A few crucial molecular signals give rise to chemical gradients that organize the developing embryo

Bears mate in wintertime. The female then retires into a cave to give birth, after several months, to three or four youngsters. At the time of birth, these are shapeless balls of flesh, only the claws are developed. The mother licks them into shape.

This ancient theory, recounted by Pliny the Elder, is one of the many bizarre early attempts to explain one of life's greatest mysteries - how a nearly uniform egg cell develops into an animal with dozens of types of cells, each in its proper place.

The difficulty is finding an explanation for the striking increase in complexity. A more serious theory, popular in the 18th and 19th centuries, postulated that an egg cell is not structureless, as it appears, but contains an invisible mosaic of "determinants" that has only to unfold to give rise to the mature organism. It is hard for us now to understand how this idea could have been believed for such a long time. To contain the complete structure of the adult animal in invisible form, an egg would also have to contain the structures of all successive generations, because adult females will in time produce their own eggs, and so on, ad infinitum. Even Goethe, the great poet and naturalist, favored this "preformation hypothesis", because he could not think of any other explanation.

About 100 years ago experimental embryologists began to realize that developmental pathways need not be completely determined by the time the egg is formed. They discovered that some experimental manipulations led to dramatic changes in development that could not be explained by the mosaic hypothesis. If an experimenter splits a sea-urchin embryo at the two-cell stage into two single cells, for example, each of the cells will develop into a complete animal, even though together the two cells would have produced only one animal if left undisturbed. When human embryos split naturally, the result is identical twins.

Slowly an important idea emerged: the gradient hypothesis. One of the proposers of this idea was Theodor H. Boveri of the University of Würzburg, the founder of the chromosomal theory of inheritance. Boveri suggested that "a something increases or decreases in concentration" from one end of an egg to the other. The hypothesis, in essence, is that cells in a developing field respond to a special substance - a morphogen - the concentration of which gradually increases in a certain direction, forming a gradient. Different concentrations of the morphogen were postulated to cause different responses in cells.

Although concentration gradients of morphogens could in principle explain how cells "know" their position in an embryo, the idea was for a long time not widely accepted. One of the difficulties lay in explaining how a morphogenetic gradient could be established and then remain stable over a sufficient period. In a developing tissue composed of many cells, cell membranes would prevent the spread of large molecules that might form a concentration gradient. In a single large egg cell, conversely, diffusion would quickly level such a gradient. Further, the biochemical nature and the mechanism of action of morphogens were a mystery.

For most biologists, the means of gradient formation remained elusive until recently, when researchers in several laboratories discovered gradients operating in the early embryo of the fruit fly, Drosophila. For most non-biologists, it is a surprise that many of the mechanisms of development are best known in Drosophila, rather than in animals more closely related to humans. The examples I shall describe illustrate the reason for the preeminence of Drosophila as an experimental subject: a lucky coincidence of advantages makes it almost ideal for studies in genetics, embryology and molecular biology.

Drosophila became the laboratory animal of choice for studying Mendelian genetics early this century because the fly is easy to handle and quick to breed in large numbers, making it possible to search through many individual flies for mutants. Studies of mutants have successfully elucidated metabolic pathways and regulatory processes in viruses, bacteria and yeast. Twenty years ago Eric F. Wieschaus, now at Princeton University, and I extended this approach to Drosophila by searching for genes that control the segmented form of the larva. The larva is relatively large - about one millimeter long - and has well-defined, repeated segments that emerge within 24 hours of the laying of the egg. These features are crucial for interpreting experimentally induced abnormalities that affect the pattern of development.

Another key advantage of using Drosophila for embryological studies is that during its early development the embryo is not partitioned into separate cells. In the embryos of most animals, when a cell's nucleus divides, the rest of the cell contents divides with it. Cell membranes then segregate the halves, yielding two cells where there was one. Hence, the embryo grows as a ball of cells. In contrast, the nucleus of the fertilized Drosophila egg divides repeatedly, but membranes do not isolate the copies. Eventually thousands of nuclei lie around the periphery of what is still, in a manner of speaking, a single cell. Only after three hours of cell division, when some 6000 nuclei have formed, do separating membranes appear.

