particularly at 85 d, suggesting that this period is vital for muscle development in fetal Texel sheep. Myofiber hyperplasia reached a peak at 100 d in Ujumqin sheep, when the number of myofibers exceeded that of Texel sheep. Notably, with development, the number of myofibers in Texel fetuses recovered to that of Ujumqin fetuses at 120 d, although myofiber hypertrophy in Texel sheep was greater than that in Ujumqin sheep at this stage. Therefore, a new wave of myogenesis most probably occurs at 120 d in Texel sheep, given that the number of myofibers increased in Ujumqin sheep during this period.

Additionally, many small-diameter fibers occurred in the muscle at 135 d in both breeds. The small fibers could be due either to intrafascicular terminations 24 or have been generated by the fusion of myoblasts during the period investigated. Therefore, to explore the myogenic potential in the two breeds of sheep, we identified Pax7-positive cells in skeletal muscle at 135 d using immnofluorescence staining (IFS) and found that the ratio of the number of Pax7-positive cells (myoblasts or satellite cells) to the fibers on each muscle slide was significantly higher in the Texel fetuses than in the Ujumqin fetuses (Figure 3). In combination with Figure 2, this observation suggests that a period of 100-120 d is another new proliferative stage for myofiber hyperplasia in Texel sheep. Higher numbers of Pax7-positive cells were correlated with a higher muscle mass in Texel sheep during both the prenatal and postnatal stages.

A mean transcript sequence identity of 97% was found between ovine and bovine orthologs, which probably highlights the utility of the microarray technique with ovine samples 25 . However, some no-reporting probe sets were found, partly because of the poor performance of the probe sets with samples from a related species. The best way to resolve this type of problem is to use a species-specific microarray. In the present study, we applied the first specialized and standardized transcriptome-wide sheep oligo DNA microarray (Agilent Sheep Gene Expression Microarray; Agilent Technologies, Santa Clara, CA, USA). Due to improvements in species specificity, the average detectable rate of all probe sets reached 87.83% (Additional file 1), which enabled us to capture the subtle changes in gene expression.

A pool of differentially expressed (DE) probes from the two groups was used for a systematic hierarchical clustering to gain insight into the transcriptome-wide similarities among all 31 individuals investigated (Figure 4). We found that individuals at the same developmental stage were clustered together regardless of their genetic background, indicating that differences derived during development were larger than those from the genetic backgrounds between the breeds.

According to the primary functions of different genes including muscle, lipid, and the immune system, we visualized the data on MA and bar plots (Figures 5 and 6). Combined with Figures 1 and 2, Figure 6 also suggests the phenotypic profile in muscle and the profile of a given category of DE genes at different developmental stages. The biological functions and canonical pathways were explored at various developmental stages in Texel and Ujumqin sheep (Tables 1, 2, 3, 4 and 5). The partial DE genes in the two breeds of sheep at various developmental stages are listed in Tables 6 and 7 (the full set of differential genes at each developmental stage between the breeds is listed in Additional files 2, 3, 4, 5 and 6).

To confirm the differential expression of the genes in the microarray analysis, we selected 15 genes to validate by quantitative PCR (qPCR). Among these, eight genes (ACTB, TNC, MYO9A, MYL6B, MYH1, CASQ1, TPM2, and TMOD4) encode myofibrillar proteins and two (EEF1A2 and LEF-1) are transcription factors that regulate the genes associated with myogenesis. PIK3R1 is involved in fat metabolism regulation, whereas TRRAP, TXNRD1, and INSIG1 participate in the regulation of gene expression and cell proliferation. Additionally, we also detected MSTN expression during skeletal muscle development. As in previous studies 26 27 28 29 , RpLP0 was chosen as an ideal reference gene to normalize the data. The ratios of 12 DE genes at particular developmental stages in the two sheep breeds in the microarray analysis agreed with the qPCR results (Table 8).

