Defining Contributions of Paternally Methylated Imprinted Genes at the Igf2-H19 and Dlk1-Gtl2 Domains to Mouse Placentation by Transcriptomic Analysis*♦

Parental genome functions in ontogeny are determined by interactions among transcripts from the maternal and paternal genomes, which contain many genes whose expression is strictly dependent on their parental origin as a result of genomic imprinting. Comprehensive recognition of the interactions between parental genomes is important for understanding genomic imprinting in mammalian development. The placenta is a key organ for exploring the biological significance of genomic imprinting. To decipher the unknown roles of paternally methylated imprinted genes on chromosomes 7 and 12 in mouse placentation, we performed a transcriptomic analysis on placentae in three types of bimaternal conceptuses that contained genomes derived from both non-growing and fully grown oocytes. Furthermore, we used the Ingenuity pathway analysis software to predict key networks and identify functions specific to paternally methylated imprinted genes regulated by the Igf2-H19 imprinting control region and Dlk1-Dio3 imprinting control region. The data suggested that dynamic conversion of the gene expression profile by restoring the expression of paternally methylated imprinted genes resulted in phenotypic improvements in bimaternal placentae. These results provide a framework to further explore the role of epigenetic modifications in paternal genome during mouse placentation.

In mammals, embryonic development and other phenotypic characteristics are influenced by the genetic constitution of zygotes. Maternal and paternal genetic contributions are usually distinctive. However, certain differences are present in the epigenetic modifications of some parental alleles, which lead to functional differences between certain homologous chromosomal regions. Genomic imprinting results in functional non-equivalence of the maternal and paternal genomes, thereby preventing the development of viable parthenotes in eutherian mammals. Usually, parthenotes die before embryonic day 9.5 (E9.5) 2 and have poorly developed extraembryonic tissues (1,2). Therefore, genomic imprinting has been regarded as an obligatory step in mouse placentation.
In mice, an extremely small number of primordial germ cells, i.e. cells from which germ cells originate, first differentiate from pluripotent epiblast cells at around day 6.5 of gestation (3,4). The primordial germ cells acquire epigenetic modifications that are differentially imposed during spermatogenesis and oogenesis (5,6). The paternal-or maternal-specific DNA methylation at differentially methylated regions associated with imprinted genes regulates their allele-specific expression. This leads to decisive functional differences between the maternal and paternal genomes; thus, both parental genomes are required for normal development to term in mammals. It is known that Dnmt3L cooperates with Dnmt3a and Dnmt3b to methylate DNA de novo (7). Studies on the extraembryonic tissues of Dnmt3L matϪ/Ϫ embryos have revealed that appropriate maternal methylation of imprinted genes is absolutely essential for mouse placentation (8). Additionally, in our previous study, we showed that when non-growing oocytes of wildtype mice (WT) (ng WT ), in which imprints are not established, were combined with fully grown (fg) oocytes, the resulting ng WT /fg bimaternal conceptuses could have placentae comprising three basic layers: the trophoblastic giant cells, spongiotrophoblast layer, and labyrinthine layer (9). However, the ng WT /fg placenta showed diagnostic defects such as severe growth retardation, disproportionate expansion of the spongiotrophoblast layer, distorted and ambiguous boundary between the spongiotrophoblast and labyrinthine layers, and the presence of abnormally enlarged giant cells (10). This probably occurs because the maternal epigenotype on chromosomes 7 and 12, the Igf2-H19 imprinting control region (ICR) and Dlk1-Dio3 ICR, causes inappropriate expression of the imprinted genes.
To further understand the roles of paternally methylated imprinted genes regulated by Igf2-H19 ICR and Dlk1-Dio3 ICR in mouse placentation, we have analyzed the ng/fg bimaternal placentae that were switched from the maternal to the paternal epigenotype in the imprinted domain at the distal regions on chromosomes 7 and 12 (10). The genetic backgrounds of these ng oocytes were as follows. One was derived from mutant mice carrying a 13-kb deletion in the H19 transcription unit including the germline-derived differentially methylated region on chromosome 7 (ng ⌬ch7 ). Another set of oocytes was derived from mutant mice carrying a 4.15-kb deletion in the intergenic germline-derived differentially methylated region on chromosome 12 (ng ⌬ch12 ) (11,12). Interestingly, ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae showed reciprocal phenotypes, i.e. the ng ⌬ch7 /fg placenta exhibited severe dysplasia, such as an expanded spongiotrophoblast layer and a malformed labyrinthine zone, but exhibited normal-sized giant cells. In contrast, the cell layers in the ng ⌬ch12 /fg placenta retained their normal structure with a normal circulatory system, but its total mass was extremely low and expanded giant cells were present. These findings strongly suggested that investigating the ng/fg bimaternal placenta was a powerful means of defining paternal genetic contribution to mouse placentation and for understanding how paternally methylated imprinted genes controlled by Igf2-H19 ICR and Dlk1-Dio3 ICR complementarily organize placentation. The manner in which the appropriate expression of imprinted genes regulated by these two ICRs leads to definitive placentation in bimaternal conceptuses is unclear, and the global gene expression dynamics underlying bimaternal placentation also remain to be elucidated.
