|Home | About | Journals | Submit | Contact Us | Français|
Evidence from mechanical, teratological, and genetic experimentation demonstrates that holoprosencephaly (HPE) typically results from insult prior to the time that neural tube closure is completed and occurs as a consequence of direct or indirect insult to the rostral prechordal cells that induce the forebrain or insult to the median forebrain tissue, itself. Here, we provide an overview of normal embryonic morphogenesis during the critical window for HPE induction, focusing on the morphology and positional relationship of the developing brain and subjacent prechordal plate and prechordal mesoderm cell populations. Subsequent morphogenesis of the HPE spectrum is then examined in selected teratogenesis mouse models. The temporal profile of Sonic Hedgehog expression in rostral embryonic cell populations and evidence for direct or indirect perturbation of the Hedgehog pathway by teratogenic agents in the genesis of HPE is highlighted. Emerging opportunities based on recent insights and new techniques to further characterize the mechanisms and pathogenesis of HPE are discussed.
While the genesis of holoprosencephaly (HPE) has been studied in a number of mammalian and non-mammalian species (e.g. amphibians [Holtfreter and Hamburger, 1955], fish [Blader and Strahle, 1998], chick [Cordero et al., 2004], rat, sheep [Keeler, 1975]), recent investigations employing mice have been particularly informative. This is true for both genetic and environmental etiologies, the former of which has recently been reviewed by Schacter and Krauss , and Geng and Oliver . Herein, the focus is on the development of teratogen-induced HPE in mice.
As opposed to most gene manipulation-based investigations, teratogenesis studies afford the opportunity to examine the result of insult occurring within a narrow window of development. In mice, the critical period for the induction of HPE is between gestational days (GD) 7 - 8.5; i.e. during gastrulation and neurulation stages. This corresponds to human development between the middle of the 3rd week to early in the 4th week post-fertilization. At these early embryonic stages, humans and mice are remarkably similar and interspecies extrapolations regarding molecular and cellular mechanisms appear well founded.
It has long been recognized that the cell populations that are initially affected to result in the median forebrain defects characteristic of HPE are those that normally lie rostral to the notochord; i.e., they are prechordal and include those endodermal and mesodermal cells that are subjacent to the rostral-most aspect of the neural plate and/or the neuroepithelial cells of the prospective forebrain, itself. As early as the 1930s, Adelman  reported cyclopia resulting from mechanical removal of the prechordal mesoderm in amphibian embryos. Holtfreter and Hamburger, in , showed that this same endpoint could result from teratogen-mediated interference with gastrulation, the developmental process that results in formation of the mesoderm and definitive endoderm of the embryo. More recently, the molecular signaling events required for the development of the prechordal tissues, including those associated with forebrain induction and subsequent early forebrain development have been elucidated [Gunhaga et al., 2000; Geng et al., 2008; Aoto et al., 2009; Maurus and Harris, 2009]. Comparable to insult by genetic manipulation, a variety of teratogens that cause HPE (though probably not all) appear to directly or indirectly affect hedgehog (Hh) signal transduction, involving either Hh generating or receiving cell populations.
Figures 1--4,4, which are comprised of light and scanning electron micrographs, are designed to provide a framework for understanding developmental events/tissues that, when disrupted, can result in HPE. A majority of the images are of mouse embryos, but a few micrographs of chicken embryos are included to provide additional pertinent information. Focus is on the cellular relationships and morphology of prechordal cell populations; i.e. cells that are located rostral to the notochord. In a recent review, Müller and O'Rahilly  point out difficulties/discrepancies in the terminology that is used to describe the cell populations that are subjacent to the developing rostral neural plate. These authors suggest abandoning the commonly (and, in many cases, interchangeably) -used term, “prochordal plate” in favor of the term “prechordal plate”. They define the prechordal plate as being a temporary cellular plaque that is situated rostral to the notochordal process at Carnegie stage 7 & 8 in the human (days 16-18 post-fertilization) and that comes to underlie the rostral part of the neural plate by Carnegie stage 9 (20-21 days; the time when the first pairs of somites are visible). While some confusion surrounds the origin of the prechordal plate cells, they are generally described as arising from the epiblast and as being endodermal. This is in contrast to the notochord, which at early stages of development is comprised of a cell layer that is laterally continuous with the endoderm, but is considered mesodermal [Sulik et al., 1994]. As illustrated in Figure 1, which shows Hamburger and Hamilton (H&H) Stage 7 and 8 chick embryos (stages when 1-4 somite pairs are present), the rostral-most tissue underlying the forebrain, i.e. the prechordal plate, is composed of an organized tall columnar cell population that is continuous with the foregut endoderm. Between the prechordal plate and the developing notochord is a mesenchymal (fibroblastic/non-epithelial) axial population of cells that is considered prechordal mesoderm. These prechordal mesoderm cells are located subjacent to the developing midbrain, while a more condensed axial population that forms the notochord is subjacent to the hindbrain (Fig. 1b). Studies employing chicks have shown that just prior to the stage shown in Figure 1, expression of molecular markers characteristic of ventral telencephalic cells are specified in response to Shh signals derived from the region of the node and (possibly) the prechordal plate region [Gunhaga et al., 2000].
