Epigenetic Interactions and Gene Expression in Peri-Implantation Mouse Embryo Development

I. INTRODUCTION

One of the central questions in vertebrate developmental biology is how genetic and epigenetic information interact to form a three-dimensional embryo. Previous studies have established the general mechanisms of this process [Edelman, 1988, 1992; Nieuwkoop et al., 1985]: 1) The expression of certain developmentally important genes is regulated epigenetically and is position dependent. This expression depends on previously formed structures and is important for the subsequent interactions leading to the final pattern of the embryo. 2) The impetus for morphogenesis is cellular in origin, because it involves cell division, cell movement, cell adhesion, and cell death. In general, isolated molecular systems are not sufficient to provide the basis for patterning. 3) Cellular forces are linked to the nucleus via molecules such as proteins, peptide growth factors, hormones, receptors, and intracellular signal transduction cascades. The major tasks remaining for a resolution of this central question are to identify the specific gene products and control loops that link the tissue, cellular, and molecular levels of embryonic regulation.

The early morphogenesis of mammalian embryos emphasizes the differentiation of cell types that are essential for early embryonic nutrition within the maternal reproductive tract. Thus preimplantation development of eutherian embryos consists essentially of forming an outer layer of cells (trophectoderm) that is involved in attachment to the uterus [reviewed by Cruz and Pedersen, 1991]. Marsupial embryos also form an outer layer of cells (protoderm), but it is less invasive than the trophectoderm layer of eutherian embryos and generally does not develop into a chorioallantoic placenta [reviewed by Selwood, 1992]. Instead, marsupial embryos derive their nutrition through a vascularized yolk sac that becomes closely apposed to the uterine walls. Eutherian mammals, especially rodents, also utilize the yolk sac for early nutrition, before the chorioallantoic placenta is fully functional [Perry, 1981]. Thus mammals become equipped during their peri-implantation development for growth within the maternal uterine environment. By contrast, oviparous vertebrates are provisioned with nutritional resources by the yolk contained within the egg and proceed directly from cleavage to morphogenesis of the embryo proper.

Despite these differences in their strategies for embryonic nutrition, all vertebrates share the same basic body plan, i.e., they are bilaterally symmetrical quadripeds and might be expected to have similar embryonic strategies for achieving this form. Mechanisms involved in vertebrate body pattern formation are particularly evident during gastrulation, when mesoderm cells initially appear between ectoderm and endoderm layers at the prospective posterior end of the embryo, thus generating its anteroposterior axis. After gastrulation, neurulation occurs as the result of inductive interactions between mesoderm and the overlying ectoderm [Spemann, 1938; Cooke, 1985; Hamburger, 1988; Jones and Woodland, 1989]. Other dorsal mesoderm eventually forms the future notochord and somites, while ectoderm and endoderm develop into axial structures of the nervous system and the gut, respectively. The mesoderm becomes further subdivided as it moves toward the prospective head, trunk, and tail regions and mediates the specification of regional tissue identities as the anteroposterior structures are formed in the adjacent neural plate and endoderm. Morphogenetic mechanisms during this period of vertebrate development have been studied mainly in amphibian and bird embryos [Keller et al., 1992].

While these extensive experimental studies of morphogenetic mechanisms in other vertebrate classes have provided an understanding of the importance of epigenetic interactions in early development, relatively little is known about the cellular or molecular bases for these processes in mammals. This is due in part to the difficulty in obtaining sufficient amounts of preimplantation or peri-implantation material for biochemical or molecular analysis. However, the application of in situ hybridization and polymerase chain reaction (PCR) technologies has facilitated the analysis of gene expression at these early stages of development [Wilkinson and Green, 1990; Rappolee et al., 1989]. Although analysis of mammalian development in vitro remains more cumbersome than that of amphibian or bird embryo systems, model cell systems have been developed for studying the differentiation of early cell lineages: These pluripotent cell types include both embryonal carcinoma (EC) cell lines and embryonic stem (ES) cell lines [Robertson, 1987]. Moreover, studies involving gene targeting and homologous recombination in ES cells are being used to produce genetically altered mice, thus providing a powerful approach for determining the role of specific molecules in mammalian development [Rossant and Joyner, 1989; Rossant and Hopkins, 1992].

In view of these advances in the field of mammalian developmental biology, we have undertaken a review of the role of epigenetic interactions in the differentiation of the extraembryonic and embryonic cell lineages that constitute the peri-implantation mouse embryo. In the first section, we review the differentiation and fate of the cell lineages that form during preimplantation mouse development (trophectoderm [TE], inner cell mass [ICM], and primitive endoderm and primitive ectoderm), emphasizing the evidence for epigenetic interactions in their development; we also consider the implications for epigenetic interactions of studies of model cell systems, EC and ES cells, and from genomic imprinting. In the second section we summarize observations of some potential epigenetic signalling molecules, or morphogens, that have been identified in mouse embryos, and we also consider the role of transcription factors that could be responsible for the influence of epigenetic signals on lineage-specific gene expression. Although evidence about the function of specific molecules in these early morphogenetic events is limited, we cite a combination of in vivo and in vitro studies to summarize the role of epigenetic interactions in mouse embryogenesis.

II. LINEAGE ANALYSIS IN PERI-IMPLANTATION MOUSE EMBRYOS

A. Differentiation of the Early Lineages

We now largely understand the fate of the primary cell types that form during the preimplantation development of eutherian mammals, but the molecular basis of their initial differentiation remains obscure. The earliest cell types, TE and ICM, arise as the embryo undergoes cleavage and compaction while it is being transported from the site of fertilization (the ampulla of the oviduct) to the uterine lumen, where implantation takes place. During this time, the cleaving blastomeres progressively increase their degree of adhesion with each other, until the outlines of individual blastomeres are not readily distinguishable. As this process of compaction continues, the outer surfaces of the blastomeres develop junctional complexes, an indication of incipient trophectoderm differentiation.

Beginning at the eight-cell stage, outer cells of mouse embryos form apical microvilli and other polarized characteristics that culminate in the differentiation of trophectoderm. The individual blastomeres of intact embryos have stereotypical fates that reflect their state of cytodifferentiation: Polarized, outer cells always contribute some descendants to TE, whereas apolar, inner cells contribute to the ICM [Ziomek and Johnson, 1982; Johnson and Ziomek, 1983]. Inner cells are first formed during the division from the 8-cell to 16-cell (morula) stage through cleavage of some blastomeres with a division plane parallel to the outer surface of the embryo; other blastomeres divide in a plane perpendicular to the surface of the embryo, generating two outer descendants at the morula stage [Sutherland et al., 1990]. During the progression from the morula to the early blastocyst (32-cell) stage, additional inner cells form, as the result of divisions of outer cells that yield an outer and an inner descendant [Pedersen et al., 1986]. The number of cells recruited to the ICM at each of these two cleavage divisions (fourth and fifth cleavages) varies from embryo to embryo, through a regulative process that adjusts the number recruited in the fifth cleavage depending on the number of inner cells present after the fourth cleavage [Fleming et al., 1987]. Although the mechanism of this regulation is unknown, it has the consequence of providing a more consistent number of ICM cells and a more constant ratio of ICM to TE cells than would occur in the absence of such regulation. Thus, as early as preimplantation stages, there is evidence for a role of epigenetic mechanisms in development of the mammalian embryo.

The determination of the fate of individual blastomeres as TE or ICM also results from epigenetic interactions. The “inside–outside” hypothesis proposes that pluripotent cells of the compacting morula respond to differences in their microenvironments, outer cells respond to their environment by forming TE, and inner cells to theirs by forming ICM [Tarkowski and Wroblewska, 1967]. The two central predictions of the “inside–outside” hypothesis are that cleavage stage blastomeres would remain totipotent at least until the morula stage and that they would respond to changes in their environment by differentiating according to their inside or outside position. Studies of disaggregated blastomeres from each of the cleavage stages have shown that they are capable of forming both TE and ICM when aggregated with other genetically marked blastomeres [reviewed by Gardner, 1983; Pedersen, 1986]. This capacity has been demonstrated for blastomeres isolated from two-cell through late morula stages. Cells placed in outer positions of aggregation chimeras tend to form TE, while those placed inside tend to form ICM. Even the inner cells at the early blastocyst stage retain the capacity for differentiation into TE if placed in outer positions [see Pedersen, 1986, for review]. Therefore, the two essential predictions of the “inside–outside” hypothesis are fulfilled.

However, there is no evidence for specific microenvironmental factors produced by outer or inner cells that influence their path of differentiation as proposed by the “inside–outside” hypothesis. When eight-cell or morulae with intact zonae pellucidae are microinjected into the blastocyst cavity of giant, chimeric blastocysts, they develop as morphologically normal blastocysts, indicating that this microenvironment does not possess diffusible factors sufficient to induce ICM differentiation [Pedersen and Spindle, 1980]. Conversely, embryos exposed to the actin microfilament inhibitor cytochalasin D fail to undergo compaction and do not form morphologically normal blastocysts, yet they synthesize polypeptides characteristic of the ICM [Surani et al., 1980]. The absence of a clear case for the molecular basis of the “inside–outside” hypothesis leaves us unclear how to evaluate it further.

The “polarization” hypothesis proposes that TE differentiates in response to asymmetric cell contacts, while inner cells with their symmetric cell contacts develop as ICM [reviewed by Johnson and Maro, 1986]. This hypothesis also postulates epigenetic interactions for the first differentiative step in mouse embryogenesis. It is apparent that cell contacts are essential for polarization and compaction to occur, and that cell–cell adhesion is implicated in these processes, as proposed by the polarization hypothesis. Compaction depends on the close contact between adjacent blastomeres and on the calcium-dependent cell surface adhesion molecule known as uvomorulin [Hyafil et al., 1981]. Interference with uvomorulin function by calcium depletion or using antibodies blocks the polarized differentiation of mouse blastomeres, as in other cell culture systems [Shirayoshi et al., 1983; Vestweber et al., 1987; McNeill et al., 1990]. However, the limited degree of polarized cytodifferentiation in rabbit morulae development raises the question of whether the polarized phenotype itself is a general phenomenon required for TE differentiation among eutherian mammals [Ziomek et al., 1990].

A genetic approach to the role of epigenetic interactions in TE and ICM differentiation would be helpful, provided that mutations affecting such early steps in mammalian development could be recovered. The earliest known mutations lead to developmental arrest during cleavage, without specifically affecting either TE or ICM [reviewed by Magnuson, 1986; Pedersen, 1988]. An informative mutation would be one that resulted in all cells of the embryo following a single differentiative pathway, which would be different from that followed by some cells under normal circumstances. Such a mutation would presumably render the mouse embryo “TE-less” or “ICM-less” and would fulfill the definition of a homeotic mutation, whether the gene was related to existing homeobox genes or had an independent molecular identity.