Recently, we used ng oocytes of double mutant mice (ng ⌬Double ) that harbored deletions in both the Igf2-H19 ICR and the Dlk1-Dio3 ICR to demonstrate the generation of bimaternal embryos that develop as normal individuals at a high success rate equivalent to that obtained by in vitro fertilization of normal embryos (13). This result was obtained by the expression of imprinted genes regulated by Igf2-H19 ICR and Dlk1-Dio3 ICR, which was appropriately corrected in the ng ⌬Double /fg bimaternal embryo. In this study, we determined the placenta-forming ability of the ng ⌬Double /fg bimaternal embryo and further investigated the genes that are dominated by the paternally methylated imprinted genes controlled by Igf2-H19 ICR and Dlk1-Dio3 ICR. First, we used the same methods as those described in our previous report to evaluate the phenotypic characteristics of the ng ⌬Double /fg placenta (10). Next, we compared the global gene expression profiles of four placenta types, namely, the ng ⌬ch7 /fg, ng ⌬ch12 /fg, ng ⌬Double /fg, and WT placentae (Fig. 1). Finally, for in-depth investigation into the roles of the corrected expression level of paternally methylated imprinted genes on chromosome(s) 7 and/or 12 in mouse placentation, we conducted the Ingenuity Pathway Analysis (IPA) to predict the key networks of genes that function during mouse placentation.
In this study, we observed that the ng ⌬Double /fg placenta showed remarkably improved phenotypes in comparison with the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae. Additionally, comparative global gene expression analyses of the ng ⌬Double /fg placenta and the other bimaternal placentae, i.e. the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae, enabled us to predict the functions of paternally methylated imprinted genes regulated by Igf2-H19 ICR and Dlk1-Dio3 ICR in the mouse placenta. They demonstrated that the restored gene expression of paternally methylated imprinted genes on chromosomes 7 and 12 led to widespread alteration of the expression of other genes and remarkable improvements in the placental phenotypes. Thus, these global gene expression profiles would be useful for defining paternal genetic contribution regulated by Igf2-H19 ICR and Dlk1-Dio3 ICR to mouse placentation.

EXPERIMENTAL PROCEDURES
Oocyte Manipulations-Fully grown germinal vesicle oocytes from the ovarian follicles of B6D2F1 (C57BL/6N ϫ DBA2) female mice were collected in the M2 medium 44-48 h after the mice were injected with equine chorionic gonadotropin. Ovulated MII oocytes were also collected from superovulated B6D2F1 mice 16 h after they were injected with human chronic gonadotropin. We collected ng oocytes that were in the diplotene stage of the first meiosis from the ovaries of 1-day-old newborn mice. Serial nuclear transfer was performed using a previously described method (14). The ng oocytes derived from double mutant females with heterozygous ⌬ch7 (Ϫ/ϩ) and heterozygous ⌬ch12 (ϩ/Ϫ) were fused with enucleated germinal vesicle oocytes. After fusion with inactivated Sendai virus, the reconstructed oocytes were cultured for 14 h in ␣-minimum essential medium (Invitrogen). A spindle from the reconstructed oocytes was again transferred into the ovulated MII oocytes and treated with 10 mM SrCl 2 in Ca 2ϩfree M16 medium for 2 h. Bimaternal embryos of four different genotypes were obtained: the ng WT /fg, ng ⌬ch7 /fg, ng ⌬ch12 /fg, and ng ⌬Double /fg embryos (13). These embryos were cultured in the M16 medium at 37°C for 3.5 days in an atmosphere containing 5% CO 2 , 5% O 2 , and 90% N 2 . The embryos that developed to the blastocyst stage were transferred into the uterine horns of recipient female mice at 2.5 days of pseudopregnancy. The placentae were recovered from the pregnant mice and used in subsequent microarray and morphometric analyses. All experiments were carried out on three individual placentae of each class: the ng ⌬ch7 /fg, ng ⌬ch12 /fg, ng ⌬Double /fg bimaternal, and WT placentae. Genotyping of the bimaternal conceptuses was performed with the use of their yolk sacs, as reported previously (13).
