Developmental Cell, Vol. 6, 133–144, January, 2004, Copyright 2004 by Cell Press

A Genome-Wide Study of Gene ActivityReveals Developmental Signaling Pathwaysin the Preimplantation Mouse Embryo

mouse is the primary model system for mammalian de-velopment and diseases, and the early mouse embryoprovides the developmental context to address manyfundamental questions, including establishment of cel-lular polarity, lineage differentiation, developmental po-

Q. Tian Wang,1,2,4 Karolina Piotrowska,3

Maria Anna Ciemerych,3,5 Ljiljana Milenkovic,2

Matthew P. Scott,2 Ronald W. Davis,1,4

and Magdalena Zernicka-Goetz 3

1Department of Biochemistry2 Departments of Developmental Biology and tency, and specification of embryonic axes. However,

the early mammalian embryo is far from amenable toGenetics andHoward Hughes Medical Institute the powerful tools of phenotype-based genetic screens,

and thus many genes are yet to be characterized forStanford University School of MedicineStanford, California 94305 their roles during early, in particular preimplantation,

embryonic development.3 Wellcome Trust/Cancer Research UK Institute andDepartment of Genetics Although the early mammalian embryo is refractory

to experimental disruptions such as aggregation or re-University of CambridgeTennis Court Road moval of blastomeres or polar cytoplasm, embryonic

polarity and spatial pattern begin to develop at the pre-Cambridge CB2 1QRUnited Kingdom implantation stage (Lu et al., 2001; Zernicka-Goetz,

2002). By the blastocyst stage, a set of distinct lineages4 Stanford Genome Technology Center855 California Avenue of cells is established that could potentially interact with

each other (Tam et al., 2001). Moreover, it has beenPalo Alto, California 94304recently shown that the axes of blastocyst polarizationcan be traced back to spatial cues and cell fate regula-tion at the beginning of development (Piotrowska andSummaryZernicka-Goetz, 2001; Gardner, 2001; Piotrowska et al.,2001; Fujimori et al., 2003). The orientation of the firstThe preimplantation development of the mammaliancleavage is regulated so that the cleavage plane roughlyembryo encompasses a series of critical events: thecoincides with one determined by the second meiotictransition from oocyte to embryo, the first cell divi-division and the site of sperm entry (Piotrowska andsions, the establishment of cellular contacts, the firstZernicka-Goetz, 2001; Plusa et al., 2002a, 2002b). Whenlineage differentiation—all the first subtle steps to-development is unperturbed, the resulting daughterward a future body plan. Here, we use microarrays tocells tend to follow different fates so as to foreshadowexplore gene activity during preimplantation develop-the embryonic-abembryonic axis of the blastocyst. Thusment. We reveal robust and dynamic patterns of stage-at least some basic aspects of the postimplantationspecific gene activity that fall into two major phases,embryo body plan may have a preimplantation originone up to the 2-cell stage (oocyte-to-embryo transi-(Weber et al., 1999). To search for regulators that maytion) and one after the 4-cell stage (cellular differentia-be involved in control of early polarity and pattern, it istion). The mouse oocyte and early embryo expressimportant to have a global picture of the regulation ofcomponents of multiple signaling pathways includinggene expression in the preimplantation mouse embryo.those downstream of Wnt, BMP, and Notch, indicating

In recent years, developmental biologists have turnedthat conserved regulators of cell fate and pattern for-to genome-scale expression analysis to facilitate genemation are likely to function at the earliest embryonicdiscovery. For example, cDNA and oligonucleotide mi-stages. Overall, these data provide a detailed temporalcroarrays have been used to investigate the develop-profile of gene expression that reveals the richness ofmental regulation of gene expression in C. elegans (Ji-signaling processes in early mammalian development.ang et al., 2001; Baugh et al., 2003) and Drosophila(White et al., 1999; Furlong et al., 2001; Arbeitman et al.,Introduction2002). In the preimplantation mouse embryo, a large-scale cDNA analysis has been performed that allowedOur current understanding of the development of multi-some insight to the “phased” gene expression patternscellular organisms has greatly benefited from molecular(Ko et al., 2000). This was achieved by single-pass se-genetic studies using model systems including C. ele-quencing of the 3� ends of clones, or expressed se-gans, Drosophila, zebrafish, and mouse. Developmentalquence tags (ESTs), from seven cDNA libraries repre-processes are routinely analyzed in terms of their under-senting different stages of development. The frequencylying genetic components. The identification of genesof ESTs provides a first approximation of the expressionfunctioning in a particular process often paves the waylevel of genes. However, in order to faithfully reflectfor in-depth biochemical and cell biological studies. Thethe relative abundance of transcripts that may differ bymany hundred fold, an astronomical number of ESTs

*Correspondence: qwa205@stanford.edu (Q.T.W.), mzg@mole.bio. need to be cloned and sequenced, which is not yetcam.ac.uk (M.Z.-G.)

practical. A less comprehensive but more feasible ap-5 Present address: Department of Embryology, Institute of Zoology,proach is suppression subtractive hybridization. In aFaculty of Biology, University of Warsaw, Miecznikowa 1, 02-096

Warsaw, Poland. recent study, this technique was used to identify oocyte-

Developmental Cell134

a high degree of reproducibility in sample collection,Table 1. Time Points of the Transcription Profile Time CourseRNA preparation, RNA amplification, and array hybrid-

Number of Range of r Values Average rization. Furthermore, these results suggest that em-Stage Replicates between Replicates Valuebryos staged by morphological criteria are also similar

GV oocyte 4a 0.982–0.996 0.988 at the level of gene expression.Metaphase II (MII) 3 0.968–0.994 0.977 In general, early cleavage-stage embryos exhibit bet-Zygote 4 0.970–0.993 0.980

ter correlation between replicates (average r valueEarly 2-cell 2 n/a 0.985�0.984) than postcompaction embryos (average r valueMid 2-cell 3 0.974–0.981 0.977�0.963). The smaller r value in later stages could beLate 2-cell 2b n/a 0.996

4-cell 3 0.988–0.994 0.991 due to increased variance between individuals, to lesser8-cell 3 0.975–0.985 0.981 synchrony in the postcompaction embryos, or to the16-cell 2 n/a 0.970 increased numbers of cell types at later stages. In con-Early blastocyst 4c 0.958–0.980 0.973

trast to the consistent spectrum of transcripts detectedMid blastocyst 2b n/a 0.948at a particular time point, the changes in gene activityLate blastocyst 3 0.957–0.970 0.962over time are significant (r values between consecutive

n/a, not applicable. time points shown in Figure 1A). During preimplantationa These four replicates were from two independent embryo collec-development, the expression levels of more than 4000tions, which were each independently profiled twice.genes varied by �5-fold. The 4000 genes are aboutb These two replicates were from one embryo collection that wasone-third of the genes examined, suggesting that tran-independently profiled twice.

c These four replicates were from three independent embryo collec- scription, adenylation, or RNA stability of a significanttions. One collection was profiled twice, and the average expression percentage of the genome is regulated during preim-values were used to calculate correlation to the other two collec- plantation development.tions.

