I started my scientific carer by investigating the development of the digestive tract in the laboratory of a well-known embryologist, Etienne Wolff, then professor at the Collège de France. My animal model was the chick embryo. The investigations that I pursued on liver development together with serendipity, led me to devise a cell-marking technique based on the construction of chimeric embryos between two closely related species of birds, the Japanese quail (Coturnix coturnix japonica) and the chick (Gallus gallus).
The possibility to follow the migration and fate of the cells throughout development from early embryonic stages up to hatching and even after birth, was a breakthrough in developmental biology of higher vertebrates.
This article describes some of scientific achievements based on the use of this technique in my laboratory during the last 38 years.
In 1969, I made an observation which led me to devise a cell marking technique that could be applied to the development of the avian embryo. This observation changed the course of my research and to a certain extent also of my life. Other important events in my career took place in 1986, when I was awarded the Kyoto Prize, and, in the same year, the Gold Medal of the Centre national de la recherche scientifique (CNRS).
One of the pleasant outcomes of the Kyoto Prize was that since then I have often visited Japan, and have had the great satisfaction of receiving talented Japanese scientists in my laboratory. The other was the friendship and mutual respect that developed between myself and Professor Tokindo Okada, a scientist who played a major role in establishing developmental biology research in Japan.
It is a great pleasure and an honor for me to be invited by my friend and former postdoctoral fellow, Dr Harukazu Nakamura, to participate in this anniversary issue of the journal Development, Growth and Differentiation.
I will, in these following pages, give a brief overview of the various subjects in which the avian embryo chimeras technique was instrumental in the research carried out in my laboratory over the last 38 years.
The years of choice
I started my professional life as a high school teacher. I was then 24 years old and already the mother of my first daughter, Claire. I had just successfully passed the concours d'agrégation de sciences naturelles, a competitive exam which opens the door for a civil service career as a lycée professor.
In the next 8 years, during which I had my second child, Laure-Anne, I taught high school children. After a few years in the Lycée however, I became concerned that teaching the basic principles of biology and geology to adolescents would ineluctably become a routine. This feeling may have pushed me to find a more rewarding and exciting occupation. It could also have been nostalgia for the university years where one follows science on the move. Which of these motivations prompted me to change directions of my professional life, I do not exactly know. Both probably have played a role in this decision. By all means, I was not fully satisfied with my job and decided to try and go back to the university to prepare a doctorate.
This was not easy in the late 1950s. Research in France had suffered from the five long years of the Second World War. The country had been exhausted of its wealth and most laboratories nearly completely devoid of resources were trying to recover from what had been a long night.
I was lucky to be introduced to Professor Etienne Wolff who was running one of the most dynamic laboratories of experimental biology at the Collège de France in Paris. I worked in his laboratory at the Institut d'Embryologie et de Tératologie expérimentales, located in a superb environment in Nogent-sur-Marne at the border of the Bois de Vincennes on the East side of Paris. For 2 years I remained a professor at the lycée and I came to the laboratory when my teaching duties allowed me free time.
I rapidly realized that research would be my choice for life. Etienne Wolff was very encouraging and wise. He had created, in Nogent, a very well organized laboratory with a pleasant atmosphere for work. At the end of these 2 years, during which I could test my motivation and he could evaluate my capacities for this activity, I was able to switch from teaching to research administration, and from 1962 I could devote myself full time to my research project.
My first project was a rather large one, since I was told to study the development of the digestive tract and associated glands. As Etienne said, it was a neglected field and not much attention had been given to the development of endoderm-derived organs. I learned from him how to handle early embryos, to make current type microsurgery such as grafting tissues into the coelomic cavity or on the chorio-allantoic membrane of host embryos. Organotypic tissue culture was among the techniques that Etienne Wolff had developed with one of his collaborators, Katy Haffen. Histology and some histochemistry were other methods regularly used by embryologists at that time.
In 1964, I defended my thesis, entitled Etude expérimentale de l'organogenèse du tube digestif et du foie chez l'embryon de Poulet, that was published in the bulletin Biologique de la France et de la Belgique (Le Douarin 1964). The highlight of this work was the analysis of liver development which revealed that the hepatic tissue develops from a few endodermal cells located in the anterior intestinal portal resulting from the coalescence of paired endodermal areas (forming the pre-hepatic endoderm) located on each side of the primitive streak and head process at the presomitic stages. These pre-hepatic areas of endoderm were juxtaposed to the future cardiac mesoderm.
I was then able to demonstrate that two successive steps of induction of the pre-hepatic endoderm were required for the differentiation of hepatocytes. The first one, which took place early (before the 5-somite stage) and arose from the precardiac mesoderm, was necessary for the efficacy of the second which led to the differentiation of the hepatocytes.
Hepatocytes are not the only cell type present in liver. In fact, they are able to develop from the endodermal hepatic rudiment only if they receive a signal from the mesenchymal tissue that they colonize. The liver mesenchyme develops at the same time as it is invaded by cords of epithelial cells from which the hepatocytes differentiate. This is why it had never been described or isolated. I designed a microsurgical technique through which the mesenchymal component of the liver develops in the absence of its endodermal component by preventing the hepatocytes from invading it. The liver mesenchyme developed as empty bags connected with the septum transversum.
It turned out that, in the embryo, the liver mesenchyme was composed of cells belonging to the vascular endothelial lineage. It thus became possible to isolate the liver mesenchyme and to put it in organotypic culture, making it possible to explore its role in the differentiation of the hepatic endoderm in depth.
This particular experiment, in which liver mesenchyme and hepatic endoderm where co-cultured, led me to devise a cell marking technique which was to be a very efficient mean to get new insights into the behavior, movements and interactions of embryonic cells in the developing organism.
