A comparative genome analysis of gene expression reveals different regulatory mechanisms between mouse and human embryo pre-implantation development

We searched GEO [8] for the gene expression profiling studies related to pre-implantation development. Finally, we chose the data set GSE18290 contributed by Xie D, containing 52 samples of early bovine embryo, human embryo, and mouse embryo [9]. As we were interested in the mouse and human, we deleted 16 bovine samples. There were 36 samples left, 18 for human (GSM456643 to GSM456660) and 18 for mouse (GSM456661 to GSM456678). Human and mouse embryos were harvested at successive stage from oocyte to blastocyst. They were both generated at one-cell stage, two-cell stage, four-cell stage, eight-cell stage, morula, and blastocyst, the number of replication is 3. Total RNAs were extracted, amplified and hybridized onto Affymetrix microarrays. The platforms were Affymetrix Mouse Expression 430A Array (MOE430A, total 22690 probe sets) and Affymetrix Human Genome U133 Plus 2.0 Array (HG-U133_Plus_2, total 54675 probe sets).

The datasets we chose (CEL files) were downloaded from GSE18290. For mouse and human respectively, 18 samples were divided into 6 groups according to different periods (the time condition). Probe intensities were then normalized, and expression signals of all genes (probe sets) were calculated using the Robust Multichip Averaging (RMA) algorithm in affy package [10, 11]. Statistical analysis was performed by one-way ANOVA with a Benjamini and Hochberg False Discovery Rate (BH-FDR = 0.05) for multiple testing correction followed by Tukey's post-hoc tests [12]. Differentially expressed genes between two neighbor stages were identified by 2.0 fold-changes. Clustering on groups and genes was performed based on the identified genes' expression using the method of Hierarchical clustering. All of these processes were performed using software packages developed in version 2.4.0 of Bioconductor and R version 2.10.0 [10, 13].

Further classifications of GO annotations and pathway analysis were performed by using the Database for Annotation, Visualization and Integrated Discovery (DAVID) [14, 15], revealing over-represented function of identified genes associated with developmental stages or time specification.

Through our analysis by ANOVA-FDR test with False Discovery Rate (FDR) ≤ 5%, 12930 probe sets were shown statistically significant changes during mouse pre-implantation development. By contrast, the number for human was 21746. To identify the gene expression patterns of these two species, we employed hierarchical clustering on the samples' groups by using the above identified genes and then had a comparison between mouse and human (Figure 1). Obviously, there was difference between mouse and human transcription profiles. For mouse, we identified two major transitions in the gene expression patterns: one-cell stage to two-cell stage and 4-cell stage to 8-cell stage. These transitions separated mouse pre-implantation embryos into three phases: one-cell stage (Mouse Phase 1); two-cell stage and 4-cell stage (Mouse Phase 2); and 8-cell stage, morula and blastocyst (Mouse Phase 3). The current microarray data were mostly consistent with the previously reported expression patterns of genes [2]. By contrast, for human, there were also two major transitions, but they were: 4-cell stage to 8-cell stage and 8-cell stage to morula. Similarly, they separated human pre-implantation embryos into three phases: one cell stage, two-cell stage and 4-cell stage (Human Phase 1); 8-cell stage (Human Phase 2); morula and blastocyst (Human Phase 3).

To obtain a perspective on global gene expression changes, we performed a pair-wise comparison of gene expression data for all pre-implantation developmental stages. The threshold of Fold Change was 2.0. The results for mouse and human are detailed in Table 1 and Figure 2. The lists of mouse and human significantly regulated genes in each stage comparison can be seen in Additional file 1 and 2, respectively. For mouse, there were two major transient waves of gene expression changes. The first wave was from one-cell stage to two-cell stage, the number of identified differently expressed probe sets was totally 4316, including 1867 down-regulated and 2449 up-regulated. The second one was from 8-cell stage to morula, a total of 3400 probe sets were identified, 819 for down-regulated and 2581 for up-regulated. In comparison, only one major wave appeared in human expression profiles, but the appearing periods focused on the later stages. It was from 4-cell stage to 8-cell stage, total number is 7278 including 2180 down-regulated and 5098 up-regulated. These found waves were mostly consistent with the result of above hierarchical clustering, also revealing different gene activations among pre-implantation developmental stages between mouse and human.