During the transition from germinal vesicle stage oo-cyte to metaphase II (MII) oocyte, a 14–15 hr processknown as meiotic maturation, the oocyte responds tohormone stimulation by exiting from cell cycle arrestversus 8-cell-embryo-specific cDNA clones (Zeng andat prophase I and reentering meiosis (Wassarman andSchultz, 2003). Clones that hybridize more strongly toDePamphilis, 1993). During these events, transcripts ofone library than to the other (more than 5-fold) werelarge numbers of genes change level in both directionsconsidered interesting and further identified by se-(737 genes have a �3-fold increase and 1082 genesquencing. Compared to single-pass EST sequencinghave a �3-fold decrease) (Figure 1B). The decreasesand suppression subtractive hybridization, microarray-are consistent with previous work showing that meioticbased transcription profiling is a more high-throughputmaturation is accompanied by the degradation and/ormethod to analyze gene expression patterns and gener-deadenylation of a large amount of maternal mRNAates more quantitative information. However, due to the(Paynton et al., 1988). While the synthesis of new tran-difficulty in obtaining sufficient intact RNA to probe mi-scripts essentially ceases after the nuclear envelopecroarrays on which most or all of the genome is repre-breaks down as the oocyte reenters meiosis (Bachvar-sented, no systematic studies of this type have beenova, 1985), the poly(A) tails of some classes of existingperformed previously in early mouse embryos. This ledtranscripts are elongated, leading to increased transla-us in this study to use microarrays to investigate thetion and protein levels (Bachvarova, 1985, 1992). Thedevelopmental regulation of gene activity in the earlyregulation of poly(A) tail length is a major mechanismmouse embryo and examine its dynamics throughoutfor controlling the activity of maternal transcripts (Bach-12 time points from the unfertilized oocyte to late blasto-varova, 1985, 1992). Because the RNA amplification usedcysts.here involves annealing between the poly(A) tail andan oligo-(dT) primer, an elongated poly(A) tail may in-

Results crease the efficiency of annealing and therefore the de-tection of such transcripts. Indeed, close examination

The Transcriptome of the Early Embryo Is Stage ofgenes that are known to undergo polyadenylation orSpecific and Evolves Over Time deadenylation during meiotic maturation (Gosden, 2002)Using improved RNA extraction and amplification tech- confirmed that the poly(A) length-mediated regulationniques (Wang et al., 2002), we have generated microar- of gene activity is reflected in our data (Supplementalray-based gene expression data representing 12 devel- Table S2).opmental time points from germinal vesicle (GV) stage Following fertilization, there is a reprogramming ofoocyte to expanded blastocyst (Table 1). Oocytes or gene expression as transcription from the zygotic ge-embryos were carefully staged based on morphological nome begins and maternal messages continue to becriteria and collected in pools of 30–134 (Supplemental degraded and/or deadenylated (Paynton et al., 1988;Table S1 [http://www.developmentalcell.com/cgi/content/ Schultz, 2002). We observed an increase in the complex-full/6/1/133/DC1]). Transcription profiles were gener- ity of the embryo’s RNA population during this timeated on Affymetrix Murine Genome Array U74Av2 and (Figure 1C). To monitor how many genes are representedanalyzed as previously described (Wang et al., 2002). by detectable transcripts at a given stage, we eitherOverall similarity in profile between replicates of the used the GeneChip Algorithm (Affymetrix, 1999) to deter-same time point is assessed by calculating Pearson mine the percentage of genes receiving “Present Calls”Correlation (r value), which ranges from 0.948 to 0.996 (P call %) (Supplemental Table S3) or counted the num-

ber of genes whose calculated transcript levels arewith a median of 0.977 (Table 1). These r values indicate

A Genomic Study of Preimplantation Development135

above a threshold (data not shown). Both analyses ledto similar conclusions. During meiotic maturation (Figure1C), the P call % decreased significantly, suggestingthat the oocyte’s RNA becomes less complex. This isconsistent with the virtual lack of new transcription andlarge-scale mRNA degradation that occur at this time(Paynton et al., 1988; Bachvarova, 1985). Upon fertiliza-tion and the onset of zygotic genome transcription(Schultz, 2002), the P call % begins to increase (Figure1C), suggesting that more genes are becoming tran-scriptionally active.

After the 8-cell stage, as the embryo embarks on com-paction, followed by differentiation of the trophecto-derm (TE) and inner cell mass (ICM) lineages, profilingof the whole embryo may fail to detect transcript levelchanges for genes that are induced or inhibited in aregional fashion, especially if the expression level is low.For example, if a gene is uniformly expressed in the8-cell embryo but is inhibited in the inner cells at the16-cell stage, profiling of the whole embryo would detecta �2-fold decrease in transcript level at most. Con-versely, if a gene is induced in a small number of cells,the transcripts may be too diluted in whole-embryo RNAto be detected with confidence. Despite the possiblyreduced sensitivity of profiling, the measurable com-plexity of the embryo’s transcripts become higher inthese later stages: the P call % averages 47.7 aftercompaction, compared to 44.2 between fertilization andcompaction (Figure 1C). This is consistent with the em-bryo’s need to fulfill increasingly diverse functions asits organization becomes more complex.

An unbiased analysis of the gene expression datareveals the transition in the molecular composition ofthe embryo as it develops. The time points were com-pared by hierarchical clustering (Sherlock, 2000) to seewhich are most similar in gene expression. This dividedthe 12 time points into two large groups (Figure 1D), oneconsisting of oocytes and the first cleavage (“oocyte-to-embryo phase”), the other consisting of later stages(“cellular differentiation phase”). Thus the molecularanalysis highlights the second cleavage as a major tran-sition point, a transition that is not discernible basedupon morphological criteria. The massive change ingene expression is consistent with the different develop-mental needs of early versus late preimplantation em-bryos. The protein activities needed for early tasks, suchas exiting meiotic arrest, are likely encoded by maternaltranscripts and by genes transcribed immediately uponthe onset of zygotic genome transcription. Later taskssuch as faster cleavage, intercellular signaling, and di-versification of cell fates are probably directed by zy-gotic genes that are induced after the embryo has suc-cessfully completed the oocyte-to-embryo transition.

