| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Research Article |
Laboratory for Mammalian Germ Cell Biology,4 Center for Developmental Biology, RIKEN Kobe Institute, Kobe 650-0047, Japan
Precursory Research for Embryonic Science and Technology,5 Japan Science and Technology Agency, Saitama 332-0012, Japan
Laboratory of Molecular Cell Biology and Development,6 Graduate School of Biostudies, Kyoto University, Kyoto 606-8502, Japan
ABSTRACT
Germ cell fate in mice is induced in proximal epiblast cells at Embryonic Day (E) 6.5 by signaling molecules. Prdm1(also known as Blimp1)-positive lineage-restricted precursors of primordial germ cells (PGCs) initiate the formation of a cluster that differentiates into Dppa3 (also known as stella)-positive PGCs from around E7.0 onwards in the extra-embryonic mesoderm. Around E7.5, these PGCs begin migrating towards the definitive endoderm, with concomitant extensive epigenetic reprogramming. To gain a more precise insight into the mechanism of PGC specification and its subsequent development, we exploited quantitative, single-cell, gene expression profiling to explore gene expression dynamics during the 36 h of PGC differentiation from E6.75 to E8.25, in comparison with the corresponding profiles of somatic neighbors. This analysis revealed that the transitions from Prdm1-positive PGC precursors to Dppa3-positive PGCs and to more advanced migrating PGCs involve a highly dynamic, stage-dependent transcriptional orchestration that begins with the regaining of the pluripotency-associated gene network, followed by stepwise activation of PGC-specific genes, differential repression of the somatic mesodermal program, as well as potential modulations of signal transduction capacities and unique control of epigenetic regulators. The information presented here regarding the cascade of events involved in PGC development should serve as a basis for detailed functional analyses of the gene products associated with this process, as well as for appropriate reconstitution of PGCs and their descendant cells in culture.
developmental biology, early development, embryo, gamete biology, gene regulation
INTRODUCTION
Segregation of the germline from somatic lineages is one of the most essential events in development, since this process ensures the acquisition, modification, and reservation of the ‘totipotent' genome for subsequent generations. There are essentially two modes, often referred to as ‘preformation' and ‘epigenesis', respectively, that guarantee this process across the metazoans [1]. In preformation, a maternally inherited localized determinant, the germ plasm, dictates germline specification so that early blastomeres that inherit these RNA-protein granules adopt the germ cell fate. In contrast, in epigenesis, germ cell fate is specified by signaling molecules at a relatively late stage of development in a potentially equivalent population of cells. Many model organisms, including C. elegans, D. melanogaster, X. laevis and D. rerio, specify germ cell fate by preformation, whereas mammals determine germ cell lineage by epigenesis [1, 2].
Classically, in the mouse, primordial germ cells (PGCs), which constitute the first population of the germ cell lineage that appears in the embryo, have been identified as an alkaline phosphatase-positive cluster of approximately 40 cells in the extra-embryonic mesoderm at around Embryonic Day (E) 7.0 [3–5]. Subsequent studies have revealed that signaling molecules of the BMP family from extra-embryonic tissues and their signal transducers are important for the establishment of germ cell lineage [6–13]. However, until relatively recently, the intrinsic molecular mechanisms underlying germ cell specification and the properties of the resultant founder population of PGCs have been poorly defined.
Using single-cell gene expression analysis, Saitou et al. have identified Ifitm3 (also known as fragilis) and Dppa3, which are highly and specifically expressed, respectively, in the founder PGCs. These authors proposed a molecular model for germ cell specification, a key event of which is the repression of Homeobox (Hox) genes in the founder PGCs at E7.5 [14]. More recently, Prdm1, which is a potent transcriptional repressor of a histone methyltransferase subfamily, was found to mark the origin of the germ cell lineage in the most proximal layer of the epiblast as early as E6.25E6.5 prior to the onset of gastrulation [15, 16]. These Prdm1-positive cells initiate to form a cluster within the extra-embryonic mesoderm at around E6.75E7.0, to differentiate into PGCs with characteristic alkaline phosphatase activity, expression of Dppa3, repressed expression of Hox genes. Importantly, genetic lineage-tracing experiments have shown that Prdm1-expressing cells contribute only to Dppa3-positive PGCs, which indicates that Prdm1-positive cells in early gastrulating embryos are most likely lineage-restricted to PGCs [15, 16]. Subsequently, Dppa3-positive PGCs start to migrate towards the definitive hindgut endoderm, with concomitant genome-wide epigenetic reprogramming, which includes substantial demethylation of DNA and Histone3/Lysine9 dimethylation (H3K9me2) at around E8.0, and the upregulation of H3K27me3 at around E8.75 [17].
Despite these essential advances, there remains much to be learned about the precise mechanisms of germ cell specification and the resultant characteristics of established PGCs. The clarification of both of these still poses a formidable challenge, mainly because of technical difficulties associated with the analysis of a small number of cells that almost constantly change their properties cell-by-cell. For a precise understanding and subsequent reconstruction of germ cell specification in vitro, an essential first step is to develop a technique that enables quantitative analysis of this in vivo process.
In the present study, we exploited a modified quantitative single-cell cDNA amplification method followed by real-time quantitative PCR (Q-PCR) analysis to better understand the gene expression dynamics involved in germ cell specification. We examined the expression of multiple genes with distinct functional categories at four different time points during the 36 hours of germ cell differentiation from E6.75 to E8.25 and compared it with the expression in somatic neighbors. Our results reveal that germ cell specification involves a dynamic, stage-dependent transcriptional control that begins with the regaining of the pluripotency-associated gene network, followed by stepwise acquisition of PGC-specific genes, differential repression of the somatic mesodermal program, and a specific modulation of signal transduction capacities and epigenetic regulators. These findings presumably reflect the complex network of signaling molecules necessary for the proper setting of this potentially immortal cell lineage, and should serve as a basis for detailed functional analyses of the gene products associated with this process, as well as for properly reconstituting PGCs and their descendants in culture.