Wilson et al. and Maier et al. showed that at least three waves of myogenesis occurred until d 76 of gestation in sheep 9 11 . We demonstrated that the myofiber numbers increase sharply with a pulsed-wave mode at a particular developmental stage in sheep. This is the first time that the mode of fetal myofiber proliferation was revealed in a large mammal during the second half of gestation. In addition to the longissimus dorsi muscle, similar patterns of myofiber proliferation were found in the semitendenosus, gluteus medius, gastrocnemius, and triceps brachii muscles in Texel and Ujumqin fetal sheep (data not shown). These findings indicate that the specific developmental stage during which myogenic surges of myofiber hyperplasia appeared is pivotal for fetal muscle development as well as postnatal muscularity potential in sheep.

Unlike the studies of Wilson et al. and Maier et al., who identified different generations of myogenic cells using electron microscopy 9 11 , we counted the muscle fibers under a light microscope and found many small-diameter fibers until 135 d in both breeds. We considered that at least part of the small fibers must have been generated by the fusion of myoblasts (or satellite cells) during the study period. Pax7 plays an important role conferring myogenic potential to these progenitor cells 30 , and Pax7 is an ideal cell marker of myogenic progenitor cells such as myoblasts and satellite cells 31 . Our findings indicate a higher potential for postnatal muscle growth that entails more Pax7-positive cells in the fetal stage.

Of the five prenatal stages, the numbers of DE genes in the two sheep breeds were highest at 70 and 120 d, indicating that rapid myofiber proliferation included numerous genes during the myogenic process. However, two surges of myofiber proliferation appeared at 85 and 120 d in Texel fetuses, probably because the expression of more genes associated with myoblast proliferation and differentiation at 70 d prepared for the myofiber hyperplasia that followed at 85 d, and more genes related to the immune system were differentially expressed between T70 and U70 (Figure 6), which led to larger total counts of DE genes. The difference between the two sheep breeds in the timing of myogenesis in terms of transcriptomic levels was similar to that found between pig breeds 32 . Cagnazzo et al. showed that myogenesis-related gene expression is greater in early Duroc (a breed with more intramuscular fat) embryos than in early Pietrain (a highly muscled breed) embryos at 14-49 d of gestation, whereas the opposite was found in late embryos (63-91 d of gestation) 32 . The present findings suggest that highly muscled breeds have a longer myogenic process during prenatal stages and that myogenesis is more intense in late-stage fetuses, which was validated by Pax7-cell staining in longissimus muscle cross sections (Figure 3). However, whether this myogenic process is the same in sheep as in pigs at earlier embryonic stages remains to be investigated.

We found that the immune and hematological systems' development and function were most overrepresented of all physiological systems and functions, accompanied by muscle system development, which was consistent with a microarray developmental analysis in cattle fetuses 21 . Recent discoveries have revealed complex interactions between skeletal muscle and the immune system that regulate muscle regeneration and myogenesis, and many immune molecules, such as tumor necrosis factor-alpha (TNF-α), NF-kB, interleukin (IL)-4, IL-6, IL-10, and leukemia inhibitory factor are involved in muscle cell proliferation and differentiation 33 34 35 36 37 38 . In most cases, whether perturbation in the same signaling pathways related to myogenesis occurs will determine normal myogenic development or muscle disorders. Therefore, the pathways associated with immune and muscle disorders in this work are valuable for human myopathy at the prenatal stage. The hematological system is also implicated in myogenesis 39 40 41 42 43 , which probably reflects the systemic requirement for muscle function during fetal development.

Nervous system development was overrepresented in the last two development stages in the two sheep breeds. Neurons are indispensable for maintaining normal muscle physiological function and also parallel the skeletal muscle in auxology. Comprehensive and close interactions occur between muscles and the nervous system during development 44 45 46 . Muscle development is regulated by the central nervous system in Drosophila and pigs 47 48 , and muscle fiber type is dependent on the pattern of innervation of a muscle established due to differential projection patterns between fast and slow motoneurons 49 50 . Animals with different muscle phenotypes undergo diverse innervation patterns during fetal development. Double-muscled cattle have an additional 13-26% increased branching in terminal axons compared to that in normal cattle caused by a real increase in the number of myofibers 51 .