Histological and Morphometric Analyses-We measured the areas of the labyrinthine and spongiotrophoblast layers using our previously described method (10). Briefly, we captured digitized images of the midline paraffin sections stained with hematoxylin/eosin. Placental entire images were saved as high quality TIF files and analyzed by the MetaMorph software (Universal Imaging Co., Downingtown, PA). The total number of pixels in each layer was calculated with the measurement tool of this software. Additionally, we analyzed the average areas of giant cells and the number of blood vessels within the labyrinth using the PALM Robo software 2.2-0103 (P.A.L.M. Microlaser Technologies AG). We calculated the number of blood vessels by tracing the area surrounding each blood vessel, i.e. within 114,550 m 2 of the labyrinth. In the case of giant cells, the average areas of five giant cells per hematoxylin/eosin section from each placental type were calculated. Morphological analyses were carried out at the end point of mouse gestation, namely E19.5, because the phenotypical gaps between the WT and the bimaternal placenta trended to expand with the advance of gestation (10).
Microarray Analysis-We used a commercially available acid-phenol reagent (TRIzol; Invitrogen) for total RNA extraction from four types of whole placentae, ng ⌬ch7 /fg, ng ⌬ch12 /fg, ng ⌬Double /fg bimaternal, and WT, at E12.5. Transcriptomic analysis was performed at E12.5, when all types of the bimater-nal conceptuses showed relatively high survivability as compared with the later embryonic days, and their placentae appeared to be appropriate for reliable evaluation (15). The total RNA from three replicates of each placenta was used for the GeneChip one-cycle eukaryotic target labeling assay (Affymetrix, Santa Clara, CA). Fifteen micrograms per replicate of fragmented cRNA samples was serially hybridized to the GeneChip mouse genome 430 2.0 array (Affymetrix) that contained 45,101 probe sets, and these were then processed according to the manufacturer's instructions (GeneChip Expression Analysis Technical Manual, Affymetrix (34)).
Briefly, the GeneChip operating software 1.3 (GCOS; Affymetrix) was first used to quantify microarray signals with default analysis parameters. The GCOS output files were then loaded into GeneSpring v7.3 (Agilent Technologies, Santa Clara, CA) with per chip normalization to the 50th percentile and per gene normalization to the average expression level of the WT placenta. The first operation simply identified transcripts with a raw signal of at least 50 in terms of the expression level of the WT placenta. A filtered list was created of all the genes detected (GCOS "P" and "M" calls) in all replicates. The GCOS output files of all the genes after GeneSpring normalization were also used in a one-way analysis of variance analysis with post hoc testing using Tukey's true test for significance (p Ͻ 0.01 at E12.5). These lists were referred to as the chr12 and chr7 gene lists. All genes in the chr12 and chr7 gene lists were integrated into the IPA pathways on the basis of the biological relationship described in previous studies.
Ingenuity Pathway Analysis-IPA version 3.0 was used to determine the possible biological pathways and the inter-relationships between network genes in subsets of candidate genes that had particularly interesting patterns. A detailed description of IPA can be found at the Ingenuity Systems site. Data sets containing the Affymetrix gene identifiers and their corresponding expression -fold change values were uploaded as Excel files. Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base. Network building was initiated with the program for querying the Ingenuity Pathways Knowledge Base for interactions between focus genes and all the other gene objects stored in the Knowledge Base. A set of networks was then generated, and the IPA computed a score for each network according to the fit of the network to the set of focus genes. Biological functions were calculated and assigned to each network.
Quantitative RT-PCR Analysis-Total RNA preparation was carried out as described in microarray analysis. The cDNAs were then synthesized using the SuperScript TM III RNase H reverse transcriptase kit (Invitrogen) in a reaction solution (20 l) containing the total RNA (1 g) prepared from each placenta. Finally, we performed a quantitative analysis of the gene expression by using real-time PCR (7500 real-time PCR system, Applied Biosystems) after preparing a reaction mixture (SYBR GREEN PCR master mix, Applied Biosystems). The primers used for the analysis were as described in supplemental Table 1. Ten transcripts were selected from the results of the array expression analysis (see Figs. 7 and 8) for verification by real-time RT-PCR; however, the expression level of the Pla2g2c gene was below the detection limit of this method of analysis.