MATERIALS AND METHODS
Isolation of Total RNA from Embryonic Stem (ES) Cells and cDNA Synthesis
E14tg2a ES cells were cultured as described [18]. Total RNA from the ES cells was isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany). For the synthesis of control cDNAs, 1 µg of total RNA was reverse-transcribed using Superscript III (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions.
Single-Cell-Level cDNA Amplification
Amplification of cDNAs from single-cell RNA or single-cell-level RNA was performed as described with modifications [19–21]. The ES cell total RNA (1000 ng/µl) was serially diluted to concentrations of 2.5 ng/µl, 250 pg/µl, and 25 pg/µl. Then, 0.4 µl (10 pg, corresponding to the typical amount of mammalian single-cell-level total RNA) of the final dilution (25 pg/µl) was added to 4.75 µl of single-cell lysis buffer (1x PCR buffer II [Applied Biosystems, Foster City, CA], 1.5 mM MgCl2 [Applied Biosystems], 0.5% NP40, 5 mM DTT, 0.3 U/µl Prime RNase Inhibitor [Eppendorf, Hamburg, Germany], 0.3 U/µl RNAguard RNase Inhibitor [GE Healthcare], 0.95 ng of the cDNA synthesis primer and 0.05 mM each of dATP, dCTP, dGTP, and dTTP [GE Healthcare]). The cDNA synthesis primer sequence was: 5'-TATAGAATTCGCGGCCGCTCGCGA(T)24-3'. All the primers described in this manuscript were purchased from Operon Biotechnology (Huntsville, AL) or Hokkaido System Science (Sapporo, Japan). The reaction mixture was heated at 70°C for 90 s and was immediately placed on ice for 1 min. After brief centrifugation, a 0.3-µl volume of RT mixture (133.3 U/µl SuperScript III [Invitrogen], 3.33 U/µl RNAguard RNase Inhibitor [GE Healthcare], and 1.1–1.3 µg/µl T4 gene 32 protein [Roche, Basel, Switzerland]) was added to each reaction tube. The reaction mixture was incubated at 50°C for 5 min and heat-inactivated at 70°C for 10 min. The tubes were immediately put on ice for 1 min, and after 15 s of centrifugation, a poly(A) tailing reaction was performed by adding 5 µl of TdT reaction buffer (1x PCR buffer II, 1.5 mM MgCl2, 3 mM dATP, 0.1 U/µl RNase H [Invitrogen], and 0.75 U/µl terminal deoxynucleotidyl transferase [Invitrogen]) and incubating at 37°C for 15 min. The reaction was inactivated by incubation at 70°C for 10 min. Then, 50 µl of the PCR mixture (1x ExTaq buffer [Mg2+ included], 0.25 mM each of dATP, dCTP, dGTP, and dTTP, 1.2 µg cDNA synthesis primer, and 0.05 U/µl ExTaq Hot Start Version [Takara]) was added, and mineral oil was overlaid. The second strand was synthesized in a thermal cycler (Applied Biosystems) according to the following protocol: 94°C for 3 min, 50°C for 2 min, and 72°C for 3 min. The PCR reaction was immediately followed by the protocol of: 94°C for 45 s, 67°C for 1 min, and 72°C for 3 min, with a 6-s extension per cycle for 24 cycles. The amplified cDNA was purified using the QIAquick PCR Purification Kit (Qiagen) and dissolved in 50 µl of EB buffer (10 mM TrisCl [pH 8.5]).
Q-PCR and Estimation of Gene Copy Number Per Cell
Q-PCR was performed using the 7900 Real-Time PCR System (Applied Biosystems) and SYBR Green PCR Master Mix (Applied Biosystems) according to the manufacturer's instructions. The sequences of the primers for the genes examined are listed in Supplemental Table S1 (available online at http://www.biolreprod.org). The doubling efficiencies of all the primers were measured using as templates pGEM-T-based (Promega, Madison, WI) plasmids that carried corresponding specific inserts. All the amplified samples were diluted so that the threshold cycle (Ct) values of the genes examined were within the range of 11 to 30, within which the Ct values of the standard plasmids were accurately correlated with their concentrations. All the experiments were carried out twice, and the average of the Ct values was then converted into approximate copy numbers according to the following formula:
where exi indicates the log10 of transcripts per sample i, Sx and Ix are respectively the slope and intercept estimated from the plot of copy numbers against the Ct values of each plasmid standard (Supplemental Table S1), and cxi is the Ct value from Q-PCR.
Single-Cell Isolation and Global cDNA Synthesis
All the animals were treated with appropriate care according to the RIKEN ethics guidelines. Embryos of the CD-1 background at E6.75, E7.25, E7.75, and E8.25 were collected in Dulbecco Modified Eagle medium (DMEM) (Gibco, Gaithersburg, MD) that was supplemented with 0.5% BSA. Embryonic fragments from the posterior extra-embryonic mesoderm (E6.75), the base of the allantois (E7.25-E7.75), and the hindgut endoderm (E8.25) were dissected out using a glass needle and incubated in 0.05% trypsin and 0.5 mM EDTA for 7 min, followed by dissociation into single cells by trituration with a mouth pipette. Dissociated single cells were randomly picked up and placed directly into a tube with 4.75 µl of the single-cell lysis buffer, as described above. The entire process was performed as quickly as possible in order to minimize the effect of trypsin/EDTA treatment on gene expression.
The single-cell cDNA synthesis was performed exactly as described above for single-cell-level cDNA amplification. The amplified samples were screened to identify the germline cells using gene-specific PCR for Prdm1, Ifitm3, and Pou5f1 (also known as Oct4) (E6.75) or Prdm1 and Pou5f1 (E7.25-E8.25). Six cells from each stage were selected and further amplified by PCR for the mass Q-PCR analysis. The amplification was carried out in buffer that contained 1x ExTaq buffer, 0.2 mM dNTP, 1 ng/µl cDNA synthesis primer, and 0.1 U/µl ExTaq polymerase with 1/20 volume of template and using the following protocol: five cycles of 95°C for 1 min, 67°C for 2 min, and 72°C for 6 min. The amplified cDNA was purified using the QIAquick PCR Purification Kit (Qiagen) and dissolved in 50 µl of EB buffer.