This peculiarity allows chemicals to spread freely through the early embryo and influence the developmental fate of large regions of it. As experimentalists, we can therefore transplant cytoplasm (the viscous fluid within cells) or inject biological molecules into various regions of a Drosophila embryo and observe the results.

The Power of gradients

In addition, Drosophila is fairly easy to study with the techniques of molecular biology. The insect has only four pairs of chromosomes, and they exist in a special giant form. The giant chromosomes make it possible to see under the microscope, in many cases, the disruptions in the genetic material caused by mutations. This fact helps a great deal when the mutations are being studied. Last but not least, by exploiting naturally occurring mobile genetic elements, it is possible to add, with high efficiency, specific genes to the genetic complement of Drosophila.

By studying mutants, researchers have found about 30 genes that are active in the female and organize the pattern of the embryo. Only three of them encode molecular signals that specify the structures along the long antero-posterior (head-tail) axis of the larva. Each signal is located at a particular site in the developing egg and initiates the creation of a different he site of the signal.

One of the signals controls the development of the front half of the egg, which gives rise to the head and thorax of the larva.

The simplest of the morphogenetic gradients initiated by these signals consists of a protein called Bicoid, which determines the pattern in the front part of the larva. My colleague Wolfgang Driever and I found that a concentration gradient of Bicoid is present in the Drosophila embryo from the very earliest stages. The concentration is highest at the head end of the embryo, and it declines gradually along the embryo's length. Mutations in the bicoid gene of a Drosophila female prevent the development of a Bicoid gradient. The resulting embryos lack a head and thorax.

Bicoid acts in the nuclei of the embryo. The protein is termed a transcription factor, because it can initiate transcription of a gene. This is the process whereby messenger RNA (mRNA) is produced from the genetic material, DNA; the cell then uses the mRNA to synthesize the gene's protein product. Transcription factors operate by binding to specific DNA sequences in the control regions, or promoters, of target genes. In order to bind to a promoter, Bicoid must be present above a certain threshold concentration.

Driever and I have investigated the interaction of Bicoid with one target gene in particular, hunchback. Hunchback is transcribed in the front half of the early embryo, and the gene's promoter contains several Bicoid binding sites. We carried out two types of experiment: in one, we changed the concentration profile of Bicoid, and in the other we changed the structure of the hunchback gene promoter.

By introducing additional copies of the bicoid gene into the female, it is possible to obtain eggs with levels of Bicoid that are four times higher than normal all along the gradient. In these embryos, the zone of hunchback gene activation extends towards the posterior, and the head and thorax develop from a larger part of the embryo than is normal. This abnormality could in principle arise either because the Bicoid concentration gradient was steeper in the manipulated embryos or because the absolute level of Bicoid concentration was higher. The correct interpretation was made clear by an experiment in which we made mutant embryos that had a level Bicoid concentration along their length, so there was no gradient at all. These embryos produced only one type of anterior structure (head and thorax); which type depended on the Bicoid concentration. The experiment shows, then, that the absolute concentration of Bicoid, not the steepness of the gradient, is important for controlling subsequent development of each region.

In the second type of experiment the Bicoid gradient was left unchanged, but the promoter region of the hunchback gene was altered. When the altered promoter bound only weakly to Bicoid, higher Bicoid concentrations were required to initiate hunchback transcription. Consequently, the edge of the zone of hunchback activity shifted forward. In these embryos, as one might predict, the head forms from a smaller than normal region. This experiment revealed that Bicoid exerts its effect by binding to the hunchback promoter.

These experiments show how a morphogen such as Bicoid can specify the position of a gene's activation in an embryo through its affinity for the gene, in this case hunchback. In theory, a large number of target genes could respond to various thresholds within the gradient of a single morphogen, producing many different zones of gene activation. In reality, however, a gradient acts directly on usually no more than two or three genes, so it specifies only two or three zones of activation.

