ÐÓ°ÉÔ­´´

There is something gratifying about the fact that one of the most influential ideas ever to enter the annals of botany should have come from an 18th-century poet. But then, Goethe was no ordinary poet. He was also an enthusiastic botanist who for much of his life was fascinated with shape and form. Occasionally he would come across a plant with a flower organ growing in the wrong place, a petal where a stamen should be, for example, or a leaf growing in place of a sepal: evidence, he argued, that the characters of different parts of a flower are inherently interchangeable; that they are all variations on a common theme.

Thus was born the phenomenon known to biologists as ‘homeosis’, the rare tendency of a plant organ or the limb of an animal to adopt the character of a quite different organ or limb during development. Examples of homeosis abound in the world of insects, the most famous being a mutant fruit fly which brandishes a leg where an antenna should be. A vast field of academic research – the genetics of fruit fly development – has grown up around such creatures. But today homeosis is helping researchers to unravel a different mystery. Two hundred years after Goethe described the curious phenomenon, it is inspiring biologists to look for the genes that control the development of flowers.

Botanists have long known that flowers originate from a dome-shaped group of dividing cells located at the tip of a plant stem. The mystery has been how these formless clumps of cells ‘know’ how to develop into organs of such elaborate beauty. What kind of genetic programming has evolved in plants to enable them to flower? In recent years answers have begun to emerge from genetic studies of mutant flowers similar to those which exercised Goethe’s mind.

Unlike animals, where there is only a limited period of embryonic development, plants never stop developing. Throughout their lives they usually retain groups of dividing cells, or meristems, in the tips of their shoots and roots. As a plant grows, the dome-like meristem at the tip of its shoot will climb steadily up from the ground by adding new cells to the stem beneath it. On the flanks of the dome, bulges appear which grow into either leaves or further meristems – the beginnings of new branches. This constant production of stem, branches and leaves is often halted by the development of the organs that make up a flower.

Typically, these organs reside in a series of concentric rings, or whorls. First there are the sepals, the leaf-like structures which encircle and protect petals. The petals in turn encircle pollen-producing stamens; and at the very centre of the flower lie the carpels whose ovules form the plant’s seed when they are fertilised by pollen. Each of these organs first emerges as a mere bulge on the flanks of a meristem, the sepals developing first and the carpels last. The hidden hand of the plant genome, therefore, must instruct the cells of each whorl not only to adopt the appropriate character of these organs but to do so at the correct time during development.

Jumping genes

To trace the genes that perform these jobs, molecular biologists have focused on two species, Antirrhinum majus (the snapdragon) and Arabidopsis thaliana (a small weed). Each has different, but complementary, advantages. Antirrhinum propagates well vegetatively and has large flowers that are easy to emasculate and cross. Also, its genome contains some well-characterised jumping genes, or transposons, which move from site to site on the genome, acting as flags for developmental genes. Arabidopsis has a rapid generation time and an unusually small and manipulable genome. Inserting foreign DNA into the Arabidopsis genome is relatively easy.

In the mid-1980s, we began to look for the genetic switch for flowering in the Antirrhinum. For five years we cultivated and bred Antirrhinums in their thousands, in the hope of finding one that was unable to flower – one whose genetic switch was somehow defective. Success finally came in 1987. In the process of screening more than 80 000 plants, we found a single Antirrhinum that was unable to flower in response to the environmental stimuli that normally promote flowering, such as changes in day length. Instead it continued to make leafy shoots. It behaved rather like a record stuck in a groove and continued to play the ‘leafy shoot tune’ rather than progressing to the ‘flower tune’.

In 1990, we traced this curious behavior to a defect in a gene called floricuala (flo). It turned out that one of the Antirrhinum’s transposons had become lodged inside the flo gene, thus disabling it. But to find out more about the gene, we had to isolate it; and here the transposon was a great advantage. Knowing its sequence, we could design a DNA probe that would fish it out from a vast pool of genetic material extracted from the mutant Antirrhinum. Being trapped inside the flo gene, the transposon emerged from this pool with flo DNA – our prize – joined to its two ends. After sequencing this DNA, we were able to tackle an important question: at what stage during development does the gene become active inside the meristem? The answer turned out to be just two days after the plant’s leaves receive the environmental stimulus to flower. In the initial stages of the Antirrhinum’s development the gene is dormant.