Statistical Analyses-Data on the mRNA expression level by real-time RT-PCR were examined using analysis of one-way analysis of variance and the Fisher's protected least significance difference test by using the statistical analysis software Statview (Abacus Concepts, Inc., Berkeley, CA). A p value of Ͻ0.05 was considered significant.

RESULTS AND DISCUSSION
Definitive Placentation of ng ⌬Double /fg Bimaternal Conceptuses-As stated previously, we observed reciprocal defects in the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae (10). Furthermore, we found that the weight of the ng ⌬Double /fg placenta, which could First of all, we precisely investigated that the ng ⌬Double /fg placenta is phenotypically identical to the WT placenta followed by transcriptomic analysis. a, investigation of differentially expressed genes in the ng ⌬Double /fg placenta relative to those in the WT placenta may lead to normality verification of the ng ⌬Double /fg placenta (Table 1; Fig. 3). b, differentially expressed genes in the ng ⌬Double /fg placenta relative to those in the ng ⌬ch7 /fg placenta may be associated with peculiar disorders to the ng ⌬ch7 /fg placenta, suggesting biological functions of paternally methylated imprinted genes on chromosome 12 in placentation (Tables 2 and 4; Figs. 3, 4, 6, and 7). c, similarly, differentially expressed genes in the ng ⌬Double /fg placenta relative to those in the ng ⌬ch12 /fg placenta may be associated with peculiar disorders to the ng ⌬ch12 /fg placenta, suggesting biological functions of paternally methylated imprinted genes on chromosome 7 in placentation (Tables 3 and 5 support term development of bimaternal fetuses, was equivalent to that of the WT placenta (13); however, in-depth phenotypic studies are required to confirm this. In this study, we observed for the first time that the ng ⌬Double /fg placenta showed proportionate ratios of the spongiotrophoblast and labyrinthine layers (Fig. 2, a and b) with normal-sized giant cells (Fig. 2, c and e). Furthermore, the average area of the vasculature in the ng ⌬Double /fg placenta was similar to that of the WT placenta (Fig. 2d), indicating that the circulatory system of the ng ⌬Double /fg placenta was normal. In the ng ⌬Double /fg placenta, malformations specific to the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae were entirely restored, as typified by the distinct boundary between the spongiotrophoblast and labyrinthine layers (Fig.  2e). These results demonstrate that the ng ⌬Double /fg bimaternal conceptuses have a definitive placenta with regard to the competency to support both term development and morphogenesis. Therefore, comparison of the gene expression profiles of the ng ⌬Double /fg placenta and other bimaternal placentae, namely, the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae that showed reciprocal defects (10), may provide further insight into the roles of paternally methylated imprinted genes on chromosomes 7 and 12 in mouse placentation.
Global Gene Expression Analysis of the Bimaternal Placentae-Global gene expression analysis was performed using a microarray method to investigate the underlying alterations in gene expression from defective placentation in the ng ⌬ch7 /fg and ng ⌬ch12 /fg conceptuses to normalized placentation in the ng ⌬Double /fg conceptuses. We constructed hierarchical clustering with all the transcripts on the Genome 430 2.0 chip set using microarray data from the WT, ng ⌬ch7 /fg, ng ⌬ch12 /fg, and ng ⌬Double /fg placentae (n ϭ 3) (Fig. 3a). As expected, each placental type was clustered depending on the genetic background, suggesting that the results reflected the gene expression pattern specific to each ng/fg bimaternal placental type. The validity of this clustering was supported by the principal component analysis (PCA) (x axis, PCA component 1, 83.83% variance; y axis, PCA component 2, 4.575% variance; z axis, PCA component 3, 3.984% variance). The results indicated that among the three variances, the x axis showed the most significant variance (Fig. 3b). In the x-y plane, the ng ⌬Double /fg placenta was relatively closer to the WT placenta, whereas the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae almost coincided with each other. Both of them were distant from the WT placenta, indicating that the expression pattern of the ng ⌬Double /fg placenta was comparatively closer to that of the WT placenta than to those of the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae. Thus, the concordance between the results from the hierarchical cluster analysis and PCA indicated that the global gene expression pattern of the ng ⌬Double /fg placenta was closer to that of the WT placenta than to those of the other bimaternal placentae, i.e. the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae.