The next cell types to form in mammalian embryos are the primitive endoderm and primitive ectoderm, both derivatives of the ICM. Primitive endoderm differentiates on the blastocelic surface of the ICM just before the embryo attaches to the uterus at 4.0 days of gestation (dg) [Nadijca and Hillman, 1974; reviewed by Gardner, 1983]. The primitive ectoderm consists of the remaining undifferentiated cells in the core of the ICM [reviewed by Rossant, 1984]. Primitive endoderm appears to differentiate because of its outer position on the surface of the ICM. When ICMs were isolated from late blastocysts they formed an entire outer layer of primitive endoderm, whereas they would have formed endoderm on only one surface in the intact blastocyst [Rossant, 1975]. Thus the blastocelic surface is an outer environment conducive to primitive endoderm formation. Since early ICM cells have the capacity to form TE, why does the inner blastocele of the ICM differentiate into primitive endoderm rather than into TE? ICM cells retain the capacity for TE development until embryos reach the expanded blastocyst stage (3.5 dg). This is approximately the time of the sixth cleavage division, when the ICM gains the capacity for primitive endoderm formation [Nichols and Gardner, 1984; Chisholm et al., 1985; reviewed by Gardner, 1983; Rossant, 1986]. Thus, although the information conveyed by the “outside” environment of the blastocyst cavity may be similar to that surrounding the embryo, by 4.0 dg the ICM cells have undergone a transition to a new state, from which they can only differentiate into primitive endoderm or primitive ectoderm. There could be a role of the internal blastocyst environment and cell–cell contacts in suppressing trophectoderm differentiation before this time, since ICMs of early blastocysts do not differentiate into TE in the intact embryo. But after the transition, primitive endoderm differentiation is no longer suppressed by (nor does it require) the blastocyst environment, because ICMs from expanded blastocysts develop primitive endoderm either in situ or in isolation.

An alternative source of epigenetic information for primitive endoderm differentiation is through cell interactions. Enders et al. [1978] reported a compaction-like process just preceding endoderm differentiation by the ICM. However, there is no body of evidence to implicate uvomorulin as in TE and ICM differentiation. The amount of laminin present may be important in primitive endoderm differentiation, however, judging from the inhibitory effects of exogenous laminin or antilaminin antisera on endoderm differentiation in EC embryoid bodies [Grover et al., 1983]. In addition, studies on EC cells have implicated cell–cell and cell–matrix interactions in the differentiation of the primitive endoderm [Casanova and Grabel, 1988], and a similar case can be made for the importance of these interactions in maintaining the differentiated state of visceral endoderm in embryos [Kalimi and Lo, 1989]. Further analysis of factors affecting primitive endoderm differentiation in isolated ICMs should be fruitful.

The differentiation of primitive ectoderm has not been studied in nearly such detail. Indeed, it is not clear from their morphology whether primitive ectoderm cells actually differentiate into a new phenotype or simply remain an undifferentiated stem cell population or an “embryoblast” at the late blastocyst stage (4.5 dg). The primitive ectoderm cells acquire a columnar phenotype and form an epithelial layer (also known as the embryonic ectoderm or epiblast) at the time the proamniotic cavity forms (5.5 dg). On the basis of culture studies using isolated cores of primitive ectoderm, it initially appeared that primitive ectoderm was capable of differentiating into primitive endoderm. However, Gardner [1985] demonstrated that pure cores of primitive ectoderm did not form primitive endoderm and attributed the previous observations to artifacts arising from a multilayered endoderm. Apparently the determination of ICM cells’ capacity for primitive endoderm versus primitive ectoderm occurs at the late blastocyst stage (4.5 dg), with the innermost cells becoming primitive ectoderm [Gardner, 1985; reviewed by Gardner and Beddington, 1988]. The nature of the factors that regulate this developmental decision remains obscure.

B. Fate of the Early Lineages

The fate of the early cell lineages in mouse embryo—TE, ICM, primitive endoderm, and primitive ectoderm—has been studied by blastocyst micromanipulation and other methods for clonal analysis, leading to a coherent view of early allocation events [reviewed by Gardner, 1983; Rossant, 1984; 1986; 1987; Gardner and Beddington, 1988; Pedersen, 1988]. The TE develops into three trophoblast cell types, including trophoblast giant cells, extraembryonic ectoderm, and ectoplacental cone; all are extraembryonic and become part of the interface between conceptus and maternal tissues during in utero development. The ICM develops into primitive endoderm and primitive ectoderm, which have very different fates: Primitive endoderm subsequently forms visceral and parietal endoderm, which contribute strictly to extraembryonic tissues, while primitive ectoderm forms a wide variety of tissues, both embryonic and extraembryonic. The principal descendants of primitive ectoderm are the primary germ layers of the embryo proper—the embryonic mesoderm, endoderm, and ectoderm—which form during gastrulation. Other products are the amniotic ectoderm and the extraembryonic mesoderm, which contributes to the yolk sac, chorion, amnion, and allantois. Together, these descendants of the early lineages constitute the peri-implantation conceptus. In the following discussion, we will examine the evidence for the fate of TE, primitive endoderm, and primitive ectoderm, focusing on the role of epigenetic mechanisms in their development.

1. Trophectoderm

The diploid trophoblast descendants of TE—the extraembyronic ectoderm of the chorion and the diploid cells of the ectoplacental cone—arise through the continued proliferation of polar TE during peri-implantation development of the mouse conceptus [Gardner et al., 1973; Rossant and Tamura-Lis, 1979; Papaioannou, 1982]. The mural TE cells differentiate during early phases of implantation (4.0–4.5 dg) to become the primary trophoblast giant cells [Dickson, 1966; Barlow et al., 1972]. The nuclei of these cells enlarge through repeated cycles of DNA synthesis without cell division (endoreduplication), thus acquiring many copies of chromatids loosely resembling polytene chromosomes [Varmuza et al., 1988]. Differentiation of cells destined to become primary trophoblast giant cells is altered, however, if a vesicle of isolated mural TE cells is injected with a donor ICM obtained from another embryo. In such “reconstituted blastocysts,” the mural TE cells in contact with the donor ICM form diploid extraembyronic ectoderm and ectoplacental cone [Gardner et al., 1973; Rossant and Tamura-Lis, 1979; Papaioannou, 1982]. Both of these trophoblast cell types are able to differentiate into trophoblast giant cells under a variety of culture conditions when they are separated from ICM derivatives; cultured ectoplacental cone rapidly differentiates into giant cells, while extraembryonic ectoderm requires longer for this transition [Rossant and Tamura-Lis, 1981). The giant cells formed in culture thus resemble the “secondary giant cells” that surround the mouse conceptus during peri-implantation development.

The blastocyst reconstitution studies indicate that inductive interactions between ICM (and its derivatives) and the TE (and its derivatives) are responsible for maintaining the diploid phenotype in closely adjacent trophoblast cell types. Moreover, the trophoblast culture experiments indicate that the entire trophoblast lineage is capable of giant cell differentiation, which is thus the default pathway for trophoblast cells when inductive signals from the ICM are absent or are distant, as in the case of primary and secondary giant cells. There is no information about the possible identity of such inductive signals, although the giant cell formation by extraembryonic ectoderm cultured in conditions that preserve its tissue integrity suggests that substances derived from the ICM rather than embryonic organization per se are responsible for the inductive effect [Rossant and Tamura-Lis, 1981].

There has been considerable controversy over whether the ICM makes a significant cellular contribution to the trophoblast lineage, in addition to its evident inductive role. Blastocyst reconstitution studies demonstrated that with few exceptions proliferating mural TE was the source of the trophoblast cell types [Gardner et al., 1973; Rossant and Tamura-Lis, 1979; Papaioannou, 1982; Rossant and Croy, 1985]. But the cell population dynamics of the mouse blastocyst indicate that despite similar mitotic indices there is a gradual decline in the proportion of ICM cells relative to TE; this could be interpreted either as differential cell death or as ICM contribution to the polar TE [Handyside, 1978; Handyside and Hunter, 1986]. Because the polar TE was discarded in the blastocyst reconstitution studies, the possibility of ICM contribution to the TE lineage through polar TE had not been addressed. Analysis of the fate of marked polar TE cells in intact blastocysts revealed that they contributed descendants to the mural TE during subsequent blastocyst growth [Copp, 1979; Cruz and Pedersen, 1985]. Cruz and Pedersen [1985] interpreted the patterns of labeled mural TE descendants as indicating movement of labeled polar TE cells away from the embryonic pole and their replacement by unlabeled cells derived from the ICM. Movement of polar TE away from the embryonic pole occurs during development of the rabbit blastocyst and in other eutherian mammals [Williams and Biggers, 1990; reviewed by Cruz and Pedersen, 1991]. Its relevance to mouse blastocyst development, however, was questioned by Dyce et al. [1987], who concluded on the basis of labeling studies using tracer microinjection and fluorescent microparticles that labeled polar TE cells were passively displaced from the embryonic pole and that the ICM contributes descendants to the polar TE only rarely and in minimal numbers. Their conclusions contrast with those of Winkel and Pedersen [1988], who used microinjection of cell lineage tracers to mark progenitor cells in the ICM and concluded that there was a substantial contribution from ICM to TE. In a recent study Gardner and Nichols [1991] addressed this controversy by replacing some of the inner or outer cells of late morulae with donor cells from synchronous embryos. They found that donor inner cells rarely colonized the polar trophectoderm of recipient blastocysts, even when the donor cells were present throughout the period of polar TE development, and that donor outer cells regularly colonized the entire trophoblast lineage. This issue clearly deserves further study to resolve these divergent interpretations of TE origin. However, based on current information, the hypothesis that ICM is the predominant source of the definitive polar TE cells must be regarded with caution.

Taking into account all of these observations on lineage relationships within the trophoblast lineage, a model emerges (Fig. 1) for the fate of polar TE [Copp, 1979; Cruz and Pedersen, 1985; Rossant and Tamura-Lis, 1981]. Mural TE cells undergo endo-reduplication as implantation begins, but additional cells are recruited to the mural TE population through continued migration from the proliferating polar TE. After implantation, the polar trophectoderm becomes multilayered as ectoplacental cone and extraembryonic ectoderm form, thus increasing the distance between ICM derivatives and trophoblast. Peripheral ectoplacental cone cells, receiving diminishing inductive signals from the ICM, cease proliferating and begin endo-reduplication as secondary giant cells, continuing to accumulate with already-formed giant cells. According to the lineage model advanced by Rossant and Tamura-Lis [1981], extraembryonic ectoderm serves as a stem cell population for the trophoblast lineage, continuing proliferation throughout the peri-implantation period and contributing descendants to the ectoplacental cone that in turn form secondary giant cells (Fig. 1D). In summary, the trophoblast lineage develops in a highly organized process involving supporting interactions with the ICM, progressive differentiation, and directional growth. Several of these morphogenetic events appear to involve epigenetic mechanisms, including effects of trophoblast derivatives on primitive endodermal derivatives.

Fig. 1.

Model depicting the fate of cells in the trophoblast lineage during peri-implantation mouse embryo development. A: Polar trophectoderm cells (top) contribute descendants to mural trophectoderm (bottom) during preimplantation blastocyst growth (arrows). B: Polar trophectoderm cells contribute to extraembryonic ectoderm as trophectoderm proliferates (thick arrows), and trophoblast giant cells (finely stippled) undergo implantation in the uterine crypt (fine arrows); uterine decidual tissue is indicated by coarse stippling. C: Proliferating diploid trophoblast cells contribute to the elongating ectoplacental cone as implantation proceeds. D: Lineage map indicating the origins of the trophoblast lineage and modeling the relationships between polar trophectoderm, extraembryonic ectoderm, ectoplacental cone, and trophoblast giant cells. Extraembryonic ectoderm and diploid ectoplacental cone are envisioned as self-renewing stem cell populations for the trophoblast lineage during postimplantation development. (A–C, adapted from Copp, 1979, and D, adapted from Rossant and Tamura-Lis, 1981, with permission of the publisher.)