How is the morphogenetic Bicoid gradient itself established? While the unfertilized egg is developing, special nurse cells connected to it deposit mRNA for Bicoid at its anterior tip. Synthesis of Bicoid, which starts at fertilization, is therefore already under way when the egg is laid. As the embryo develops, the protein diffuses away from the site of its production at the head end. Bicoid is unstable, however, so its concentrations at remote points - that is, at the end that will become the abdomen - never reach high levels. The resulting concentration gradients persists until cell membranes form.

This simple diffusion mechanism is accurate enough to meet the requirements of normal development. Remarkably, even substantial changes in Bicoid levels - doubling or halving - result in normally proportioned larvae. It appears that mechanisms operating at later stages of development can correct some errors in the early stages. If a researcher transplants bicoid mRNA into the posterior pole of a normal embryo, an additional Bicoid protein gradient arises, oriented opposite to the natural one. The resulting embryo displays a duplicate head where the abdomen should be. This experiment shows conclusively that bicoid mRNA is by itself sufficient to determine polarity.

Other work has revealed how the bicoid mRNA is positioned precisely within the egg cell. Paul M. Macdonald of Stanford University has identified a large section of the bicoid mRNA molecule that contains all the information required for a cell to recognize it, transport it and anchor it. Daniel St. Johnston and Dominique Ferrandon, while working in my laboratory, found that a molecular complex consisting of bicoid mRNA and a protein known as Staufen will move in one direction along structural elements in cells called microtubules. It seems likely that this effect explains the localization of bicoid mRNA, although other proteins are certainly also involved.

Whereas Bicoid is determining the anterior section of the larva's long axis, a different morphogenetic gradient is determining the posterior part. The gradient in this case is composed of the protein Nanos. Nanos mRNA localizes in the cytoplasm at the posterior end of the egg. This occurrence depends critically on another molecular complex consisting of the Staufen protein and mRNA from a gene named oskar. Anne Ephrussi and Ruth Lehmann, then at the Whitehead Institute for Biomedical Research in Cambridge, Mass., demonstrated the crucial role of oskar by replacing the section of mRNA required for localization with that section of bicoid mRNA. This hybrid molecule behaved like bicoid mRNA, collecting at the anterior pole rather than at the posterior one. The manipulation misdirected the nanos mRNA to the anterior pole, causing the embryos to develop with two abdominal ends in mirror symmetry.

Getting around Cell Membranes

The mechanisms that produce the morphogenetic gradients of Bicoid and Nanos, both of which are large molecules, can operate only when there are no cell membranes to hinder diffusion. In most animals, however, early development creates cell membranes between different regions of the egg, so these mechanisms cannot work. It is notable, then, that the dorsoventral (top-bottom) axis of the Drosophila embryo, unlike the anterior-posterior axis, is defined by a single gradient that could develop even in the presence of cell membranes. This mechanism may thus be more typical of those found in other creatures.

The first embryonic pattern along the dorsoventral axis is determined by the gradient of a protein called Dorsal. Like Bicoid, Dorsal is a transcription factor, and it controls the activity of several target genes in a concentration-dependant manner. The Dorsal protein acts as both a transcriptional activator and a repressor - inside cell nuclei, it turns genes on or off. When its concentration in the cell nucleus exceeds a particular threshold, Dorsal activates the transcription of a pair of genes that play important roles in subsequent development. Whenever Dorsal's nuclear concentration exceeds a lower threshold, it represses the transcription of two different genes. If the concentration of Dorsal in the various cell nuclei is arranged as a gradient, each of these pairs of genes will subsequently be expressed on a different side of the embryo.

The formation of the nuclear concentration gradient of Dorsal protein is, however, entirely different from the formation of the Bicoid gradient. Overall, the concentration of Dorsal protein is actually level throughout the embryo. Christine W. Rushlow and Michael S. Levine of Columbia University, along with my colleague Siegfried Roth and me, have shown that what does vary along the dorsoventral axis of the embryo is the degree to which Dorsal protein is sequestered in nuclei. Close to the dorsal side of the embryo, the protein is found increasingly within the cytoplasm; on the ventral side it is found mainly within nuclei.