The quail–chick chimeras
The experiment which put me on the track to devising a cell marking technique which, over the ones previously used, had the advantage of being stable and fully reliable, consisted in associating the liver mesenchyme from quail embryos with the hepatic endoderm from chick. The aim was to see if the inductive influence exerted by the mesenchyme on the endoderm could work across species. The answer was “yes”: a lobe of liver, in which the mesenchymal cells of the quail species were closely associated with the hepatocytes derived from the chick endoderm, developed in the culture dish. Although this seems evident today, since we are aware of the conservation over evolution of the living processes, it was not so at that time.
But the most interesting result of the experiment was not the answer to the question raised. It came from a fact that had escaped the attention of the rare biologists who had previously used quails as an experimental material. I noticed that the nucleolus was particularly large and conspicuous in quail mesenchymal cells. This was striking to me, since one of the characteristics of the differentiation of hepatocytes was the enlargement of the nucleolus. It was thus intriguing to see a large nucleolus not only in the hepatocytes, but also in the mesenchymal cells of the chimeric liver lobes that developed in the cultures. I decided to look more closely into the peculiarity of quail cell nuclei, using histochemical techniques, such as the Feulgen-Rossenbeck's procedure that stains DNA, and also a method for staining the RNA components of the nucleolus.
It appeared that the nucleolus of all embryonic or adult cells of the quail was enlarged and contained a large amount of heterochromatic DNA, making this organelle strongly Feulgen-positive. This unusual characteristic attracted my attention since the nucleolus normally contains only a small amount of DNA and is not stained after the Feulgen-reaction. Cell types in which protein synthesis is active, such as hepatocytes, have a large nucleolus mainly composed of RNA.
Thus, as far as the structure of its nucleolus was concerned, the quail species appeared as an exception. This particularity made quail cells easily recognizable from chick cells at the single cell level and at any developmental stage. This prompted me to devise a cell-marking technique by constructing chimeras between quail and chick embryos. (Fig. 1) When the monoclonal antibody technology became available, species-specific antibodies were prepared against either quail or chick antigenic determinants and were used for the analysis of the chimeras (Fig. 2).
I advocated combinations of quail and chick cells, a now widely used method, which makes it possible to follow, in quail–chick chimeras, the fate of any particular cell type during its whole ontogenetic history (Le Douarin 1969). The remarkable advantage of this cell-marking technique over those used before (like labeling dividing cells with tritiated thymidine) is its stability. For this reason, it has opened new avenues for research on the morphogenetic processes taking place during development. With this tool in hand, I decided to orient my research activity toward the study of cell migrations during embryogenesis, retaining, however, a central interest for the role and mechanisms of cell–cell interactions in morphogenesis and differentiation.
The quail–chick marker system enhanced significantly the value of the avian embryo as a model for embryological research. The combination of its advantages such as the availability of the embryo to observation and manipulations during the entire period of development with molecular methods, have also increased its usefulness for modern studies in developmental biology. An important step toward making molecular studies possible on this material has been the capacity to electroporate nucleic acids into defined territories of the embryo in ovo.
Combinations of the quail–chick marker system with these genetic manipulations allow the refined developmental analyses (e.g. Creuzet et al. 2006) currently carried out in my and other laboratories.
During the course of the past 35 years of a research which is still ongoing, I have investigated in depth two embryonic systems in which cell migrations play a major role: the neural primordium including the central nervous system and the neural crest with its multiple derivatives, and the hematopoietic/immune system. I have also been interested in the development of other tissues and organs, such as the paraxial mesoderm, which forms the vertebral column, since the differentiation of vertebrae is closely related to that of the spinal cord.
The neural crest
The neural crest is a transitory and pluripotent structure of the vertebrate embryo. It was discovered by Wilhem His in 1868 and mostly studied in lower forms of vertebrates (e.g. amphibians) during the first half of the 20th century. This early work was summarized in a well acknowledged monograph by Sven Hörstadius (1950) entitled The Neural Crest that appeared in 1950.
When I became interested in this structure, in 1968, only a little information was available about the contribution of the neural crest to embryogenesis in higher vertebrates (birds and mammals). Moreover, knowledge about the neural crest in lower vertebrates was only partial.
Identification of the derivatives of the neural crest in the amniote vertebrates
With my colleagues, Marie-Aimée Teillet, Christiane Le Lièvre and Josiane Fontaine, then at the University of Nantes, we used the quail–chick chimera system to unravel the ontogeny of the neural crest in the avian embryo. Studies carried out in the mouse embryo, that can now be maintained in culture for 2 to 3 days during early organogenesis, and on which one can also apply genetically engineered cell marking techniques, have confirmed the assumption that the general conclusions drawn from our quail–chick chimeras experiments are valid in mammals.
The cells which compose the neural crest lose their epithelial arrangement and disperse within the embryo along definite pathways, at precise times during development to finally settle in elected points where they differentiate into a large variety of cell types and participate in a number of tissues. It is thus an interesting developmental model to investigate the mechanisms of cell migration and histogenesis in vivo. These studies can shed light on the process of tumor cell metastasis later in life. Furthermore, since the neural crest is a highly pluripotent structure, it is suitable to investigate the problem of cell lineage segregation during ontogeny. Moreover, since it appeared with vertebrates in the group of chordates, it is also interesting to study it from an evolutionary point of view.
By constructing quail–chick chimeras in which part of the neural primordium of the host embryo was substituted by its counterpart taken from a donor of the other species, the migration and fate of the neural crest cells could be followed during the entire embryonic life and even after birth thanks to the stability provided by this cell labeling technique. Such chimeras are able to hatch and display normal growth and behavior before being subjected to graft rejection (see below the section on immunology).
The main findings from these experiments were the following:
1The demonstration of the considerable contribution of the neural crest to the vertebrate head – to the facial and visceral arch skeletal and connective structures, the skull – and also the cardiovascular system. These notions were new. Early work had only shown that the neural crest contributes to the facial skeleton in amphibians. The fact that the entire facial skeleton and a large part of the skull were of neural crest origin (Couly et al. 1993) had not been shown (Fig. 3).
2The demonstration that the calcitonin-producing cells and the chemoreceptor cells of the carotid body originate from the neural crest. The implication of the neural crest in the endocrine system was only known for the adrenomedullary cells.