Regarding development of prechordal and notochordal cell populations in mice, Sulik et al,  showed that on the ventral surface of GD 7 embryos (Theiler Stage 10, corresponding to approximately 16-18 days post-fertilization in humans, when gastrulation is initiated and the first mesodermal cells can be identified), the anterior midline is identifiable by the presence of small clusters of mono-ciliated cells (Fig. 2 a, b). As development progresses, a larger grouping of these cells are present near the apex of the cup-shaped embryo and comprise the notochordal plate. The notochordal plate extends rostrally from the node (Fig. 2 c, d), the latter of which is at the anterior-most aspect of the primitive streak (Fig. 2d, i). By GD 7.5 (Theiler stage 11; presomite; corresponding to approximately 19-20 days post-fertilization in humans) the prechordal plate is readily identified as a circle of mono-ciliated cells that are subjacent to the rostral-most portion of the anterior neural plate (Fig. 2e, g). Caudal to the prechordal plate and extending to the position of the node, a line of mono-ciliated cells is intercalated into the ventral cell layer in the midline (Fig. 2e, g). Shh is expressed in a temporally-dependent manner in these axial populations [Echelard et al., 1993; Gunhaga et al., 2000]. Notably, the prechordal plate can be identified morphologically prior to the time in development when in situ hybridization indicates Shh gene expression. An early genetic marker of the prechordal plate is gooscoid (gsc) [Filosa et al., 1997; Aoto et al., 2009] (Fig. 2f).
Figure 2h shows a dorsal view of a GD 7.5 mouse embryo from which the amnion has been removed to allow visualization of the neural plate. At this stage in development, the neural plate comprises nearly the entire anterior half of the embryo, with the prospective forebrain (prosencephalon) being at the rostral rim. Examining a midsagittal cut through an embryo of this stage allows appreciation of the relationships of the ventro-axial cell populations to the developing brain (Fig. 2i). Additionally informative is removal of ventral cell populations (endoderm, including the prechordal plate), which allows visualization of the mesenchyme (i.e., a non- epithelial, fibroblastic cell population) that is ventral to the developing brain. These mesenchymal cells are considered to be mesodermal (i.e. “of the middle germ layer”) and are derived, in part, from the primitive streak via gastrulation [Tam et al., 1993]. Some authors indicate that additionally, the prechordal plate is a source of cranial mesenchyme. For example, Müller and O'Rahilly  state that the human prechordal plate transitions from a solid median structure to laterally migrating cells that form bilateral premandibular condensations.
Between the end of the 7th to the middle of the 8th day after fertilization in the mouse (Theiler stage 12; corresponding to approximately 20-22 days post-fertilization in humans) the first 7 pairs of somites form and the optic sulci that develop within the diencephalic portion of the forebrain become clearly distinguishable (Fig. 3). As previously shown in comparably-staged chick embryos (Fig. 1), the prechordal plate in early somite-stage mice presents as a compact mass of cells that is subjacent to the median aspect of the forebrain. This is readily seen in Figure 3b, which is an image of a mouse embryo from which a portion of the cranial neuroepithelium has been removed. Also illustrated is the association of the prechordal plate with a more laterally positioned mesenchymal cell population. At the late end of this stage of development, the prechordal plate is subjacent to the median diencephalic floor (Fig. 3c), with the developing eyes (optic sulci) positioned more laterally. At this time, the telencephalic portion of the forebrain (that part from which the cerebral hemispheres arise) is a very small component of the developing brain, making up only its rostral-most rim (Fig. 3c).
