The ABC model of flower development: then and now

It is straightforward to review the state of the field of genetic approaches to flower development in the late 1980s, as we published a historical review as a background to our work – like the 1991 paper (Bowman et al., 1991), this was also in Development (Meyerowitz et al., 1989). We pointed out that what later would be considered to be mutants with altered flower development – in types, numbers and positions of the floral organs (sepals, petals, stamens and carpels) – were known from ancient Greece, and had been studied in great detail in the 19th century, as had normal flower development, the microscopic study of which was first extensively published by Payer in 1857 (Payer, 1857). Plant genetics was of course well developed at the time, as it had been in progress since the original discoveries of Mendel in 1865. Homeotic flower variants, in which one floral organ type is replaced by another, were specifically recognized and studied in the 19th century (e.g. Masters, 1869) under the term ‘metamorphy’, while Bateson introduced the concept of homoeosis in 1894. A collection of homeotic mutants of snapdragon existed by the 1930s in the laboratory of Erwin Baur and co-workers (Stubbe, 1966). Thus, both the materials (at least in other species than the one we used) and methods (not scanning electron microscopy, but microscopic and detailed analysis of early flower development) used in our 1991 paper were available more than 100 years earlier, and the exact materials with which cognate experiments could have been carried out existed more than 50 years earlier. The key to our model was the use of double and triple mutants. The Baur laboratory made many mutants, but none (of which we have found a record) that involved more than a single homeotic mutant.

Our paper therefore serves as a counter-example to the notion that advances in biology derive only from advances in methodology. Our 1991 work could have been carried out a century earlier, except that the conceptual framework of biology at that time (especially the notion of a regulatory gene), was such that no one had thought to do it.

We began with a set of four homeotic mutants of Arabidopsis in which fairly normal floral organs were found in floral whorls where they would not be expected in wild-type flowers. The mutants were obtained from the generosity of colleagues, particularly Maarten Koornneef, then at the University of Wageningen (The Netherlands), and had already been described in detail as single and double mutants in the first issue of The Plant Cell (Bowman et al., 1989). We concluded that the genes function in overlapping fields that occupy two adjacent floral whorls, and that they ‘act in allowing cells to recognize their position in the developing flower’.

Subsequently, additional mutant alleles were obtained and analyzed, revealing the likely null phenotypes, and triple mutants were generated (Fig. 1; for further commentary, see Bowman, 2010). The new data were the foundation on which we built ‘A simple model’, now known as the ABC model of floral organ identity. The letters came from the three overlapping fields, named A (APETALA2 gene function, AP2), B (APETALA3 and PISTILLATA, AP3/PI) and C (AGAMOUS, AG). AP2 was proposed to function in whorl 1 to define sepals, and AG in whorl 4 to control carpel identity. The overlaps explained the combinatorial roles of AP2 and AP3/PI in whorl 2 that normally defined petals, and AP3/PI and AG in whorl 3 that specified stamens. A key new component, without which the model does not work and that distinguishes this model from others, was the proposal that genes acting in the A and C fields, AP2 and AG, were mutually antagonistic, something shown at the molecular level beginning very shortly afterwards (Drews et al., 1991). It was aesthetically satisfying to be able to account for the eight floral phenotypes in terms of the model, and its attraction lay in its simplicity and predictive power. Even so, it only explained organ identity, and we listed a series of ‘unexplained complications’ associated with mutant phenotypes, including mosaic organs, secondary flowers, floral indeterminacy and missing organs.

We chose to send the manuscript to Development rather than to a plant-oriented journal on the encouragement of editor Keith Roberts, who seemed to share our holistic view that developmental principles are shared across kingdoms. And, color illustrations were free and there were then no page limits. The two reviewers were positive, with one saying ‘This is a stimulating paper which should have a large impact on the field. Interest rating 9.’ The other reviewer thought it ‘potentially an excellent paper’, but ‘the writing style could be improved’ and wanted it shortened by half. However, the first reviewer, and the editor, disagreed, and its 20 pages stood. We suggested a ‘fruit salad’ cover photo of the wild-type and all mutant phenotypes that was also accepted (Fig. 2).