Normalization of Gene Expression Measurements
The log10-expression levels (exi) of genes in single germline or somatic cells measured by Q-PCR were normalized to minimize the technical variations in each experiment, as follows:
where Ti is the normalization parameter for sample i, and exi and Exi are the log10-expression levels of gene x before and after normalization, respectively. For normalization standards, we used three housekeeping genes, Gapdh, Arbp, and Ppia, the expression levels of which were assumed to vary little among cell types. The normalization parameter Ti was calculated so that the variance of the normalized expression levels of the standard genes was minimized using the least-squares method:
where N is the number of samples (42; 6 independently amplified samples at each of four and three developmental stages of the germline and somatic cells, respectively), and M is the number of genes used for normalization (three).
Since Exi indicates the log10-expression levels after 24 cycles and/or five additional cycles of PCR amplification, Exi was transformed into the log10 of copy numbers per sample Cxi using one of the two following formulae:
where D24 and D5 are the log10-amplification coefficients for 24 cycles or five additional cycles of PCR, respectively. D24 was calculated by comparison of the expression levels of the housekeeping genes Gapdh, Arbp, and Ppia in nonamplified cDNA sample (1µg of ES cell total RNA) with those in samples amplified for 24 cycles from a single-cell-level RNA (10 pg). Under our conditions, D24 was calculated to be 4.97 ± 0.13. D5, which was calculated by comparison of the expression levels of the same three housekeeping genes before and after 5 cycles of PCR amplification of ES cell samples amplified for 24 cycles, was calculated as 1.40 ± 0.06.
Evaluation of the Statistical Significance of the Single-Cell Gene Expression Differences
The significance of the difference between the estimated gene expression levels at a certain stage of the germ cell and somatic cell development was evaluated by analysis of variance (ANOVA) for the seven groups analyzed (E6.75, E7.25, E7.75, and E8.25 germline cells, and E6.75, E7.25, and E7.75 somatic cells). For the genes that showed a statistically significant difference in expression levels (P < 0.05 by ANOVA) at least in one group, the difference in the expression levels between the germline and somatic cells at each of the corresponding stages was further examined by the Student t-test.
Heat Map
The Microsoft Excel Visual Basic macro was utilized to make a heat map with a gradient of blue, green, yellow, and red colors, according to the logarithm of the expression levels. The genes were grouped according to the results of ANOVA (P < 0.05) and the Student t-test (P < 0.05) (Supplemental Table S2).
Hierarchical Clustering Analysis
Unsupervised hierarchical clustering of the Q-PCR data obtained from all 42 single cells examined was performed using the GenePattern software (http://www.broad.mit.edu/cancer/software/genepattern/) with the default setting. The analysis was conducted for the 57 genes that were significantly detected in at least one cell type (i.e., categories A-D of Fig. 4; see Results).
|
RESULTS
Simple Semi-Quantitative Single-Cell cDNA Amplification Method
We developed a simple, semi-quantitative, single-cell cDNA amplification method based on the original PCR-based protocols described previously (see Materials and Methods) [19–21]. We isolated total RNA from ES cells (2.5 x 106) and diluted it to the single-cell level (10 pg) for use as templates for amplification. Using Q-PCR, we evaluated the relative abundances of the transcripts in amplified samples. As shown in Figure 1, A and B, in our amplification protocol, five genes that were expressed at widely different levels in ES cells (Gapdh, Pou5f1, Sox2, Nanog, and Dppa5, which is also known as Esg1) were amplified exponentially and relatively proportionally at least until the 24th cycle of PCR amplification, indicating a powerful and proportional amplification of a small amount of synthesized cDNA in our amplification procedure after fewer than 24 cycles. We then evaluated the representation and reproducibility of amplification from the single-cell level cDNA at the 24th cycle by comparing the relative abundances of as many as 21 genes [Ppia, Arbp, Gapdh, Dppa5, Zfp42 (also known as Rex1), Pou5f1, Sox2, Ifitm3, Eed, Nanog, Dnmt3b, Yy1, Dnmt1, Espl1, Ezh2, Ehmt2 (also known as G9a), Eras, Ehmt1 (also known as Glp), Nodal, Dppa3, and Tial1] in amplified cDNAs from nine independent experiments and in nonamplified cDNAs synthesized directly from undiluted (1 µg) total RNA. As shown in Figure 1, C and D, most of the 21 genes, especially the ones expressed at more than 100 copies per cell, were represented correctly and with high reproducibility in the amplified products. The majority of the genes were plotted between 4-fold difference lines compared with the nonamplified control in each experiment (for all expression level ranges, R2 = 0.884, with 62% and 84% of the genes plotted between the 2.0-fold and 4.0-fold difference lines, respectively; and for the genes with more than 100 copies per cell, R2 = 0.883, with 81% and 96% of the genes plotted between the 2.0-fold and 4.0-fold difference lines, respectively). Although it should be noted that some genes showed relatively poor amplification, which might be due to their specific 3'-end sequences, these results collectively indicate that our amplification procedure detects most if not all of the genes expressed at the single-cell level in representative and reproducible manners.