Figure 1. Global Changes in Gene Expression during Preimplanta-tion Development

(A) Pearson correlation (r ) between consecutive time points reflectsthe degree of change in transcriptome. The plot serves as a visual (B) The numbers of probe sets whose expression increases or de-guide to the rate of transcriptome change through preimplantation creases by more than 3-fold between consecutive time points.development: the smaller the r value, the steeper the line connecting (C) The P call percentages vary greatly between different time points.the two time points, the bigger the change in transcript composition. The error bars reflect variation between replicates.The diagrams of embryos are, from left to right, precompaction (D) Hierarchical clustering of time points divides preimplantation8-cell, compacted 8-cell, 16-cell, and blastocyst. The time windows development into two phases, the “oocyte-to-embryo phase” andof several major developmental processes are diagrammed below. the “cellular differentiation phase.”

Developmental Cell136

Figure 2. The Data Set Recapitulated the De-velopmental Regulation of Gene Expressionat Various Stages during Preimplantation De-velopment

(A) Expression patterns of several maternaltranscripts during the transition from oocyteto embryo: Zp1-3, mos, tPA, and SLBP. Foreach gene, the peak expression level is setas 100% and the expression levels at othertime points are normalized to percentage ofpeak level (%peak).(B) Expression curves of ZO-1 and Cx43, twocell adhesion genes functioning during andafter compaction.(C) Expression curves of Eomes and Crtr-1,two genes with lineage-specific functions.

The Data Set Has Recapitulated the Temporal abundant in GV-stage oocyte to low in MII oocyte andto background level after fertilization (Figure 2A).Regulation of Gene Expression

at Various Stages In contrast, tPA, Moloney sarcoma oncogene (mos),and stem-loop binding protein (SLBP) exhibit an in-Inspection of the array data for a number of known genes

showed that the results are consistent with previously crease in expression during oocyte maturation. tPA en-codes a secreted serine protease that is a componentreported expression data and with the specific times of

early development when the genes are needed (Figure 2 of the proteolytic activity required for the rupture ofselected follicles during ovulation (Ny et al., 2002). Theand Supplemental Table S2). First, we looked at several

maternal transcripts whose expression changes during protein product of the mos gene, a serine-threoninekinase, is a key regulator of oocyte maturation and isoocyte development and maturation have been well

documented. Before the mouse oocyte embarks on a specifically required for meiosis II in the mouse (Col-ledge et al., 1994; Hashimoto et al., 1994). In the MIIdiscontinuous meiosis, it accumulates a large amount of

mRNAs during a growth phase. The synthesis of mRNA oocyte, the expresssion of both tPA and mos are in-creased from an already significant level in the GV oo-virtually ceases after prophase I, but the posttranscrip-

tional regulation continues to ensure the timely expres- cyte (Figure 2A). This induction is transient, and theexpression quickly decreases to undetectable levelssion of various genes (Bachvarova, 1992). A maternal

mRNA may be translated along with cellular housekeep- during the first cleavage. While SLBP is also induced inthe MII oocyte, its inhibition comes later and slowering genes throughout oogenesis and degraded after

meiosis and fertilization, such as transcripts encoding (Figure 2A), consistent with immunoblotting assays ofits protein abundance and with its proposed roles in thethe three zona pellucida glycoproteins (Zp1-3), or it may

be stored in dormant or “masked” forms for future ex- translational activation of histone mRNPs (Allard et al.,2002). The expression curves of Zp1-3, mos/tPA, andpression, such as tissue-specific plasminogen activator

(tPA). The Zp1-3 genes encode three zona pellucida SLBP fall into three distinct classes in good accordancewith the difference in their time windows of function.glycoproteins that are critical for fertilization, the post-

fertilization block to polyspermy, and protection of the These observations suggest that the array data recapitu-lates the transcription-independent regulation of genepreimplantation embryo (Epifano et al., 1995). In the

array data, the transcript levels of Zp1-3 change from expression during the transition from oocyte to embryo.

A Genomic Study of Preimplantation Development137

We next examined a number of zygotically expressed of a correctly polarized microtubule cytoskeleton re-quired for proper localization of the anterior and poste-genes functioning at different stages of preimplantationrior determinants Bicoid and Oskar and for the asymmet-development (Figure 2B). ZO-1 and Cx43 encode com-ric positioning of the oocyte nucleus (Roth et al., 1995).ponents of tight junctions and gap junctions, respec-The mouse genome contains two genes homologoustively (De Sousa et al., 1993; Sheth et al., 1997). Theto Drosophila cni, cnih, and cnil, both of which wereassembly of tight and gap junction complexes occursexpressed as maternal transcripts (Figure 3A and dataduring and after compaction, when the blastomeres losenot shown). The finding that murine celsr1, dag1, cnil,their round shape and become tightly packed with eachor cnih is expressed at this stage identifies them asother (Fleming et al., 2001). The array data showed thatcandidate regulators of the polarity of the oocyte, andboth genes are abruptly induced prior to compactionsubsequently of the zygote.(Figure 2B). Eomes and Crtr-1 encode transcription fac-

Consistent with the knowledge that fertilization initi-tors that regulate cell fate and pluripotency in the blasto-ates a cascade of cell cycle events including exit fromcyst (Russ et al., 2000; Pelton et al., 2002). The arraysoocyte arrest, completion of the second meiotic division,indicate the induction of these genes during and afterand the subsequent mitotic cleavage, we also spottedcompaction, with the highest transcript levels found atmaternal transcripts for several genes with unique func-the blastocyst stage (Figure 2C).tions in cell cycle regulation (Figure 3A, highlighted inThe genes examined for validation are expressed atblue). Nuclear protein 95 is an essential player in thedifferent times, and also at different levels (Supplemen-resumption of mitosis in terminally differentiated cellstal Table S2). Our data accurately reflect the transcript(Bonapace et al., 2002). Similarly, thymidylate kinase isfluctuations observed in previous studies. For example,regulated at both the mRNA and enzymatic activity levelZp3 transcripts are abundant in preovulation oocytesas mammalian cells exit quiescence and enter cell cycle(Epifano et al., 1995), while transcription from the Eomes(Liang et al., 1995). The gene G1 to phase transition 1locus is first detectable at a very low level in the blasto-is essential for the G1-to-S phase transition in yeast andcyst (Russ et al., 2000). In our data set, the peak expres-human (Hoshino et al., 1989). Transcripts of all threesion level of Zp3 in GV-stage oocyte (Supplemental Ta-genes are present until the zygote stage and are thenble S2) is 13–22 times higher than that of Eomes insignificantly reduced as the embryo undergoes its firstthe blastocyst (Supplemental Table S2). Although thiscleavage, suggesting that they may play important rolesdifference might not be a linear reflection of the absolutein the resumption of oocyte meiosis after arrest and/orabundance of the two transcripts, it shows that the datain the initiation of the first mitosis of the embryo.set encompasses a wide range of expression levels.