The ABC model was popularized in a review in Nature published later in 1991 by the senior author and Enrico Coen, whose group had been making parallel new findings of similar homeotic mutants in snapdragon (Antirrhinum majus) (Coen and Meyerowitz, 1991). In this review, functions were distinguished from regions (or fields) by using a, b and c for functions and A, B and C for regions, although it is now customary to capitalize the functions. The importance of comparative findings in Antirrhinum was highlighted by the earlier contribution of Zsuszanna Schwarz-Sommer and colleagues, who independently generated a floral organ identity model that proposed the equivalent of B and C functions, but lacked A function and was not tested using multiple mutants (Schwarz-Sommer et al., 1990).

The ABC model is still widely used as a framework for understanding floral development today (Krizek and Fletcher, 2005; Causier et al., 2010). The impact of the Development paper is reflected in its continued high citation rate (it still gathers over 30 citations a year according to the Web of Knowledge, http://wokinfo.com). Significant advances since 1991 include findings that: (1) all genes, except AP2, encode MADS transcription factors (e.g. Weigel and Meyerowitz, 1994); (2) another A function gene, APETALA1, exists in Arabidopsis (Bowman et al., 1993; Gustafson-Brown et al., 1994); (3) four other MADS genes (SEPALLATAs) are involved in establishing the floral nature of flower organs in Arabidopsis (Pelaz et al., 2000) (often called E function, although using our terminology these would be meristem identity, not organ identity, genes); (4) SEP proteins likely act in multimeric combination with A, B and C function MADS proteins (the quartet model) (e.g. Melzer and Theissen 2009); (5) AP2 transcripts are regulated post-transcriptionally by microRNAs (Aukerman and Sakai, 2003; Chen, 2004); and (6) AP2 is a direct negative regulator of AG expression, a very recent finding (Dinh et al., 2012).

Perhaps owing to its relative simplicity, and to the ubiquity of the appreciation of flowers in human society, the ABC model was rapidly introduced into university textbooks, not only those focused on developmental biology (e.g. Wolpert and Tickle, 2011), but also to first year general biology textbooks (e.g. Campbell et al., 1999; Freeman, 2008), cell biology texts (e.g. Alberts et al., 1994) and those focused on genetics (e.g. Griffiths et al., 1993; Sanders and Bowman, 2012). The ABC model is even being taught to high school students in some locales (Fig. 3).

Given the conservation of floral organ position across angiosperms, it was a natural question as ask whether the ABC model could be applied to all flowering plants – and not only to Arabidopsis and snapdragon, where it was developed. Its integration with new findings that were, at that time, defining the course of angiosperm evolution was catalyzed by two Keystone Symposia on evolution and plant development (early evo-devo) held at Taos (New Mexico) in 1993 and 1997. Two early tests of its general applicability were observations that showed the occurrence of B-class mutants in monocots (Ambrose et al., 2000; van Tunen et al., 1993). Such observations formed the foundation for continuing studies into the diversity of flower architecture, from orchids where the complex perianth is patterned by differential expression of multiple B-class gene paralogs (Mondragon-Palomino and Theissen, 2011), to the ‘inside-out’ flowers of Lacandonia schismatica, where central stamens are surrounded by carpels (Alvarez-Buylla et al., 2010).

The ABC model has also illuminated the evolutionary appearance of flowers, a puzzle that has occupied botanists since Charles Darwin. Both B- and C-class orthologs are found in gymnosperms, with C-class genes expressed in both male and female reproductive structures and B-class expression in male reproductive tissues (Mouradov et al., 1999; Shindo et al., 1999; Sundstrom et al., 1999; Tandre et al., 1998). Thus, the specification of stamens and carpels seems to have its origin in the common ancestor of seed plants, and the mystery of flower evolution lies in ‘tinkering’ (Jacob, 1977) with pre-existing genetic machinery. None of the ABC-type MADS box genes appears to exist outside angiosperms and gymnosperms, suggesting that their origin lies in extensive gene duplications of an ancestral MADS box gene in the lineage leading to seed plants (Floyd and Bowman, 2007).

In summary, althought the relatively simple ABC model we proposed in 1991 has grown in complexity over the past 21 years – particularly with increasing knowledge of its molecular mechanism – it still provides a basis for our understanding of flower development and morphology in model systems, and across evolution.