|
Population Analysis of Prdm1-Positive Cells during the Specification of Germ Cell Fate
Using the simple, quantitative, single-cell cDNA amplification method, we set out to explore precisely the expression dynamics of key genes associated with the specification of germ cell fate in mice. We dissected out the smallest possible regions from embryos of CD1 background, in which PGC precursors (defined as Prdm1-positive and Dppa3-negative cells) or PGCs (defined as Prdm1-positive and Dppa3-positive cells) were most likely localized, at the following stages: E6.75 [early/mid-streak (E/MS) stage], E7.25 [not/early bud (0/EB) stage], E7.75 [early head fold (EHF) stage], and E8.25 (810 somite stage) (Fig. 2A). For embryos from E6.75 to E7.75, we removed the visceral endoderm from the fragments, and for embryos at E7.25 and E7.75, we removed the extending allantois as much as possible. We then dissociated them into single cells, which we picked up randomly to synthesize single-cell cDNAs. The number of single cells isolated at each stage is summarized in Figure 2B. All the amplified samples were screened by gene-specific primers for Prdm1 [15], Ifitm3 [14, 22, 23], and Dppa3 [14, 24] for E6.75 embryos, and for Prdm1 and Dppa3 for the embryos from other stages, so as to identify PGC precursors (E6.75) and PGCs (E7.25, E7.75, and E8.25), respectively. To examine the precise character of Prdm1-positive cells during the course of PGC differentiation, all the Prdm1-positive cells from all stages were also examined for the expression of Pou5f1 [25] and Sox2 [26], which are master regulators for pluripotency [27], and Hoxb1, which is a key gene that is repressed in founder PGCs [14], and the approximate copy number of each gene expressed in single cells was estimated (see Materials and Methods, Figs. 3 and 4, and data not shown). Almost all the Prdm1-positive cells we analyzed were Pou5f1-positive from E6.75 to E7.75 (86/89, 97%), whereas 8/30 Prdm1-positive cells were negative for Pou5f1 at E8.25. The identities of these Prdm1-positive, Pou5f1-negative cells at E8.25 remain to be determined.
|
|
Figure 2, B-D summarizes the transitions of the proportions of the sub-populations of Prdm1-positive and Pou5f1-positive cells at the four developmental time-points. These data indicate potentially important events for understanding the germ cell specification process. First, Hoxb1 mRNA is present initially in at least some PGC precursors. At E6.75, as many as 9/26 (35%) of the Prdm1-positive population were Hoxb1-positive (sum of the blue and sky blue circles in Fig. 2C), and at E7.25, 12/40 (30%) of the Prdm1-positive population were Hoxb1-positive. The percentage of Hoxb1-positive cells declined after E7.75 in the populations we analyzed, although approximately 10% of the Prdm1-positive cells still expressed Hoxb1. Second, almost all (57/61, 93%) of the cells with Dppa3 expression lacked Hoxb1 expression, irrespective of the original embryonic stage (Fig. 2, B and C, and data not shown). Third, the expression of Sox2 was detected only in a subpopulation of Prdm1-positive cells at E6.75; only 7/26 (27%) of the Prdm1-positive cells at this stage expressed Sox2 at more than 10 copies per cell (data not shown) (54% in all Sox2 expression level ranges). The proportion of Sox2-positive cells increased gradually during the course of PGC specification, and most (20/22, 91%) of the cells we analyzed had become Sox2-positive at E8.25 (Fig. 2, B and D). Interestingly, similar to the case with the mutually exclusive expression of Dppa3 and Hoxb1, almost all the cells that were strongly positive for Sox2 were Hoxb1-negative (data not shown). Fourth, Sox2 expression in the Prdm1-positive populations preceded Dppa3 expression in the same populations (Fig. 2D). Collectively, these data suggest that Prdm1-positive, lineage-restricted PGC precursors [15], at least some of which are initially positive for Hoxb1 and may have gene expression properties similar to those of neighboring somatic cells, upregulate Sox2 followed by Dppa3, to acquire the PGC property (see the Discussion). The established PGCs did not show Hoxb1 expression.
Gene Expression Dynamics Associated with PGC Specification
To investigate more extensively how the PGCs acquire their characteristic properties, we analyzed the expression of various functional categories of genes during the specification of PGC fate. Six amplified cells at each of four stages were selected for these analyses: two Prdm1-positive and Hoxb1-positive and four Prdm1-positive and Hoxb1-negative cells at E6.75, and six Prdm1-positive and Dppa3-positive cells for all the other stages. At the same time, the gene expression of neighboring somatic cells (Prdm1-negative and Hoxb1-positive cells), which probably share a common origin with PGCs at E6.75, E7.25, and E7.75, was analyzed for comparison. We did not include somatic cells at E8.25 because many of the PGCs at E8.25 migrate in the hindgut [17], and neighboring somatic cells at this stage are no longer the same population as they are at E6.75-E7.75. The 42 selected cells (six PGCs at each of four stages, six somatic cells at each of three stages) were amplified quantitatively for five additional PCR cycles (see Materials and Methods), and the expression levels of multiple genes, including the three housekeeping genes Gapdh, Ppia, and Arbp, were analyzed by Q-PCR (Fig. 3).
Housekeeping Genes
Initially, we examined the expression profiles of the housekeeping genes. The mRNAs for Gapdh, Arbp, and Ppia were estimated to be present relatively constantly in the order of approximately 103 to 104 molecules per cell in all the single-cell cDNAs examined. This agreed with the previous report on the transcript number estimation of housekeeping genes within a single cell [28], although the cells used were different, demonstrating further the quantitative amplification achievable with our method. Therefore, we used the expression levels of these three genes to normalize all the gene expression data in this manuscript (see Materials and Methods).
Key Molecules Associated with Germline Specification
We also quantified the expression of four PGC markers, Ifitm3, Dppa3, Prdm1, and Akp2 (also known as tissue non-specific alkaline phosphatase) [29]. Consistent with previous observations [14, 22], Ifitm3 was detected to a relatively weak extent in somatic cells but was expressed much more strongly (from 102 to >103 copies per cell) in the PGC precursors and PGCs. Although Dppa3 expression was detected only in a small population (2/26, 7.7%) of cells at E6.75 (Fig. 2B), which we consider to be the first few cells that acquire PGC characteristics, this gene was dramatically upregulated at E7.25 (around 103 copies per cell at E8.25), whereas no expression was observed in Prdm1-negative somatic neighbors. Prdm1 was expressed by the PGC precursors (E6.75) and the PGCs in the range of 101–102 copies per cell using this method. Akp2, which is a classical marker for PGCs, was detected not only in the PGC precursors and PGCs, but also in somatic neighbors, with PGCs having the highest expression (5 x 102 copies per cell). These results are highly consistent with previous in situ hybridization and transgenic analyses [14, 15, 22, 29], which supports the idea that our simple single-cell cDNA amplification method successfully quantifies gene expression levels in single cells, and reinforces the earliest and specific onset of Prdm1 expression in the germ cell lineage.