2. Primitive endoderm

In studies of cell fate beginning at the late blastocyst stage (4.5 dg) Gardner and coworkers used blastocyst injections to analyze the fate of primitive endoderm, showing that descendants of this cell lineage populate the visceral extraembryonic endoderm and the parietal endoderm. In initial studies, visceral endoderm cells from 4.5 dg embryos were injected into expanded blastocysts (3.5 dg), and their descendants were analyzed at midgestation, showing that primitive endoderm cells contributed to yolk sac endoderm but not to definitive embryonic (gut) endoderm [Gardner and Rossant, 1979]. Subsequent studies led to the discovery that visceral endoderm was able to contribute to the parietal endoderm as well as yolk sac endoderm; parietal endoderm, on the other hand, had descendants only in the parietal endoderm population [Gardner, 1982; Cockroft and Gardner, 1987].

These results are consistent with other studies on 6.5 and 7.5 dg egg cylinders, indicating that the visceral endoderm layer has the capacity for differentiation into parietal endoderm. In experiments involving dissection of 6.5 dg mouse embryos, Hogan and Tilly [1981] found that visceral extraembryonic endoderm cells differentiated into parietal endoderm when they were cultured in contact with extraembryonic ectoderm undergoing the transition from diploid trophoblast to giant cells. They proposed that these transitions also occur in the intact embryo, where extraembryonic ectoderm may differentiate into trophoblast giant cells coordinately with the differentiation of visceral endoderm to parietal endoderm [Hogan and Tilly, 1981; Hogan and Newman, 1984]. Accordingly, their model (Fig. 2) envisions a movement of visceral embryonic endoderm towards the extraembryonic region, concomitantly with movement of visceral extraembryonic endoderm toward the parietal endoderm regions and is consistent with the progenitor–descendant relationship between the visceral and parietal endoderm found in the blastocyst reconstitution studies [Gardner, 1982; Cockroft and Gardner, 1987].

Fig. 2.

Model depicting the fate of cells in the primitive endoderm lineage during postimplantation development of the mouse egg cylinder. At 6.5 days (top), outer cell layer is composed entirely of primitive endoderm-derived cells, consisting of visceral embryonic endoderm (VE End), visceral extraembryonic endoderm (VEX End), and parietal endoderm (P End). α-Fetoprotein-synthesizing cells are indicated by open circles (VE End), and laminin and collagen IV–synthesizing cells are indicated by closed circles. Embryonic ectoderm (E Ect) and extraembryonic ectoderm (EX Ect), trophoblast giant cells (GCs) and Reichert’s membrane (RM) are also indicated. At 7.5 days (bottom), extraembryonic mesoderm (Meso) has accumulated in the area formerly occupied by extraembryonic ectoderm, and definitive endoderm cells have accumulated at the distal tip of the egg cylinder (left) as a result of gastrulation. Visceral extraembryonic endoderm cells of the yolk sac (VYSac) synthesize α-fetoprotein at this stage. Also indicated is movement of cells from visceral embryonic endoderm to visceral extraembryonic endoderm, which is envisioned as the stem cell population for the parietal endoderm. (Reproduced from Hogan and Tilly, 1982, with permission of the publisher.)

This model has not been evaluated in the intact conceptus, because all experimental approaches that have been devised for studying cell lineage relationships in mouse embryos involve dissection of the peri-implantation embryo and culture in vitro. The studies by Gardner and coworkers reveal the fate of cells placed in a blastocyst environment, but do not directly address the fate of cells residing in the visceral endoderm of the intact egg cylinder. Analysis of lineage relationships within the primitive endoderm and trophectoderm lineages, in particular, would require an approach that allows cell marking after implantation to determine the fate of these cells. Creation of transgenic lines of mice possessing appropriate indicator genes under the regulation of transcriptional regulator genes such as flp might facilitate such experiments [O’Gorman et al., 1991]. Without such studies, we can only roughly estimate the limits of visceral endoderm contribution to the parietal endoderm population. Cockroft and Gardner [1987] found that the incidence of parietal endoderm colonization by descendants of visceral endoderm declined precipitously from 5.5 to 6.5 dg. These observations indicate that by 6.5 dg donor visceral endoderm no longer has the capacity to contribute substantially to the parietal endoderm, even though the visceral endoderm of the egg cylinder still contains many cells capable of parietal endoderm differentiation at this stage [Hogan and Tilly, 1981]. In studies of the fate of visceral embryonic endoderm cells of 6.5 and 7.5 dg embryos, Lawson and coworkers [Lawson et al., 1986; Lawson and Pedersen, 1987] found that labeled descendants in yolk sac endoderm were spread in an arc roughly perpendicular to the anteroposterior embryonic axis and moreover tended to remain in coherent patches [Lawson and Pedersen, 1991], as described by Gardner [1984]. These results suggest that the contribution to parietal endoderm may diminish because the differentiation of visceral endoderm after the onset of gastrulation encourages coherent clonal growth rather than cell mixing or migration.

The synthetic pattern of α-fetoprotein (AFP) in the visceral endoderm layer shows that changing associations between early cell lineages alters their expression of lineage-specific genes. Dziadek and Adamson [1978] observed that AFP was first synthesized in the visceral embryonic endoderm of 6.5 dg embryos but was not synthesized in the visceral extraembryonic endoderm until later stages (7.5 dg). The onset of AFP synthesis in visceral extraembyronic endoderm coincided with the migration of extraembryonic mesoderm into space formerly occupied by extraembryonic ectoderm, leading to the close association of this mesoderm with the visceral extraembryonic endoderm in this region (Fig. 2). Further experimental studies showed that AFP synthesis by visceral endoderm was inhibited by the proximity of the extraembryonic ectoderm to the visceral extraembryonic endoderm. Visceral extraembryonic endoderm cells reassociated with embryonic ectoderm expressed AFP, whereas the same endoderm cells reassociated with extraembryonic ectoderm did not [Dziadek, 1978]. Moreover, extraembronic fragments that contained visceral extraembryonic endoderm cells in close association with extraembryonic ectoderm synthesized large amounts of collagen IV and laminin, the extracellular matrix components characteristic of parietal endoderm [Dziadek and Adamson, 1978]. Together, these observations suggest that epigenetic signals emanating from trophoblast lineages suppress the visceral extraembryonic endoderm phenotype and induce differentiation into parietal endoderm. The molecular identity of such morphogens is still unknown, but some ideas about the signalling pathways that can affect differentiation within the primitive endoderm lineage have emerged from studies of model cell systems, as discussed subsequently.

Because numerous studies of inductive interactions have demonstrated that there are mutual interactions between adjacent tissues, it is reasonable to expect that visceral endoderm cells also generate signals that affect the development of their neighbors. Visceral extraembryonic (yolk sac) endoderm has been shown to be necessary for blood island development in the extraembryonic mesoderm of chick embryos [Wilt, 1965; Miura and Wilt, 1969]. Similarly, the chick hypoblast layer (equivalent to visceral embryonic endoderm) appears to have a decisive role in the differentiation of the epiblast layer during gastrulation [Azar and Eyal-Giladi, 1981].

3. Primitive ectoderm

Blastocyst injection studies by Gardner and coworkers have demonstrated conclusively that the somatic and germ cell lineages of the fetus, as well as extraembryonic mesoderm and amniotic ectoderm, are descendants of the primitive ectoderm [Gardner and Papaioannou, 1975; Gardner and Rossant, 1979; reviewed by Rossant, 1984; Gardner, 1983]. This cell lineage is thus equivalent to the epiblast layer of the chick embryo, which gives rise to all fetal lineages, including the embryonic gut endoderm [Vakaet, 1985]. Mouse epiblast becomes recognizable as an embryonic epithelium upon formation of the proamniotic cavity at 5.5 dg [Snell and Stevens, 1966]. Thereafter the epiblast population expands by rapid proliferation [Snow, 1977] and contributes cells to the endodermal, mesodermal, and ectodermal lineages beginning at the onset of gastrulation, approximately 6.5 dg. The fate of mouse epiblast cells has been examined by microinjecting individual cells with lineage tracers at 6.7 dg [Lawson et al., 1991] and by transferring groups of donor epiblast cells labeled with gold-conjugated wheat germ agglutinin to unlabeled host embryos at 7.5 dg [Tam and Beddington, 1987; Tam, 1989], followed by culture for 1 to 2 days.

For the purpose of analyzing the fate of epiblast cells, Lawson et al.[1991] divided the cup-like embryonic portion of the mouse egg cylinder (containing the epiblast) into 11 arbitrary sectors (Fig. 3A). The first tier of sectors (I–V) passes from the anterior of the epiblast to its posterior, the base of the primitive streak. The second tier of sectors also passes from anterior to posterior, from sector VI to the tip of the primitive streak in sector X. The distal end of the egg cylinder forms sector XI. The analysis of epiblast cell fates and allocation to the primary germ layers and organ rudiments was performed by microinjecting a single progenitor cell of each embryo with a combination of horseradish peroxidase and rhodamine-conjugated dextran. The distribution of descendants in the neuroectoderm shows that this tissue is derived primarily from the anterior axis of the epiblast (sectors VI and XI). A smaller contribution was derived from sectors I and VII. The distribution of descendants in the embryonic gut endoderm indicates that this germ layer is derived from a region extending from sector VIII to sector X. This region overlaps the region that gives rise to the head process, which contributes to the notochord of the embryo [Lawson and Pedersen, 1986]. The descendants in the embryonic mesoderm are derived from the lateral and posterior regions (including sectors II–V and VII–X), but not from the anterior-most sectors (sectors I and VI) or from the tip of the epiblast (sector XI). The distribution of descendants in yolk sac mesoderm indicates that this tissue is derived mainly from the posterior epiblast region near the border of the embryonic and extraembryonic portions of the conceptus (sectors III–V). The allantois was derived mainly from sector II, in the anterior epiblast, with an occasional contribution from sector V, in the posterior epiblast. Epiblast progenitors that gave rise to the blood islands in the yolk sac also produced other yolk sac mesoderm cells. This indicates that the hematopoietic stem cell progenitors are derived from a common ancestor with extraembryonic mesoderm (KA Lawson and RA Pedersen, unpublished observations). Together with other examples of single epiblast cells that contributed to two or more primary germ layers, this observation indicates that neither determination of cell fate nor cell allocation in the epiblast of mouse embryos is complete before epiblast cells enter the primitive streak [Beddington, 1983; Lawson et al., 1991]. This conclusion is not consistent with the prediction from a study on gastrulating chick embryos describing the epiblast as a mosaic of pure clones destined to generate ectoderm, mesoderm, or endoderm [Stern and Canning, 1990]. Rather, germ layer fate appears to be determined in mouse embryos by epigenetic signals that reach the epiblast cells during or after the process of gastrulation.

Fig. 3.

A: Diagram of pregastrula/early gastrula stage mouse egg cylinder (6.7 day), identifying sectors (I–XI) marked by lineage tracer microinjection. The anterior of the prospective craniocaudal axis is oriented to the left, the posterior to the right. The typical extent of the primitive streak in early gastrula stage embryos is indicated as a vertical rectangular box (right). The position of the injected epiblast cell is estimated along the H axis (dorsoventral height) and D axis (egg cylinder diameter) as the distance (h) from the embryonic/extraembryonic border and distance (d) from the anterior margin. Bar-0.1 mm. B: Fate map of the epiblast of the prestreak and early streak stages of the mouse embryo (6.7 days) showing the derivation of the primary germ layers, extraembryonic mesoderm, and amniotic ectoderm when cultured from prestreak to midstreak or from early streak to neural plate stages. (Adapted from Lawson et al., 1991, with permission of the publisher.)