All this strengthens the suspicion that flo acts as a genetic masterswitch in flowering. But how exactly? Here the picture begins to blur. The problem is that the protein encoded by flo shows little resemblance to any known protein, so predicting its cellular function is difficult. Nevertheless, based on its effect on plants, flo probably works by regulating other genes. Its activation is probably one of the first steps in a cascade of genetic events which ultimately produce the flower’s organs.

If this is true, it begs a further question: what are these other genes and how do they specify the different organs of the flower? Some clues have come from genetic studies of flowers with developmental abnormalities. Our group in Norwich, Elliot Meyerowitz and his colleagues at Caltech in Pasadena and a group at Cologne’s Max Planck Institute, led by Hans Sommer and Zsuzsanna Schwartz-Sommer, have spent the last few years analysing the genetic origins of some of these abnormalities. We now know that they reside in defects in the genes that act after flo.

Mutant flowers

Plants carrying these defects are still able to make flowers, but some of their whorls appear to contain inappropriate organs. The abnormalities fall into three classes: in class a, carpels grow in place of sepals and stamens in place of petals; in class b, sepals replace petals and carpels replace stamens; and in class c, petals replace stamens and sepals replace carpels (see Figure 1). Our research is based on the assumption that each of these abnormalities is caused by a defect in one of three cellular ‘functions’: a, b and c. These functions stem from the collective action of a suite of genes and act in combination to specify the type of organ made in each whorl. A simple analogy illustrates how this works.

Imagine that the three functions correspond to three notes on a piano and that a specific note or chord (if more than one note is played) signifies a particular type of floral organ. When the four whorls of the flower are made, note a is played in the outermost, signifying sepals, a and b are played together in the second whorl to signify petals, b and c in the third whorl give stamens and c alone in the central whorl gives carpels. These three notes and their combinations thus provide enough information to specify the fate of each whorl. Starting from the outer whorl and moving towards the centre, the combinations are a, ab, bc, c (see Figure 2).FIG-mg18184702.jpg

Now suppose that one of the notes, b say, does not sound when the corresponding piano key is struck. The outermost whorl will still make sepals because this requires only the a note, but when the second whorl is made, even though both the a and b keys are struck, only a will sound. As a result, sepals will be made in the second whorl, too, where the petals would normally form. Similarly, in the third whorl, only c will sound even though b and c are both struck, giving carpels where stamens normally occur. The central fourth whorl will be normal (carpels) because only c is required there.

The end result is a flower ‘sounding’ like a, a, c, c, rather than the normal a, ab, bc, c and therefore comprising whorls of sepal, sepal, carpel, carpel – the archetypal class b mutation. The class a and c mutations can be explained in a similar way, although there is an additional complication. To predict the correct pattern of organs, one must assume that a and c notes are not completely independent. If for some reason a does not sound, then c must be played instead, and if c does not sound then a must be played.

So much for the analogy, how are these floral ‘notes’ actually encoded in the plant genome? Once again, the breakthrough came in 1990, when the Cologne group isolated a gene needed for b and the Pasadena group isolated a gene needed for c. The sequences of these genes vary, but – and this is the important thing – each possesses a stretch of code similar to stretches found in regulatory genes discovered in humans and yeast. Biologists know that the proteins encoded by these genes act by binding to DNA, as do some of the genes isolated in Cologne. Their job is probably to regulate genes lower down in the cascade of genetic events that produces the flower: the genes which ring the cellular changes that seal the fates of each whorl.

An interesting question raised by this model is what happens to the a, b and c notes after they have been struck. Are the notes held down for the whole of the musical piece or do they sound for only a short period? We do not yet have a full answer. However, we do know that in the case of Antirrhinum petals, at least one b gene, called def, is needed for the duration of development. Once again, it was a mutant Antirrhinum, this time one with an especially informative abnormality, that provided the evidence.