We identified several valuable canonical pathways directly associated with muscle development and function, myogenesis, myoblast proliferation, and the cell cycle in muscle at various developmental stages, such as calcium 52 , CXCR4 53 54 55 , and VEGF signaling 56 . Furthermore, other pathways involved in adipose and muscle metabolism, such as inositol metabolism, steroid biosynthesis, EIF2 signaling, glycerolipid metabolism, the glycolysis/gluconeogenesis pathway, and N-glycan biosynthesis, are significantly affected during development. Steelman et al. suggested that Wnt signaling is a potential downstream target of myostatin for postnatal skeletal muscle growth and hypertrophy in mice 57 . However, we focused on the transcriptome during prenatal stages rather than postnatal stages and found that various canonical pathways are predominant in skeletal muscle development at different stages due to a myostatin mutation in sheep.

STAT3, a member of the STAT family and a cooperator in the Janus kinase pathway, plays a dual role in the regulation of myoblast proliferation and differentiation, but is dependent on interactions with various cofactors 58 59 60 61 62 . We found that STAT3 expression was higher in U70 than T70, which suggests that STAT3 is most probably associated with the differentiation rather than proliferation of myoblasts. The GO analysis of DE genes between T70 and U70 confirmed our presumption, and muscle fiber hyperplasia was more intense in T70 (Figure 2) as well.

ALDOA is responsible for significant activation during the differentiation of chicken primary myoblasts and plays an important role in muscle gene transcription 63 64 . In our study, a higher ALDOA expression level in U70 indicated that more myoblasts exited the cell cycle and entered differentiation in U70 than those in T70. The leptin receptor (LEPR), a protein secreted from adipocytes, is responsible for fat mass regulation via leptin in the hypothalamus 65 66 . LEPR expression was higher in U70 than in T70. Whether the higher LEPR expression at prenatal stages is associated with high postnatal adipose deposition remains to be investigated.

Both hyperplasia and hypertrophy continued in skeletal muscle at 85 d. FBP2, which encodes a gluconeogenesis regulatory enzyme that catalyzes the hydrolysis of fructose 1,6-bisphosphate to fructose 6-phosphate and inorganic phosphate, was expressed higher in T85 than in U85. A previous study suggested that FBPase participates in some nuclear processes during the development and regeneration of skeletal muscle 67 68 . Additionally, PRG3 expression was also higher in T85. Previous studies have shown that PRG3 plays a role in some early aspects of skeletal myogenesis 69 70 71 72 .

The expression levels of PIK3R1, TSHR, and PON1 were higher in U85 and U100 than that in T85 or T100. Previous studies have demonstrated that these genes are involved in fat metabolism and adipocyte development 73 74 75 76 .

PDK4, a gene that was upregulated in T100, is a key regulatory enzyme involved in switching the energy source from glucose to fatty acids in response to physiological conditions. The pyruvate dehydrogenase complex occupies a central and strategic position in muscle intermediary metabolism and is primarily regulated by phosphorylation/dephosphorylation 77 . In addition, PDK4 is significantly associated with intramuscular fat and muscle water content 78 , indicating that PDK4 is involved in meat quality.

TXNRD1 is a variant of thioredoxin reductase, an important selenoprotein that maintains cellular redox balance and regulates several redox-dependent processes during apoptosis, cell proliferation, and differentiation 79 80 81 . LEF1 is a transcription factor involved in the regulation of myogenesis. Loss of Lef1-mediated repression results in an increased number of cells expressing Pax-7 and Pax-3, suggesting that Wnt signaling via Lef1 acts to regulate the number of premyogenic cells in somites 82 . In our study, LEF1 expression was higher in U100 than that in T100, which is consistent with the rapid proliferation of myofibers in U100. Although transcription factor activity is predominantly manifested at the protein level, the difference in LEF1 expression level indicated a potential difference of LEF1 between breeds.

OVAR-DQB1 (sheep MHC class II) was expressed much higher in T100 and T120, particularly in T120 (Table 7), compared with U 100 and U120. Results from Karpati et al. indicated that the class I molecule may be involved in the fusion of myogenic cells during muscle regeneration 83 . Honda and Rostami demonstrated that the expression of class I antigens on muscle cells is not only immunologically modulated but also developmentally regulated, and that these antigens may play a role in cell recognition and interaction during the myogenic fusion process. The presence of the antigens, however, was transitory, and they disappeared as myoblasts fused and differentiated into multinucleate myotubes 84 .