Studies of the gastrula fate map of the mouse embryo have revealed striking topological similarities to the avian gastrula fate map. Not only do the migratory movements of mouse epiblast converge on the primitive streak as in the chick embryo, but also the germ layer precursors are arrayed in the same successive order of convergence on the streak in the two species (Fig. 3B): Precursors of notochord and gut endoderm are the first to enter the streaks, followed by those for embryonic and extraembryonic mesoderm, then for neuroectoderm and surface ectoderm, which do not leave the epiblast layer [Lawson et al., 1991; Vakaet, 1984]. There is also a strong similarity in the pattern of surface ectoderm and brain segment precursors in the mouse epiblast as compared with the chick, with anterior epiblast forming nonneural ectoderm as well as telencephalon and diencephalon [Tam, 1989; Couly and LeDouarin, 1987]. These similarities imply that fundamental mechanisms of organ formation are conserved among vertebrates. Thus, in searching for epigenetic mechanisms in mammalian development, it is important to take into account evidence from other vertebrate models, in addition to mammalian embryos and model cell systems.

C. Model Cell Systems

EC cell lines established from teratocarcinomas have been used extensively as models for the early cell lineages of mouse embryos [reviewed by Martin, 1980; Hogan et al., 1983]. After in vitro exposure to retinoic acid, these cells differentiate into derivatives of primitive endoderm [Strickland and Mahdavi, 1978]. Treatment with cyclic AMP shifts their differentiation toward a parietal endoderm phenotype, while culture in suspension induces embryoid bodies to form an outer layer with a visceral endoderm phenotype [Strickland et al., 1980; Hogan et al., 1981]. These observations may suggest a role for protein kinase A in the signal transduction pathway leading to parietal endoderm differentiation. While the regulated differentiation of EC cells provides a useful model for analyzing the primitive endoderm lineage, they have been less satisfactory models for primitive ectoderm derivatives. The cells that remained at the core of embryoid bodies of the F9 line of EC cells did not differentiate into embryonic gut endoderm or mesoderm, although they could form neural-like cells [Kuff and Fewell, 1980; Hogan et al., 1983]. Other EC lines varied markedly in their capacity to contribute to somatic tissues of blastocyst injection chimerae, but rarely generated germline chimerism [Papaioannou and Rossant, 1983]. The isolation of pluripotent human EC cell lines has raised expectations that they would provide similar models for studying early differentiative processes in human development [Andrews, 1984; Pera et al., 1989; Wiles, 1988].

The greatest progress in developing model cell systems for early mouse embryogenesis has been establishment of ES cells by direct culture of inner cell masses from mammalian embryos [Martin, 1981; Evans and Kaufman, 1981; reviewed by Robertson, 1987]. These cell lines have been shown to differentiate spontaneously in culture and must in fact be prevented from differentiating, using the cytokine leukemia inhibitory factor differentiation inhibitory activity (LIF/DIA). The endoderm that forms around the embryoid bodies of these cell lines has been shown to express genes characteristic of visceral extraembryonic endoderm, although only in a subset of the total endodermal cells [JJ Latimer, CA Burdsal, and RA Pedersen, unpublished observations]. The other cells in the endoderm layer probably represent primitive and parietal endoderm. Furthermore, these cell lines show distinct patterns of differentiation when allowed to differentiate spontaneously rather than under the control of retinoic acid in aggregation culture. The spontaneous differentiation of the ES cells is more progressive and less uniform than that of the retinoic acid–induced differentiation and differs significantly from that seen in F9 cultures induced to differentiate with retinoic acid (JJ Latimer, CA Burdsal, and RA Pedersen, unpublished observations). Interestingly, the expression of certain related gene families such as the globin gene family is induced in these differentiated cells in the correct temporal pattern (i.e., with fetal globins actively expressed first and adult globin last), although the early genes do not seem to be switched off at the time the adult genes become active [Lindenbaum and Grosveld, 1990] ES cells cultured in medium supplemented with methylcellulose and peptide growth factors can undergo hematopoiesis [Wiles and Keller, 1991], and embryoid bodies cultured for prolonged periods can form structures resembling blood islands [Risau et al., 1988; Wang et al., 1992]. Embryoid bodies derived from ES cells can also transcribe genes that characterize cardiac muscle [Robbins et al., 1990; Sanchez et al., 1991]. A more compelling observation, however, is the absence of axial development in ES as well as EC embryoid bodies, indicating that gastrulation does not typically occur. These data suggest that although a more progressive type of differentiation is manifested by ES cells in culture, some of the proper signalling required for axial formation is missing and the resultant differentiation is consequently disorganized. Future studies involving the definition of serum-free media that contain the appropriate growth factors and cytokines for specific types of differentiation need to be performed. Growth on defined extracellular matrix components may also be required to induce differentiation of specific cell types.

Because of their high rates of contribution to both somatic and germ line lineages in chimerae, ES cells are being used extensively as vehicles for introducing transgenes into mice [see Rossant and Hopkins, 1992, for review]. In this context, it has been shown that the cells must be maintained in undifferentiated form to retain their pluripotency. When LIF/DIA is added to culture medium it prevents differentiation, although culture of ES cells on feeder layers of embryonic fibroblasts or STO cells is also useful in maintaining their pluripotency [Robertson, 1987].

ES cell lines have also been derived from both parthenogenetic and androgenetic blastocysts [Robertson et al., 1983; Stewart, et al., 1991] (KS Sturm, JJ Meneses, and RA Pedersen, unpublished observations) as well as normally fertilized diploid blastocysts. These ES cells have been valuable for studying the mechanism and consequences of genomic imprinting, a perturbation of epigenetic mechanisms that provides additional models for studying early developmental events.

D. Genomic Imprinting

Experimental evidence with mice indicates that normal prenatal development of mammalian embryos requires both maternal and paternal genetic contributions [reviewed by Surani, 1986; Solter, 1988]. This specialization of egg and sperm nuclei, defined as genomic imprinting, is thought to occur during gametogenesis. Imprinting seems to act through the expression of specific genes, which are located on six different mouse autosomes [reviewed by Cattanach and Beechey, 1990]. The morphological and developmental phenotypes of parthenogenetic (diploid maternal genotype) and androgenetic (diploid paternal genotype) embryos have been studied extensively, and the identities of several imprinted genes are now known [see Pedersen et al., 1992 for review].

Diploid parthenogenotes are capable of preimplantation development, but die at midgestation stages with extensive perturbations in the differentiation of their extraembryonic and embryonic cell lineages [Sturm et al., 1992, reviewed in Pedersen et al., 1992]. The most conspicuous abnormalities are the tendency for trophoblast to differentiate into giant cells, with few diploid cells capable of proliferation in this lineage, and for the primitive endoderm lineage to differentiate completely into parietal endoderm, with few morphologically recognizable visceral extraembryonic endoderm cells. Parthenogenesis also causes defects in the embryo proper, including abnormalities of mesodermal and ectodermal lineages ranging from their complete absence to slight morphological defects in the heart, somites, and other organ rudiments [Sturm et al., 1992]. When diploid parthenogenotes are combined with normally fertilized embryos as aggregation chimerae, they contribute extensively to many tissues of adult mice, including the germline, but are notably absent from skeletal muscle; descendants of parthenogenetic cells are selectively eliminated from the trophoblast and primitive endoderm lineages during early embryonic development from the trophoblast and primitive endoderm lineages [Fundele et al., 1989; Nagy et al., 1989; Clarke et al., 1988; Thomson and Solter, 1989]. The deficiencies in parthenogenetic embryos thus seem to be of two kinds: 1) cell autonomous defects in extraembryonic lineages, which are not rescued by combination with normal cells in aggregation chimerae; and 2) defects in embryonic lineages (other than skeletal muscle) that involve epigenetic signals, because they are largely correctable by normal embryonic cells in aggregation chimerae. Chimerae produced by reconstituting blastocysts so they have normal trophoblast and primitive endoderm cells, but have parthenogenetic primitive ectoderm, also die at midgestation stages, although they reach slightly more advanced stages than pure parthenogenotes [Gardner et al., 1990]. These observations further indicate that embryonic lineages are perturbed by parthenogenesis and suggest that they are deficient for diffusible factor(s) normally generated either by embryonic sources or by extraembryonic sources that are not accessible to the parthenogenetic embryo, even in such chimerae.

Although less extensive studies have been done on androgenetic embryos, descendants of their cells also can be sustained in chimerae beyond the midgestation time when pure androgenotes die, suggesting perturbation of some diffusible component(s) in androgenotes [Barton et al., 1991; Mann and Stewart, 1991]. In these cases, however, the chimerae containing descendants of androgenote embryos or ES cells display skeletal defects, suggesting that androgenotes may generate an excess of diffusible factors responsible for these and other abnormalities [Mann et al., 1990; Mann and Stewart, 1991; Barton et al., 1991].

The genes that have been identified as imprinted endogenous genes include growth and differentiating-regulating factors and their receptors. The first endogenous gene to be identified was insulin-like growth factor (IGF)-II [DeChiara et al., 1991], followed by the IGF-II/mannose-6-phosphate receptor [Barlow et al., 1991] and the H19 gene, with no known function [Bartolomei et al., 1991]. Recently, Rappolee et al. [1992] observed that the IGF-I receptor gene is not transcribed in preimplantation mouse parthenogenotes, suggesting that its expression is affected by imprinting at early stages of parthenogenetic development. The identities of these genes, taken together, provide evidence that genomic imprinting can act through perturbations of growth and differentiation, as suggested by Surani et al. [1988] and more recently by Haig and Graham [1991] and Sturm et al. [1992]. However, deficiencies in the known genes cannot account for all the abnormalities observed in parthenogenotes, androgenotes, or other maternal duplication/paternal deficiency and paternal duplication/maternal deficiency phenotypes; therefore, other imprinted endogenous genes must be involved in generating the biological effects of imprinting observed during mouse development [see Cattanach and Beechey, 1990; Pedersen et al., 1992, for review]. The rationale for continued examination of these model systems is that they will reveal the identities of additional genes that are integral components of epigenetic mechanisms that act during early mammalian development.

III. GENE EXPRESSION IN THE PERI-IMPLANTATION EMBRYO

Some of the seminal epigenetic interactions that drive inductive developmental processes have been described morphologically in the previous sections. Although the subject of this section is gene expression in the peri-implantation embryo that contributes to these cell–cell interactions, a discussion of extracellular matrix (ECM)–involved interactions has been omitted. This subject has been recently reviewed [Edelman, 1988; Jessel and Melton, 1992]. We have chosen instead to restrict our discussion to the expression of trans-active effector molecules that impact more directly on the nucleus either as mitogenic agents or transcriptional regulators.

Although a discussion of the morphogenetic properties of retinoic acid is included in the section on mammalian homeotic genes, we have not included a section on retinoic acid receptor molecules. These have been recently reviewed [Linney, 1992; Ruberte et al., 1991; Dolle et al., 1989b].

A. Selected Cytokines

Several cytokines have been identified as having a role in mammalian embryonic development. Among these is a cytokine known as leukemia inhibitory factor (LIF) or differentiation inhibitory factor (DIA). This cytokine was originally purified, characterized and cloned because of its ability to induce the differentiation as well as suppress the clonogenicity of mammalian myeloid leukemia cell lines [Tomida et al., 1990; Maekawa and Metcalf, 1989]. However, LIF/DIA ligand, like certain other factors active with myeloid cells (interleukin-6, [IL-6], granulocyte colony stimulating factor [G-CSE]; and macrophage colony stimulating factor [M-CSF]), seems to be involved in the hepatic acute phase reaction as it is released into the serum by the liver in response to acute inflammation [Metcalf, 1988; Baumann and Wong, 1989; Kordula et al., 1991]. In addition to these functions, LIF/DIA promotes bone resorption in vitro [Abe et al., 1986] and is identical to the melanoma-derived lipoprotein lipase inhibitor [Mori et al., 1989] and cholinergic neuronal differentiation factor from heart cells [Yamamori et al., 1989]. LIF/DIA receptors are present on monocyte macrophage cells of the hematopoietic system although their biological significance is not clear.