By the time that mouse embryos have approximately 8 somite pairs (Theiler stage 13; gestational day 8.5), in situ hybridization studies show that Shh expression is initiated in the neuroepithelium [Echelard et al., 1993]. It is first expressed in the ventral midline at the level of the midbrain, extending into the diencephalic portion of the forebrain and caudally into the hindbrain, such that by the time 15 somite pairs are present (approximately 12 hours later), expression is continuous in the ventral midline from the rostral limit of the diencephalon to the presumptive spinal cord.
Figure 4 includes images of both chick (a-c) and mouse embryos (d-i) at developmental stages that correspond to those occurring in humans during approximately the 24th to 26th days post-fertilization. During this time, the neural folds fuse, with the rostral closure site occurring at the junction of the telencephalic and diencephalic portions of the prosencephalon. Relative to the diencephalon, the telencephalon remains relatively small. In chick embryos, near the time of rostral neural tube closure (H&H 10-11), the prechordal plate is positioned at the apex of the foregut pocket, remaining continuous with the foregut endoderm and in close proximity to the floor of the diencephalon (Fig. 4a-c). Also in close proximity to the prechordal plate is the thickened surface ectoderm of Rathke's pouch, the progenitor of the anterior pituitary gland (Fig. 4c). By this time, the notochord has separated from the underlying endoderm, appearing as a median cellular cord that extends rostrally to the prechordal mesoderm. This morphology is very similar to that in comparably-staged mouse embryos (Theiler stage 14; corresponding to human embryos at approximately 24 days post-fertilization and occurring at 9 days of development in mice; Fig. 4d-f). By the time that mouse embryos reach Theiler stage 15 (9.5 days post-fertilization; corresponding to humans at approximately 26 prenatal days), 21-29 somite pairs are present, the oropharyngeal membrane has broken down and no longer separates the oral and foregut cavities, and the prechordal plate cannot be identified (Fig. 4g-i). In the midline, mesenchymal cells fill the space between the mesencephalic flexure and the roof of the foregut, and there is close apposition between the ectoderm of Rathke's pouch and the floor of the diencephalon.
A major aim of this overview is to illustrate the morphogenesis of a wide range of teratogen-induced brain and facial abnormalities that comprise the HPE spectrum. As described in a recent review by Cohen and Shiota , HPE has been experimentally produced in a wide variety of non-mammalian and mammalian models by a number of different teratogenic agents. Dysmorphogenesis resulting from exposure of mouse embryos to a selected group of these teratogens during the critical period for HPE induction (GD 7-8.5) is described.
Among the earliest reports of experimentally-induced HPE were those in which ethanol was administered to chicks [Féré, 1899], frogs [LePlat, 1913] and fish [Stockard, 1910]. More recently, the genesis of a HPE spectrum resulting from exposure of mice to this teratogen has also been described [Sulik and Johnston, 1982; Sulik and Johnston, 1983; Webster et al., 1983; Sulik, 1984; Schambra et al., 1990; Godin et al., 2010]. In the mouse, treatment of dams with high doses of ethanol administered as 2 doses given 4 hours apart early on the 7th gestational day (when embryos are at the stages shown in Fig. 2a-d) results in the HPE phenotypes shown in Figures 5 and and6.6. In normal embryos having 29 somite pairs (late on GD 9), the olfactory placodes/developing nasal pits are evident and widely spaced on the frontonasal prominence (the tissue surrounding the telencephalon) (Fig. 5a). In ethanol-exposed embryos at this same developmental stage, the olfactory placodes are too closely spaced, accompanied by narrowing of the frontonasal prominence (Fig. 5 d, g). On GD 11, a time when the nasal pits should be surrounded by well developed lateral and medial nasal prominences (the progenitors of the nasal alae, and the nasal tip along with the intermaxillary segment, respectively), affected embryos present with medial nasal prominences that are abnormally small and closely approximated to the point of appearing as a single band of median tissue between nostrils that are too closely spaced (Fig. 5b, e, h). Deficiency in the median aspect of the forebrain accompanies that of the upper midface and is evidenced by abnormally close proximity (and in severe cases, median confluence) of the ganglionic eminences (Fig. 5f, j). Figure 6 illustrates a spectrum of the ethanol-induced facial phenotypes as they appear on GD 14 in mice. Mildly affected animals have facial features consistent with those in fetal alcohol syndrome, i.e. a small nose, deficient philtrum, and a long upper lip, all of which are indicative of median tissue deficiencies (Fig. 6a, b). More severely affected mice may present with a single nostril, as in human cebocepahly (Fig. 6c, d), or with a median cleft lip associated with an absent intermaxillary segment (Fig. 6 e, f). Notably, in teratogen-exposed mice, even in inbred animals, widely ranging degrees of effect commonly occur within single litters. This is attributable, at least in part, to the fact that there is significant intralitter variation in developmental stages [see Sulik and Johnston  for details regarding the interrelationship of the developing face and brain in ethanol-induced HPE in the mouse].