We also determined the onset of expression of several genes that are known to have critical roles in PGC development. Nanos3, which is a mouse homolog of the evolutionarily conserved RNA-binding protein Nanos [30–32], is known to be important for PGC development, since the loss of its function leads to the disappearance of PGCs by E12.5 [32]. Since PGCs that lack Nanos3 are apparently affected as early as E8.5, Nanos3 is believed to be expressed from earlier stages, although its expression has been confirmed only after E9.5 [32]. We found that Nanos3 expression was indeed initiated specifically in PGCs at E7.25 (101–102 copies per cell). This finding is consistent with the phenotype of the Nanos3 mutant and designates Nanos3 as an early germ cell-specific transcript similar to Dppa3. We also examined the expression of Dead end (Dnd) [33], which is the mouse orthologue of the Zebrafish RNA-binding protein Dead end [34]. Dnd has been recently identified as the gene responsible for the Ter mutation, in which PGC development is severely affected as early as E8.0 [35], and is suggested to have a role in RNA metabolism in the germ cell lineage [33]. However, as is the case for Nanos3, the precise modes of Dnd expression in early PGCs and somatic neighbors have not been fully analyzed. We found that at E6.75, Dnd was expressed both in the PGC precursors and somatic neighbors at a certain frequency, whereas this gene was sharply upregulated only in PGCs after E7.25. The expression level of Dnd increased to about 2 x 102 copies per cell at E8.25 in PGCs. We also examined the expression of Tial1, which encodes another RNA-binding protein, the loss of function of which leads to early (E9.5) PGC death through unknown mechanisms [36]. We found that Tial1 was expressed at similar levels in the single-cell cDNAs of both PGCs and somatic neighbors, which suggests that Tial1 is upregulated subsequently in PGCs. These findings demonstrate the power of our single-cell cDNA quantitative gene expression analyses for the accurate description of gene expression profiles.
The Kit tyrosine kinase receptor, which is known to be expressed in PGCs as early as E7.5 [37], is involved in germ cell development [38, 39], together with its ligand, Kitl [39]. Therefore, we monitored the expression levels of these genes. Consistent with the previous in situ hybridization data, Kit expression was specifically upregulated in PGCs at E7.25, and its expression was maintained throughout the observed stages. On the other hand, Kit was not detected between E6.75 and E7.75 in the neighboring somatic cells. In contrast, Kitl was temporarily expressed at high levels in PGCs at E7.25, and then gradually decreased within 24 h. It is of note that PGCs at E7.25 that form a tight cluster expressed both Kit and Kitl at high levels (102 copies per cell in both cases). Cxcl12 encodes a chemokine that regulates PGCs to colonize gonads [40]. We found that the expression of this gene did not differ significantly between PGCs and their somatic neighbors during the period of investigation.
Since Prdm1 encodes a protein that is a member of the PR domain-containing protein family, we examined the expression profiles of the other family members (Prdm1-Prdm16). The expression levels of Prdm4, Prdm5, and Setd8 were similar between PGCs and somatic cells at the stages examined. Interestingly, Prdm14 was upregulated exclusively in the PGCs, in a manner similar to those of Dppa3, Nanos3, and Kit, which suggests the involvement of Prdm14 in PGC specification.
Pluripotency-Associated Genes
The expression levels of genes related to pluripotency, i.e., Pou5f1, Sox2, Nanog, Dppa5, and Eras, were monitored. Pou5f1, Sox2, and Nanog are reported to be required for the pluripotency of the inner cell mass cells of blastocysts, as well as for the derivation and maintenance of ES cells [25, 26, 41, 42]. We found that Pou5f1 was expressed in the Prdm1-positive PGC precursors and PGCs at similar levels (100 copies per cell) and at all the observed stages, while the neighboring somatic cells gradually downregulated the expression of this gene. Despite consistently high expression levels of Pou5f1 in the PGC precursors and PGCs, Sox2 expression was low or absent at E6.75 (see also Fig. 2, B and D, as well as the above section on population analyses), but was progressively upregulated in the subsequent stages. Importantly, Sox2 expression was barely detectable in the neighboring somatic cells, which indicates that the POU5F1-SOX2 complex exists only in PGC precursors after E6.75 in the posterior extra-embryonic mesodermal regions. Nanog was found to be expressed continuously in the PGC precursors but was downregulated sharply in somatic neighbors after E6.75. In contrast, the expression of Dppa5, a gene that is highly expressed in early embryos, ES cells, and PGCs in the genital ridges [43], was detected only at E6.75 and at low levels in both the PGC precursors and somatic cells, and was not upregulated until at least E8.25 in the PGCs. We also examined the expression of Eras, which is involved in the tumorigenic activity of ES cells [44], and found that this gene was not expressed in any of the cells we examined. These results indicate that PGC specification requires the expression of key pluripotency-associated transcription factors but does not require upregulation of all the genes associated with ES cells. By further monitoring the expression of other Sox genes, we found that Sox3 and Sox17 were transiently and specifically upregulated in PGCs at around E7.25.