Of greatest significance to early mouse development is the differentiation inhibitory activity of LIF/DIA on ES cells derived from the inner cell mass of the mammalian blastocyst [Williams et al., 1988; Gough et al., 1989]. This functional role coupled with the fact that expression of the Lif gene varies at different stages of embryogenesis [Murray et al., 1990] suggests that it plays an important role in embryogenesis. In fact Lif expression has been shown to coincide with blastocyst formation and always precedes implantation [Bhatt et al., 1991]. These results may indicate that one of the principal functions of LIF/DIA is to regulate blastocyst growth and initiate implantation.

Although there is a single copy of the Lif gene in both mouse and humans, two independently transcribed forms of this mRNA exist, one that is secreted and diffusible, and one that is extracellular matrix-bound [Rathjen et al., 1990a,b]. In view of the pleiotropic effects of this molecule in many different organ systems, such differential transcription could provide at least two distinct types of cellular regulation.

LIF/DIA and IL-6 are both cytokines that have pleiotropic effects on many cell types and systems [Hilton and Gough, 1991; Allan et al., 1990; Gough and Williams, 1989; Reid et al., 1990]. Although in many cases the activities of these two molecules seem to be similar, the structures of the two proteins are unrelated. However, the cDNA of the Lif receptor has recently been isolated and characterized [Gearing et al., 1991] and has been shown to be related to the gp130 signal-transducing component of the IL-6 and G-CSF receptors. The transmembrane and cytoplasmic regions of the LIF/DIA receptor and gp130 are most closely related. These structural similarities may suggest a common signal transduction pathway existing for the two receptors and may explain the similarities in biological function. Furthermore, it has now been shown that both LIF/DIA and IL-6 inhibit the differentiation of mouse EC F9 cells when induced to differentiate by retinoic acid alone or in combination with dibutyryl cAMP [Hirayoshi et al., 1991].

B. Peptide Growth Factors in Development

The concept of induction in experimental embryology was illustrated elegantly by Spemann and Mangold [1924] when they demonstrated that the dorsal lip of the blastopore of a gastrula stage amphibian embryo had special trans-active properties; i.e., when taken from one gastrula and transplanted into the ventral side of another, this piece of dorsal mesoderm could induce a secondary neural plate, leading to a twinned tadpole. It is now known that many factors operate in trans to induce proliferation of specific cell types as well as differentiation of others in the vertebrate embryo. These secreted proteins include peptide growth factors such as the growing family of fibroblast growth factors (FGFs; aFGF, bFGF, WNT-2, HST/k-FGF, FGF-5, FGF-6), transforming growth factors (TGFs), and TGF-related factors, i.e. the activins and the oncogenes of the WNT family; and the insulin-like growth factors (IGFs) [for review, see Nilsen-Hamilton, 1990]. The patterns of expression of these genes as well as those of the genes of their receptors help to elucidate the complex nature of their respective mechanisms during development. Increasing evidence suggests that peptide growth factors have crucial roles in development [Davidson, 1990; Whitman and Melton, 1989]. Polypeptide growth factors mediate many cell–cell interactions that presumably occur during mammalian development [Mercola and Stiles, 1988].

FGF is involved in various developmental events in amphibia, including mesoderm induction, induction of homeobox genes, establishment of anteroposterior polarity [Ruiz i Altaba and Melton, 1989], neuronal differentiation and survival [Morrison et al., 1988], and angiogenesis [Folksman and Klagsbrun, 1987]. Originally FGF was isolated in two forms in mammals, acidic (aFGF) and basic (bFGF). Both forms act via the same cell surface receptor and are potent fibroblast mitogens [Caday et al., 1990]. The ability of both forms to interact with a common receptor presumably enables aFGF and bFGF to exert similar effects on a wide range of mesodermal and neuroectodermal cell types in amphibia to control their proliferation and differentiation [Gospodarowicz, 1987].

The best known effect of bFGF is its ability to act as a mitogen for cells of mesenchymal origin [Gospodarowicz et al., 1987]. Both aFGF and bFGF exert a wide range of effects on different adult mammalian cell types in vitro. These effects include stimulation of endothelial cell migration and proliferation, neurite outgrowth, and enhanced nerve cell survival and differention [Gospodarowicz et al., 1987]. A possible role for the FGF family in pathological processes such as cancer has become evident with the identification of oncogenes that encode proteins having a 40%–50% sequence homology to aFGF and bFGF [Dickson and Peters, 1987]. These oncogenes include Wnt-2, Hst, Fgfk, and Fgf, and they are discussed in detail below. Another member of this family, keratinocyte growth factor (KGF), is a specific mitogen for epithelial cells [Finch et al., 1989].

In mammalian embryos the presence of FGF protein and mRNA during development has been detected in several ways including a fibroblast mitogenesis assay [Caday et al., 1990], immunoprecipitation [Seed et al., 1988], immunocytochemistry [Gonzalez et al., 1990] and Northern blotting analysis [Hebert et al., 1990]. Recently, Gonzalez et al. [1990] reported the protein localization of bFGF in 18-day rat fetus. The distribution of this protein was found to be widespread, suggesting multiple functions in various developing organ systems.

In an attempt to understand further the roles of FGF in development, another study involving identification of FGF-responsive or FGF receptor (FGF-R)–bearing cells was done to determine spatial and temporal patterns of expression. Transcripts were localized with varying intensities to all organs, with the possible exception of the liver [Wanaka et al., 1991]. Relatively strong signals were located in the mesoderm-derived tissues such as perichondrium, metanephros, prevertebral column, and myotome. Since the purification and cloning of the chicken bFgf-r [Lee et al., 1989], several laboratories have isolated Fgf-r cDNAs [Reid, 1990; Mansukhani et al., 1990]. This work suggests that the FGF-R belongs to a family of tyrosine kinase receptors based on highly conserved tyrosine kinase domains and relatively unconserved extracellular domains.

In amphibia FGF is probably a vegetalizing factor, working in combination with another growth factor, transforming growth factor-β (TGF-β) to induce ventral mesenchyme induction in early embryos [Slack et al., 1987]. In the chick it may have a similar function [Mitrani et al., 1990a]. Functional testing of FGF in Xenopus has been performed using a dominant negative approach to prove that it functions in mesoderm formation [Amaya et al., 1991]. TGF-β has been shown to be capable of positively or negatively regulating aFGF or bFGF, depending on the cell type involved [Frater-Schoder et al., 1986].

The FGFs are expressed differentially in cell culture systems [Hebert et al., 1990]. Since EC and ES cell lines are believed to be model systems of the preimplantation mammalian embryo, it may be possible to draw some limited conclusions about expression of these genes in the early embryo from these types of studies. These conclusions can now be verified using PCR or in situ hybridization on embryos. PSA-1 cells can be grown as aggregates that differentiate into simple embryoid bodies after 3–6 days in culture or allowed to progress to cystic embryoid bodies after 12–15 days in culture. Finally, embryoid bodies that have been replated in tissue culture dishes for an additional 10 days have also been analyzed for Fgf expression. The most restricted pattern of expression was shown by Fgfk, as this gene was expressed only in undifferentiated EC or ES cells or in early differentiation stages of PSA-1 cells. Transcripts from the three other genes Fgfb, Fgfa and Fgf5 were detectable by Northern blot analysis from whole embryos or various embryonic tissues ranging from 10.5 to 15.5 days. In PSA-1 cells, which were assumed to represent earlier cell types, Fgfb, Fgfa, and Fgfk were expressed in embryoid bodies, regardless of their state of differentiation. Fgf5 mRNA expression underwent a dramatic 15-fold increase in steady-state levels when PSA-1 cells differentiated into embryoid bodies. Interestingly, the level of Fgf5 did not increase when PSA-1 cells were specifically differentiated into visceral or parietal endoderm, so the changes in Fgf5 mRNA upon differentiation of these cells were presumed to be due to changes in the inner core cells of the embryoid bodies rather than the endoderm cells themselves [Hebert et al., 1990]. Although EC cells are not analogous to preimplantation embryos in the sense that they are transformed, and also lack the three-dimensional cell–cell contacts present in the embryo, studies like these are useful in providing clues as to the cell type exclusivity of expression one might be able to expect of a gene within the mammalian embryo.

Another gene that shares homology with Fgf is the Wnt-2 gene, which has no homology with the Wnt-1 gene. The WNT-2 protein has an atypical hydrophobic leader sequence and may be proteolytically processed at dimers of basic residues [Nusse, 1988]. In the mouse embryo expression of the Wnt-2 gene is found in the primitive streak from 7.5 to 9.5 days [Wilkinson et al., 1988]. Since the embryonic mesoderm arises from the primitive streak, this is the area where one would expect to find a response to a mammalian mesoderm-inducing factor, although at this point there is no direct evidence that Wnt-2 is such a factor in mammals. At later stages of development Wnt-2 mRNA is detectable in the endoderm of the first three pharyngeal pouches, the hindbrain adjacent to developing otocyst, and the mesenchyme of developing teeth. Although these are not areas associated with mesoderm formation, they are places where differentiation is modulated by epithelial mesenchymal interactions. Therefore Wnt-2 may function as an extracellular inducer at this time in development [Wilkinson et al., 1988, 1989].

The TGF-β family of proteins has also been implicated as playing a role in mammalian development. Originally in vitro studies showed that TGF-βs 1–3 were mitogenic for cells derived from supporting tissues such as bone and cartilage. They have, however, also been shown to inhibit proliferation of many cell types. The TGF-βs have a similarly complex regulation of differentiation, having a stimulatory effect on certain cell types and an inhibitory effect on others. In addition, TGF-βs stimulate extracellular matrix deposition, are chemotactic for particular cells, and in the amphibian embryo induce mesoderm formation [Nilsen-Hamilton, 1990].

The TGF-βs 1–3, like the FGFs, are involved in complex cell–cell interactions occurring in the adult as well as the embryo [Akhurst et al., 1991]. Although five cDNA clones have been isolated in the Tgf-β family, only TGF-βs 1–3 have been found in mammals. TGF-βs 1–3 are similar in that each polypeptide is synthesized as a precursor monomeric protein that is subsequently cleaved to form a 112 amino acid polypeptide. This polypeptide remains associated with the latent “pro” portion of the molecule [Lyons and Moses, 1990; Miller et al. 1990a,b]. Biological activity in all three forms of TGF-β is associated with dimerization of the monomers, generally as homodimers, and subsequent release of the latent “pro” peptide. The fully processed form of the TGF-β 3 protein has approximately 80% identity with the analogous regions of both β 1 and β 2, although there is only 27% identity at the amino-terminal regions [ten Dijke et al., 1988; Derynck et al., 1988]. In general, TGF-β 1–3 have qualitatively similar activities when added to cells in culture and seem to interact with the same cell surface molecules [Graycar et al., 1989], although some differences in their biological activities have been reported [Merwin et al., 1991; Rosa et al., 1988].