How does this strange gradient of Dorsal concentrated in nuclei arise? Normally, what stops Dorsal from entering nuclei is a protein called Cactus, which binds to it. On the ventral side of the embryo, however, Dorsal is released from this bound state by an activation pathway involving at least 10 proteins.

The ventral signal that starts this process originates early in egg development inside the female. Yet its effect - the importation of Dorsal to the nucleus - takes place several hours later, in embryos with rapidly dividing nuclei. Thus, the signal must be very stable. The signal's exact nature remains unclear, but it is concentrated in the specialized membrane - known as the vitelline membrane - that surrounds the egg after it is laid.

Painstaking experiments by my colleagues David Stein and me and by Kathryn V. Anderson and her colleagues at the University of California at Berkeley have established that some early components of the activation pathway are produced in the mother's follicle cells, which surround the unlaid egg. Others are produced in the egg cell and then deposited either in the egg's cytoplasm or in its cell membrane or secreted into the space surrounding the egg.

Initially, the protein components of this pathway are evenly distributed, each in its proper compartment. Then, the signal, which identifies the ventral side, becomes active. This signal seems to determine the Dorsal gradient by triggering a cascade of interactions among the proteins of the activation pathway; the cascade conveys into the egg the information about which side will be ventral.

This message relay system probably relies on gradients of its own. It seems likely that a true gradient first appears in the space surrounding the egg cell, because large proteins can easily diffuse through this region. The gradient signal is thought to cause graded activation of a receptor molecule in the egg's cell membrane; that is, the receptors may become either more or less active depending on how ventral their position is. The receptors could then transmit a similarly graded signal into the egg cytoplasm, and so on.

Thus, the signal that initiates the formation of the embryo's dorsoventral pattern circumvents the obstacle to diffusion. In order to do this, it relies on a message relay system that, through a variety of protein molecules, carries the gradient information from one compartment to another. (A similar mechanism for carrying a signal across the egg cell membrane operates in the terminal pathway, which is the system that controls structures at both ends of the antero-posterior axis.) In this manner, signals from outside an egg, where a gradient can easily form by diffusion, can be transmitted to the inside. The result is the graded importation into the nuclei of a protein that was initially evenly distributed.

Patterns in common

What conclusions can we draw from these investigations? Before gradients were identified, biologists believed that morphogens might constitute a special class of molecule with unique properties. This is clearly not the case. In the early Drosophila embryo, many "ordinary" proteins that can serve different biochemical functions can convey positional information.

In some instances, such as the process determining the dorsoventral pattern, a gradient arises first by diffusion and is then copied down a molecular chain of command by activation of successive proteins. In other cases, gradients have inhibitory effects. The Nanos gradient, for example, represses the cell's use of one type of evenly distributed mRNA, thereby creating a gradient of the opposite orientation.

In all the pathways so far investigated, the final result is a gradient of a morphogen that functions principally as a transcription factor, initiating or suppressing the transcription of one or more target genes in a concentration-dependant manner. These gradients are sometimes quite shallow: Bicoid and Dorsal decline in concentration only slowly along the length of the embryo. Yet they somehow cause the protein products of their target genes to have sharp cutoff points. How can this happen?

One way this might occur is if several molecules - either different ones or multiple copies of the same one - cooperated to bring about transcription. The dynamics often result in a steep dependence on the concentration of one or more of the components. It is noteworthy, then, that genes activated by Bicoid or Dorsal proteins contain multiple adjacent binding sites, often for different transcription factors that may modulate the genes' activity.

Some morphogenetic gradients apparently yield but a single effect: if the concentration of the morphogen in particular place is above a critical threshold, a target gene is activated; otherwise, it is not. In other cases, different concentrations of morphogen elicit different responses, and it is this type of gradient that is most important for providing an increase in the complexity of the developing organism.