While cyclopia/synophthalmia have not been noted to result from ethanol treatment on GD 7 in mice, mild to severe microphthalmia or anophthamia occurs. While the eyes and forebrain remain sensitive to ethanol's teratogenicity when exposure occurs at later stages (extending through GD 9) [Kotch and Sulik, 1992; Parnell et al., 2009], HPE is an endpoint that appears to be specific to ethanol exposures that occur when gastrulation is just being initiated (GD 7 in the mouse). This temporal specificity for the induction of HPE with ethanol in mice is consistent with that for other species. Studies employing zebrafish as conducted by Blader and Strahle  are particularly informative with respect to identification of a narrow window of vulnerability. These investigators showed that exposure of zebrafish embryos to ethanol over a time period as brief as 3 hours, encompassing late blastula and early gastrula stages, caused severe HPE. In the fish embryos, synophthalmia was a common endpoint.
Blader and Strahl's work has also shed light on the mechanisms involved in HPE induction. Using gooscoid as a molecular marker for the prechordal plate (defined as that tissue that is derived from axial hypoblast cells that involute/ingress at the onset of gastrulation and migrate anteriorly ahead of the chordal mesoderm), they illustrated an abnormal caudal placement of this tissue and suggested that ethanol arrests its migration. Thus, absence of inducing tissue for the forebrain was considered to be the primary cause for the ethanol-induced HPE. Others have shown that in zebrafish, shh mRNA injection can prevent ethanol-induced cyclopia, suggesting that abrogation of Hedgehog (Hh) signaling is one of the major effects of ethanol exposure [Loucks and Ahlgren, 2009]. The specific cellular mechanism(s) by which ethanol perturbs this signaling remains to be definitively established as does the initial cell population and developmental event involved.
Another teratogenic agent that, when acutely administered on GD 7 in mice causes HPE is retinoic acid [Sulik et al., 1995]. As with ethanol, the result of exposure to this agent is a wide range of severity within the HPE spectrum. As shown in Figure 7, in addition to median upper face deficiencies, some of which are severe enough to yield proboscis formation (union of the lateral nasal prominences with absence of intermediate tissues), the developing lower jaw may be hypoplastic. Lower jaw deficiencies can also result from ethanol exposure on GD 7 [Godin et al., 2010], but are not illustrated herein. Typically, the maxillary component of the developing upper jaw appears to be much less affected than the mandibular region. That the mesenchyme of the maxillary prominences is almost entirely derived from neural crest cells, while that of the mandibular prominences has a large mesodermal contribution may account for this differential effect. The neural crest cells that contribute to the maxillary and mandibular prominences begin to migrate into those regions from the mid and upper hindbrain levels of the neural folds on GD 8 in the mouse, while the mandibular mesoderm is laid down at very early gastrulation stages; i.e. at or very near the GD 7 time of insult. Figure 7e illustrates an extreme example of loss of prechordal tissues. In this retinoic acid-exposed mouse fetus, the entire forebrain and associated facial tissues, along with the mandibular prominences are absent. However, the mid and hindbrain, along with the maxillary prominences are preserved.