3The localization along the neural axis of the level of origin of the various derivatives of the neural crest and the establishment of the pathways and chronology of neural crest cell migration.
4The demonstration of the plasticity of the neural crest cells fated to build up the ganglia and nerves of the peripheral nervous system (PNS) and the fact that their differentiation largely depends upon environmental cues arising from the tissues in which they differentiate at the term of their migration. (see Le Douarin 1982 for a review, and Le Douarin and Teillet 1974; Le Douarin et al. 1975b).
This was the first in vivo demonstration of the influence of environmental factors on neuronal differentiation.
The segregation of the cell lineages arising from the neural crest: first demonstration of a “neural stem cell”
One of the most striking features of the neural crest is the fact that it gives rise to a large number of different cell types. This being established, my next undertaking was dedicated at deciphering how and when the different cell lineages arising from this structure become segregated. I first focused my investigations on the PNS which itself contains highly diversified cell types. Our experiments were first carried out in vivo. It turned out that, although the neural crest is regionalized into several distinct areas yielding different PNS structures in normal development, spatial disturbances of this pre-existing order did not result in major abnormalities in PNS ontogeny. This meant that one neural crest area can be substituted for another to provide the embryo with sensory, sympathetic, parasympathetic and enteric ganglia (Le Douarin & Teillet 1974; Le Douarin et al. 1975a).
These results also showed that each level of the neural crest contains a range of developmental potentialities which is larger than those actually used during development. The role of the microenvironment on the developing neural crest cells thus appeared decisive in determining their fate. Along the same line, we could also demonstrate that, once differentiated, the peripheral ganglia contain undifferentiated and still highly pluripotent precursors, in their non-neuronal population. These precursors can be led to differentiate if they are transplanted into the appropriate sites of a younger host. Thus, cells that keep the pluripotent character of their early neural crest progenitors are set aside during the development of the PNS ganglia where they remain quiescent (Le Lièvre et al. 1980). Whether these cells are recruited later in life for neuronal or glial cell renewal seemed likely, meaning that these undifferentiated cells might form a population of resting stem cells. This result takes on a novel significance in the light of experiments showing the presence of pluripotent stem cells in most tissues of the adult body. In the 1980s, they opened up new avenues of research which required investigations to be pursued on the neural crest in vitro. We thus developed a method that allowed clonal cultures of neural crest cells in which no restrictions were imposed to the full expression of their differentiating capabilities. This led us to show that, as early as the migratory stage, neural crest cells are a heterogeneous population of differently committed cells. Fully committed precursors, giving rise to clones with only one cell type, were found, as well as pluripotent progenitors, in the respective proportions of about 20% and 80% (Baroffio et al. 1988; Baroffio et al. 1991; and see Le Douarin et al. 2004 for references). The putative neural crest stem cell able to give rise to virtually all the phenotypes derived from this structure was not found until recently (Calloni et al., 2007).
From these results we were able to propose a cell lineage model that accounts for the diversification of neural crest derivatives from a multipotent stem cell (Fig. 4).
Among its derivatives, the cephalic neural crest includes not only melanocytes, neuronal, glial and endocrine cell types, like its truncal counterpart, but also the so-called mesectoderm, from which connective and adipose tissues, smooth muscle cells (myofibroblasts), cartilage and bone differentiate. Clonal cultures of cephalic neural crest cells revealed that the early migrating population contains cells endowed with the capacity to yield neural, endocrine and melanocytic derivatives together with mesenchymal cells; this meant that, during neural crest development, these different lineages segregate late during the ontogenetic process, a fact that contradicted the then current opinion (for discussion, see Le Douarin & Kalcheim 1999).
Moreover, the cells able to yield, in their progeny, all the derivatives of the neural crest, are highly pluripotent. They can be considered as the “stem-like” cells of the neural crest. Whether they deserve to be called real “stem” cells awaits the demonstration that they are endowed with self-renewal capacity. Interestingly, we have recently shown that the cloning ability of such cells in early quail cephalic neural crest migrating cells increases considerably (from 6.5% to 18.5%) in the presence of the morphogen Shh (Calloni et al. 2007).
Pluripotent precursors intermediate between stem cells and the fully committed progenitors were identified (e.g. the GM precursors which yield clones containing glia and melanocytes or the GMF precursors which produce glia, melanocytes and myofibroblasts). Recently these progenitors were shown to be able to self-renew and thus deserve to be considered as stem cells (Trentin et al. 2004).
Neuropoiesis: A parallel between neural crest development and hemopoiesis
We therefore started to look for growth and survival factors able to promote the proliferation and differentiation of a selected set of precursor cells of each type of neural crest derivatives.
When considering the cell lineages hierarchy of neural crest and hematopoietic cells, the parallel is evident; this is why the term neuropoiesis was coined (Anderson 1989).
This notion led me to search for secreted signaling molecules playing roles in neural crest cell differentiation, similar to the cytokines in blood cell development.
The identification of molecular markers specific for the different cell types arising from neural crest progenitors was a prerequisite to the analysis of the phenotypes present in the cultures. From the 1980s onward, my laboratory was engaged in the production of monoclonal antibodies (mAbs) directed against molecular markers for the avian neural crest cells and their derivatives. Some of them have been studied in depth.
For example, the SMP mAb led to the purification and cloning of the gene of the Schwann cell myelin protein, which exhibits a strict specificity for oligodendrocytes in the central nervous system (CNS) and Schwann cells in the PNS (Dulac et al. 1988; Dulac et al. 1992). Its expression by PNS glial cells turned out to be constitutive during the process of differentiation, and strongly repressed by the gut environment in enteric glial cells and in the peripheral ganglia, in satellite cells (Dulac & Le Douarin 1991). Another example is the glycoprotein BEN identified by means of the mAb technique (Pourquiéet al. 1992a,b).