Recent studies have shown that expression patterns of the Tgf-β mRNAs in the murine embryo are overlapping but distinct [Pelton et al., 1990]. Furthermore, these patterns were shown to change over the course of development and were most often found to be present in tissues that were in the process of undergoing morphogenetic alterations, such as in the developing whisker follicle [Lyons et al., 1990]. All three Tgf-βs were expressed similarly in the mature follicle, whereas they were all different in the immature follicle. Similar trends in expression have been seen in the human embryo [Gatherer et al., 1990; Sandberg et al., 1988]. In terms of the protein expression of these genes, a study utilizing isoform-specific antibodies has shown that the Tgf-βs 1–3 are expressed in unique temporal and spatial patterns relative to one another in a wide range of tissues [Pelton et al., 1991]. These results may suggest that all three mammalian forms of TGF-β act through both paracrine and autocrine mechanisms.

Activins are also members of the TGF-β family. Activins were first described as factors that stimulate the release of follicle-stimulating hormone (FSH) from the anterior pituitary [Mason et al., 1985; Ling et al., 1986]. They have been purified in vivo as a homodimer of two βA chains (activin A) or a heterodimer of βA and βB chains (activin AB) [Vale et al., 1986]. Activin A has also been shown to stimulate erythroid differentiation [Yu et al., 1987]. Within the activin family is a related molecule, inhibin, that blocks the FSH-releasing capacity of pituitary cells and consists of an inhibin-specific A chain paired with either of the B chains [Mason et al., 1985].

Both mammalian activin A and recombinant Xenopus activin B induce a variety of dorsal anterior mesodermal and neural structures, including eyes, in Xenopus. Xenopus animal caps cultured in the presence of activins will differentiate into structures that possess well-defined head structures, notochord, and muscle [Asashima et al., 1990a,b; Sokol et al., 1990]. These results suggest that activins induce presumptive ectoderm to form mesoderm dorsoanterior structures [Sokol et al., 1991]. Another piece of evidence supporting this hypothesis is that activin mRNA injected into Xenopus embryos causes the formation of a partial second dorsal axis [Thomsen et al., 1990]. However, evidence against the latter proposal includes the fact that activins (βA and βB) are not expressed at the stage of embryonic development in vivo (32–64 cell stage) when the mesoderm-inducing capacity is present (both activins βA and βB are expressed later in development) and the fact that the endogenous inducer (but not activins) is capable of respecifying the fate of ventral ectoderm, resulting in the formation of a dorsal axis on the ventral side of the embryo [Ogi, 1967; Nieuwkoop, 1969]. In addition, vegetal blastomeres, which contain an endogenous dorsal inducer, are capable of rescuing axial formation (of neural or dorsal mesoderm tissues) after embryos have been UV irradiated. Activins are not able to achieve either of the two latter feats [Sokol and Melton, 1991].

The timing of activin gene expression suggests that in Xenopus activin may play an important role in mesoderm induction and axial patterning [Thomsen et al., 1990; Smith et al., 1990]. Furthermore, activins have also been shown to induce axial formation and been shown to be expressed in the hypoblast of the chick [Mitrani et al., 1990b]. At present there is no published evidence that activins are analogous in function with those of Xenopus in the context of the mammalian embryo.

Some studies have been performed on EC cells in culture using combinations of activins and other growth factors. In one study, activin A and bFGF were shown to prevent the death of P19 cells, which normally die within 2 days grown in serum-free N2 medium on tissue culture plastic [Shuber and Kimura, 1991]. Furthermore, the substratum to which these cells were exposed was capable of mediating their mitogenic response to growth factors, i.e., when P19 cells were cultured on extracellular matrix components laminin and fibronectin, they responded to activin A and bFGF by actively dividing. This study may indicate that, to evaluate the role of activins in mammalian cells, investigators will have to provide more complex culture conditions that include extracellular matrix components that may be present during embryonic development in utero. This idea may also serve to improve and extend embryo culture [Copp and Cockroft, 1990].

Another group of peptide growth factors that is involved in mammalian embryogenesis is the insulin and insulin-like growth factor (IGF) family. The IGFs, like the growth factors previously discussed, also participate in a wide variety of biological responses. One of the most intriguing observations on the effects of the insulin family in the preimplantation period of embryogenesis is that the treatment of blastocysts in culture with insulin leads to the formation of mice that are larger than controls after transfer to a pseudopregnant foster mother [Gardner and Kaye, 1991]. During preimplantation development, the major responses to insulin develop when there are numerous receptors on the cell surface during the morula and blastocyst stages. It has been shown that at these times insulin stimulates the incorporation of glucose into nonglycogen macromolecules and increases the incorporation of labeled precursors into RNA, DNA, and protein [Rao et al., 1990]. The presence of transcripts for Igf I, Igf II, and their receptors in early mouse embryos provides additional evidence for an autocrine role in growth regulation during peri-implantation stages [Rappolee et al., 1992; Telford et al., 1990].

The IGFs are mitogenic growth factors whose activities are modulated by the presence of accessory binding molecules. Both IGF I and IGF II are held on binding proteins in serum that increase their half lives and alter their access to cells. These high-molecular-weight complexes are unable to pass through the walls of capillaries and therefore limit the access of IGFs to tissues. These binding complexes do not appear, however, until late in fetal development, and are therefore not involved in embryogenesis, although they may restrict IGF access to the placenta [Daughaday, and Rotwein, 1989; Froesch et al., 1985].

Although the IGFs are encoded by single-copy genes in mammals, there is considerable heterogeneity in their expression at the level of the proteins. All of the members of this family are produced as precursor prohormones that are proteolytically cleaved to their mature size by peptidases. Different proteins are produced by alternative splicing of mRNAs or posttranslational processing [Gammeltoft, 1989].

The IGFs share receptors with insulin. The receptor for IGF I with highest affinity is IGF I receptor. IGF II interacts with the mannose-6-phosphate receptor as well as the IGF I receptor with lower affinity. The mannose-6-phosphate receptor also has the ability to bind glycosylated TGF-β 1 precursor [Clairmont and Czech, 1989].

IGF II has recently been shown to play a role in embryonic growth [Gray et al., 1987]. The Igf II gene uses at least three different promoters and expresses several transcripts in many tissues during embryogenesis. Abundant IGF II mRNA expression is found in all the trophectoderm derivatives after implantation [Lee et al., 1990]. Later it is detected in the extraembryonic mesoderm at the early primitive streak stage. IGF II mRNA is also expressed transiently in the primitive endoderm, disappearing after yolk sac formation. Expression in the embryo proper appeared first at the late primitive streak/neural plate stage in lateral mesoderm and in the anteroproximal cells located between the visceral endoderm and the most cranial region of the embryonic ectoderm.

The first direct evidence that IGF II plays a role in embryonic development comes from experiments by DeChiara et al., in which disruption by gene targeting was performed in ES cells. Subsequent reintroduction into mouse host blastocysts gave rise to heterozygous mutant mice that were smaller than wild-type ES-derived littermates, leading to the identification of Igf II as an endogenous imprinted gene. These growth-deficient mice were apparently normal and fertile [DeChiara et al., 1990, 1991]. The Igf II receptor gene has also been shown to be imprinted [Barlow et al., 1991].

C. Oncogenes

It has been assumed for some time that mechanisms of carcinogenesis could be related to control of normal embryogenesis. Indeed, a tumor resembles an undifferentiated, rapidly proliferating embryonic cell, and, conversely, an embryo placed into certain ectopic nonuterine sites eventually develops into a tumor [Silver, 1983]. Many tumors display embryonic antigens such as the carcinoma embryonic antigens [Huang, 1990b]. Some forms of cancer are thus postulated to be the consequence of aberrant expressions of genes whose normal function is to modulate proliferation during development.

Immortalized and fully transformed cells frequently transcribe genes that are expressed in, and presumably influence, normal mammalian development [Ruddon, 1987]. In some cases these oncofetal genes do not appear to contribute to the neoplastic phenotype; for example, AFP is expressed by trophoblast cells and by many tumor cells. In other cases developmentally regulated genes play a primary role in the conversion of cells to the transformed phenotype; for example, the protooncogenes: c-Myc, c-Src, c-Fos, and c-Fms are all expressed during embryonic development and have been shown to regulate developmental steps in vitro [Adamson, 1987]. Another example of these types of genes is illustrated by the gene Pem, derived from a T lymphoma, which is not expressed in adult tissue but is first expressed in early embryos and subsequently in extraembryonic tissues.

Many protooncogenes are expressed during specific stages of mammalian development. These include Wnt-1 (formerly int-1), Wnt-2, Wnt-3, c-Fms, c-Myc, N-Myc, c-Jun, and junB which are all expressed in particular temporal and spatial patterns in the developing mammalian embryo and presumably play a role in embryonic cell growth or differentiation [Wilkinson et al., 1987, 1988, 1989; Regenstreif and Rossant, 1989; Schmid et al., 1989; Mugrauer et al., 1989]. It is generally assumed that these genes do not play an exclusive role in control of development, but are performing the same essential function in the embryo or fetus as they do in the adult.

Oncogenes can be divided into several categories. Certain oncogenes are growth factors. One of the best examples of oncogenes of this type is v-Sis, which is nearly identical to the β-chain of platelet-derived growth factor (PDGF). Cells overexpressing the cellular homolog of PDGF can grow in the absence of PDGF in the medium [Owen et al., 1984]. Other examples include the Wnt-2 gene, which was identified by its activation in mouse mammary tumors after insertion of a proviral DNA [Moore et al., 1986]. This gene has been shown to have 40%–60% amino acid identity with the aFGF and bFGF proteins [Dickson and Peters, 1987]. Further examples are provided by the Hst/ks (Fgfk) oncogene, which shares similar amino acid identity with the 120 amino acid core characteristic of the FGFs [Delli-Bovi et al., 1987], and Fgf5, originally identified as a human oncogene [Zhan et al., 1988].

The Wnt-1 gene encodes a protein with a hydrophobic leader sequence that is presumably a signal peptide. Furthermore, it is cysteine rich, a fact that is consistent with its putative identity as a growth factor or receptor, although there is no direct functional evidence for such extracellular action or cell surface properties [Nusse, 1988]. The Wnt-1 gene is 54% identical with the Drosophila wingless gene, which plays a role in segmentation. It is expressed at the mRNA level in the mouse embryo between 8 and 14 days, within the developing nervous system, specifically in areas of the neural plate, anterior head folds, neural tube, and spinal cord [Shackleford and Varmus, 1987; Wilkinson et al., 1987]. The spatial and temporal patterns of expression of this gene in the mouse suggested that it may be involved in neural tube maturation, although it is now known that targeted disruption of both alleles of this gene gives rise to mice with severe abnormalities in the development of the mesencephalon and cerebellum and that the recessive mutation in mice known as swaying is a mutant allele of the Wnt-1 gene [Thomas and Capecchi, 1990; Thomas et al., 1991; McMahon and Bradley, 1990].

The Wnt-1 and Xwnt-8 (Xenopus wnt-8) genes have been shown to be important in Xenopus axial formation [see Jessell and Melton, 1992, for review]. In Xenopus, mesoderm is induced in an area known as the marginal zone of the blastula stage embryo. This marginal zone is located at the border of the animal and the vegetal hemispheres, destined to become ectoderm and endoderm, repectively. Factors responsible for the induction of mesoderm apparently arise from the vegetal cells [Nieuwkoop, 1973; Boterenbrood and Nieuwkoop, 1973; Gimlich and Gerhart, 1984; Dale et al., 1985] and act upon the marginal zone. This vegetal dorsalizing region is also known as the Nieuwkoop center [Gerhart et al., 1989], which acts to stimulate the exposed marginal zone cells to form dorsal mesoderm. This dorsal mesoderm, known as the Spemann organizer [Spemann and Mangold, 1938], forms the dorsal lip of the blastopore, a place where gastrulation occurs earlier and more extensively than on the ventral side of the embryo [Gerhart and Keller, 1986].