Next, one-way analysis of variance post hoc testing results after GeneSpring normalization revealed that in terms of the number of genes that were significantly expressed, there was a smaller difference between the WT and ng ⌬Double /fg placentae than between the WT and other bimaternal placentae (Fig. 3c). We also prepared a list of genes that were differentially expressed in the ng ⌬Double /fg and WT placentae. Of the transcripts that were significantly altered in the ng ⌬Double /fg placenta, 2 and 13 genes were up-and down-regulated by at least 2-fold each (Table 1). However, the biological function of most of these genes in the placenta is unknown; for example, Ser-pinb1 was up-regulated by more than 6-fold in the ng ⌬Double /fg placenta. This gene encodes an efficient inhibitor of neutrophil serine proteases and is required to protect the neutrophil and collectin surfactant protein-D from excess neutrophil serine proteases in the lung (16). However, the function of this gene in the placenta is unknown. Furthermore, Dnmt3a was down-regulated by more than 2-fold in the ng ⌬Double /fg placenta; its gene product functions as a de novo methyltransferase that is essential for normal development (17). Down-regulation of Dnmt3a by ϳ2-fold is likely to cause little change in early mouse placentation, and this observation was supported by the viability of mutants carrying heterozygously deleted Dnmt3a from E8.5 to E15.5 (17). However, at present, the reason for Dnmt3a downregulation is unclear. Nevertheless, by analogy with the phenotypical similarity between the WT and ng ⌬Double /fg placentae, genes that were altered in the ng ⌬Double /fg placenta including Serpinb1 and Dnmt3a do not appear to be involved in placental morphogenesis, at least, with regard to the phenotypes examined in Fig. 2.
Furthermore, 138 and 159 genes in the ng ⌬Double /fg placenta were differentially expressed as compared with those in the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae, respectively. Using these genes, we constructed a Venn diagram with the Gene Spring software, and gene lists that represented differentially and specifically expressed genes in the ng ⌬Double /fg placenta relative to the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae were prepared (Fig. 3d). Identifying differentially expressed genes in the ng ⌬Double /fg placenta relative to those in the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae may assist in elucidating the peculiar roles played by paternally methylated imprinted genes on chromosomes 7 and 12 in mouse placentation. The results indicated that between the ng ⌬Double /fg and the ng ⌬ch7 /fg placentae (ch12 gene list), 96 differentially expressed genes were present, whereas between the ng ⌬Double /fg and the ng ⌬ch12 /fg placentae, 117 differentially and specifically expressed genes were present (ch7 gene list) (Fig. 3d).
Complementary Functions of Paternally Methylated Imprinted Genes on Chromosomes 7 and 12 in Mouse Placentation-To understand the biological and molecular processes represented in the ch12 and ch7 gene lists, we per- formed IPA. On the basis of their known biological relationships, which have been described previously (13, 18 -20), the genes in these two lists were integrated into the IPA pathways. The result of the integration showed that six networks were produced from the ch12 and ch7 gene lists (Tables 2 and 3; Figs. 4 and 5). In the ch12 gene list, we investigated the biological functions of six networks that resulted from globally altered gene expression due to paternalization of the gene expression pattern in the Dlk1-Dio3 region in the ng ⌬Double /fg placenta; this expression pattern differed from that in the ng ⌬ch7 /fg placenta (Table 2; Fig. 4). This list of genes includes members of gene families that have been identified in previous studies and genes involved in the cell cycle, organ morphology, cellular development, etc. In the ch7 gene list, the biological functions of the six constructed networks largely showed genes involved in the cell cycle, carbohydrate metabolism, and cellular growth   (n ‫؍‬ 3). a, hierarchical clustering of all the samples from different genetic backgrounds. Colors correspond to the relative RNA abundance for more than 39,000 transcripts. b, principal component analysis of gene expression in all the groups subjected to the hierarchical clustering analysis. c, one-way analysis of variance post hoc testing analysis of the following four placenta types: WT, ng ⌬ch7 /fg, ng ⌬ch12 /fg, and ng ⌬Double /fg placentae. Each box shows the number of genes that are statistically similar (green) or different (red) in a group-to-group comparison. The intensity of the green and red colors indicates whether the value is more or less than the sum of the gene number. d, Venn diagrams constructed with the GeneSpring software show genes that were differentially and specifically expressed in the ng ⌬Double /fg placenta in comparison with the ng ⌬ch7 /fg (red, ch12 gene list) and the ng ⌬ch12 /fg (blue, ch7 gene list) placentae individually. These gene lists were integrated into the IPA pathways (Figs. 4 and 5). and proliferation (Table 3). In addition, as expected, Dlk1 and Igf2 were incorporated into network 3 in Table 2 (Fig. 4) and networks 1 and 4 in Table 3 (Fig. 5). The biological functions of network 3 in Table 2 were cellular development, connective tissue development, and functional and developmental disorders; this result is consistent with our previous observations of severe dysplasia in the ng ⌬ch7 /fg placenta (10). Genes in networks 1 and 4 in Table 3 are involved in the cell cycle, cancer, reproductive system diseases, cellular movement, and cell death. Cell cycle and cell death may be responsible for the phenotypes of the ng ⌬ch12 /fg placenta, namely, extremely low placental mass and the presence of expanded giant cells. Furthermore, the constructed networks were bound in the commonly appearing genes as shown in Fig. 6. Thus, the present transcriptomic study suggests that Dlk1 and Igf2 play pivotal roles in improving the placentation of bimaternal conceptuses. Both genes appear to have propagating effects on other multiple genes involved in mouse placentation.
