Günter Theißen

Flowers are complex structures. Theytypically consist of four types of organarranged in four whorls: sepals, petals,

stamens and carpels. No wonder that theirdevelopment is complicated. Nevertheless,there seem to be simple rules that underliethis process, as was realized when it becamepossible to analyse ‘floral homeotic mutants’— plants with flowers that have seeminglynormal floral organs in places where organsof another type are normally found.

In the model plant thale cress (Arabidopsisthaliana), homeotic mutants are categorizedas A, B and C. Ideal class A mutants havecarpels in the first whorl instead of sepals, andstamens in the second whorl instead of petals.Class B mutants have sepals rather than petalsin the second whorl and carpels rather thanstamens in the third; and class C mutants havepetals instead of stamens in the third whorland sepals instead of carpels in the fourth.

The existence of defined classes of mutantssuggested simple combinatorial models, suchas the ‘ABC’ model of flower development.This proposes three different floral homeoticfunctions to explain how the different floralorgans adopt their unique identities duringdevelopment. These functions are termed A, Band C, with A specifying sepals in the first floralwhorl; A&B, petals in the second; B&C, sta-mens in the third; and C, carpels in the fourth.Because they had been identified by mutantanalysis, it was clear that the homeotic func-tions are provided by sets of floral homeoticgenes. In Arabidopsis there are at least two A-function genes, two B-function genes andone C-function gene, all of which encode

transcription factors — proteins that recog-nize specific DNA motifs and thereby regulatethe expression of the genes that contain them.

Besides elegance, simplicity was certainlyamong the most attractive features of theABC model, which found its way into modern textbooks and numerous reviews.However, it was soon realized that the modelis incomplete. Expression of ABC genesthroughout a plant does not transformleaves into floral organs. Thus the ABC func-tions, although necessary, are not sufficientto superimpose floral organ identity on a leaf development programme.

Another class of genes (SEPALLATA) isnow known to be required and, togetherwith the ABC genes, is sufficient for specifi-cation of petals, stamens and carpels. Howdo these SEP genes fit into the ABC model?Some have argued that they contribute to Band C functions, but they are more likely tohave an additional function.

Indeed, despite these insights, I wonderwhether floral homeotic functions are still auseful concept at all. Now that we know thatthere are at least as many floral homeoticfunctions as there are types of floral organ,the concept no longer provides a useful sim-plification. Wouldn’t it suffice to refer to thefour different states of floral organ identityon the one hand, and to the combinations offloral homeotic genes that specify theseorgan identities on the other? This seems tobe a particularly useful way of thinking aboutthe control of floral organ identity in thelight of new evidence about how floralhomeotic genes interact and exert their functions at the molecular level.

Proteins that are encoded by the floral

homeotic genes of Arabidopsis are nowknown to bind to DNA as multimeric complexes that contain B-function and SEPproteins, as well as either an A-function or aC-function protein. These are exactly thecombinations of proteins that are sufficientto specify petal or stamen identity, respec-tively. This finding immediately suggestsfuture research goals: to define the exactstructures of the transcription-factor com-plexes inside the living plant cell; to identifythe target genes for which transcription isregulated by the binding of the complexes; toexplain the gene specificity of that binding;and to study how target-gene functionbrings about floral organ identity. Neitherthe concept of floral homeotic functions nor the ABC model will offer any help inachieving these exciting ambitions.

Was the ABC model an intellectual detour,or even an obstacle to scientific progress? Notat all. It provided a working hypothesis withnumerous implications that could be experi-mentally tested, so it was an incredibly usefulconcept for a while. But scientific terms andconcepts should be continually evaluatedwith respect to their heuristic value. Now thatwe have a direct understanding of the molecu-lar mechanisms involved, there may no longerbe a need for the abstract concept of floralhomeotic functions to be squeezed betweenfloral homeotic genes and floral organ identi-ty. ABC functions may now obscure, ratherthan enlighten, our understanding of the linkage between molecular genetic events andfloral phenotype.

The ABC model undoubtedly attractedmany enthusiastic researchers to studyflower development, and as such formed thebasis for rapid progress in the field. So itcould be that, ultimately, the success of theABC model contributed to its abolition. ■

Günter Theißen is at the Institute for Genetics,Friedrich Schiller University, Philosophenweg 12, D-07743 Jena, Germany.

FURTHER READINGEgea Gutierrez-Cortines, M. & Davies, B. Trends PlantSci. 5, 471–476 (2000).Goto, K., Kyozuka, J. & Bowman, J. L. Curr. Opin. Genet.Dev. 11, 449–456 (2001).Jack, T. Trends Plant Sci. 6, 310–316 (2001).Theißen, G. & Saedler, H. Nature 409, 469–471 (2001).

Genetics of identityconcepts

NATURE | VOL 414 | 29 NOVEMBER 2001 | www.nature.com 491

FlowerdevelopmentThe story of the outmoded ABCmodel shows that even the mostsuccessful concepts must becontinually re-evaluated.

Floral combinations: the wildflower Hepatica nobilis

DIE

TM

AR

NIL

L/B

BC

WIL

D

© 2001 Macmillan Magazines Ltd