Recent studies have shown that the Wnt oncogene families and the activins can induce certain aspects of dorsal axis formation in Xenopus embryos. When Xenopus embryos are injected with Wnt-1 and Xwnt-8 mRNAs, dorsal axial structure bifurcation occurs. This result resembles that observed in Xenopus embryos injected with activin mRNAs. The hypothesis formed from these results was that the ectopically expressed WNT proteins were interfering with the Spemann organizer and subsequently causing it to split [Sokol et al., 1991]. However, in a recent study by Smith and Harland the vegetal blastomeres exposed to exogenous Xwnt-8 mRNA contributed progeny to the endoderm rather than the induced dorsal axis, indicating that the Xwnt-8 mRNA may instead cause cells to act as a Nieuwkoop center rather than a Spemann organizer. Furthermore, both Wnt-related mRNAs were shown to have the ability to rescue UV-treated axis-deficient embryos, whereas activins could provide only partial rescue [Sokol et al., 1991; Smith and Harland, 1991]. Recently an additional insertion site for MMTV proviruses has been identified, Wnt-3. This gene encodes a protein that is 47% identical to WNT-1 at the amino acid level. The expression of this gene has not yet been characterized at the biochemical or molecular level in the embryo [Roelink et al., 1990].

Another class of oncogenes includes that of growth factor receptors. Especially interesting are the receptors with tyrosine kinase activity that include epidermal growth factor (EGF) and HER 2/Neu, both of which can become oncogenes if properly activated, i.e., a point mutation in the transmembrane domain of HER2/Neu receptor causes this receptor to become oncogenic [Bargmann et al., 1986]. Other receptor-derived oncogenes possess different structural lesions such as point mutations in the cytoplasmic regions or deletions and carboxy-terminal truncations that appear to enhance and modulate the transforming signal [Studzinski, 1989].

D. Homeobox Domain Genes

The homeobox is a highly conserved 180 bp sequence that was first described in genes controlling pattern formation in Drosophila melanogaster [Akam, 1987; Gehring, 1987]. This sequence is normally located close to the C terminus of the protein, encoding a 60 amino acid homeodomain containing a helix–turn–helix motif that function as a sequence-specific DNA-binding domain. More than 30 murine homeobox genes (Hox) have been isolated by homology to Antennapedia or other members of this highly conserved gene family. In the mouse four clusters of these genes have been identified: Hox-1 (Chrm.6), Hox-2 (Chrm.11), Hox-3 (Chrm.15) and Hox-4 (Chrm.2) [Colberg-Poley, 1985]. The identical order of paralogous genes [Schughart, 1988] in all four clusters suggests that they emerged by complete or partial duplication of a common ancestral cluster. Alignment can be extended to the Drosophila ANT-C and BX-C complexes [Gaunt, 1988]. A strong correlation exists between the position of a gene within the cluster and its expression pattern along the anteroposterior (A–P) axis in both Drosophila and vertebrates [Akam, 1989]. In the mouse, this is most obvious in the central nervous system where the distinct anterior boundaries of expression are exhibited [Gaunt, 1988].

The Hox-5 gene complex exhibits an interesting pattern of expression in murine limb buds after the ninth day of gestation [Dolle et al., 1989a], although no function for the encoded genes has been defined. There are few other genetic markers known that characterize cells in early and intermediate stages of murine development; for example, no gene has been associated with the initial process of segmentation which occurs at 8dg in the mouse [Rossant and Joyner, 1989].

Many of the known murine homeobox genes encode multiple transcripts during embryogenesis [Odenwald et al., 1987]. Tissue-specific differences are seen in the sizes of transcripts produced by respective homeobox genes such as Hox-3.2 and Hox-2.6. It is possible that alternative splicing plays an important role in tissue-specific regulation of many of the vertebrate Hox genes [Graham et al., 1989].

Most murine homeobox genes can be induced during differentiation of F9 EC cells by retinoic acid. Furthermore, the genes of the human Hox-2 cluster are differentially activated by retinoic acid in the EC line NT2/D1, depending on their position in the cluster [Simeone et al., 1990], with genes at the 5′ end requiring much higher levels of retinoic acid (up to 10⁻⁵M) for activation than genes in the 3′ part of the cluster (10⁻⁸M). These results are remarkably similar to the colinear patterns detected in the mouse embryo. An analogous result was achieved in a similar study using Xenopus embryos treated with retinoic acid [Papalopulu et al., 1990, 1991]. Although a series of comparable experiments has not been performed in F9 EC cells, it has been shown that the Hox-3.2 gene lying in the 5′ region of the cluster is not induced during differentiation of F9 cells, whereas the next 3′ gene, Hox-3.1, is activated by retinoic acid (5 × 10⁻⁷M) in combination with 10⁻³M cAMP [Breier et al., 1986].

The colinearity of response to retinoic acid and gene order of Hox-2 genes could arise from a graded morphogen signal in the embryo. This morphogen could then induce a differential homeobox gene response. As has been cited, retinoic acid stimulates expression of many of the homeobox genes in teratocarcinoma cells. Experimental evidence in chicken embryos suggests that retinoic acid is important in patterning in the limb [Brockes, 1989]. Therefore, the differential response of Hox-2 genes to retinoic acid could be one of the means that is used during embryogenesis to set up a gradient of homeobox expression and ordered, partially overlapping domains of expression.

The patterns of expression of homeobox domain genes suggest that they play a role in positional specification in vertebrate development, particularly in axis specification, where they may establish an “address” or identity through an overlapping “combinatorial code” [Sham et al., 1991]. In the absence of naturally occurring mouse mutants, homologous recombination and gene targeting will presumably address the problems of assigning function to each of these genes and determining functional redundancy where homologies between different Hox clusters is apparent. Indeed such studies have already begun with targeted disruption of the mouse homeobox gene Hox-1.5 [Chisaka and Capecchi, 1991], which has led to the formation of homozygous mice that exhibit reduced thyroid and submaxillary tissue and a wide range of throat abnormalities, but no apparent failure of peri-implantation embryogenesis.

E. POU/Oct Genes

The octamer motif (ATGCAAAT) is a cis-active promoter element required for both ubiquitous and cell-specific type expression of certain histone and immunoglobulin light and heavy chain genes. The obvious paradox in the same element being required for two different types of expression was explained by the result that two different proteins were able to interact with this sequence in the different cell types tested [Kemler and Schaffner, 1990]. Additional proteins that bind directly to this octamer motif have now been identified at specific developmental stages and in various model systems of development. These proteins all belong to the POU family, as defined by the sequence homology with three mammalian transcriptional factors and one nematode regulatory protein (Pit-1/GHF-1, Oct-1, Oct-2, and Unc-86) [Herr et al., 1988].

POU proteins have two conserved domains: a domain specified by the POU-specific domain and a homeodomain that is distantly related to the prototype Antennapedia homeodomain and nearby on the amino-terminal side. Outside of these two domains the amino acid sequences are highly divergent and contain sequences required for transcriptional activation. The POU domain is 81 amino acids long and contains two subdomains of high sequence homology. Each of these subdomains shares two features: a cluster of basic amino acids in the center and a predicted α-helix at the carboxy-terminal end. The “recognition helix” or the most conserved part of the POU domain contains 11 invariant residues. The RVFCN motif within this recognition helix has a cysteine residue at position 9 that is specific for the POU family, whereas the other residues are extremely well conserved in all homeodomain proteins.

The homeodomain found in POU proteins is 60 amino acids in length. It contains a DNA-binding domain as well as three well-defined helices, two of which form helix–turn–helix motifs. An intact homeodomain is required for DNA binding of all POU proteins, whereas the contribution of the POU-specific domain varies, depending on the DNA-binding site and on the identity of the POU protein.

Comparison of all POU homeodomain and POU subdomain sequences divides the POU family into five classes. This classification is based on the similarity of the variable linker sequences that flank the POU-specific domain and the POU homeodomain. This linker area contains 14–26 amino acids.

Most POU genes are differentially expressed throughout embryogenesis (see Table I). However, unlike the Hox and the Pax genes, their expression patterns are not regular and not explainable with any sort of unifying hypothesis.

TABLE I.

Summary Of Octamer Proteins

OCT protein*	Embryonic expression	Adult expression
OCT-1	Ubiquitous	Ubiquitous
OCT-2	Neural tube, entire brain except telencephalon	Lymphoid cells, nervous system, intestine, testis, kidney
N-OCT-2	Nervous system	Nervous system: astrocytes, glioblastoma, and neuroblastoma cell lines
MiniOCT-2	Nervous system, developing nasal neuroepithelium	Nervous system, primary spermatids
N-OCT-3	Nervous system	Nervous system, glioblastoma, and neuroblastoma cell lines
OCT-4A, OCT-4B, OCT-5	Totipotent stem cells: pregastrulation embryonic ectoderm, primordial germ cells, testis, ovaries	Oocytes
OCT-6	Blastocyst, ES and EC cells, brain	Nervous system, testis
OCT-7, 8	Nervous system	Nervous system

^*OCT-9 and OCT-10 proteins have also been isolated, but their expression patterns are presently unknown.

Two of the POU genes are expressed in the early embryo, Oct-4 (also called Oct-3 or NF-A3) and Oct-6. Oct-4 expression seems to be correlated with an undifferentiated or stem cell population. It is first detectable in the totipotent and pluripotent stem cells of the pregastrula stage embryo, is subsequently downregulated during differentiation of these cells, and eventually becomes confined to the germ cell lineage [Okamoto et al., 1990]. In ES and EC cells, both Oct-4 and Oct-6 are downregulated when the cells are induced to differentiate with retinoic acid. Oct-6 is also expressed later in development in specific neurons of the developing and adult brain and also in testis. SCIP/tst-1 is the rat homologue of Oct-6, which is expressed transiently during the period of rapid cell division separating the premyelinating and myelinating phases of Schwann cell differentiation. This gene may therefore play a role in the progressive determination of Schwann cells [Monuki et al., 1989, 1990].

Although most murine developmental control genes that have been identified by homology with Drosophila genes are expressed at postimplantation stages, several of the Oct genes are expressed at pre-implantation stages. The analysis of POU genes in model systems of early embryo ES and EC cells may indicate their putative roles in early development. Activation and repression via the octamer motif have been documented in EC cells, depending on the regulatory elements present and on the amount of Oct factors present [Lenardo et al., 1989; Scholer et al., 1989]. Several developmental control genes contain octamer motifs in their promoters and are developmentally expressed after Oct-4 and Oct-6 (e.g., Hox-1.3), and these may be acted upon by the octamer genes [Scholer et al., 1991].

In summary, based on the present evidence, POU domain genes seem to be involved in cell proliferation. Some POU genes are clearly more highly expressed during increased cellular proliferation, suggesting a possible role in replication (e.g., SCIP (tst-1/Oct-6). The importance of Oct-4 (Oct-3) in particular at early embryonic stages has been demonstrated by injection of antisense Oct-4 oligonucleotides into fertilized oocytes, resulting in an inhibition of DNA synthesis and arrest of the embryo at the one-cell stage [Rosner et al., 1991]. Future studies using this technique and others will elucidate the role of these interesting gene products.