Further Investigations into the ch12 and ch7 Gene Lists-We also screened the genes in the ng ⌬Double /fg placenta whose differential expression was more than 2-fold that of the genes in the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae (Tables 4 and 5). Furthermore, to evaluate the definite effects of paternalization in the Dlk1-Dio3 and Igf2-H19 regions, we noted the experimental variability after determining the expression levels in three microarray trials (Figs. 7 and 8). If the expression level of a given gene remains within the range between the levels of the WT and ng ⌬Double /fg placentae, the gene should be regarded as being within the range of normality. This is because genes that are differentially expressed between the WT and ng ⌬Double /fg placentae are apparently not directly/indirectly associated with mouse placentation, at least in placental morphogenesis analyzed in this study (Fig. 2).
In comparison with the ng ⌬ch7 /fg placenta, the most altered gene in the ng ⌬Double /fg placenta was Dlk1 (Table 4; Fig. 7). This gene is paternally expressed in the endothelial cells of fetal

TABLE 2
Networks generated by IPA for transcripts that were specifically and differentially expressed in the ng ⌬Double /fg placenta in comparison with the ng ⌬ch7 /fg placenta Statistically significant genes were specifically and differentially expressed in the ng ⌬Double /fg placenta in comparison with the ng ⌬ch12 /fg placenta (ng ⌬Double /fg versus ng ⌬ch12 /fg). These were used as the input in the IPA. blood vessels and is present on chromosome 12. It encodes a transmembrane protein containing epidermal growth factor repeats and is a member of the Notch/Delta/Serrate family of developmental signaling molecules. It is involved in several differentiation processes and is expressed in the fetal endothelial cells of the murine placenta; however, its precise function in the placenta is yet to be determined (21)(22)(23)(24)(25)(26). After Dlk1, the gene that was most up-regulated and showed more than a 9-fold change in the ng ⌬Double /fg placenta was Ser-pina6 (Fig. 7). This gene encodes an ␣-globulin protein with corticosteroid binding properties and is the major transport protein for glucocorticoids and progestins in the blood of most vertebrates. The function of Serpina6 in the placenta remains unknown; however, it is known that its corresponding protein localizes to the placenta, particularly to the maternal and fetal blood vessels (27). A low level of Serpina6 expression may be a possible cause of vascular defects in the ng ⌬ch7 /fg placenta. In contrast, the Syne2 mRNA was overexpressed in the ng ⌬ch7 /fg placenta. In humans, the SYNE2 mRNA is expressed in vascular smooth muscle cells and skeletal muscle cells, which are associated with the sarcoplasmic reticulum, and it may provide a network of scaffolds that spatially orients the myofibrils, sarcoplasmic reticulum, and cytoskeleton (28). However, there are no further details on the specific functions of SYNE2. Syne2 transcription is positive in the mouse placenta (29), but further studies are required to determine whether Syne2 overexpression in the ng ⌬ch7 /fg placenta is involved in a spectrum of placental vascular and angiogenic defects.