CEBPB, an important transcription factor, acts as a indispensable regulator of adipocyte differentiation during adipogenesis 85 86 87 . We found that CEBPB expression was higher in skeletal muscle in T120 than in U120, indicating a key difference in adipogenesis between Texel and Ujumqin sheep. However, which breeds contain more adipose mass than the other at this developmental stage remains to be investigated.

TRRAP, which was downregulated in T120, in contrast to that in U120, is vital for embryonic survival and control of the mitotic checkpoint 88 89 90 91 . Deletion of TRRAP leads to a reduced level of beta-catenin ubiquitination, a lower degradation rate, and accumulation of the beta-catenin protein, whereas TRRAP knockdown results in abnormal retention of beta-catenin at chromatin and concomitant hyperactivation of the canonical Wnt pathway 92 . However, the canonical Wnt pathway is downregulated in the absence of myostatin through beta-catenin, whereas the Wnt/calcium pathway is upregulated 57 , suggesting that TRRAP may negatively regulate skeletal muscle development between the canonical Wnt and the Wnt/calcium pathways through beta-catenin at 120 d. Further efforts are warranted to test the effect of TRRAP on skeletal muscle development.

Other than the genes discussed above, we found that MSTN expression was lower in Texel than in Ujumqin sheep through the five development stages (Table 8). This result indicated that MSTN downregulation contributed to the hyperplasia and hypertrophy of myofibers in Texel sheep through the second half of gestation. Moreover, the genes discussed above are potential myostatin targets for further investigation.

All experimental and surgical procedures were approved by the Biological Studies Animal Care and Use Committee, Shanxi Province, Peoples Republic of China. Seventy-eight Ujumqin and 54 Texel ewes were prepared. These purebred female animals involved were selected based on their age (3-5 years old), body weight (50-55 kg), and body size. After these animals were subjected to pre-feeding for 45 days, estrus was synchronized in all ewes using an implanted controlled internal drug release device (Pharmacia & Upjohn Pty Limited, Parramatta City, NSW, Australia) and intramuscular injections of pregnant mare serum gonadotropin (Ningbo Renjian Pharmaceutical Co., Ltd, Zhejiang, China), according to the manufacturers' protocol, followed by artificial insemination using the corresponding breed sire's sperm. Three pregnant ewes from each breed were subject to caesarean section to collect the fetuses at 70, 85, 100, 120, and 135 d of gestation, and then the 12 different anatomic skeletal muscles from each fetus were dissected and weighed (longissimus dorsi, semitendenosus, semimembranosus, gluteus medius, femoral quadriceps, gastrocnemius, serratus ventralis thoracis muscle, biceps femoris, adductor, supraspinatus, infraspinatus, and triceps brachii). Two samples were taken from each muscle at standardized anatomical sites. The first sample (3 g) was dissected and rapidly frozen whole in isopentane chilled over liquid nitrogen for histological examination. The second sample (5 g) was snap-frozen in liquid nitrogen for gene expression analysis.

To examine the development of skeletal muscle in both sheep breeds, we measured the number and the diameter of the longissimus dorsi muscle fiber in Ujumqin and Texel sheep during five development stages using hematoxylin and eosin staining. Serial cross sections of 10 μm thickness were cut at -20°C using a cryostat. Photomicrographs of five to eight randomly selected areas were used to estimate the the diameter and the total number of myofibers in the cross sections of muscle examined. Muscle fiber measurements were conducted using DT2000 (V2.0) image analysis software (Nanjing East-image Ltd., Nanjing, China).

The primary IFS antibodies were rabbit-laminin antibody (Sigma, St. Louis, MO, USA) and mouse monoclonal anti-pax7 (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, USA). The corresponding secondary antibodies were anti-mouse immunoglobulin (ALEXA488) and anti-rabbit immunoglobulin (ALEXA594) (Invitrogen, Carlsbad, CA, USA). Staining was performed according to the manufacturer's instructions. The IFS slides were analyzed using a system of epi-fluorescence microscopes (TE2000U; Nikon, Tokyo, Japan). The satellite cell (SC) nuclei were stained green and the basal lamina, red. Images were subsequently analyzed using Image Pro Plus (v5.1) image analysis software (Media Cybernetics, Inc., Bethesda, MD, USA). The numbers of SCs and myofibers in each section were counted, and the SC concentration within a given area of view was also calculated for each slide.