IV. DISCUSSION

It is apparent from this perspective on the cell lineage relationships during early mouse embryogenesis that epigenetic interactions play a substantial role in peri-implantation mammalian development. We summarize below the time, place, and direction of these interactions; we also consider the nature of disturbances to these epigenetic interactions that may be involved in the aberrant development of model cell systems and as a result of imprinting. Finally, we evaluate the possible roles of specific peptide growth factors and other agents in the epigenetic interactions that occur during early mouse embryogenesis.

A. Sites of Epigenetic Interactions in Normal Development

The cell lineage relationships described above provide evidence for several distinct epigenetic interactive processes in peri-implantation mouse embryogenesis (Fig. 1): 1) outer cells at the morula stage interact with the inner cells, suppressing their tendencies for trophectoderm differentiation or preventing exposure to the outside environment; 2) ICM cells interact with the overlying polar TE of the early blastocyst, suppressing their tendencies for mural TE differentiation and sustaining the proliferation of diploid cells of the trophectoderm lineage; 3) newly differentiated primitive endoderm cells of the late blastocyst interact with the remaining core cells of the inner cell mass (primitive ectoderm), suppressing their tendencies for primitive endoderm differentiation or preventing exposure to the blastocele environment; 4) trophoblast giant cells of the late blastocyst and later stages interact with adjacent cells of the primitive endoderm lineage, inducing their differentiation as parietal endoderm; 5) visceral embryonic endoderm cells (primitive endoderm) of the early egg cylinder interact with the adjacent epiblast (primitive ectoderm) layer, inducing their differentiation into mesoderm, leading to formation of the primitive streak; 6) extraembryonic ectoderm of the early egg cylinder interacts with the adjacent visceral extraembryonic endoderm, suppressing the synthesis of AFP; and 7) visceral extraembryonic endoderm of the late egg cylinder interacts with the adjacent extraembryonic mesoderm, inducing the differentiation of blood islands. There are undoubtedly other epigenetic interactions between other early cell lineages of the mouse embryo, and some of these interacting tissues may have bidirectional signalling. However, these examples are sufficient to establish the importance of such signaling mechanisms even at the earliest stages of mammalian development.

The evidence for such epigenetic interactions in normal mouse embryogenesis may provide clues for understanding the etiology of the developmental perturbations observed in model cell systems. The earliest stages of mouse embryogenesis are not applicable to EC and ES differentiation in vitro, because neither of these cell systems generates trophectoderm in isolation. However, ES cells transferred to mouse blastocysts were able to contribute to trophectoderm and primitive endoderm populations at a low frequency [Beddington and Robertson, 1989], showing that some cells were capable of responding to signals for such differentiative events. Although EC cells do not spontaneously differentiate to form primitive endoderm cell types, these are induced by exposure to retinoic acid. The tendency for retinoic acid–induced EC cells to differentiate into parietal endoderm after exposure to cyclic AMP indicates that they are capable of responding to their molecular signals and suggests that such signals could act through the protein kinase A signal transduction pathway. The capacity that is conspicuously absent from most EC cells is mesodermal differentiation, and neither EC nor ES cells establish an embryonic axis by primitive streak formation at gastrulation despite their capacity for expression of cardiac muscle-specific genes [Robbins et al., 1990; Sanchez et al., 1991]. This may reflect a deficiency in signals responsible for mesoderm induction, which (by analogy with chick and amphibian embryos) probably arise from primitive endoderm cells at the posterior of the prospective embryonic axis. The capacity for EC and ES cells to participate in axial development when transferred to the blastocyst microenvironment indicates that their capacity to respond to such signals is intact. In mice, the source of these signals for axial development may originate in the relationship between the primitive ectoderm and the extraembryonic lineages trophectoderm and primitive endoderm, which are absent from embryoid bodies. Moreover, the site of primitive streak formation in the mouse embryo in utero appears to reflect either the contact point of the trophectoderm with the uterine wall, or a precocious differentiation of trophoblast that presages anteroposterior axis formation and determines the orientation of implantation [Smith, 1985].

Another feature in the development of embryoid bodies is their failure to form well-differentiated blood vessels, despite the capacity of ES cells to differentiate into hematopoietic cells [Doetschman et al., 1985; Wiles and Keller, 1991; Lindenbaum and Grosveld, 1990; Wang et al., 1992]. This may also reflect a deficiency in signals involved in angiogenesis [Klagsbrun and D’Amore, 1991].

The abnormalities of mouse isoparental embryos (parthenogenotes and androgenotes) that arise from genomic imprinting may also be partially accounted for as deficiencies in epigenetic interactions. Both parthenogenotes and androgenotes are capable of forming morphologically normal preimplantation embryos and thus appear competent in generating signals required for early lineage differentiation. However, the morphological abnormalities of parthenogenotes may indicate deficiencies in the epigenetic signals that regulate polar trophectoderm differentiation, since parthenogenotes form an excess of giant cells and retain few diploid cells in the trophoblast lineage. Similarly, there may be deficiencies in epigenetic mechanisms regulating the differentiation of primitive endoderm, because parthenogenotes form an excess of parietal endoderm [Sturm et al., 1992; reviewed by Pedersen et al., 1992]. While parthogenotes were capable of forming both embryonic and extraembryonic mesoderm, they often lacked an embryonic axis, suggesting eviatic function of the epigenetic processes leading to axial mesoderm induction. On the other hand, the more advanced parthenogenotes were fully capable of forming blood islands, indicating that they could generate any epigenetic signals necessary for this process.

B. Role of Specific Factors in Epigenetic Interactions

A wide variety of scenarios can be envisioned for the inductive interactions between the various cell lineages of mammalian embryos. General properties of epigenetic signalling are the synthesis of ligand(s) by an inducing cell population and the reception of such signals by the responding cell population. The ligand may be secreted as a diffusible factor or may be retained in a membrane-bound form. The capacity for numerous ligands to interact with extracellular matrix materials provides additional possibilities of reservoirs, barriers, or concentration of such factors and blurs the distinction between diffusible and static factors [see Jessell and Melton, 1992, for review]. Because many of the potential ligands known to be synthesized in early embryos share these properties and because other ligands may well be produced by the interacting tissues, it is probably not fruitful to speculate at this point on the identity of ligands responsible for each of the epigenetic interactions that characterize early mouse development. However, further consideration of the effects of certain ligands or receptors allows us to exclude a role for the epigenetic interactions during peri-implantation mouse embryo lineage differentiation. The relevance of specific peptide growth factors and receptors to mouse development can be evaluated on the basis of 1) direct evidence for a function or, more commonly, 2) circumstantial evidence indicating the synthesis of the transport or protein in appropriate tissues and stages of development or a biological response to exogenously added agent.

We can conclude on the basis of direct functional evidence that certain genes are essential for early events of lineage differentiation and fate in mouse embryos. The role of stem cell factor [Steel locus; Copeland et al., 1990; Zsebo et al., 1990; Huang et al., 1990a], c-Kit [W locus; Chabot et al., 1988; Geissler et al., 1988], PDGF-alpha [Patch locus; Mercola et al., 1990], IGF-II [De Chiara et al., 1990] and Wnt-1 [McMahon and Bradley, 1990; Thomas and Capecchi, 1990] have been examined experimentally by gene targeting using homologous recombination in ES cells or by analysis of existing mutations. In every case, the phenotype of the homozygous mutant shows effects later in fetal development rather than at peri-implantation stages. A possible explanation for the lack of an effect of these mutations is that there are other redundant functions that mask their consequences. Such explanations must be invoked to account for the lack of effect of these mutations in tissues that express the gene in order to reconcile this with the circumstantial evidence for a developmental role. An example of this situation is stem cell factor, which is strongly transcribed in the embryonic endoderm at 7.5 dg and the ectoplacental cone at 9.0 dg [Motro et al., 1991], but shows no mutational effects until germ cell and hematopoietic development begins.

The stage and tissue-specific expressions of peptide growth factors and cytokines and their receptors have been interpreted as circumstantial evidence for a developmental role in mammals. Genes in this category include c-Fms, Egf, Fgfs, Igf-I, Lif, Pdgf-A, Tgf-α, and Tgf-β. The effect of exogenous ligand on in vitro embryonic development provides another line of indirect evidence for developmentally active epigenetic signals. Genes in this category include b-Fgf, insulin, Pdgf-A, and others. A thorough evaluation of the epigenetic role of each of these products will require analysis by antisense oligonucleotides, dominant negative transgenics, function-perturbing antibodies, gene targeting, or other novel methods for disrupting the function of the endogenous gene and its products.

A similar evaluation of specific transcription factors, including Hox genes and POU genes, suggests on the basis of their transcriptional patterns that they function during early mouse development, but there is no decisive evidence for their role in lineage differentiation and fate. We conclude that, despite the evidence for numerous epigenetic processes in early mammalian development, there is no documented role of known signalling factors in the emergence of the tissues that comprise the midgestation mouse conceptus. The most extensive analysis of epigenetic processes during early vertebrate development have been carried out using amphibian embryos. While these studies have implicated members of the Tgf-β and Wnt families as inducers of mesoderm, only b-Fgf has been shown to have a role by disrupting the endogenous gene product, leading to abnormalities in axial development [Amaya et al., 1991; Jessell and Melton, 1992; see Slack and Tannahill, 1992, for review]. Thus there is still no consensus on the identity of the factors responsible for vertebrate epigenetic interactions. Because of their unique strategy of devoting early stages to differentiation of extraembryonic membranes, mammalian embryos may have expression patterns and roles for these genes that are distinct from those found in amphibian embryos. Further work is therefore needed to determine which, if any, of the gene products known to be synthesized in early mammalian embryos generate the epigenetic phenomena summarized here.

Biographies

JEAN J. LATIMER is a postdoctoral fellow in the laboratory of Roger Pedersen of the Laboratory of Radiobiology and Environmental Health at the University of California, San Francisco. Dr. Latimer received her B.A. in Cell Biology from Cornell University in 1982, and her Ph.D. in Molecular and Cellular Biology from the State University of New York at Buffalo for her work with Heinz Baumann at the Roswell Park Cancer Institute. Her thesis involved characterization of the molecular evolution of the α₁-antitrypsin gene in the mouse genus, particularly the unusually abundant renal expression of this gene in the Asian wild mouse species Mus caroli. Her current work in the field of developmental biology is focused on the establishment and utilization of embryonic stem (ES) cell lines as models of early mouse development. Besides establishing such cell lines herself, Dr. Latimer has used ES cells to define molecular markers of early differentiation and to isolate pure populations of differentiated cell types via fluorescence-activated cell sorting. She is currently using these cells to create mouse models of human DNA repair-deficiency syndromes using the technique of targeted homologous recombination, beginning with ERCC1, a gene implicated in repair of lesions induced in DNA by ultraviolet light.

ROGER A. PEDERSEN is Professor of Radiology and Anatomy at the University of California at San Francisco, where he teaches developmental biology and mammalian embryology. He received his B.A. degree from Stanford University in 1965 and his Ph.D. under Clement Markert at Yale University in 1970. He pursued postdoctoral research with John Biggers at the Johns Hopkins University, beginning his studies using the mouse embryo as a model system for mammalian embryonic development. Since 1971, Dr. Pedersen has been a member of the Laboratory of Radiobiology and Environmental Health at the University of California. His research has involved mechanisms of embryotoxicity and repair in early mammalian development, analysis of cell potency and fate in pre- and postimplantation mouse embryos, and mechanisms of genomic imprinting in mammals. He has written numerous reviews on early mouse development, and co-produced an instrumental videotape on the use of mice in transgenic research. Since 1991 he has served as Series Editor of Current Topics in Developmental Biology. His research papers have appeared in Development, Developmental Biology, Genes and Development, Nature, Science, and other journals.