Acute exposure to other teratogens on GD 8-8.5, when mice are at very early somite stages, also can yield defects in the HPE spectrum [Wei and Sulik, 1993; Lipinski et al., 2008]. Figure 8 illustrates defects caused by acute maternal treatment on GD 8 with Ochratoxin A (OA), a food-born mycotoxin. Some of the affected animals present with severe microphthalmia, while in others the eyes are less severely diminished in size, but very closely approximated, to the point of union (synophthalmia). In each of the specimens shown, virtually all of the medial and lateral nasal prominence tissues are absent. This is evidenced by median union of the maxillary prominences, producing a snout (not a proboscis) that is surrounded by rows of hair follicles. In some specimens with this facial morphology there was also failure of anterior neural tube closure, resulting in anencephaly (not shown). Study of the OA-mediated pathogenesis illustrated that within 6 hours of maternal drug treatment, a remarkable amount of cell death occurred in the presumptive telencephalon and forebrain floor [Wei and Sulik, 1993]. Excessive cell death in selected cell populations/regions was also notable at 24 hours after maternal treatment. The frontonasal prominence was particularly affected.
OA has also been employed is gene/environment interaction studies. Ohta et al  have shown that in mice carrying a mutation in Gli3 (a downstream gene in the Hh signaling cascade) which results in HPE, polysyndactyly, and neural tube defects, OA treatment on GD 7.5 increases the incidence of abnormalities. Additionally this group showed that in these mice, maternal folinic acid administration at times surrounding the OA treatment reduces the incidence of teratogen-induced defects.
Also studying gene/environment interactions, Lanoue et al  showed that in mutant mice in which cholesterol levels are low due to mutation of the apolioprotein b gene, further reduction resulting from treatment with a cholesterol biosynthesis inhibitor induces HPE. In some individuals the forebrain and facial defects are accompanied by hindbrain and limb reduction abnormalities (Fig. 9) and in others the cranial neural tube fails to close (Fig. 10). In this model, the drug-induced hypocholesterolemia was accomplished by treating dams on their 4th through 7th day of pregnancy with the biosynthesis inhibitor. That the hindbrain and limbs (whose critical periods are after GD 7) were affected indicates that cholesterol remained reduced to teratogenic levels for some time beyond GD 7. Recognizing the requirement of cholesterol for normal Hedgehog signaling [Incardona and Roelink, 2000] Lanoue et al  pointed out the similarities between the defects observed in their model and in Shh knockouts [Chiang et al., 1996]. More recently, Li et al , have shown that in zebrafish, supplementing ethanol-exposed embryos with cholesterol rescues a loss of Shh signal transduction, and prevents embryos from developing HPE.
More directly examining teratogen-mediated insult to the Hh signaling pathway in mice, Lipinski et al [2008 and submitted] have employed cyclopamine and a potent synthetic analogue. These compounds inhibit the morphogenetic activity of the Hh pathway by binding to and preventing activation of the transmembrane protein, Smoothened (Smo) [Chen et al., 2002]. In the absence of Hh ligand, its receptor, Patched (Ptc1) inhibits Smo activity, presumptively through a small molecule mediator [Taipale et al., 2002; Bijlsma et al., 2006]. Upon Hh binding to Ptc1, inhibition of Smo is relieved, triggering a complex downstream signaling cascade that culminates in target gene activation via the Gli family of transcription factors [Ingham and McMahon, 2001]. As in the classic studies in which Binns and Keeler exposed sheep to cyclopamine [reviewed in Keeler, 1975] this agent and the analogue also cause the HPE spectrum in mice. In the Lipinski et al studies, maternal drug treatment was initiated on GD 8.25 - 8.5; i.e. close to the end of the known critical period for HPE induction. In addition to the range of defects described herein for other teratogens (e.g. cebocephaly; Fig. 11a), unilateral and bilateral clefts of the lip as well as clefts of the secondary palate were induced (Fig. 11b, c). In some of the mice with cleft lip, the intermaxillary segment is notably reduced in size and in others it appears normal. These phenotypes are also seen in clinical populations. In individuals in which intermaxillary segment reduction is recognized, median forebrain deficiencies are expected. However, in those that have what appears to be typical unilateral or bilateral cleft lip and palate, this is not the case. Noteworthy in this regard, is that histological sections of the cyclopamine-exposed mouse fetus featuring cleft lip and palate and shown in Figure 11 c revealed agenesis of the anterior pituitary (Fig. 12 f).