Data between the two breeds were analyzed using the t-test procedure in SPSS 13.0 (SPSS, Inc., Chicago, IL, USA). A p-value of < 0.05 was considered statistically significant.

We made five independent comparisons at the same development stages in both breeds using a t-test and detected 207, 115, 90, 151, and 72 DE probes with a greater than twofold change at 70, 85, 100, 120, and 135 d of gestation, respectively. The new specialized and standardized transcriptome-wide sheep microarray (Agilent Technologies) contained 15,008 sheep probes sourced from RefSeq and UniGene. The oligo microarray was 15 K in size. The longissimus dorsi muscles from ovine fetuses were investigated for gene expression. The microarray of each development stage for each breed contained three biological replicates (n = 3) except the T100 team (n = 4). Thus, 31 hybridizations in total were performed. Data were deposited in the NCBI database (GSE23563).

Total RNA was extracted from the frozen longissimus dorsi muscles using an RNeasy Mini Kit with DNase treatment (Qiagen, Valencia, CA, USA) according to the manufacturer's protocol. The total RNA from each individual was labeled with Cy3 fluorescence. The generated targets were mixed and subjected to hybridization with the Sheep Gene Expression Microarray (Agilent Technologies), according to the manufacturer's protocol.

Scanning of the microarrays was performed using a DNA microarray scanner (Agilent Technologies). The scanning parameters were a 5 μm scan resolution, PMT with 100%, and 10% once, respectively. All probe sets whose signals were reported as "absent" during the flag call in all 31 arrays were filtered out, and those from probe sets called "present" on at least one chip were used for further analysis. Scanner output image files were normalized and filtered using Feature Extraction Software v10.5 (Agilent Technologies).

The raw signal intensity data were normalized using the quantile method, and the data were transformed to a base-2 logarithm for further statistical analysis. To explore the difference in gene expression between breeds at each developmental stage, we conducted a comparative analysis with the unpaired unequal variance t-test (Welch's t-test) method in GeneSpring10.0 (Agilent Technologies). Five t-test contrasts were performed independently, based on the premise that no relationships existed between the data at different time points. The thresholds for significance were p < 0.05 and a fold change of twofold or more with FDR < 0.4.

Only 1508 of the 15,008 probe set were definitive in the Agilent Sheep Gene Expression Microarray. To further clarify the function of the DE genes in this study, we conducted a BLASTN search for highly homologous human sequences (search setting: query coverage not less than 50%, E value less than 1.00E-100). DAVID Bioinformatics Resources (http://david.abcc.ncifcrf.gov) 93 94 and IPA (http://www.ingenuity.com) 95 were both used for further functional analysis to identify which molecular events or cascades were involved during skeletal muscle development. Significance was expressed as a p-value, which was calculated using the EASE score (p-values < 0.05 were considered significant).

Total RNA was used to make cDNA using the PrimeScript^® 1st Strand cDNA Synthesis Kit (Takara Bio, Shiga, Japan). q-Real-time PCR was performed on an ABI 7500 instrument (Applied Biosystems, Foster City, CA, USA) using the Fast EvaGreen^® Master Mix (Biotium, Hayward, CA, USA). Thermal cycling consisted of an initial step at 95°C for 10 min followed by 40 cycles at 95°C for 30 s and 62°C for 30 s. Fifteen DE genes identified in the microarray experiment were selected for validation by q-real-time PCR; RpLP0 was chosen as the reference gene for normalization of all data because it was expressed more stably 26 27 28 29 . Each q-real-time PCR reaction (in 20 μl) contained 10 μl of Fast EvaGreen^® Master Mix, 0.5 μl of each primer, 2 μl of normalized template cDNA, and 7 μl water. The q-real-time PCR measurements were performed in triplicate on each cDNA sample (n = 31), and gene expression was quantified relative to RpLP0 expression using the 2-ΔΔCt method. Gene expression ratios were normalized to the RpLP0 gene in the same sample.