Our present finding suggests that in 1- and 2- cell stages of humans, there are not many genes that get regulated. But subsequently, numerous of the differentially expressed genes have been identified in both 4-cell stage and 8-cell stage, thus these two stages may play essential roles in human pre-implantation development. The most early and widely cited report has also shown that some of the major qualitative changes which occur between the four- and eight-cell stages are dependent on transcription and cleavage is not sensitive to transcriptional inhibition until after the four-cell stage [16]. Moreover, according to more biological processes participating in Human Phase 3 (morula and blastocyst), it suggests that more preparations should be needed for implantation of human. On the contrary, in the case of mouse though most of the genes are regulated (up or down) between 1- and 2-cell stages, there are fewer but essential genes up/down regulated during later stages of pre-implantation development. The transition from the two-cell to four-cell stage is particularly important in pre-implantation mouse embryonic development, as it involves transcriptional reprogramming and cellular differentiation [17]. It is well known that the transcriptome of the mammalian embryo is comprised of maternally deposited transcripts after fertilization, but maternal transcripts are degraded and replaced by zygotic transcripts after several cell divisions, the transition is called zygote genome activation (ZGA) [2, 3]. Our findings suggest that the timing of ZGA and maternal transcripts degradation is different between mice and human, it occurs between 1-cell and 2-cell stage in mouse, but between 4-cell and 8-cell stage in human development. They are supported by previous reports [2, 16, 18].

In order to provide information about the expression changes of individual genes over time, we further respectively analyzed 12930 and 21746 statistically significant genes by a k-means clustering method for mouse and human pre-implantation development [19]. Although the dynamics of actual gene expression changes of individual genes was very complex, the k-means clustering provided a good overview of expression trends and formed a wave-like expression pattern. As a result, six clusters were identified respectively for mouse and human (Figure 3 and 4). The significant genes from each cluster were shown in Additional File 3 and 4. Several regulated genes specific to each cluster were different between mouse and human.

For mouse, the identified 6 clusters can be assigned to four main groups. The first group appears to represent genes that showed a steady increase and decrease throughout pre-implantation stages, including cluster 1 (containing 5749 probe sets) and cluster 2 (2197 probe sets). In this group, it contains Esrra and Esrrb respectively in cluster 2 and cluster 1, which have been reported as associated with obesity and type 2 diabetes [20, 21]. The second group appears to represent genes that showed a dramatic decrease throughout the stages, abundant in one cell stage but degraded during pre-implantation development, which is cluster 3 (1883 probe sets). The third group is exactly opposite to the second one, it appears to represent genes that showed a dramatic increase throughout the developmental stages, including cluster 5 (690 probe sets). The last group includes cluster 4 (1721 probe sets) and cluster 6 (690 probe sets), the genes here were firstly activated from one cell stage to 8-cell stage, peaked at 8-cell stage but degraded at later stages. In fact, the last group contains Mapk1 and Mapk3, respectively in cluster 4 and cluster 6, which are both well-known genes having central roles in diverse cellular functions [22‐25].

For human, these 6 clusters can also be divided into 4 groups. The first group appears to represent genes that showed a steady increase and decrease through the stages, except from 4-cell stage to 8-cell stage. The cluster 1 (containing 9169 probe sets) showed a dramatic increase from 4-cell stage to 8-cell stage, whereas the cluster 2 (5751 probe sets) showed a dramatic decrease during the period. In this group, there are our interesting genes like ESR1 and ESR2. As in the mouse, The second group appears to represent genes that showed a dramatic decrease throughout the stages, abundant in one cell stage but degraded during pre-implantation development, which is cluster 4 (1907 probe sets). The third group appears to represent genes that showed a dramatic increase throughout the developmental stages, including cluster 3 (2214 probe sets) and cluster 6 (894 probe sets). The last group only includes cluster 5 (1811 probe sets), also like the mouse, appearing to represent genes that were firstly activated from one cell stage to 8-cell stage, peaked at 8-cell stage but degraded at later stages. For central genes, we also take the mitogen activated protein kinase (MAPK) members for example. Mapk1 includes in cluster 2, 3 and 5; there are Mapk7 in cluster 2, Mapk8 in cluster 4, Mapk9 in cluster 3, Mapk12 in cluster 2, Mapk13 in cluster 1 and Mapk14 in cluster5. Almost all of MAPK members involve in the regulation of human pre-implantation development.

During mouse pre-implantation development, the exclusive expression of the Xist gene from the paternally inherited allele is thought to play a role in the inactivation of the paternally inherited X chromosome in the extra-embryonic cell lineages of the developing female embryo. The previous study using human cleavage-stage embryos derived by in vitro fertilization revealed that a pattern of XIST expression different from that in the mouse [26]. In our study, the profiles of XIST expression in both mouse and human pre-implantation development were known. As a result, the expression of Xist was significantly regulated at mouse 2-cell stage and blastocyst, but almost significantly regulated through human pre-implantation development (from 2-cell to blastocyst). On the other hand, in the k-means clustering of significant genes, the mouse Xist gene was in Cluster 5 but the human XIST gene was in Cluster 3. The details were shown in Additional file 1 and 2, colored in red. Thus, all of these demonstrate that not as in the mouse, the pattern of human XIST expression is not consistent with a role for early expression in the choice of paternal X inactivation but participates in regulating the whole human early embryo development.