Thus, the data are of sufficient scope and accuracy toTemporal Patterns of Gene Expression Pointserve as a systematic gene expression exploration ofto Ordered Transcription Networkspreimplantation development.As the examples in Figure 2 have illustrated, in a devel-opmental system, the time window of a gene’s expres-sion, in particular its induction, often precedes and over-Maternal Transcripts of Polarity-Regulatinglaps that of its function. Moreover, molecular geneticGenes in the Mouse Oocytestudies in various model systems have found that whenGenes that are expressed in both GV and mature (meta-two genes are coregulated, there is a good chance thatphase II) oocytes but become inhibited by the late 2-cellthey participate in the same developmental process. Thestage were analyzed by hierarchical clustering. Twocomprehensive genomic view of transcription allows usknown genes with stage-specific expression and func-to inspect and compare the expression patterns of multi-tion during oocyte maturation, mos and tPA, were usedple functionally related genes, for example components

as landmarks to help narrow down our focus to a clusterof transcription networks. The POU-domain transcrip-

of 57 genes represented by 62 probe sets (Figure 3A).tion factor Oct4, encoded by the Pou5f1 gene (Nichols

Noticeable within this cluster were three genes whose et al., 1998; Niwa et al., 2000), is required for the mainte-homologs have been shown to regulate cellular polarity nance and proliferation of pluripotent cells in both thein Drosophila: flamingo, dystroglycan 1, and cornichon ICM and the ICM-derived embryonic stem (ES) cells.(Figure 3A, highlighted in red). The murine celsr1 gene Expression of Pou5f1 peaks in the 16-cell morula andencoding a cadherin EGF LAG seven-pass G-type re- gradually decreases to an intermediate level in the lateceptor is the homolog of Drosophila flamingo, which is blastocyst (the absolute expression level is still veryinvolved in planar polarity and asymmetric division of high) (Figure 4A and Supplemental Table S5). This ob-sensory organ precursors (Lu et al., 1999; Usui et al., served decrease in overall expression level validates1999). The dystroglycan 1 gene encodes a transmem- previous studies showing that Pou5f1 is ubiquitouslybrane receptor that provides a structural link between expressed in cleavage-stage embryos but becomes re-extracellular matrix components and cytoskeletal pro- stricted to the ICM at the blastocyst stage (Nichols etteins, and is required for the microtubule organizing al., 1998).center to move to the posterior of the early Drosophila Like other POU family transcription factors, the activ-oocyte, a critical step in the establishment of AP polarity ity of Oct4 is regulated by cofactors. The expression ofin the oocyte and the future embryo (Deng et al., 2003). defined and putative Oct4 targets and cofactors hasImmunostaining of the mouse GV-stage oocyte con- been previously investigated in blastocyst-stage em-firmed the expression of DAG1 protein (Figure 3C). Muta- bryos and/or cultured cells, but knowledge of the devel-

opmental regulation of their expression is fragmentary.tion in Drosophila cornichon (cni) prevents the formation

Developmental Cell138

Figure 3. Maternal Transcription of Genes Implicated in Setting Up Polarity

(A) A 57-gene cluster of maternal transcripts, represented by 62 probe sets, that are most abundant during oocyte maturation. Relativeexpression levels are in color: blue indicates low levels and red indicates high levels.The average trend of expression of the cluster is shown in (B). Consistent with the oocyte’s unique need to regulate a discontinuous meiosis,several transcripts encode proteins with roles in the resumption of cell cycle (highlighted in blue). Highlighted in red are the murine homologsof three Drosophila genes implicated in establishing cellular polarity: flamingo, dystroglycan 1, and cornichon. The complete list of the geneswithin this cluster and their expression can be found in Supplemental Table S4.(C) Immunofluorescent staining of ovarian sections confirming the expression of the DAG protein in GV stage oocytes. The oocyte membraneis prominently stained. Cumulus cells also express DAG. The inserts shows phase contrast picture.

Our data provide an opportunity to systematically exam- interaction with Oct4 in phage display assays (Butteroniet al., 2000). Tera, a gene expressed in pluripotent tera-ine and compare the temporal expression patterns of

these genes in vivo. The best characterized Oct4 cofac- tocarcinoma cells and a putative Oct4 target identifiedalong with Upp (Saijoh et al., 1996), is coexpressed withtor is Sox2, a high mobility group (HMG) box-containing

protein required to maintain pluripotency in the epiblast HMG2 in vivo (Figure 4C). This makes HMG2 a candidateas the Oct4 cofactor in the regulation of Tera.(Botquin et al., 1998; Niwa et al., 2000; Avilion et al.,

2003). The role of Sox2 in the Oct4-mediated regulation Taken together, the genomic analyses link four Oct4target genes with two distinct cofactors, based on strik-of Opn and Utf1 has been demonstrated in vitro using

isolated enhancer elements (Botquin et al., 1998; Nishi- ing correlations between the temporal expression pat-terns of the cofactors and the target genes. The putativemoto et al., 1999). Sox2 is strongly induced during the

fifth cleavage along with two Oct4 target genes, Opn target gene Upp is induced along with the Oct4 cofactorSox2 and two other Oct4 targets, and another putativeand Utf1 (Figure 4B). A third gene, Upp, has been postu-

lated to be an Oct4 target based on evidence from cul- target gene Tera is coexpressed with the putative co-factor HMG2.tured cells and embryos (Saijoh et al., 1996; Niwa et al.,

2000). The array data show close temporal correlation The results suggest that there are two groups of Oct4target genes, and each group is induced at a distinct time.between the expression of Upp and that of defined Oct4

target genes (Figure 4B), supporting this hypothesis and HMG2 and its target gene(s) are expressed throughoutpreimplantation development. Genes under the regula-suggesting that Sox2 is a cofactor in Upp regulation.

Another HMG box-containing protein, HMG2, has tion of Sox2 are induced at the time of lineage different-iation, when Oct4 expression and pluripotency bothbeen postulated to be an Oct4 cofactor based on its

A Genomic Study of Preimplantation Development139

If multiple components of a pathway are expressedand if the expression is developmentally regulated, thereis a high chance that the pathway is active, as has beenthe case in numerous previous studies. For the Wntpathway (Figures 5A and 5B and Supplemental TableS6) we detected transcripts encoding ligands (Wnt1,Wnt3, Wnt3a, and Wnt7b), receptors (Fz2 and Fz4), intra-cellular signal transducers and modifiers (Dsh, APC,Axin), and nuclear effectors such as homologs of Dro-sophila arm, Tcf, and groucho. Expressed N pathwaygenes include homologs of Drosophila N, Delta, deltex,fringe, serrate, Enhancer of split, and presenilin (Figures5C and 5D and Supplemental Table S6). For the BMPpathway, we found multiple BMPs, multiple MADs, aswell as homologs of the receptors and signal transduc-tion components saxophone, punt, pointed, and thick-vein (Figures 5E and 5F and Supplemental Table S6). Incontrast, we detected little or no expression of Hedge-hog pathway components.