Factors controlling the development of neural crest derivatives
Several mutations affecting the development of neural crest derivatives have been observed long ago in the mouse and the genes affected in some of these mutants have been identified. Such is the case for the mutants W and Steel and Piebald-lethal and Lethal spotting. W and Steel present the same phenotype which includes pigmentation defects, due to inactivation of the genes encoding the c-kit tyrosine kinase receptor in W and its ligand, the Steel Factor (SF), in Steel. In Piebald lethal, the mutated gene is EDNRB encoding the receptor for the ligand Endothelin 3 (EDN3), which is not produced in Lethal spotting. These mutations result in agangliogenesis in the terminal colon (Hirschprung disease in humans) and pigmentation defects. The effect of EDN3 and SF were tested on neural crest cultures. EDN3 turned out to have a strong proliferative effect on neural crest cells in culture (Lahav et al. 1996), whereas SF acts as a survival factor and promotes the differentiation of cells already engaged in the melanocyte differentiation pathway (Lahav et al. 1994). The proliferation effect of EDN3 is exerted on the GM bipotent precursors that we previously identified in clonal cultures of neural crest progenitors (Lahav et al. 1998). We showed also that, in avian species, the effect of this factor on the GM precursors, and differentiated melanocytes as well, is related to the presence of a distinct receptor, EDNRB2 (Lecoin et al. 1998).
The reversibility of the differentiated cells
Apart from exceptions (see G. Eguchi in this issue), the differentiated state is considered as being stable. We have however been able to demonstrate that both glial cells and melanocytes can reverse their phenotypes if they are prompted to divide by the growth factor EDN3.
EDN3 turned out to be a powerful factor for proliferation of the cells belonging to the glial-melanocytic lineages. The presence of EDN3 in clonal cultures of differentiated melanocytes results in intensive proliferation and in transitory dedifferentiation of the cells which, after a while, exhibit either the glial or the melanocytic phenotypes that coexist in the same cultures and even sometimes in the same cells. EDN3 is therefore able to induce the transdifferentiation of the melanocytes that, while dividing, reverse to the GM progenitor state with ability to self-renew (Dupin et al. 2000; Real et al. 2006). This is a clear demonstration of the reversibility of the differentiated state. Similar results have been obtained in clonal cultures of glial, SMP-positive, cells that could give rise to clones containing both melanocytes and glial cells as well as undifferentiated bipotential GM progenitors (Dupin et al. 2003; Real et al. 2005).
The neural crest and the vertebrate head
As already shown by my work with C. Le Lièvre in 1975 (Le Lièvre & Douarin 1975), the neural crest plays an important role in the development of the vertebrate head. This problem was further documented in the work that we have been pursuing ever since in my laboratory (see Le Douarin 1982; Le Douarin & Kalcheim 1999 for reviews).
In vertebrates, the mesoderm is at the origin of the so-called mesenchyme from which a number of tissues differentiate: connective, adipose, cartilaginous tissues, bones and smooth muscles. It turned out, from the results of our studies that, in this respect, the vertebrate head has an original status, since the contribution of the mesoderm to the head skeleton is limited to the occipital region and to a part of the otic capsule.
The neural crest cells form not only the facial skeleton and connective tissues, but also covers the forebrain whose development increases considerably along the vertebrate phylum. All the facial tissues originate from the neural crest at the exclusion of two cell types: striated myocytes (except for iridal muscles [Creuzet et al. 2005]) and vascular endothelium (Fig. 4). The mesenchyme of neural crest origin is designated as mesectoderm.
In view of these embryological results, Gans and Northcutt (1983), in 1983, put forward their concept of the “new head” on the basis of the conclusions that we had reached in the 1970s. According to these authors, due to its contribution to the head, emergence of the neural crest was essential for the evolutionary transition from protochordates to vertebrates. Our recent findings, which have significantly renewed the current notions on the embryology of the vertebrate head, further substantiate this view.
From the early neural plate to the brain
By following the development of the early neural primordium (the anterior neural plate), it turns out that, in the original configuration visible in the jawless vertebrates (hagfish and lampreys), the anteriormost part of the brain corresponds to the diencephalon (thalamus, hypothalamus and pituitary gland) with only a modest dorsal development of the telencephalon. The diencephalon corresponds to the anterior end of the notochord, a vertebrate organizing center that plays an important morphogenetic role in patterning the body plan at different steps. We have shown that the differentiation of the paraxial mesoderm (cephalic and somitic) into cartilage and bone depends upon a signal arising from the notochord (mediated by the secreted glycoprotein sonic hedgehog). The presence of the notochord, up to the mesencephalon–diencephalon junction, thus accounts for the formation of the vertebral column and the occipital region of the skull (Pourquiéet al. 1993).
The evolution of the vertebrate phylum is characterized by the development of the cerebral hemispheres which reach their maximal size in humans and other primates.
We focused our attention on the early developmental steps of the cerebral hemispheres and showed (by following the fate of the various regions forming the early neural plate through the quail–chick chimera system) that the cerebral hemispheres arise from the anterior and lateral areas of the neural plate. After the fusion of the neural folds and the formation of the encephalic vesicles, these lateral areas are the site of an intensive growth. Thus they develop rostrally beyond the anterior tip of the notochord and of the adenohypophysis. The latter becomes “buried” inside the stomodeal cavity while maintaining its close relationships with the floor of the diencephalon (i.e. the hypothalamus; Fig. 5). Since both notochord and mesoderm are absent at the telencephalic level, no skeleton was available to enclose this “new brain” whose development increased during the course of evolution and which became covered by cells of neural crest origin, forming both the forebrain meninges and the frontal and parietal skull. Thus, co-evolution of the anterior brain and of the neural crest-derived skeleton occurred. This process must have been critical for the development of higher cognitive functions in the more recent forms of vertebrates.
It has to be noted that this considerable development of the forebrain has been accompanied by the emergence of sense organs: the eyes, that originate from the diencephalon, and the smell organs, the precursors cells of which have been localized in the anterior neural fold (see Le Douarin & Kalcheim 1999).
