which is called SQUAMOSA, does not carry out 

 function A. Inactivating the SQUAMOSA gene leads 

 to fewer flowers and more inflorescence branching 

 in the snapdragon, but it does not interfere with the 

 formation of sepals or petals. 



Why should API in arabidopsis act differently 

 than its homologues do in other species? Vivian 

 Irish and I examined API, SQUAMOSA, and their 

 homologues in fifty-four disparate angiosperm 

 species, then constructed an evolutionary tree of 

 all the homologues. The tree showed at what point 

 gene duplications had taken place during the evo- 

 lution of flowering plants, much the way the evo- 

 lutionary tree of a family of animal or plant species 

 shows at what point new species arise. 



We discovered that an important genetic duplica- 

 tion had taken place in a large assemblage of related 

 species called the core eudicots. Many familiar 

 plants belong to the core eudicots, including arabi- 

 dopsis, daisies, oaks, roses, snapdragons, and toma- 

 toes. Because of the gene duplication, all the species 

 of core eudicots carry API homologues that belong 

 to two groups; for simplicity, I'll call them group X 

 and group Y. The API gene of arabidopsis and the 

 SQUAMOSA gene of the snapdragon belong to 

 group X. The other homologues fall into group Y, 

 in which we also discovered a second doubling. 

 That brings the total number of predicted AP I ho- 

 mologues in the genome of each species of core eu- 

 dicots to three: one homologue from group X and 

 two from group Y. 



But the mustard family, including arabidopsis, is 

 different from other eudicots. Mustards, it seems, 

 have lost one gene from group Y, and duplicated 

 API itself, the gene from group X. Thus, ara- 

 bidopsis still has three API homologues, like the 

 other core eudicots, but two belong to group X 

 and only one belongs to group Y. Because copies 

 of genes can take on new roles, it is likely that the 

 group X genes and the group Y gene in arabidop- 

 sis have divvied up the tasks of 

 flower formation differently than 

 the three API homologue genes 

 have in other core eudicots. 

 That probably explains why 

 API acts differently in ara- 

 bidopsis than its homo- 

 logues do in other species. 



Bird of paradise 



38 NATURAL history June 2006 



Thus, the evolutionary history of the API ho- 

 mologues shows that the A part of the ABC 

 model may exist only in the mustard family. But a 

 closer look at the evidence suggests that the A part 

 may not even exist there. The API gene in ara- 

 bidopsis performs not just the official role ascribed 

 to A genes, forming petals and sepals. API also 

 plays a more basic role, one that is outside the 

 scope of the ABC model: directing the meristem 

 to form a flower. Recall that, as I mentioned ear- 

 lier, when API is experimentally inactivated, leaves 

 and branches form instead of sepals and petals. The 

 resulting structure resembles a branched inflores- 

 cence more than it does a flower. 



The other two API homologues in arabidopsis 

 also share in directing the meristem to form a 

 flower. For example, if API and the other gene 

 belonging to group X, which is called CAULI- 

 FLOWER, are inactivated, flowers just don't form. 

 Instead, the inflorescence proliferates a dense head 

 of branches, like a tiny cauliflower. In fact, a study 

 of cultivated cauliflower (also a member of the 

 mustard family) has shown that a defect in its ho- 

 mologue of the CAULIFLOWER gene is probably 

 responsible for the characteristic dense white curds. 



In species outside the mustard family, when API 

 homologues are inactivated, only the sepals, and not 

 the petals, often fail to form properly. But in nearly 

 all species, the inactivation reduces flowering and 

 increases branching. That finding is strong evidence 

 that the fundamental role of AP I homologue genes 

 is helping direct the flower to form in the first 

 place — not directing sepal and petal formation. API 

 homologues are therefore more accurately de- 

 scribed as floral-meristem-identity genes — genes 

 that direct a meristem to become a flower. 



In fact, it appears that whenever sepals fail to 

 form properly, flowering itself is reduced. Forming 

 flowers and forming sepals, therefore, seem to be 

 controlled by the same API homologues, acting 

 early in the development of the meristem. Once 

 the API homologues (with the help of several 

 other floral-meristem-identity genes) instruct an 

 early meristem to grow into a flower, the forma- 

 tion of sepals is also set in motion. Later, in the de- 

 veloping floral meristem, B, C, and E genes kick 

 in to make petals, stamens, and carpels. 



In short, no genes divvy 

 up the developmental work in 

 way function A was originally de- 

 fined. Furthermore, the genes that were 

 thought to provide function A 

 act well before the organs 

 begin to specialize, during a 

 developmental stage that the model 