ID Molecules in network
In the ch7 gene list, Igf2 was the most altered gene in the ng ⌬Double /fg placenta in comparison with the ng ⌬ch12 /fg placenta. This gene encodes a member of the insulin family of polypeptide growth factors, which is involved in placental development and growth (30) ( Table 5; Fig. 8). It is an imprinted gene on chromosome 7 and is expressed only from the paternally inherited allele in the placenta and fetal tissues, excluding the brain (31). Igf2 mRNA expression level in the ng ⌬Double /fg placenta was equivalent to that in the WT placenta (Fig. 8) and, therefore, the normal-sized ng ⌬Double /fg placenta was attributed to restored Igf2 mRNA expression. In our previous study, the ng ⌬ch7 /fg placenta consistently showed a greater mass than the ng ⌬ch12 /fg placenta; however, these masses were significantly lesser than that of the WT placenta (10). In the ng ⌬ch7 /fg placenta, the Igf2 mRNA expression level did not reach the range between the levels of the WT and ng ⌬Double /fg placentae, although it was much higher than the levels of the ng ⌬ch12 /fg placenta. This suggests that appropriate Igf2 expression requires proper expression of paternally methylated imprinted genes on chromosome 12. The appropriate expression pattern of paternally methylated imprinted genes on chromosome 12 leads to normal differentiation of the trophoblast, resulting in orderly histogenesis (10). This histological normality may be one of the conditions required for appropriate Igf2 mRNA expression in the placenta. The slight hypoplasia of the ng ⌬ch7 /fg placenta with a disproportionally layered structure may be due to the improper expression pattern of paternally methylated imprinted genes on chromosome 12. The observation that the mass of the ng ⌬Double /fg placenta was completely normalized provided further evidence for this (Fig. 2).
After Igf2, the genes that were up-regulated by more than 2-fold in the ng ⌬Double /fg placenta were Pla2G2C, Tacc2, and Txn (Fig. 8), although the biological functions of these genes in the placenta have not yet been established. Meanwhile, Aldh1a3 and Cxcl10 were down-regulated by more than 2-fold in the ng ⌬Double /fg placenta. The function of Aldh1a3 in the placenta is also unknown, but it is known that Cxcl10 encodes a chemokine of the CXC subfamily, and the binding of this pro- Networks generated by IPA for transcripts that were specifically and differentially expressed in the ng ⌬Double /fg placenta in comparison with the ng ⌬ch12 /fg placenta Statistically significant genes were specifically and differentially expressed in the ng ⌬Double /fg placenta in comparison with the ng ⌬ch7 /fg placenta (ng ⌬Double /fg versus ng ⌬ch7 /fg). These were used as the input in the IPA analysis.      . circle and diamond) indicates whether the protein is a structural protein, transcription factor, etc. (refer to supplemental Fig. 1).       FIGURE 5. Functional networks of genes that are differentially and specifically expressed in the ng ⌬Double /fg placenta in comparison with the ng ⌬ch12 /fg placenta (refer to the legends for Fig. 4 and supplemental Fig. 1).

ID Molecules in network
tein to CXCR3 results in pleiotropic effects. Additionally, Cxcl10 overexpression stimulates cell proliferation (32,33), which may be associated with the gain of placental mass. In fact, we had observed that the placental mass in the ng ⌬ch12 /fg conceptus increased as compared with that in the ng WT /fg conceptus. Therefore, this result indicates that up-regulation of Cxcl10 compensates for the gain in placental mass in the ng ⌬ch12 /fg conceptus lacking Igf2 expression. However, in the ng ⌬Double /fg placenta, Cxcl10 overexpression was entirely restored. This is attributed to the fact that Igf2 transcription level in the ng ⌬Double /fg placenta is normalized.
Validation of the Microarray Data-Ten transcripts were selected from the array expression profile to validate the microarray data by quantitative RT-PCR (Fig. 9). Among them, nine transcripts were detectable as follows: Dlk1, Serpina6, Tram2, Syne2, Igf2, Tacc2, Txn, Aldh1a3, and Cxcl10. The expression level of the Pla2g2c gene was undetectable with little expression; therefore, we could not verify the further analysis by quantitative RT-PCR. The expression patterns of Dlk1 and Igf2 genes were identical to those by the microarray expression analysis (Fig.  9, a and b). Furthermore, in the ch12 gene list, the Serpina6 and the Syne2 genes were significantly down-and up-regulated in the ng ⌬ch7 /fg placenta as expected (p values; 0.0061 and 0.0258, respectively) (Fig. 9a).
In the ch7 gene list, the anomalous expression patterns of the Tacc2, Aldh1a3, and Cxcl10 RNAs in the  Tables 4 and 5.

TABLE 4
Genes showing a 2-fold or higher expression difference between the ng ⌬Double /fg and ng ⌬ch7 /fg placentae Thirteen representative genes in the ng ⌬Double /fg placenta were expressed at a level that was 2-fold higher than that in the ng ⌬ch7 /fg placenta (bold, up-regulated; italic, down-regulated). TCDD, 2,3,7,8-tetrachlorodibenzo-p-dioxin; POZ, poxvirus and zinc finger; BMP, bone morphogenetic protein.