The fate map of the neural plate, that we published in 1987 (Couly & Le Douarin 1987), has served as a reference for many investigations on the forebrain carried out by other investigators in birds, mammals and Xenopus.
Acquisition of higher brain functions in vertebrates has been accompanied by a change in their life style as compared to their filter-feeder relatives, the cephalocordates. Vertebrates became able to seek for their food and later became predators. Their facial skeleton, which is entirely derived from the neural crest (see Le Douarin 1982) involves the first organ of predation, the jaw, which is particularly developed in certain teleost fish.
In conclusion, the embryological analysis of neural crest ontogeny has shed a new light on the evolution of the vertebrate phylum and of the vertebrate head, which is characterized by an increased participation of the ectoderm via the neural primordium. The latter not only generates the brain, but also the neural crest, a critical structure for the construction of the face, yielding a large part of the heart and of the head vasculature and, as recently demonstrated in my laboratory, regulating the development of the pre-otic brain. (Creuzet et al. 2006; N. M. Le Douarin and S. Creuzet, unpubl. data, 2007).
Genetic control of the development of the head skeleton from the neural crest
From 1995 until today, we have been analysing the role of Hox genes in the development of neural crest derivatives. Among the number of genes expressed during the development of the head and face, Hox genes have attracted much attention. The anteriormost level of expression of the four first paralogous groups of Hox genes corresponds to the limit between two rhombomeres. No Hox gene expression is observed anteriorly to the r1–r2 limit. Each rhombomere (or pair of rhombomeres) is thus characterized by a combination of Hox gene expression (Hox-code). Mutational analyses carried out in the mouse have shown that Hox genes are critical for neural crest derivatives patterning, for vertebral column development from the paraxial mesoderm, and for that of the brain stem from the rhombencephalon.
As far as the development of the head skeleton is concerned, Hox genes have focused my attention for the following reasons. Hox gene expression defines two domains in the cephalic neural crest: a Hox-negative anterior domain extending from the mid-diencephalon down to the third rhombomere, from which the facial skeleton arises, and a Hox-positive domain which yields the hyoid cartilage and the neural crest cells that participate in cardiac histogenesis, and which provides mesenchyme to pharynx-derived glands (thyroid, thymus, parathyroid). The Hox genes are expressed transiently and in a segmented pattern in the rhombencephalon. Knocking out some of these genes (Hoxa2, Hoxa3) results in abnormalities of branchial arch derivatives (for references see the review by Le Douarin and Kalcheim ).
We can demonstrate by in vivo transplantation experiments that Hox-positive neural crest cells transplanted to the Hox-negative domain of the cephalic neural crest do not differentiate into cartilage and bones. Inversely, neural crest from the Hox-negative domain transplanted to the rhombencephalon (Hox-positive) participates in the formation of the hyoid cartilage (Couly et al. 1996, 1998; Grapin-Botton et al. 1997).
Moreover, within the Hox-negative domain including the diencephalic, mesencephalic and anterior rhombencephalic levels (r1–r2), the neural crest possesses a high capacity for regeneration. For example, a quarter of the endogenous crest is able to reconstitute the entire facial and lower jaw skeleton, whatever its level of origin (e.g. diencephalic, mesencephalic or rhombencephalic). This indicates that the information to form any particular element of the facial skeleton does not reside in the neural crest itself but is imposed to it by extrinsic cues.
The complete removal of the facial skeletogenic neural crest (FSNC; from the mid-diencephalon down to the level of rhombomere 3 [r3] exclusively) results in the complete absence of the facial development and in highly abnormal fore- and midbrain (Creuzet et al. 2002, 2004, 2006).
Interestingly, gain of function of the Hoxa2 gene by the FSNC prevents both facial and brain development, so that the phenotype of these embryos is similar to that resulting from the complete excision of the FSNC (Couly et al. 2002; Creuzet et al. 2002, 2004, 2006).
Recent studies ongoing in the laboratory are in the process of deciphering the signaling cascade through which the cephalic neural crest cells influences brain development.
Our investigations have also revealed that the foregut endoderm plays a decisive role in patterning the facial skeleton. Before the formation of the branchial pouches, the ventral endoderm of the foregut is regionalized in areas which can induce the formation of a supernumerary facial cartilage or bone if it is grafted at the appropriate level of the head. One of the signaling pathways through which the pharyngeal endoderm mediates its effect upon the Hox-negative neural crest cells was shown to be sonic hedgehog (Brito et al. 2006).
Skeletogenesis from neural crest cells yields cartilage and membrane bones. Our work has shown that, in the vertebrate body, the membranous bones of the skull arise exclusively from the Hox-negative neural crest cells.
Moreover, we have shown that the genetic pathways involved in the development of the head skeleton from the neural crest and in the development of the vertebral body from the sclerotomes of the somites are different. In neural crest-derived dermal bones, the transcription factor genes involved are Msx1- 2, whose expression depends on members of the TGFb family of secreted proteins, like Bmp4 and Bmp7. In contrast, the cartilage of the vertebral body develops if Pax1 and Pax9 genes are induced by the secreted protein sonic hedgehog produced by the notochord and the floor plate. (Takahashi et al. 1991; Monsoro-Burq et al. 1994, 1995 1996; Watanabe et al. 1998)
Toward the end of the 1980s, I decided to apply the quail–chick system to the study of some problems concerning brain development. It is possible to construct chimeric birds in which elected regions of the neural epithelium which forms the neural plate is exchanged between embryos of the two species at embryonic day 2 (E2) before it is vascularized. The embryos subjected to grafts of any region of the neural primordium, whether it belongs to the brain, the spinal cord or the neural crest, are able to hatch. Although part of their brain or spinal cord belongs to another species they are viable for several weeks, until delayed graft rejection takes place (see below the immunological studies on chimeras and the following articles: Kinutani et al. 1986; Kinutani et al. 1989; Balaban et al. 1988).