Hox Genes and Mesodermal Markers
Previous reports have suggested that Hox cluster genes are repressed during the specification of PGC fate, whereas neighboring somatic cells upregulate the expression of these genes to acquire the somatic mesodermal fate [14, 15]. Therefore, the expression of several genes related to mesodermal differentiation was monitored in the single-cell samples. As shown in Figure 2 and described in the above section, the time-course analysis indicated that somatic neighbors expressed Hoxb1 already at E6.75 (101–102 copies per cell) and continued to express this gene throughout the stages examined, whereas PGCs downregulated the expression of Hoxb1. Somatic neighbors started to express Hoxa1 at E7.25 and continued to express it at later stages. On the other hand, although some PGCs showed temporary low-level expression of this gene at around E7.25, they essentially shut off its expression at subsequent stages. Regarding the expression of Evx1, T, and Fgf8, which are typical markers for posterior mesoderm [45–48], PGC precursors at E6.75 showed expression levels as high as those of somatic neighbors, whereas PGCs from E7.25 onwards gradually downregulated these genes, which suggests that some of the initially acquired mesodermal properties of the PGC precursors are eventually extinguished in the germline. It is also of note that the expression of Fgf8 was more tightly associated with PGCs from E6.75 to E7.75. Wnt5b, another gene that is a marker for the posterior mesoderm, including the allantois and primitive streak [49], showed rapid downregulation at E7.25 in PGCs (from 20 copies per cell at E6.75 to 0 copy at E7.25), whereas it continued to be expressed in somatic neighbors at relatively constant levels. Wnt5a [50] was expressed constantly at low levels (10 copies per cell) in both the germ and somatic lineages during the period examined. We also examined the expression of Snai1, which is a key gene for the epithelial-mesenchymal transition [51–53]. As expected, we found that somatic mesodermal cells expressed this gene at relatively high levels, especially at E7.25. In contrast, PGCs strongly repressed the expression of Snai1 as early as E7.25. In addition, we found that the expression of Myc, which is a prototypical oncogene involved in cell proliferation and a target for Prdm1-mediated repression in plasma cell differentiation [54], was sharply repressed only in the germ cell lineage after E7.25.
Epigenetic Regulators
Next, we focused on the expression of epigenetic regulators. As we have recently reported, PGCs undergo extensive reprogramming of their epigenomes soon after their specification, which includes the erasure of DNA methylation and H3K9me2, as well as the upregulation of H3K27me3 [17]. Therefore, we first examined the expression of DNA methyltransferases (DNMTs) [55, 56]. We found that Dnmt1, which encodes a maintenance methyltransferase that is essential for early embryogenesis [55], was constantly expressed at the mRNA level in both PGCs and somatic neighbors at the stages examined (102 copies per cell). Remarkably, we found that the expression of the gene for a de novo DNA methyltransferase (Dnmt3b) was strongly suppressed in PGCs at E7.25, which is in sharp contrast with its relatively constant expression in somatic neighbors. The expression of Dnmt3a was constantly low in both the germ and somatic cell lineages during the period examined.
We also monitored the expression of histone methyltransferases. Ehmt2 is a histone methyltransferase that is involved mainly in the dimethylation of H3K9, and its activity is essential for early embryonic development [57]. EHMT1, a protein that is highly similar to EHMT2, forms a complex with EHMT2, and both of these components are essential for the H3K9 dimethylation activity of the complex [58]. We found that Ehmt2 was expressed relatively constantly in both the germ and somatic lineages during the period examined, whereas Ehmt1 was apparently downregulated significantly in the PGCs after E7.25. Ezh, Eed, and Suz12 are members of the polycomb group (PcG) complex, which is involved in the trimethylation of H3K27 [59–62]. These three genes were expressed continuously at similar levels in both the germ and somatic cell lineages during E6.75-E8.25. YY1 is also a member of the PcG complex [63, 64], and its transcript was expressed continuously at all stages in both PGCs and somatic cells.
Signaling Molecules
Next, we scrutinized the expression of genes that are involved in cellular signal transduction [65, 66]. Although the expression levels for this category of genes measured by our method were generally low and it was somewhat difficult to identify significant differences between the germ and somatic lineages, we nonetheless found some statistically significant differences that may reflect the different properties of these spatially neighboring populations. It is known from gene knockout studies that molecules involved in BMP signal transduction [Bmp4, Bmp8b, Bmp2, Smad1, Smad5, Acvr1 (also known as Alk2)] [6–13] are important for the specification of germ cell fate. However, there has been very little information regarding the intrinsic expression of these molecules in early germline cells. We first examined the expression of the type I [Acvr1 and Bmpr1a (also known as Alk3)] and type II (Acv2b and Bmpr2) BMP receptors. We detected Acvr1 at low levels in both the germ and somatic lineages from E6.75 onwards. In contrast, we found that Bmpr1a expression was relatively high (102 copies per cell) in the PGC precursors at E6.75 but was decreased gradually to a low level in the PGCs, whereas in somatic neighbors, Bmpr1a was expressed at a similar level throughout the examined stages. Acvr2b was expressed constantly in both germ and somatic cells, whereas we could not detect a significant expression of Bmpr2.
We examined the expression of the Smad genes, which encode intracellular mediators of BMP signaling [65]. In the germ cell lineage, Smad1 was significantly expressed at E6.75 (102 copies per cell) but was progressively downregulated thereafter, which is consistent with the results from a previous study using the Smad1-LacZ allele [10]. Smad5 expression was only sporadically observed in the germ cell lineage from E6.75 to E7.75, whereas somatic neighbors expressed this gene more consistently throughout the period of investigation. The expression of Smad4, which is a co-SMAD that forms heterodimers with other SMAD proteins to act as transcription factors in the nucleus, was detected more consistently in the germ cell lineage and somatic neighbors, although it appeared to be slightly downregulated in the germ cell lineage at E7.25 and E7.75. These results are consistent with the expression of BMP receptor molecules. In contrast, we found that PGCs at E7.25 and E7.75 significantly and specifically upregulated the gene for SMAD3, which mediates TGFbeta signaling through heterodimerization with SMAD4. Collectively, these results suggest that PGCs decrease sensitivity to BMP signaling and increase sensitivity to TGFbeta signaling at around E7.25 to E7.75. Furthermore, we found that the germline cells expressed nodal specifically at E6.75 and E7.75, albeit at low levels.