One effect of the mutation was to disrupt the second whorl. During the early stages of flower growth the whorl developed as sepals rather than petals. Its cells were prevented from differentiating into petal cells, because a transposon had become wedged inside the def gene. Yet – and this was the crucial observation – at least in some plants, this barrier to petal development appeared to collapse during development. Why? The answer lies in the fact that transposons can jump to new locations in the chromosomes of dividing cells. This was occasionally the case for the transposon inside the def gene. Now and then it would abandon the gene in one sepal cell, leaving the descendants of that cell to sound the b note and develop as petal cells (see Figure 3). The result was a flower whose second whorl contained both petal and sepal characteristics. But the broader implication was that the second whorl needs the b note to sound throughout its development in order to retain its sense of identity.

War of the whorls

One prediction of this model is that the a, b and c sets of genes must have restricted domains of action: each must act in only two out of the four whorls of the flower. The results of further experiments bore this out. In a flowering Antirrhinum, RNA copies – the telltale signs of gene activity – of certain b genes are found mainly in whorls 2 and 3, while RNA copies of a c gene are confined to whorls 3 and 4. Furthermore, the Pasadena group has found that in mutant Arabidopsis plants that carry a defective a gene, at least one c gene is active in all four whorls. Ordinarily, therefore, the activation of a genes must prevent c genes from being switched on in the outer two whorls of the flower. This explains the rule that when a does not sound, c is sounded instead.

How do these principles of flower development compare to those governing the development of animals? Enter the fruit fly. Like all insects, Drosophila melanogaster is a segmented organism, and each of its segments carries specific structures – wings, legs, antennae and so on (see Figure 4). Remarkably, the genetic origins of this segmental pattern are broadly similar to those of the whorl pattern of flowers. In fruit flies, developmental genes act in combinations to specify the characters of the various segments of the organism, just as they do in plants to specify the characters of whorls. And in both cases, moreover, many of the genes encode DNA-binding proteins.

Yet there some important differences. Many of the genetic events that fashion the fruit fly’s body plan occur before the egg has divided into clusters of cells. Gradients of gene activity are set up at an early stage in the cytoplasm of the egg and these then ‘label’ regions of the egg that are destined to produce different segments. We know much less about the early stages of flower development. In particular, it is unclear how the pattern of the a, b and c gene activites is initially set up in the flower meristem. As the meristem comprises many cells this pattern is unlikely to be the outcome of a system of gradients within individual cells. More probable is some kind of signal that passes between cells, turning on different combinations of genes in different sectors of the meristem. One of the biggest challenges for the future is to find out what this signal is. Might genes such as flo trigger specific interactions between cells which subsequently switch on the a, b and c events in the appropriate domains?

There is another important distinction between the fruit fly and the flower. Whereas the segments of a fruit fly all begin development at the same time, the whorls of a flower emerge sequentially. In fruit flies each segment starts to play its own ‘tune’ at about the same time, but with flowers the ‘tune’ of the outer whorl comes first. The fruit fly may be somewhat exceptional in this respect: many other insects, such as grasshoppers, produce their segments sequentially.

Whether a body plan is established synchronously or sequentially has important implications for the types of mutations one sees. Hence, the scientific literature is full of examples of mutant fruit flies with fewer segments than normal but devoid of reports of flies boasting more than the usual number of segments. In plants, by stark contrast, mutations can readily cause whorls to proliferate – witness the fulsome flowers of mutant Antirrhinums carrying a defective c gene (see Figure 1). Their centres are filled with extra whorls of petals or sepals.

One explanation for this difference is that segmentation in fruit flies proceeds by subdivision of an egg of limited size. Mutants that fail to subdivide correctly tend, therefore, to give fewer segments. Quite the opposite is true for flowers: because they grow sequentially, there is always the possibility of producing more and more whorls. This is likely when any one of the suite of genes required to bring the process to an end is hit by a disabling mutation. How deep the differences between animal and plant development run, only future research will tell.

Finally, what implications does our research have for agriculture and horticulture? Because they are the source of fruit, grain and seed for propagation, flowers are of immense agronomic importance. Identifying the genes that control their development brings us a step closer to being able to manipulate flowering by genetic engineering. One approach would be to engineer genes such as flo so that they can be activated either earlier or later in the year than usual. Controlling flowering in this way might enable farmers to produce more varieties of fruit and seed for longer periods of the year.

Enrico Coen and Rosemary Carpenter are at the John Innes Institute, Colney Lane, Norwich NR4 7UH