Constructing brain chimeras was directed to the following goals: (i) studying the migrations of cells and the morphogenetic movements of the neural epithelium which take place during neurogenesis; (ii) evidencing the role of cell–cell interactions on gene regulations within the neural epithelium; and (iii) transferring behavioral traits and genetic diseases by neuronal grafts.
Following cell migrations and morphogenetic movements during neurogenesis in the brain
In the late 1980s, the dominant thinking was that cell migrations in the encephalic vesicles were essentially radial; tangential migrations had not been described whereas radial migration of the neuroblasts had been clearly demonstrated by labeling the still dividing neuroblasts in mammalian fetuses by tritiated thymidine.
In 1988, two groups using the quail–chick chimera system (see Balaban et al. 1988) or LacZ carrying retroviruses (Connie Cepko's group at Harvard) as cell markers, published side by side two articles in Science showing that extensive tangential migrations actually take place during neurogenesis in the mammalian and avian forebrain including cerebral hemispheres. This notion is now widely accepted.
In my laboratory, these explorations were further pursued on the cerebellum and they modified significantly the current view on the development of this part of the brain. First, it was established that, in contrast to what was believed, the cerebellar cortex did not originate entirely from the metencephalic vesicle (corresponding to the two first rhombomeres). An important contribution from the mesencephalon was demonstrated, such that the cerebellar primordium encompasses the territory expressing the homeobox gene En2, thus accounting for the cerebellar agenesis observed in the mouse following the targeted mutation of this gene.
Moreover, these results showed that the cerebellum arises from a territory equally distributed on each side of the constriction between the encephalic vesicles recognized as midbrain and hindbrain, the so-called midbrain–hindbrain junction (Hallonet et al. 1990; Hallonet & Douarin 1993).
Moreover, the assumption according to which the cells of the molecular layer originate from the external granular layer was shown to be incorrect. We could show that in fact these cells undergo an inward–outward migration from the ventricular epithelium as do the Purkinje cells precursors.
The role of cell–cell interactions in gene regulation within the differentiating neural epithelium
A Japanese group headed by Harukazu Nakamura, a former postdoctoral fellow of my laboratory, discovered, by using the quail–chick transplantation technique, that the midbrain–hindbrain junction is an important organizing center in brain development. They showed that transplantation of neuroepithelial grafts from this region to the diencephalon or rhombencephalon resulted in the induction of the En2 gene in the host's neighboring neural epithelium and in the differentiation of tectal or cerebellar structures depending on the site of implantation. This prompted a number of studies, first by a French group headed by Dr Alvarado-Mallard, and the discovery by Salvador Martinez and his co-workers that the critical factor in this induction is Fgf8 naturally produced in the midbrain–hindbrain junction also called the isthmus (see Le Douarin  for references).
The fact that transcription factors expressed in brain during development can be environmentally regulated prompted me to investigate the regulation of Hox gene expression in the rhombencephalon.
When we started this investigation in the hindbrain, the general view was that expression of Hox-genes in the various levels of the brain was cell-autonomous. By a series of heterotopic transplantations of rhombomeres at the early stages of neurulation between quail and chick embryos we demonstrated that in fact Hox gene expression in the various levels of the rhombencephalon obeys extrinsic cues and is regulated by signals transmitted through the neural epithelium itself and by the paraxial mesoderm. Retinoic acid is a factor able to “posteriorize” in this way the anterior levels of the brain by means of Hox gene induction.
Our experiments have so far revealed that the potentiality to express Hox genes extends up to the prosencephalon. Thus prosencephalic neuroepithelium transplanted posteriorly is induced to express Hox-genes corresponding to the position of the transplant along the neuraxis. This means that in normal development either the inductive signal is not present in the prosencephalic and mesencephalic vesicles or that it is inhibited by antagonistic cues. Several experimental results point to the distribution of this signal as a postero-anteriorly decreasing gradient, a notion compatible with the hypothesis that retinoic acid could be involved in mediating Hox gene induction.
Transposition of rhombomeres along the neural axis does not modify Hox gene expression in transplants grafted more rostrally than their normal position. In contrast, caudal transposition results in “posteriorization” of the neuroepithelium as far as expression of Hox genes is concerned thus illustrating in mammals the “posterior dominance” demonstrated in insects. In case of transposition of rhombomeres 5–6 to the level of rhombomeres 7–8 this change in Hox-code is followed by a homeotic transformation of the phenotype of the transplant. The latter differentiates neural structures corresponding to its new antero-posterior position (Grapin-Botton et al. 1995).
These experiments thus indicated that Hox code regulation at each antero-posterior level of the neural axis depends upon positional cues and showed the crucial role of Hox genes in neural specification at the rhombencephalic level.
Transfer of behavioral traits and genetic diseases of the nervous system
The finding that birds which received xenogeneic brain or spinal cord transplants were able to hatch and to survive, in good health, for 2 to 3 months before the graft is recognized as non-self by the host's immune system prompted us to try and localize in the brain the regions responsible for certain functions or for certain diseases. We have thus taken advantage of this transient state of tolerance to show that certain behavioral traits, such as genetically determined species-specific song pattern, can be transferred from the quail to the chick by grafts of definite areas of the neuroepithelium.
Transfer of species-specific song from quail to chick by transplants of mesencephalic neural epithelium
This work was carried out by Evan Balaban who spent 3 years in the laboratory as a postdoctoral fellow in the late 1980s. For his PhD he had been trained in bird ethology by Professor Peter Marler at Rockefeller University and he was initiated to embryology in Nogent. It turned out that the juvenile crow induced by testosterone in newborn birds differs by several well-defined parameters in quail and chick. The substitution of a definite territory of the mesencephalic vesicles of the chick by its quail counterpart results in the adoption by the chimera of the quail-specific crow. In quails, the crow is interrupted, several notes being separated by silences, and is accompanied by typical head movements. After returning to the US and in his own laboratory, E. Balaban could show, by the same transplantation technique, that the head movements which regularly accompany the crows are generated in the brainstem, that is, in a part of the brain distinct from the one responsible for the quail-specific crow. He then showed that these two components of the crowing behavior can be transferred independently by distinct neural transplants. More recently he demonstrated that the auditory preference of the chicken for the alarm call of the adult female of the same species can also be transferred from quail to chick by early neural grafts (Balaban et al. 1998).