We examined whether the JAK/STAT signaling pathway [66] plays a role in germ cell specification. Towards this end, we analyzed the expression of genes for the intracellular signal transducers Jak1, Jak2, Tyk2, Stat2, Stat3, Stat5
, Stat5ß, and Irf8, as well as the genes for interferon receptors, which included Ifnar1 and Ifnar2. This analysis revealed that although Ifnar1 was expressed consistently in both germ and somatic lineages, none of the signal transducers examined was detected at significant levels in either the germ or somatic mesodermal lineage, which suggests that the JAK/STAT signaling pathway is not important either for germ cell specification or for the formation of the extra-embryonic mesoderm.
Telomere-Related Genes and RNAi Machinery
Telomere maintenance is one of the key functions by which germ cells inherit a complete genome for subsequent generations [67]. Therefore, we examined the expression of genes associated with telomere maintenance, which included Tert, Terf1, Terf2, Pot1, Rif1, and Wrn, and found no clear differences between early germline cells and somatic neighbors in terms of the expression of these genes (Fig. 4). In addition, we examined the expression of Dicer1 and Eif2c2 (also known as Ago2), which are key components of the RNAi machinery [68]. We found that Dicer1 was expressed more constantly in the germline cells and that Eif2c2 was barely detectable with our method in either the germ or somatic lineage (Fig. 4).
Summary of Gene Expression Profiles
Figure 4 summarizes the gene expression profiles in the form of a heat map, based on the statistical analysis of the gene expression profiles. We have classified the analyzed genes into the following five categories based on their expression patterns and levels: group A, very high expression levels in both the germ and somatic cell lineages (housekeeping genes); group B, exclusive or higher expression in the germ cell lineage during at least one stage of development; group C, downregulated expression in the germ cell lineage but upregulated or more constant expression in the somatic cells; group D, relatively constant expression in both the germ and somatic lineages; and group E, no significant detectable expression. Group B includes Ifitm3, Prdm1, Dppa3, Sox2, Prdm14, Nanos3, Kit, and Dnd. Prdm1 and Sox2 were two of the genes that were expressed specifically in PGCs throughout E6.75–8.25. Dppa3, Kit, Nanos3, and Prdm14 were more consistently upregulated after E7.25, and Sox3 and Sox17 were transiently upregulated at around E7.25. Group C includes Hoxb1, Hoxa1, Dnmt3b, Snai1, and Myc. These genes showed complete transcriptional repression in PGCs at E7.25–7.75, whereas they were constantly expressed in the neighboring somatic cells. Group D includes Ezh2, Eed, Suz12, Ehmt2, Prdm4, Setd8, Dnmt1, and Smad4. Group E contains genes that showed almost no detectable expression at the observed stages, and possibly includes genes with expression levels lower than those of group D members.
We performed unsupervised hierarchical clustering of the Q-PCR data obtained from all 42 single cells using the 57 genes that were detected at significant levels in at least one cell type (i.e., categories A-D in Fig. 4). The results (Fig. 4F) show that 1) PGC precursors at E6.75 are more similar to the clusters of somatic neighbors, 2) PGCs at E7.25 constitute an independent cluster, to which three PGCs at E7.75 show similar properties, and 3) PGCs at E8.25 constitute another independent cluster, to which three PGCs at E7.75 exhibit high similarity. These data suggest that a key property of PGCs is acquired initially between E6.75 and E7.25, and that PGCs progressively alter their gene expression from E7.25 to E7.75, and to E8.25, which indicates that our present study identifies dynamic gene expression transitions during the early development of the germ cell lineage.
DISCUSSION
Using a simple, quantitative, single-cell cDNA amplification method followed by Q-PCR, we have analyzed gene expression dynamics during the 36 h of germ cell differentiation from E6.75 to E8.25. The precise mode of the specification of germ cell fate is an active area of research. It has not yet been elucidated how the Prdm1-positive cells increase in number or for how long the recruitment of the germ cell lineage from the precursors continues (see below). Nonetheless, our present study, which involved gene expression analysis of multiple Prdm1-positive cells at four different time-points, not only provides a precise view of the dynamic process through which PGCs eventually acquire their characteristic profile, which thus far has been reported only in part [14, 15], but also presents novel profiles, including those of signal transduction capacity and epigenetic regulation, for the PGCs. Since our method is very simple and straightforward, it should be generally applicable to any biological situation in which accurate single-cell analysis is required.
Our analyses reveal that in the CD1 background, at least a certain proportion of the Prdm1-positive cells at E6.75 and at E7.25 (approximately 35% and 30%, respectively) show expression of Hoxb1. Even at E7.75 and E8.25, approximately 10% of the Prdm1- and Pou5f1-positive populations we analyzed in this study expressed this gene. On the other hand, a previous report involving genetic lineage-tracing experiments showed that Prdm1-expressing cells contribute only to Dppa3-positive PGCs when examined between the LB and 3-somite stages, which indicates that Prdm1-positive cells are most likely lineage-restricted to the PGC fate [15, 16]. Therefore, it is reasonable to assume that Prdm1-positive and Hoxb1-positive cells develop at least until E7.25, to become Prdm1- and Dppa3-positive PGCs with the repression of Hoxb1. This idea is supported by the finding that other key mesodermal markers, such as T, Fgf8, Wnt5b, and Snai1, are also initially expressed at E6.75 in the majority of the Prdm1-positive cells we analyzed, while these genes are repressed along with the progression of PGC specification, apparently in an orderly and gene-dependent manner (Figs. 3 and 4, and data not shown). Since PGC specification in mice takes place in a region that is potentially influenced by numerous mesoderm-inducing activities, the initial upregulation of mesodermal markers in PGC precursors would not be surprising.