Recent technological advances are greatly facilitating analysis of HPE-associated dysmorphology. Exemplary is high resolution magnetic resonance imaging (magnetic resonance microscopy; MRM) [Petiet et al., 2008]. With isotropic resolution in the range of 20-39 microns (near histological levels), a relatively rapid method for morphological assessments from individual scans is provided. Additionally allowed is creation of accurate 3-D reconstructions, facilitating shape and volumetric analyses as well as correlation of brain and facial dysmorphology in individual specimens [Parnell et al., 2009]. Using this methodology to study ethanol-induced defects in fetal mice, Godin et al  have not only illustrated a broad spectrum of median forebrain insult, some examples of which are shown in Figure 13, but have also discovered that acute GD 7 ethanol exposure induces formation of cerebral leptomeningeal heterotopias (Fig. 14). This type of cortical abnormality is correlated to seizure activity in clinical populations [Verrotti et al., 2009]. It is remarkable that acute insult at a time when the neural plate is just being induced can yield this endpoint.
Another magnetic resonance-based methodology that offers considerable promise for advancing our understanding of HPE-associated defects is diffusion tensor imaging (DTI). This technique allows visualization of the pattern and integrity of CNS fiber tracts and is based on identifying the rate and direction of water diffusion through the tissue. Even in the absence of gross structural brain malformations, microstructural changes within the fiber tracts can be determined. Currently, DTI is being applied to the study of the brains of mice acutely exposed to ethanol on GD 7 [O'Leary-Moore, personal communication]. While, undoubtedly, a wealth of information will be provided from these DTI studies, technological advances that will allow reduction in the time required for small animal imaging are expected and will provide the potential to extend the analyses from fixed to live specimens, thus allowing correlation of morphological and behavioral phenotypes.
In addition to now being able to better define the features of the full HPE spectrum, the availability of a variety of teratogen-induced and genetically-based models along with advances in developmental biology research methodologies are making it possible to better understand the pertinent molecular and cellular mechanisms. With respect to teratogen-induced HPE, it will be of particular interest to further explore various mechanisms by which Hh signaling is directly or indirectly disrupted. While events occurring upstream (e.g. failure of Shh-expressing or responding cells to form) and downstream (e.g. failure of signal transduction) of Hh signal production, as well as those directly affecting the latter, may yield comparable endpoints, at least subtle differences are expected. Further studies to examine temporal specificities, the effects of drug exposure combined with gene manipulation, and insult resulting from combinations of HPE-inducing drugs are expected to provide important new insights into this etiologically complex spectrum.
For teratogen-induced HPE, understanding that the critical period for induction is in the 3rd to early in the 4th week of human gestation is, indeed, critical. Avoidance of the teratogenic agents, as well as initiation of preventive measures, such as nutritional supplementation must occur prior to the time that most pregnancies are recognized.
This work was supported by NIH/NIAAA grants AA007573, AA11605, AA017124; and was done in conjunction with the Collaborative Initiative on Fetal Alcohol Spectrum Disorders (CIFASD). Additional information about the CIFASD can be found at http://www.cifasd.org.
Robert Lipinski is a postdoctoral fellow at the University of North Carolina at Chapel Hill. He is interested in understanding how genetic and environmental/teratogenic factors interact in the genesis of complex birth defects.
Elizabeth Godin is a Ph.D. candidate in the Curriculum in Toxicology at the University of North Carolina at Chapel Hill. Her primary interest is in developmental toxicology.
Shonagh O'Leary-Moore is a postdoctoral fellow in the Fetal Toxicology Division of the Bowles Center for Alcohol Studies at the University of North Carolina at Chapel Hill. Her work focuses on neuroimaging in teratogen-exposed animal models.
Scott E. Parnell is a Research Assistant Professor at the University of North Carolina at Chapel Hill. He conducts basic research on the neuroanatomical and neurobehavioral effects of early gestational ethanol exposure.
Kathleen Sulik is a Professor in the Department of Cell and Developmental Biology and a member of the Bowles Center for Alcohol Research at the University of North Carolina, Chapel Hill. Her interests are in craniofacial and neuroteratology.