Description -Fold change Networks Location
Dlk1

TABLE 5
Genes exhibiting a 2-fold or higher expression difference between the ng ⌬Double /fg and ng ⌬ch12 /fg placentae Fourteen representative genes in the ng ⌬Double /fg placenta were expressed at a level that was 2-fold higher than that in the ng ⌬ch12 /fg placenta (bold, up-regulated; italic, down-regulated).

Igf2
Insulin-like growth factor 2 (somatomedin A)   Table 4. The y axes show the raw expression values from microarray analysis. If the expression level of a focused gene in only the ng ⌬ch7 /fg placenta is afield from the range between the levels of the WT and ng ⌬Double /fg placentae (the realm between the two broken lines), the gene expression is considered to be altered in the ng ⌬Double /fg placenta due to the restored expression of paternally methylated imprinted genes on chromosome 12 (colored in yellow). WT ng ∆ch7 /fg ng ∆ch12 /fg ng ∆Double /fg FIGURE 8. This figure shows the expression level of genes in the ng ⌬Double /fg placenta that are differentially expressed at a level higher than 2-fold that of the ng ⌬ch12 /fg placenta for each placenta (n ‫؍‬ 3). Refer to Table 5. If the expression level of a focused gene in only the ng ⌬ch12 /fg placenta is afield from the range between the levels of the WT and ng ⌬Double /fg placentae (the area between the two broken lines), we considered the gene expression to be altered in the ng ⌬Double /fg placenta due to the restored expression of paternally methylated imprinted genes on chromosome 7 (colored in yellow). However, Igf2 was an exception because the Igf2 expression level was remarkably higher than that of the other genes, although expression in the ng ⌬ch7 /fg placenta narrowly missed the range between the two broken lines (shown in light blue).  3). Nine transcripts were selected for quantitative RT-PCR analysis from the array expression profile of the ng ⌬Double /fg placenta to the ng ⌬ch7 /fg or ng ⌬ch12 /fg placentae ( Fig. 7 and 8). a, four transcripts that were selected from 13 transcripts shown in Fig. 7, namely, Dlk1, Serpina6, Tram2, and Syne2. b, five transcripts that were selected from 14 transcripts shown in Fig. 8 ng ⌬ch12 /fg placenta were also similar to those in microarray analysis, namely, Tacc2 transcript was characteristically downregulated, and Aldh1a3 and Cxcl10 were overexpressed in the ng ⌬ch12 /fg placenta (p values; 0.0147, 0.0386, and 0.0138, respectively) (Fig. 9b). Although of the nine transcripts, the Tram2 and Txn expression levels in the ng ⌬ch7 /fg and ng ⌬ch12 /fg placentae showed no significant differences as compared with those in the ng ⌬Double /fg placenta, the expression patterns of the other seven genes were consistent with the result of microarray analysis. We also confirmed that the expression patterns of these genes in the ng WT /fg placenta containing wild-type ng genomes did not denote the same tendency of those in ng ⌬ch12 /fg placentae (supplemental Fig. 2). This might be because grossly inequable tissues in cell composition in the ng WT /fg placenta caused inexplicable downstream effects on transcription levels of these genes. Consequently, judging by the results from quantitative RT-PCR by using the ng ⌬ch7 /fg, ng ⌬ch12 /fg, and ng ⌬Double /fg placentae, the concordance of results obtained from these two independent lines of experimentation strongly suggests that the expression profiles derived from the microarray analysis precisely represent the qualitative changes in gene expression that occur during improved placentation in the ng ⌬Double /fg conceptus.

Conclusions
Bimaternal mouse conceptuses in which the expression pattern of the paternally methylated imprinted genes in the Igf2-H19 and Dlk1-Dio3 regions are inappropriate resulted in lethality before E13.5, accompanied by acute placental disorders. Experiments involving paternalization of both the Igf2-H19 and the Dlk1-Dio3 regions with the use of mutant mice clearly led to definitive placentation. The global gene expression data sets in these bimaternal placentae verified the rigorous complementary correlation of paternally methylated imprinted genes on chromosomes 7 and 12. The results indicated that the contributions of paternally methylated imprinted genes on chromosomes 7 and 12 to placentation were distinctive; the former mainly controlled cell proliferation-and cell growth-related genes. In contrast, the latter controlled tissue differentiationrelated genes. Furthermore, the networks could be used to identify the functions of imprinted genes regulated by either of the two paternally methylated imprinting-control regions that have been annotated as hypothetical or as function unknown. Analysis of our transcriptome with bimaternal placentae lacking the paternal genome can help in deciphering the unknown functions and pathways of paternally methylated imprinted genes during mouse placentation.