On the other hand, the fate of Prdm1-positive and Hoxb1-positive cells at E7.75 and E8.25 is not known and deserves careful further investigation. The expression of Prdm1, as detected by in situ hybridization and in Prdm1-mEGFP transgenic mice, continues in the most proximal posterior epiblast cells and in some nascent mesodermal cells that apparently originate from these epiblast cells, at least until E8.25. Some of these nascent mesodermal cells show clear expression of DPPA3 (data not shown). Therefore, it is possible that Prdm1-positive and Hoxb1-positive cells at E7.75 and E8.25 still go on to form Dppa3-positive and Hoxb1-negative PGCs, as is the case for earlier epiblast cells. Alternatively, these cells fail to contribute any further to the germ cell lineage and adopt somatic mesodermal fates. It is important to determine the stage until which all the Prdm1-positive epiblast cells contribute exclusively to the germ cell lineage and for how long the most proximal Prdm1-positive epiblast cells produce Dppa3-positive PGCs.
Our analysis also identifies an exclusive re-activation of Sox2 expression in the Prdm1-positive population. Most of the cells (66/76, 87%) that acquired Sox2 expression were negative for Hoxb1, and almost all the cells (57/61, 93%) that were positive for Dppa3 showed complete repression of Hoxb1 expression (Fig. 2). Furthermore, the percentage of Dppa3-positive cells in the Sox2-positive population increased during the PGC specification period (14% at E6.75, 83% at E7.25, 94% at E7.75, and 95% at E8.25). Given that somatic neighbors are negative for Sox2 (Figs. 3 and 4), the upregulation of Sox2 that preceded Dppa3 expression and the repression of Hoxb1 may be key steps in the specification of PGC fate. Since SOX2, which forms a heterodimeric complex with POU5F1, has been shown to be a key player in the pluripotency of ES cells [26], the exclusive re-acquisition of Sox2 expression in the germ cell lineage may herald the regaining of potential pluripotency in this lineage. We also found that another key player in pluripotency, Nanog, is maintained specifically in PGCs among extra-embryonic mesodermal cells after E7.25. A previous report has shown that Nanog expression is initially downregulated in the epiblast after implantation [41], which is consistent with our recent single-cell microarray analysis [69]. More recent reports [70, 71] have shown that in early post-implantation embryos, NANOG expression persists in the epiblast cells at E6.5 and E7.5, which suggests that NANOG expression may be upregulated in the epiblast cells at these later stages. Interestingly, in these reports, NANOG protein was detected in PGCs only after E7.5. However, we observed continuous expression of Nanog mRNA in the Prdm1-positive cells from E6.75. This apparent discrepancy may be due to post-transcriptional regulation of Nanog mRNA in these cells. Collectively, these data provide the first clear indication that only the germ cell lineage acquires the ability to regain the expression of key transcription factors for pluripotency, at least at the mRNA level, after E6.75.
This study also shows that Nanos3 and Dnd, which are potential RNA-binding proteins that are essential for PGC development [32, 33], Kit, which is a receptor tyrosine kinase that is essential for PGC survival, and Prdm14, which is a PR-domain-containing gene with unknown function, show specific upregulation in PGCs, at least by E7.25. Therefore, we now know that Prdm1, Prdm14, Sox2, Dppa3, Nanog, Kit, Nanos3, and Dnd are among the genes that specifically associate with the PGC specification process. Moreover, from our time-course analysis, Sox2 expression is considered to occur prior to Dppa3 expression. Precise determination of the onset of the expression of these genes in the Prdm1-positive cells and the mechanism through which the expression of each gene is affected in mutants of these genes would clarify the molecular pathways and networks that lead to the formation of functional germ cell lineages.
It is noteworthy that some of the general epigenetic regulators are repressed in PGCs. The expression levels of both Dnmt3b and Ehmt1 were repressed after E7.25, which may be important for the observed genome-wide demethylation of DNA and H3K9me2 in PGCs at around E8.0 [17]. In contrast, the expression levels of all the known components of H3K27 trimethylation remained constant in the germ cell lineage, which may be essential for the exclusive upregulation of H3K27me3 in PGCs at around E8.75. These findings suggest that specific epigenetic reprogramming in the germ cell lineage is preceded by, and a consequence of, transcriptional regulation of epigenetic regulators in this lineage. It seems reasonable to assume that factors involved in germ cell specification systematically set up this event.
Our results also reveal that germ cell specification is accompanied by specific control at the transcriptional level of signal transduction machineries. We found that PGCs apparently reduce their competence to respond to BMP signals at E7.75, by repressing the expression of the Bmpr1a-encoded receptor and the intracellular signal transducers encoded by Smad1 and Smad5 at the transcriptional level. Instead, Smad3 expression is specifically upregulated, which transmits TGFbeta signaling. A recent study has indicated that BMP4 signaling from the extra-embryonic ectoderm is mediated through the visceral endoderm, and has suggested that some unidentified factor from the endoderm modulates germ cell fate [13]. Investigations of the intrinsic expression of signal transducing molecules in Prdm1-positive cells during PGC specification period may lead to the identification of a pathway for the induction of the germ cell lineage.
In summary, our single-cell quantitative gene expression profiling of the PGC specification process has identified the precise dynamics of activation and repression of a specific set of genes associated with PGC development. It will be important to validate the functional significance of the transcriptional changes seen for these genes, since there is little information available regarding the functional relevance of these transcriptional changes [15, 32, 33, 38, 39]. This study should be a key step towards the full elucidation of the PGC specification process using the single-cell microarray technology we have recently developed [69], and may provide information that is essential for the appropriate reconstitution of PGCs and their descendants in culture.
ACKNOWLEDGMENTS
We thank all the members of our laboratory for their discussion of this study.
FOOTNOTES
2 Correspondence: Mitinori Saitou, Laboratory for Mammalian Germ Cell Biology, Center for Developmental Biology, RIKEN Kobe Institute, 2-2-3 Minatojima-minamimachi, Chuo-ku, Kobe, Hyogo 650-0047 Japan. FAX: 81 78 306 3377; saitou{at}cdb.riken.jp 
3 These authors contributed equally to this work.
1 Supported in part by a Grant-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology of Japan, and by a PRESTO project grant from the Japan Science and Technology Agency.
Received: 8 May 2006.
First decision: 30 May 2006.
Accepted: 14 July 2006.
REFERENCES
