In the present study we report the transcriptomic profiling of blastocysts and degenerative embryos and identification of candidate genes and pathways involved in early embryonic development. While global gene expression in blastocysts and degenerative embryos was largely similar, 67 (33 with greater than 2 fold difference) transcripts were significantly different between these two groups of embryos. In addition, several signaling pathways were found to be altered in degenerative embryos as compared to blastocysts. Although there has been a number of reports on dynamics of transcriptomes in IVF embryos [5, 6], this study, to the best of our knowledge, reports the first direct and comprehensive comparison between blastocysts and degenerative embryos produced by IVF. We believe transcriptomic alteration characterized through this comparison could provide insights into mechanisms of early embryonic development and may help identify biomarkers for growth defect in IVF and for infertility in cattle.
Validity of the experimental design and microarray analysis
Pooling of embryos and amplification of RNA in this study were necessitated by scarcity of RNA present in embryos [5, 6]. These approaches have been well documented in the literature [5, 18, 19]. In this study, we pooled a relatively large number (n = 20) of independent embryos in each pool to achieve a sufficient accuracy of biological pooling. In order to validate the results of microarray experiment and analysis, a total of 16 different pools (eight for each of blastocysts and degenerative embryos) from two different sources of RNA were used as biological replications in the real-time RT-PCR experiments (Figure 3 and Additional File 1). Importantly, eight out of the nine differentially-expressed genes identified through the microarray experiment and SAM were validated in real-time RT-PCR, testifying to the validity of the experimental design and the analysis used in this study.
It is possible that some of the degenerative embryos could be in different developmental stages because of the three-day time window of embryo collection. However, we prefer not to narrow down this window into one or two days because shorter times would result in substantial disturbance to embryo culture and mischaracterization of the two groups of embryos. Thus, we assume that there is some variation in gene expression within the degenerative embryos and that this variation would be reduced by pooling the embryos for expression analysis. Poor synchronization of embryos and integrity of RNA extracted from embryos could potentially introduce errors to the experiment. RNA integrity was checked before every major step in the microarray experiment to ensure sample quality. Moreover, although pooling of embryos removes variation between individual embryos from expression measurements, correlations between gene expression of pools can be used to assess synchronization in the same group of embryos and identify outliers due to compromise of RNA integrity. In fact, correlations of gene expression between samples were 0.979-0.995 within the group of blastocysts and 0.951-0.990 among degenerative embryos. This is an evidence of synchronization and integrity of the biological samples used in this study.
A comparison between gene expression profiles of embryos in this study with embryos at the same developmental stage from other studies would also indicate whether RNA and data quality was compromised in our samples. In fact, high correlations between gene expression profiles of our samples and those of Kues et al.  using the same microarray platform were observed. The correlations between our IVF blastocysts and those of Kues at al. were 0.942-0.970, and the correlations between our degenerative embryos and IVF blastocysts of Kues et al. were 0.925-0.957. Collectively, these results suggest that blastocysts and degenerative embryos used in the present study were largely synchronized to the same stage.
Biologically sensible results
We identified 67 differentially-expressed transcripts and several candidate pathways associated with abnormal early embryonic development. The identification of previously known candidate genes or pathways is also an important aspect of microarray experiments. Interestingly, a number of genes and pathways identified in this study fall into this category of biologically sensible results. PHLDA2 (also known as TSSC3) was found to be upregulated in degenerative compared to blastocysts by both microarray and real-time RT-PCR (Table 1 and Figure 3). PHLDA2 is an apoptosis-related gene that maps to a paternally-imprinted region involved in cancer development . The imprinting status of bovine PHLDA2 is not yet known. However, two known imprinted genes H19 and IGF2 are located nearby on bovine chromosome 29 [21, 22], and the whole conserved cluster is imprinted in human and mouse [20, 23]. Interestingly, overexpression of PHLDA2 in mice caused placental growth retardation  while PHLDA2 knock-out mice showed placental overgrowth , indicating that proper PHLDA2 expression is required for normal placental growth. Thus, our result offers further support for the importance of tightly regulated expression of PHLDA2 and may indicate its involvement in earlier stages of development.
Another differentially-regulated gene in degenerative embryos versus blastocysts is TGFBR3, one of the three types of receptors for TGF beta and one that regulates ligand binding of TGF beta to type I and type II receptors . In addition, although not meeting our FDR cutoff, the TGF beta signaling pathway was significantly (p = 0.046) associated with the developmental status of embryos. Collectively, these two lines of evidence suggest an important role of TGF beta signaling pathway in normal embryonic development, which has been reported in other studies . Cell communication and steroid biosynthesis pathways identified by both GSEA and GO enrichment analyses are of particular interest (Table 2) and their roles in early embryonic development have been studied extensively. For example, gap junctions and cell communication have been well documented to have profound influence on early embryonic development [28, 29], while several steroid hormones are required for transition from morula into blastocyst stage [30, 31].
Small change, large effect
Although the phenotypic outcomes of degenerative embryos and blastocysts were distinct, we did not observe dramatic transcriptomic changes differentiating these two groups of embryos. The correlation between gene expression of blastocysts and degenerative embryos was relatively high (r = 0.986). Among the 67 differentially- expressed transcripts, 33 were changed by more than two fold while only eight of them differed by more than four-fold, and two of them by eight-fold. One may argue that the sample size in this study is not large enough to detect small changes, so that many genuine differentially-expressed genes were missed. This is true for most microarray experiments, which normally do not involve many samples. However, our pooling strategy that reduces variation between samples presumably should alleviate this problem. Indeed, about half (34/67) of the differentially-expressed genes were less than two-fold different. These results suggest that small transcriptomic changes can lead to the distinct phenotype observed in the degenerative embryos and that the high degree of similarity observed between degenerative embryos and blastocysts was a results of true effects rather than of insufficient experimental power.
The ability to detect differentially expressed genes can also be limited by completeness of transcripts manufactured on the microarray platform. The Affymetrix Bovine Genome Array has 24,128 probe sets representing over 23,000 bovine transcripts. Contents of the array were based on GenBank and UniGene databases. Although it is possible that some transcripts exclusively expressed during early development are not represented on the array, it is unlikely that too many transcripts are missed.
The size of the differences in expression may be a specific characteristic of these genes, yet small changes in gene expression can lead to pronounced phenotypic change. For example, silencing by microRNAs has been shown to be less than two-fold [32, 33], yet they have been suggested to regulate a wide range of developmental processes to a large degree. Thus, our results suggest that small transcriptomic changes could lead to the abnormal development of degenerative embryos.
Influence of in vitro culture
There have been several reports comparing genome wide mRNA profiles between IVF and in vivo blastocysts [7, 34]. These studies demonstrated that expressions of a number of genes were changed in IVF blastocysts as compared to embryos produced in vivo. Identification of genes affecting quality of IVF embryos due to culture system is undoubtedly important. Nevertheless, comparison between blastocysts and degenerative embryos in this study is also important because only 30%-35% zygotes can successfully develop to blastocyst stage in vitro, a large source of economic loss. There are likely to be various reasons for unsuccessful development in vitro but genetics seems to play an important role [8, 10]. Embryos cultured in vitro are in a unified environment; therefore significant differential expressions detected are likely to be associated with developmental defect rather than culture system. In fact, we compared our list of differentially expressed genes to the 200 genes that showed expression differences between IVF embryos and embryos produced in vivo by artificial insemination . Importantly, among the 67 differentially expressed genes identified in our study, only one gene (DAPP1) showed in vitro/in vivo difference . This result suggests that the differentially expressed genes identified in this study were likely due to developmental defect rather than culture system.