Some signal transduction genes are expressed atmore or less stable levels throughout preimplantationdevelopment, such as Apc2, Notch2, and SMAD6 (Sup-plemental Table S6). Many others exhibit stage-specificinduction or inhibition. Most of the developmentally reg-ulated genes fall into the two periods revealed alreadyby the clustering analyses: one early, in the oocyte andup to the 2-cell stage (Figures 5A, 5C, and 5E), and theother late, in the morula and the blastocyst (Figures 5B,5D, and 5F). For example, Deltex2 and Smad1 RNAs aremost abundant before the first cleavage (Figure 5C),while Wnt3A is induced as the blastocyst forms (Figure5B). Thus it appears that these signaling pathways areparticularly active (1) during the transition from the oo-cyte to the embryo, when spatial cues first occur andthe orientation of cleavage has an impact on future blas-tocyst polarity (Piotrowska and Zernicka-Goetz, 2001;Gardner, 2001; Piotrowska et al., 2001; Plusa et al.,2002a, 2002b; Zernicka-Goetz, 2002; Fujimori et al.,2003), and (2) just before implantation, when the blasto-cyst is organized in a way that its axial information ap-

Figure 4. Comparison of the Temporal Expression Patterns of Oct4pears to provide proximal-distal polarity information inCofactors and Targets Reveals Correlation between Distinct Cofac-postimplantation embryos (Weber et al., 1999).tors and Targets

Each of the signaling pathways we examined is active(A) The temporal expression pattern of Pou5f1, which encodes thein many developmental contexts, and the pathways of-transcription factor Oct4.ten work in partnership with each other. For example,(B) Coregulation of Oct4 cofactor Sox2 with targets Opn and Utf1

and with putative target Upp. Wnt and N signals act together to regulate different(C) Coregulation of putative Oct4 cofactor HMG2 with putative tar- processes, such as vertebrate somite segmentationget Tera. where Wnt acts upstream of Notch to regulate the rhyth-

mic expression of target genes (Pourquie, 2003). In Dro-become restricted in the ICM lineage. Thus it seems sophila sensory organ precursors, Wnt and N signalsthat the two cofactors and their respective target genes act in sequential fashion: Wnt signals orient the firstdefine two branches downstream of Oct4. division of the sensory organ precursor (SOP) along the

anterior-posterior axis (Gho and Schweisguth, 1998).The Mouse Oocyte and Preimplantation Embryo Subsequently, lateral inhibition signaling by N consoli-Express Multiple Components of Several dates molecular differences between the two daughterDevelopmental Signaling Pathways cells so that one differentiates into hair and socket, theTo search for regulators that may be involved in develop- other into sheath and neuron. The lateral inhibition isment of spatial patterning in the preimplantation em- dependent on attenuation of N signaling in one of thebryo, we examined the production of the components daughter cells by the asymmetric distribution of Numbof several signaling pathways. All are evolutionarily con- and �-Adaptin (Berdnik et al., 2002). Similar sequentialserved regulators of pattern formation and polarity, in- roles of Wnt and N occur during Drosophila musclecluding those transducing Wnt, Notch (N), and BMP progenitor cell development (Brennan et al., 1999).

In the mouse zygote, as in the Drosophila SOP andsignals (Figure 5 and Supplemental Table S6).

A Genomic Study of Preimplantation Development141

Figure 6. The Wnt Signaling Pathway Is Active at the Blastocyst Stage

RNA encoding a �-catenin-GFP fusion protein was injected into one blastomere at the 16-cell stage; the subcellular localization of �-catenin-GFP was examined at the blastocyst stage in progenies of the injected blastomere.(A–C) The fusion protein is localized in the nuclei of blastomeres in the ICM. (A) Phase contrast image showing the position of the ICM. Thepositions of two cells expressing �-catenin-GFP are marked by arrows. (B) Fluroscent image showing the expression and subcellular localizationof �-catenin-GFP. (C) Merged image of (A) and (B).(D–F) The fusion protein is localized in the cytoplasm and membrane of blastomeres in the TE. (D) Phase contrast image showing the positionof the TE. A blastomere expressing �-catenin-GFP is marked by an asterisk. (E) Fluroscent image showing the expression and subcellularlocalization of �-catenin-GFP. In addition to the marked cell, several more TE cells also express �-catenin-GFP but none of the cells exhibitnuclear localization of the fusion protein. (F) Merged image of (D) and (E).

muscle progenitor cells, orientation of the cleavage determined, studies have shown that signals such asWnt and BMP in early postimplantation embryos playplane is regulated and the resulting daughter cells tend

to adopt distinguishable fates (Gardner, 2001; Piotrow- essential roles in body axis formation (Lu et al., 2001).For example, the loss of Wnt3 disrupts gene expressionska et al., 2001; Plusa et al., 2002a). In addition to pro-

ducing many components of Wnt and N pathways (Fig- in the ICM derivatives at the egg cylinder stage, pre-venting the expression of posterior markers (Liu et al.,ures 5A and 5C), the mouse oocyte and zygote express

Numb and several homologs of �-Adaptin (Figure 5C, 1999). To examine the potential impact of Wnt signalingin the blastocyst, we have expressed a �-catenin-GFPSupplemental Table S6); both proteins are important

for N-mediated lateral inhibition in the Drosophila SOP fusion protein in cleavage stage mouse embryos. Theseexperiments have shown that nuclear localization of(Berdnik et al., 2002). These observations are consistent

with a model in which sequential actions of Wnt and N �-catenin, which indicates activation of Wnt target genetranscription (Behrens et al., 1996), appears to be re-set up the initial asymmetry in the mouse zygote and

generate the bias in cell fate that later influences polarity stricted to cells of the ICM (Figures 6A–6C) whereas inthe TE the protein remained concentrated at the cellof the developing embryo.

Many components of the examined signaling path- surface (Figures 6D–6F). This could suggest not onlythat Wnt signaling may be indeed initiated before im-ways are induced at the blastocyst stage, thus shortly

before implantation (Figures 5B, 5D, and 5F). While the plantation, but also that this could occur first in the ICMlineage of cells. Thus, the patterns of Wnt activity seenfunction of these genes in the blastocyst remains to be

Figure 5. Multiple Components of the Wnt, N, and BMP Signaling Pathways Are Expressed in a Developmentally Regulated Fashion in theMouse Oocyte and Preimplantation Embryo

Components that are inhibited after the first cleavage are shown in (A), (C), and (E); those that are induced in the morula and blastocyst stagesare shown in (B), (D), and (F). The Drosophila homologs of some components are indicated in parentheses.(A and B) Expression of selected Wnt pathway components. (C and D) Expression of selected N pathway components. Numb, a gene importantfor asymmetric attenuation of N signals during Drosophila SOP development, is expressed in the mouse oocyte and zygote (green curve in[C]). (E and F) Expression of selected BMP pathway components.