Although each morphogenetic gradient seems to control only a few target genes directly, interactions between cofactor molecules that effect transcription can radically change responses to the gradients. These mechanisms of combinatorial regulation open the way to the formation of patterns of great complexity from an initially simple system. Proteins acting as cofactors can modify a morphogen's affinity for a gene's promoter region, thus shifting a critical threshold up or down. A cofactor might even turn an activating transcription factor into a repressor. The potential for creating complex patterns becomes apparent when one considers that the cofactors may themselves be distributed in a graded fashion.

Superposing several gradients onto an embryonic region can subdivide it even more and generate additional complexity. The three pathways that define the antero-posterior axis of the Drosophila embryo together give rise to four separate and independent gradients (the terminal pathway produces two gradients, of an unknown protein). Each gradient has one or two thresholds. At least seven regions are thus defined by a unique combination of target gene expression. At the anterior end, where the gradient of the as yet unidentified terminal protein and the Bicoid gradient overlap, the combination leads to the development of the foremost extreme of Drosophila, a part of the head. The gradient of the unknown protein acting alone, in contrast, produces the structures of the opposite end, at the tip of the abdomen.

Combinatorial regulation as a principle of pattern formation is even more apparent later in fruit-fly development. For example, the gradients of transcription factors along the long axis of the embryo affect genes that, in most cases, encode other transcription factors. Those secondary factors, in turn, diffuse out into gradients of their own. At various threshold concentrations, each factor acts on its own gene targets; sometimes these thresholds are altered by other transcription factors with overlapping spheres of influence.

Concentration dependence and combinatory regulation together open up a versatile repertoire of pattern-forming mechanisms that can realize the designs encoded in genes. In Drosophila, the initial patterns generate transverse stripes of gene expression covering the part of the egg to be segmented in the larva.

This pattern in turn directs the formation of an even more finely striped pattern, which then directly determines the characteristics of each segment in the embryo. As soon as the embryo partitions itself into cells, transcription factors can no longer diffuse through the cell layers. The later step of pattern refinement therefore rely on signaling between neighboring cells, probably with special mechanisms carrying signals across cell membranes.

Many more details remain to be discovered before we have a complete picture of how the Drosophila embryo develops. Yet I believe we have now uncovered some of the principal features. This accomplishment can illuminate much of zoology, because one great surprise of the past five years has been the discovery that very similar basic mechanisms, involving similar genes and transcription factors, operate in early development throughout the animal kingdom.

Basic research on a good model system has thus led to powerful insights that might one day help us understand human development. What these insights have already provided is a satisfying answer to one of the most profound questions in nature - how complexity arises from initial simplicity.

The Author

CHRISTIANE NÜSSLEIN-VOLHARD started her academic career studying biochemistry and gene transcription in bacteria. She turned to Drosophila at the University of Basel in the mid-1970s, where she initiated the research program described in this article. In 1978 she and Eric Wieschaus became group leaders in the European Molecular Biology Laboratory in Heidelberg, Germany, where the two studied genes affecting embryonic pattern formation. For the past 10 years, Nüsslein-Volhard has been director of the genetics division of the Max-Planck-Institute for Developmental Biology in Tübingen. She is the recipient of several scientific awards and last year shared with Wieschaus and the Drosophila geneticist Edward B. Lewis the Nobel Prize for Physiology or Medicine.

Further Reading

MUTATIONS AFFECTING SEGMENT NUMBER AND POLARITY IN DROSOPHILA. C. Nüsslein-Volhard and E. Wieschaus in Nature, Vol. 287, pages 795-799; October 30, 1980.

THE MAKING OF A FLY; THE GENETICS OF ANIMAL DESIGN. Peter A. Lawrence. Blackwell Science, 1992.

THE ORIGIN OF PATTERN AND POLARITY IN THE DROSOPHILA EMBRYO. Daniel St. Johnston and Christiane Nüsslein-Volhard in Cell, Vol. 68, No. 2, pages 201-209; January 24, 1992.

THE DEVELOPMENT OF DROSOPHILA MELANOGASTER. Edited by Michael Bate. Cold Spring Harbor Laboratory Press, 1993.