A model to study a genetic form of epilepsy
Genetic forms of epilepsy are known in humans and are characterized by seizures triggered by visual or auditory stimuli (e.g. intermittent light). They are considered as belonging to the sub-cortical type of epilepsy. However, how seizures are generated and what is (are) the brain area(s) involved in this process was not well known, although it had been proposed that their focus was located in the brainstem.
As mentioned above, the embryonic grafts of neural epithelium are subjected to immune rejection in the interspecific chimeras 1–3 months after the young bird has become immunologically competent (i.e. about 2 weeks after birth). How this immune attack against the nervous tissue develops was interesting in itself and was the subject of a series of studies that will be mentioned below. It results in the destruction of the foreign neural tissue, hence constitutes a limitation of the model for the behavioral studies in interspecific chimeras that we had undertaken. In the course of this work, however, we made an observation that opened up a novel research avenue: we found that neural grafts between major histocompatibility complex (MHC)-different chickens are permanently tolerated. This led our group to develop, in collaboration with Professor R. Naquet, a well-recognized specialist in epilepsy, and Dr C. Batini, a neurophysiologist, a model for the study of a genetic form of chicken epilepsy by using a Mendelian recessive mutation discovered by Crawford (1970) in Canada. This disorder is characterized, in the homozygous epi/epi chicken, by seizures of the “grand mal” type following either light or sound stimulation. By transplanting either the prosencephalic, mesencephalic or rhombencephalic brain vesicles (entirely or partly), the neural pathways involved in the onset of the photic or audiogenic forms of epilepsy could be determined. Moreover, it was shown that the nucleus mesencephalicus (pars dorsalis) triggers the seizures after processing either stimulus (sound or light). The “epi” chicken mutant can thus be considered as a unique animal model for the so far poorly understood subcortical types of human epilepsy.
The work on chicken epilepsy has been reviewed in Trends in Neurosciences (Batini et al. 1996).
Neurulation in amniotes revisited
From the work by Hans Speman, neurulation is classically considered as the result of an induction of the ectoderm by the “organizer” and its derivative, the notochord. The resulting neural plate was thought to be formed by a homogeneous sheet of epithelial cells in which the notochord was supposed to introduce heterogeneity by inducing the neuroepithelial cells to become the floor plate, a medial structure that plays an important role in patterning the neural tube.
The classically accepted view was that the floor plate is induced by the notochord via a secreted protein encoded by the gene sonic hedgehog (shh). This induction would lead the floor plate to produce a transcription factor of the forkhead family, HNF3b, which in turn would control the production of Shh protein by floor plate cells. Shh is further considered as in important molecule for specifying the various types of neurons that differentiate in the spinal cord. By using the quail–chick marker system in the embryo in ovo, it was possible to trace the fate of the avian organizer (Hensen's node cells) throughout ontogeny. It turned out that the cells of the organizer which express HNF3b, from the onset of gastrulation, yield not only the notochord, as previously thought, but also the floor plate and a stripe of cells in the dorsal endoderm. During gastrulation and neurulation, Hensen's node moves along the antero-posterior axis according to the process called “regression” and leaves in its wake these three medio-dorsal structures. This and other experiments thus challenge the role of the notochord in floor plate induction during normal development. In fact, the floor plate develops in a cell autonomous manner and the notochord is not required for inducing the floor plate. Those observations are supported by several genetic data particularly by the existence of mutants in zebrafish (like floating head) in which a floor plate develops in a normally patterned neural tube in the total absence of a notochord.
Our results therefore led to reconsider the classical notions about the process of gastrulation and neurulation in the vertebrate phylum (Charrier et al. 1999, 2001, 2002).
They have also disclosed a novel role for the protein Shh during neurulation. If the development of the notochord and floor plate is prevented by extirpation of Hensen's node, the neural tube and the paraxial mesoderm are the site of massive cell death. The Shh morphogen thus acts as a survival factor and one of the primary role of these midline structures in the vertebrate embryo is to prevent the neural and mesodermal cells to trigger their cell death program.
Ontogeny of the hemopoietic and immune system
The ontogeny of the primary lymphoid organs, in which T and B lymphocytes differentiate, can be investigated in a privileged manner in the avian embryo, i) because of the accessibility of the embryo to experimentation at the critical stages of development; and (ii) because birds are the only vertebrates with a specialized organ for B cell differentiation, the bursa of Fabricius.
By constructing chimeric thymuses and bursas between quail and chick embryos by means of grafts or tissue culture, we showed that the hemopoietic precursor cells (HPC) which differentiate in the primary lymphoid organs are all of extrinsic origin and invade the thymic and bursic primordia at precise times of development. The thymus was shown to be seeded by waves of HPC lasting about 24 h in the quail during the embryonic and perinatal periods. These phases of receptivity are separated by refractory periods of about 5 days. In contrast to this cyclic activity of the thymus, the bursa of Fabricius is the site of a single influx of HPC, lasting 5 to 6 days during embryogenesis. The hemopoietic cells which enter the bursa at that time provide the animal with B cells for the rest of its life.
The study that we undertook concerning the mechanisms underlying the seeding process was mainly carried out on the thymus. To account for the above mentioned observations, I formulated the hypothesis that the onset of thymus seeding by HPC takes place with the onset of a chemotactic attractant mechanism originating from the thymic epithelium. According to this hypothesis, HPC sensitive to the chemoattractant home to the thymus and, through a feedback effect, stop its production. The chemoattraction hypothesis has been fully confirmed since and some chemoattractive molecules have been characterized.
The embryonic origin of the T and B lymphocyte precursors, as well as of the hemapoietic stem cells (HSC) themselves could be traced, by a group of researchers in the laboratory, back to blood islands developing within the embryo in close association with the dorsal aorta and not in the yolk sac blood islands which generate exclusively embryonic hemopoietic cells as shown in both the avian and mammalian embryos (Dieterlen-Lièvre 2007).