Developmental Cell142

amplification is linear have been reported previously (Baugh et al.,after implantation may arise from signaling that is initi-2001). For each replicate, 10 �g aRNA was fragmented and hybrid-ated during preimplantation stages. An intriguing ques-ized to Murine Genome U74Av2 GeneChips following Affymetrixtion is whether the activation of the Wnt pathway ininstructions. The entire data set has been submitted to the public

selected cells in the blastocyst influences the signaling database ArrayExpress (http://www.ebi.ac.uk/arrayexpress).events necessary for the postimplantation developmentof embryo polarity. If so, this may lead us in future toward

Data Analysisa molecular mechanism behind the observation thatAnalyses of the raw data were performed with dChip, a model-basedsome aspect of axial polarity of the early postimplanta-method for expression analysis, as described previously (Wang et

tion embryo can be traced back to that in the blastocyst. al., 2002). Data were normalized using the invariant set method ofIn a developmental system, the progressive change/ the dChip software. The calculated expression levels reflect the

restriction in cell fate and identity is likely accompanied relative amounts of RNA species. In addition to using P call % toassess the change in the complexity of the embryo’s transcriptomeby progressive changes in the transcription patterns(Figure 1C), we also examined the number of genes whose calcu-in cells and tissues. Although microarray-based genelated expression level is above 1000 or 5000. The trend is similarexpression studies have been successfully applied toto that in Figure 1C when either of the threshold was used (data

study developmental regulation of gene expression in not shown).C. elegans (Jiang et al., 2001; Baugh et al., 2003) and To classify the different stages (Figure 1D), unsupervised hierar-Drosophila (White et al., 1999; Furlong et al., 2001; Ar- chical clustering was performed. The filtered gene list was generated

by setting variation across samples (after pooling replicate arrays)beitman et al., 2002), this had not been done in theat 0.5 � standard deviation/mean � 10, P call % was set at � 8% (1mouse. Our use of microarrays has generated gene ex-out of 12 stages). The distance metric (1�correlation), and averagepression data that are both comprehensive and quanti-linkage method were used in the calculation.

tative. The quality of the data set was tested in multiple Gene-filtering criteria to generate the cluster shown in Figure 3Aways including the consistency between replicates, the were as follows: (1) calculated expression level is above 900 in atglobal patterns of gene activity, and the behavior of least one time point; (2) only annotated genes are included; (3) the

maximum expression level prior to genome activation [max (GV, MII,individual known genes. Sample exploration of the datazygote)] is at least twice the expression level in late 2-cell stage;has permitted the postulation of plausible hypothesesand finally, after hierarchical clustering of genes meeting the abovethat was not previously possible. Analysis of the Oct4three criteria, (4) the subcluster containing tPA and mos was chosen

transcription network identified two putative branches to be examined in more detail.that have distinct cofactors, targets, and temporal pat- On the Affymetrix GeneChip that we used, genes are representedterns. Analyses of several conserved signaling pathways by probe sets. Many genes are represented once, but some are

represented several times by different probe sets. In addition, manysuggested that Wnt, BMP, and N signals may be activeprobe sets represent EST clones. Different EST clones may corre-at critical times to regulate axial polarity developmentspond to the same gene. Thus there is no straightforward way tobefore implantation. The dozens of genes that we pres-convert the number of probe sets to that of genes. Although it would

ent in these examples are only a small percentage of be most accurate to use the word “probe set” to describe our globalthe �12,000 genes represented on the arrays. The rich- observations, for simplicity we use “gene” instead.ness of the data awaits further exploration in the futureas more genes are analyzed from different perspectives.

�-Catenin Localization in the BlastocystWe hope that the data will serve as a valuable resourceSynthetic RNA were transcribed in vitro from a construct encodingto the community of scientists endeavoring to under-a fusion protein of Xenopus �-catenin and MmGFP (Zernicka-Goetz

stand the earliest stages of mammalian life. et al., 1997) (Xlb-catenin-MmGFPLT). Approximately 12 pl at con-centration 1�g/1�l in water was microinjected into one blastomere

Experimental Procedures of the 16-cell stage embryo as previously described (Weber et al.,1999). Control embryos were injected with synthetic mRNA for

Embryo Collection MmGFP. Injected embryos were cultured in KSOM medium supple-F1 (C57BL6 � CBA) females were superovulated by injection of 10 mented with amino acids and 4 �g/ml of BSA. Embryos were scoredIU of PMSG, followed by injection of 10 IU of hCG 48–56 hr later. for the expression of GFP approximately 1–2 hr after injection; GFP-Each collection was made from five to six mice. MII oocytes were positive embryos were cultured further until they reached the blasto-collected 14–15 hr after hCG treatment. To collect embryos, females cyst stage. The subcellular localization of �-catenin-GFP was exam-were mated with F1 males after hCG treatment. Collections were ined by confocal microscopy (Biorad).made at 20–24 hr after hCG (zygotes), 31–32 hr (early 2-cell), 46–48hr (late 2-cell), 54–56 hr (4-cell), 68–70 hr (8-cell), 76–78 hr (16-cell),86–88 hr (early blastocysts), 92–94 hr (mid blastocysts), and 100–102 Acknowledgmentshr (late blastocysts). The collected embryos were further selectedand assigned to appropriate samples based on morphology. Early This work was supported by NIH grant to R.W.D., and a Wellcomeblastocysts contain approximately 32 cells and a small blastocoel. Trust project grant and Senior research fellowship to M.Z.-G. M.P.S.Mid-stage blastocysts contain 32–64 cells and a blastocoel that is is an Investigator of the Howard Hughes Medical Institute. K.P. isapproximately the same size as the ICM. Late-stage blastocysts a recipient of Marie Curie fellowship. We thank Randall Moon for(approximately 64 cells) are those with expanded blastocoels. Total �-catenin-GFP construct, Emma Cadwell for help with embryo col-RNA was isolated from whole embryos two to four times for each lections, the Marco Conti lab for help with immunostaining of ovarianstage. Sample collection, RNA preparation, RNA amplification and sections, David Glover, Karin Lykke-Andersen, Lars Steinmetz,labeling, and array hybridization were carried out independently for Adam Deutschbauer, Tavvi Neklesa, and Jed Dean for comments,each replicate except when noted in Table 1. and Wenzhong Xiao for discussions.