We then sought to determine the functions associated with the identified significant genes between each adjacent stage, analysis of over-represented annotations and pathways were performed by using DAVID [14, 15]. If p value was less than 0.01, it was considered as significant GO annotation or pathway. The significant GO terms of over-represented genes were shown in Table 2 and 3 respectively for mouse or human. The significant pathways of over-represented genes were shown in Table 4 and 5 respectively for mouse or human.

With comparison to mouse, due to the less different identified genes from one-cell stage to two-cell stage and from two-cell stage to 4-cell stage for human, there were less GO terms during these periods. The GO terms for human focused on such as "cell division", "cellular metabolic process", "primary metabolic process" and "RNA processing". These were all obtained for mouse, except for the GO term "death". Just like to be delayed, most of previous GO terms for mouse were appearing from 4-cell to 8-cell stage for human, such as "macromolecule metabolic process", "macromolecule localization", "establishment of protein localization" and "establishment of RNA localization". These findings support the idea that mid-preimplantation gene activation (MGA) drives the overt morphological changes in the subsequent stages, compaction and bifurcation into two-cell lineages. With the preparation of basic cellular machinery, dramatic biological processes were happening during Human Phase 3 (morula and blastocyst), which is related to the event of implantation.

Briefly, through our pathway analysis, most of the identified pathways were involved in after human 4-cell stage, but the number of identified pathways shared in mouse pre-implantation development. Moreover, many of the features and signaling pathways that are required during human pre-implantation development are also active during tumourigenesis [27]. According to our analysis based on human data, p53 signaling pathway, the pathways of Eenteropathogenic E. coli (EPEC) infection and Enterohemorrhagic E. coli (EHEC) infection were identified separately in 4-cell and 8-cell stages. These pathways were also thought to be related to the formation of cancer and the regulation of human reproduction [28‐31]. But these above pathways were not identified from microarray analysis in mouse 4-cell and 8-cell stages. On the contrary, the pathway of oxidative phosphorylation was mostly identified through mouse pre-implantation development, which hardly appeared in human data. The oxidative phosphorylation pathway was reported as one of the obligatory energy metabolism pathways in most species throughout pre-implantation development [32]. And the pathway of Citrate cycle (TCA cycle) was identified after mouse 8-cell stage, which was not identified in human data. It is well know that the mitochondrial TCA cycle is the major source of reducing equivalents in the cytosol so that any change in mitochondrial function in the embryo will be reflected in changes in the intracellular redox state. In the mouse, the metabolic substrates used by the oocyte and early embryo each have a different impact on the intracellular redox state [33]. Surprisingly, the common identified pathway from 4-cell to 8-cell stage between human and mouse was only the Ubiquitin mediated proteolysis pathway, which was reported as an important role in eukaryotic cellular processes [34]. Thus it is advisable to choose the Ubiquitin mediated proteolysis pathway as the candidate pathway for further study on reproductive biology and regenerative medicine using mouse model. Interestingly, some of the known metabolic pathways e.g. biosynthesis of steroids pathway and fatty acid metabolism pathway were shown significant in both human and mouse blastocysts. All of these suggest that unlike in mouse, of which most of pathways found were related to energy, RNA and protein metabolism, the identified pathways in human were mostly disease-related and associated with human pre-implantation embryonic development. By contrast, there are some common metabolic pathways participate in regulating the mammal early embryonic development.

Furthermore, in view of the fact that many researchers working in mice interpret their results in relation to human reproduction, mice and human reproduction differ distinctly in many aspects; e.g. the reproduction in mice is characterized by a very short oestrus cycle and thus, displays a distinctly different endocrine dynamic pattern if compared with humans or domestic animals [35‐38]. On the other hand, due to the difference in the inner environment of embryo and the external environment of uterus, there are species differences in the implantation process between mice and human [39‐41]. The same applies for the regulatory mechanisms of early embryonic development that can be not simply transferable to the human species.

In sum, pre-implantation development involves a number of biologically significant events, such as compaction and blastocyst formation, which represent morphologically dynamic changes, especially for mouse and human [27]. Although the models of mouse have been well examined, molecular mechanisms regulating the early embryo development of human have been scarcely reported and the credibility of using the mouse model to explain the regulatory mechanism of human pre-implantation development remains unclear. Briefly, in this study, we have shown differences between mouse and human pre-implantation developments both in the global gene expression pattern and the expression changes of individual genes at each stage, including different major transient waves of transcription profiles and some stage-specific genes and pathways. Undeniably, Studies in mice have provided insights into the molecular basis of human pre-implantation development because of their shared features. In both species, early embryo development leads to a complex regulatory mechanism. However, the nature of human embryonic signals that influence uterine functions is more difficult than rodents', especially at the later stages of pre-implantation development (morula and blastocyst). Thus, it limits the availability of adequate amounts of tissues for mouse analysis. Moreover, the quality and quantity of samples chosen in microarray experiments also play essential roles in effecting the accuracy of mouse model, especially the conditions for in vitro maturation and fertilization are crucial for the proper early embryo development and may result in developmental aberrations.