We have developed a series of studies aimed at the identification of surface antigens for hemopoietic cells by means of the monoclonal antibody technology. Several clones with strict cell type or cell lineage specificities have been obtained such as antibodies to T cell antigens, for class II MHC products, PNS neurons and glia, either restricted to quail or chick or common to both species. One reagent called MB1/QH1 recognizes all blood and endothelial cells of the vascular system as well as their ancestors in the mesodermal blood island from which the pluripotent hemopoietic stem cells and the endothelial cells of the blood vessels originate. Owing to its strict quail species specificity MB1/QH1 is a precious tool to analyze quail–chick chimeric tissues (Péault et al. 1983, 1987; Labastie et al. 1986; Pardanaud et al. 1987).
During the search for tyrosine-kinase receptor genes transcribed during early embryogenesis, two were found exclusively expressed in the endothelial cells of the blood vessels. Since none of these two sequences had been previously described I thought that they might represent receptors for a putative vascular endothelial growth factor. I then encouraged Ann Eichmann, then the student who had been pursuing this work, to concentrate on these two genes which she cloned, produced the protein product and prepared monoclonal antibodies against it. Some months later, similar molecules were found in mammals and my hypothesis was confirmed when the ligand, vascular endothelial growth factor (VEGF), was discovered. From that time on, the receptors were designated as VEGFR2 and VEGFR3 (respectively for Quek 1 and Quek2, the receptor that had been cloned in the quail) (Eichmann et al. 1993, 1996, 1997a,b, 1998a,b; Marcelle et al. 1994).
We found that VEGFR2 is expressed very early in the mesoderm during gastrulation. With the anti-VEGFR2 mAbs it was possible to sort out the mesodermal cells expressing this antigen by using a fluorescence activated cell sorter (FACS). The VEGFR2+ cells, cultured clonally in the absence of VEGF, differentiates into hemopoietic cells. If VEGF is added to the medium, the cells are induced to differentiate into endothelial cells. The mesodermal cells expressing VEGFR2 at these early stages of gastrulation are therefore bipotent precursors able to yield cells of both the hemopoietic and endothelial lineages. The existence of such a common precursor, termed “hemangioblast”, has long been proposed but such cells had so far failed to be characterized.
Quail–chick chimaeras as a tool to study the mechanisms of self and non-self recognition
About 10 years ago I thought that the immunological status of the grafts implanted between quail and chick embryos at early developmental stages could be an interesting model to study the problem of immunological tolerance.
The quail to chick spinal cord chimaeras proved particularly interesting inasmuch as the chickens hatching with a quail component in their CNS walk, fly and behave normally (Fig. 6). This healthy state lasts for 2 to 4 months, whereupon the chimeras develop a neurological syndrome characterized by paralyzed wings then legs. The anatomopathological signs accompanying this disease show many similarities to those of the active plaques of multiple sclerosis (MS) and of the lesions of experimental allergic encephalomyelitis (EAE). The quail to chick spinal cord chimera appears therefore as an interesting model to study the immunological aspects involved in MS in mammals (Kinutani & Le Douarin 1985; Kinutani et al. 1986, 1989).
The neurological syndrome that develops in these birds can be interpreted as a manifestation of immune rejection of the grafted tissue by the host. I interpreted the long delay between the maturity of the host's immune system and the occurrence of this rejection as a consequence of the absence of MHC molecule expression by neural cells and of the protection of nervous tissue by the blood–brain barrier. I then investigated the response of the chick host to embryonic grafts of tissues that do not have these characteristics. The graft of a quail wing bud into a chick at embryonic day 3 (E3; i.e. before it is vascularized) was followed by an acute rejection of the wing beginning on the second posthatching week. We found that, by grafting the donor thymic epithelial rudiment (not yet colonized by hemopoietic cells) at the time of limb bud transplantation, a total tolerance of the graft was obtained for the entire life span of the animal. I then decided to study this phenomenon in the mouse where genetic and molecular tools are available. The thymic epithelial rudiment, still uncontaminated by hemopoietic cells, can be isolated from E10 mouse embryo and grafted into an athymic nude mouse at birth in either iso- or allogeneic combinations. In both cases the T-cell compartment of the recipient is reconstituted. The allogeneic graft of a thymic epithelium was shown to induce full tolerance of skin grafts of the donor haplotype. However clonal elimination of T cells reactive against the graft was not (or very imperfectly) achieved in the grafted thymus, as shown by results obtained from a number of convergent approaches. Tolerance thus generated in the thymic epithelium engrafted nude mouse can be transferred to a naïve nude mouse by spleen cells. These and other experiments suggest that, besides negative selection of autoreactive clones, another mechanism is responsible for tolerance itself. My view is that the thymus is responsible for the differentiation of regulatory T cells able to suppress, in an antigen-specific manner, the activation of potentially harmful T cells that escaped elimination during the intrathymic phase of differentiation. In the thymus, it is the thymic epithelium which is responsible for this mechanism of self-tolerance.
After these seminal observations (Ohki et al. 1987; Belo et al. 1989; Salaün et al. 1990) that went against the generally accepted dogma that self-tolerance relies on intra-intrathymic elimination, a new line of research has emerged showing more and more clearly the existence and important role of these regulatory T cells, in particular in mammals.
This brief survey of nearly 50 years of research in the field of developmental biology shows the progression and diversification of the subjects that have been under scrutiny during this period. Although these biological problems were studied essentially on the avian embryo, the findings are of a general interest and particularly are also valid for mammals. Most of the problems were tackled with the powerful tool of the quail–chick chimera system. It has to be underlined that the relevance of this cell marking technique was significantly enhanced by its combination with the molecular and genetic approaches that became available during the second half of the 20th century. This is why it is presently used to carry out more and more refined analyses of complex developmental processes.
Conflict of Interest
No conflict of interest has been declared by N. M. Le Douarin.