Generation of Microarray DataAll amplifications started with �10 ng whole-embryo total RNA. Two Received: November 11, 2003

Revised: December 17, 2003rounds of amplifications were performed for each replicate using amodified aRNA protocol as described previously (Wang et al., 2002). Accepted: December 17, 2003

Published: January 12, 2004The demonstration that the methods are quantitative and that the

A Genomic Study of Preimplantation Development143

References Gardner, R.L. (2001). Specification of embryonic axes begins beforecleavage in normal mouse development. Development 128,839–847.Affymetrix (1999). GeneChip Expression Analysis Technical Manual.

Allard, P., Champigny, M.J., Skoggard, S., Erkman, J.A., Whitfield, Gho, M., and Schweisguth, F. (1998). Frizzled signalling controlsM.L., Marzluff, W.F., and Clarke, H,.J. (2002). Stem-loop binding orientation of asymmetric sense organ precursor cell divisions inprotein accumulates during oocyte maturation and is not cell-cycle- Drosophila. Nature 393, 178–181.regulated in the early mouse embryo. J. Cell Sci. 115, 4577–4586. Gosden, R.G. (2002). Oogenesis as a foundation for embryogenesis.Arbeitman, M.N., Furlong, E.E., Imam, F., Johnson, E., Null, B.H., Mol. Cell. Endocrinol. 186, 149–153.Baker, B.S., Krasnow, M.A., Scott, M.P., Davis, R.W., and White,

Hashimoto, N., Watanabe, N., Furuta, Y., Tamemoto, H., Sagata, N.,K.P. (2002). Gene expression during the life cycle of Drosophila

Yokoyama, M., Okazaki, K., Nagayoshi, M., Takeda, N., Ikawa, Y., etmelanogaster. Science 297, 2270–2275.

al. (1994). Parthenogenetic activation of oocytes in c-mos-deficientAvilion, A.A., Nicolis, S.K., Pevny, L.H., Perez, L., Vivian, N., and mice. Nature 370, 68–71.Lovell-Badge, R. (2003). Multipotent cell lineages in early mouse

Hoshino, S., Miyazawa, H., Enomoto, T., Hanaoka, F., Kikuchi, Y.,development depend on SOX2 function. Genes Dev. 17, 126–140.Kikuchi, A., and Ui, M. (1989). A human homologue of the yeast

Bachvarova, R. (1985). Gene expression during oogenesis and oo- GST1 gene codes for a GTP-binding protein and is expressed in acyte development in mammals. Dev. Biol. 1, 453–524. proliferation-dependent manner in mammalian cells. EMBO J. 8,Bachvarova, R.F. (1992). A maternal tail of poly(A): the long and the 3807–3814.short of it. Cell 69, 895–897. Jiang, M., Ryu, J., Kiraly, M., Duke, K., Reinke, V., and Kim, S.K.Baugh, L.R., Hill, A.A., Brown, E.L., and Hunter, C.P. (2001). Quantita- (2001). Genome-wide analysis of developmental and sex-regulatedtive analysis of mRNA amplification by in vitro transcription. Nucleic gene expression profiles in Caenorhabditis elegans. Proc. Natl.Acids Res. 29, E29. Acad. Sci. USA 98, 218–223.Baugh, L.R., Hill, A.A., Slonim, D.K., Brown, E.L., and Hunter, C.P. Ko, M.S., Kitchen, J.R., Wang, X., Threat, T.A., Wang, X., Hasegawa,(2003). Composition and dynamics of the Caenorhabditis elegans A., Sun, T., Grahovac, M.J., Kargul, G.J., Lim, M.K., et al. (2000).early embryonic transcriptome. Development 130, 889–900. Large-scale cDNA analysis reveals phased gene expression pat-

terns during preimplantation mouse development. DevelopmentBehrens, J., von Kries, J.P., Kuhl, M., Bruhn, L., Wedlich, D.,Grosschedl, R., and Birchmeier, W. (1996). Functional interaction of 127, 1737–1749.beta-catenin with the transcription factor LEF-1. Nature 382, Liang, P., Averboukh, L., Zhu, W., Haley, T., and Pardee, A.B. (1995).638–642. Molecular characterization of the murine thymidylate kinase gene.Berdnik, D., Torok, T., Gonzalez-Gaitan, M., and Knoblich, J.A. Cell Growth Differ. 6, 1333–1338.(2002). The endocytic protein �-Adaptin is required for numb-medi- Liu, P., Wakamiya, M., Shea, M.J., Albrecht, U., Behringer, R.R.,ated asymmetric cell division in Drosophila. Dev. Cell 3, 221–231. and Bradley, A. (1999). Requirement for Wnt3 in vertebrate axisBonapace, I.M., Latella, L., Papait, R., Nicassio, F., Sacco, A., Muto, formation. Nat. Genet. 22, 361–365.M., Crescenzi, M., and Di Fiore, P.P. (2002). Np95 is regulated by

Lu, B., Usui, T., Uemura, T., Jan, L., and Jan, Y.N. (1999). FlamingoE1A during mitotic reactivation of terminally differentiated cells and

controls the planar polarity of sensory bristles and asymmetric divi-is essential for S phase entry. J. Cell Biol. 157, 909–914.

sion of sensory organ precursors in Drosophila. Curr. Biol. 9, 1247–Botquin, V., Hess, H., Fuhrmann, G., Anastassiadis, C., Gross, M.K., 1250.Vriend, G., and Scholer, H.R. (1998). New POU dimer configuration

Lu, C.C., Brennan, J., and Robertson, E.J. (2001). From fertilizationmediates antagonistic control of an osteopontin preimplantationto gastrulation: axis formation in the mouse embryo. Curr. Opin.enhancer by Oct-4 and Sox-2. Genes Dev. 12, 2073–2090.Genet. Dev. 11, 384–392.

Brennan, K., Baylies, M., and Arias, A.M. (1999). Repression by NotchNichols, J., Zevnik, B., Anastassiadis, K., Niwa, H., Klewe-Nebenius,is required before Wingless signalling during muscle progenitor cellD., Chambers, I., Scholer, H., and Smith, A. (1998). Formation ofdevelopment in Drosophila. Curr. Biol. 9, 707–710.pluripotent stem cells in the mammalian embryo depends on the

Butteroni, C., De Felici, M., Scholer, H.R., and Pesce, M. (2000). POU transcription factor Oct4. Cell 95, 379–391.Phage display screening reveals an association between germline-

Nishimoto, M., Fukushima, A., Okuda, A., and Muramatsu, M. (1999).specific transcription factor Oct-4 and multiple cellular proteins. J.The gene for the embryonic stem cell coactivator UTF1 carries aMol. Biol. 304, 529–540.regulatory element which selectively interacts with a complex com-

Colledge, W.H., Carlton, M.B., Udy, G.B., and Evans, M.J. (1994).posed of Oct-3/4 and Sox-2. Mol. Cell. Biol. 19, 5453–5465.

Disruption of c-mos causes parthenogenetic development of unfer-Niwa, H., Miyazaki, J., and Smith, A.G. (2000). Quantitative expres-tilized mouse eggs. Nature 370, 65–68.sion of Oct-3/4 defines differentiation, dedifferentiation or self-Deng, W.M., Schneider, M., Frock, R., Castillejo-Lopez, C., Gaman,renewal of ES cells. Nat. Genet. 24, 372–376.E.A., Baumgartner, S., and Ruohola-Baker, H. (2003). DystroglycanNy, T., Wahlberg, P., and Brandstrom, I.J. (2002). Matrix remodelingis required for polarizing the epithelial cells and the oocyte in Dro-in the ovary: regulation and functional role of the plasminogen acti-sophila. Development 130, 173–184.vator and matrix metalloproteinase systems. Mol. Cell. Endocrinol.De Sousa, P.A., Valdimarsson, G., Nicholson, B.J., and Kidder, G.M.187, 29–38.(1993). Connexin trafficking and the control of gap junction assemblyPaynton, B.V., Rempel, R., and Bachvarova, R. (1988). Changes inin mouse preimplantation embryos. Development 117, 1355–1367.state of adenylation and time course of degradation of maternalEpifano, O., Liang, L.F., Familari, M., Moos, M.C., Jr., and Dean, J.mRNAs during oocyte maturation and early embryonic development(1995). Coordinate expression of the three zona pellucida genesin the mouse. Dev. Biol. 129, 304–314.during mouse oogenesis. Development 121, 1947–1956.Pelton, T.A., Sharma, S., Schulz, T.C., Rathjen, J., and Rathjen, P.D.Fleming, T.P., Sheth, B., and Fesenko, I. (2001). Cell adhesion in the(2002). Transient pluripotent cell populations during primitive ecto-preimplantation mammalian embryo and its role in trophectodermderm formation: correlation of in vivo and in vitro pluripotent celldifferentiation and blastocyst morphogenesis. Front. Biosci. 6,development. J. Cell Sci. 115, 329–339.D1000–D1007.

Piotrowska, K., and Zernicka-Goetz, M. (2001). Role for sperm inFujimori, T., Kurotaki, Y., Miyazaki, J.I., and Nabeshima, Y.I. (2003).spatial patterning of the early mouse embryo. Nature 409, 517–521.Analysis of cell lineage in two- and four-cell mouse embryos. Devel-

opment 130, 5113–5122. Piotrowska, K., Wianny, F., Pedersen, R.A., and Zernicka-Goetz, M.(2001). Blastomeres arising from the first cleavage division haveFurlong, E.E., Andersen, E.C., Null, B., White, K.P., and Scott, M.P.

(2001). Patterns of gene expression during Drosophila mesoderm distinguishable fates in normal mouse development. Development128, 3739–3748.development. Science 293, 1629–1633.

Developmental Cell144

Plusa, B., Grabarek, J.B., Piotrowska, K., Glover, D.M., and Zernicka-Goetz, M. (2002a). Site of the previous meiotic division defines cleav-age orientation in the mouse embryo. Nat. Cell Biol. 4, 811–815.

Plusa, B., Piotrowska, K., and Zernicka-Goetz, M. (2002b). Spermentry position provides a surface marker for the first cleavage planeof the mouse zygote. Genesis 32, 193–198.

Pourquie, O. (2003). The segmentation clock: converting embryonictime into spatial pattern. Science 301, 328–330.

Roth, S., Neuman-Silberberg, F.S., Barcelo, G., and Schupbach,T. (1995). cornichon and the EGF receptor signaling process arenecessary for both anterior-posterior and dorsal-ventral pattern for-mation in Drosophila. Cell 81, 967–978.

Russ, A.P., Wattler, S., Colledge, W.H., Aparicio, S.A., Carlton, M.B.,Pearce, J.J., Barton, S.C., Surani, M.A., Ryan, K., Nehls, M.C., et al.(2000). Eomesodermin is required for mouse trophoblast develop-ment and mesoderm formation. Nature 404, 95–99.

Saijoh, Y., Fujii, H., Meno, C., Sato, M., Hirota, Y., Nagamatsu, S.,Ikeda, M., and Hamada, H. (1996). Identification of putative down-stream genes of Oct-3, a pluripotent cell-specific transcription fac-tor. Genes Cells 1, 239–252.

Schultz, R.M. (2002). The molecular foundations of the maternalto zygotic transition in the preimplantation embryo. Hum. Reprod.Update 8, 323–331.

Sherlock, G. (2000). Analysis of large-scale gene expression data.Curr. Opin. Immunol. 12, 2001–2005.

Sheth, B., Fesenko, I., Collins, J.E., Moran, B., Wild, A.E., Anderson,J.M., and Fleming, T.P. (1997) Tight junction assembly during mouseblastocyst formation is regulated by late expression of ZO-1 alpha

isoform. Development 124, 2027–2037.

Tam, P.P., Gad, J.M., Kinder, S.J., Tsang, T.E., and Behringer, R.R.(2001). Morphogenetic tissue movement and the establishment ofbody plan during development from blastocyst to gastrula in themouse. Bioessays 23, 508–517.

Usui, T., Shima, Y., Shimada, Y., Hirano, S., Burgess, R.W., Schwarz,T.L., Takeichi, M., and Uemura, T. (1999). Flamingo, a seven-passtransmembrane cadherin, regulates planar cell polarity under thecontrol of Frizzled. Cell 98, 585–595.

Wang, Q.T., Xiao, W., Mindrinos, M., and Davis, R.W. (2002). YeasttRNA as carrier in the isolation of microscale RNA for global amplifi-cation and expression profiling. Biotechniques 4, 788, 790, 792passim.

Wassarman, P.M., and DePamphilis, M.L. eds. (1993). Guide to Tech-niques in Mouse Development. (San Diego: Academic Press).

Weber, R.J., Pedersen, R.A., Wianny, F., Evans, M.J., and Zernicka-Goetz, M. (1999). Polarity of the mouse embryo is anticipated beforeimplantation. Development 126, 5591–5598.

White, K.P., Rifkin, S.A., Hurban, P., and Hogness, D.S. (1999). Mi-croarray analysis of Drosophila development during metamorpho-sis. Science 286, 2179–2184.

Zeng, F., and Schultz, R.M. (2003). Gene expression in mouse oo-cytes and preimplantation embryos: use of suppression subtractivehybridization to identify oocyte- and embryo-specific genes. Biol.Reprod. 68, 31–39.

Zernicka-Goetz, M. (2002). Patterning of the embryo: the first spatialdecisions in the life of a mouse. Development 129, 815–829.

Zernicka-Goetz, M., Pines, J., McLean Hunter, S., Dixon, J.P., Siem-ering, K.R., Haseloff, J., and Evans, M.J. (1997). Following cell fatein the living mouse embryo. Development 124, 1133–1137.