The ABC Model of Flower Development

A flower develops from a floral meristem, a small dome of undifferentiated cells at the tip of a floral shoot. Instead of continuing to make leaves indefinitely, this meristem lays down four types of lateral organs in four concentric rings, called whorls, and then stops growing. Whorl 1 (outermost) produces sepals, whorl 2 produces petals, whorl 3 produces stamens (the male, pollen-producing organs), and whorl 4 (the innermost, centermost region) produces the carpels that fuse into the pistil. This 1-2-3-4 arrangement — sepal, petal, stamen, carpel — is so conserved across angiosperms that its rare exceptions (like double roses, where extra petaled whorls displace stamens) were noted by naturalists long before anyone understood the genetics. The ABC model, developed principally by Enrico Coen, Elliot Meyerowitz, and colleagues in the late 1980s, explains this pattern with three classes of gene function, called A, B, and C, each expressed in a specific, overlapping domain across the four whorls:

Organ identity is thus not assigned by any single gene but by which combination of A, B, and C activity a cell experiences. Only three overlapping binary variables (A present/absent, B present/absent, C present/absent) are needed to generate four distinct organ identities — the same logic by which a small combinatorial code can specify a large number of outcomes, echoed later in this article's discussion of the animal Hox system. A second, equally important rule is that A and C are mutually antagonistic: wherever A-function is active it represses C-function, and vice versa. This antagonism is what keeps the two functions confined to non-overlapping regions (A in whorls 1–2, C in whorls 3–4) in the wild-type flower, and it is precisely this antagonism whose loss produces the classic homeotic mutant phenotypes described below. A later addition, E-function, provided by the redundant SEPALLATA1/2/3/4 genes, was found to be required as a partner for A, B, and C in all four whorls — without it, all organs revert to leaf-like structures. This extension is now called the ABCE model.

Step 1 — the floral meristem is patterned into four whorls. Long before organ identity genes act, "meristem identity" genes (such as LEAFY and APETALA1 in Arabidopsis) convert a vegetative or inflorescence meristem into a floral meristem, committing it to produce a finite, whorled set of organs rather than an indeterminate series of leaves. This is the "decision to flower" governed by the photoperiod and vernalization pathways; the ABC genes act downstream, patterning the organs within the meristem that this earlier decision has already created.

Step 2 — A, B, and C genes are transcribed in overlapping radial domains. In Arabidopsis thaliana, the classical assignments are:

AP2 is unusual among the A-function genes: it does not encode a MADS-box protein but an AP2/ERF-domain transcription factor, and its activity is regulated post-transcriptionally, largely through repression of its mRNA by the microRNA miR172, rather than by exclusion of its transcript from whorls 3–4. All the other genes listed — AP1, AP3, PI, and AG, together with the SEP genes — belong to the MADS-box family of transcription factors, named for the shared MADS DNA-binding domain first recognized in MCM1 (yeast), AGAMOUS (plant), DEFICIENS (snapdragon), and SRF (human).

Step 3 — MADS-box proteins dimerize and then assemble into higher-order "floral quartet" complexes. A single MADS-box monomer binds DNA weakly. Two MADS proteins first form a dimer through their MADS and adjacent K (keratin-like coiled-coil) domains, and this dimer binds a specific DNA sequence called a CArG box (consensus roughly CC-A/T-rich(6bp)-GG) in the promoters of downstream target genes. Crucially, two such dimers then associate into a tetramer — the floral quartet — which loops the intervening DNA into a small chromatin domain and switches target genes on or off combinatorially. The composition of the tetramer differs whorl by whorl:

• Whorl 1 (sepal): AP1–AP1–SEP–SEP-type complexes (A + E, no B, no C)
• Whorl 2 (petal): AP1/AP3/PI/SEP tetramer (A + B + E)
• Whorl 3 (stamen): AP3/PI/AG/SEP tetramer (B + C + E)
• Whorl 4 (carpel): AG–AG–SEP–SEP-type complexes (C + E, no A, no B)

This "floral quartet model," proposed by Günter Theißen and colleagues, explains mechanistically why organ identity is truly combinatorial at the level of protein complex composition, not just at the level of gene expression maps — different partner proteins genuinely assemble into different quaternary complexes that then bind different sets of target-gene enhancers.

Step 4 — A and C repress each other's expression domains. AP2 (A-function) represses AG (C-function) in whorls 1 and 2; AG represses AP2's activity (and more directly represses AP1) in whorls 3 and 4. This mutual, spatially restricted repression is what keeps A confined to the outer two whorls and C confined to the inner two whorls in a wild-type flower. Because the repression is mutual, loss of one function is not simply "no A" or "no C" — it is loss of A and consequent unopposed expansion of C into the A domain (or the reverse), which is exactly why loss-of-function A mutants show C-function phenotypes throughout the flower, and vice versa.

Step 5 — downstream target genes execute the organ-building program. The floral quartets act as transcriptional switches on batteries of downstream genes controlling cell division, cell expansion, pigment biosynthesis (in petals), pollen development (in stamens), and ovule formation (in carpels). The ABC genes themselves are "master regulators" — they do not build the organ directly but license entire downstream gene-expression programs appropriate to that organ's identity.

The ABC model rests on three converging lines of evidence, developed largely between the mid-1980s and early 1990s.

1. Homeotic mutant phenotypes in Arabidopsis thaliana and Antirrhinum majus. Classical and induced mutations in both species produced flowers with organs in the "wrong" identity, but always in a systematic way — never randomly. In Arabidopsis, loss-of-function mutations in apetala2 (ap2) (A-function) produce flowers in which whorl 1 organs become carpel-like and whorl 2 organs become stamen-like; the flower reads carpel–stamen–stamen–carpel instead of sepal–petal–stamen–carpel. Loss-of-function mutations in agamous (ag) (C-function) do the mirror-image transformation: whorl 3 becomes petal-like and whorl 4 becomes a new flower nested inside the old one (sepal–petal–petal–sepal, repeating indefinitely, because carpel identity — which normally terminates meristem growth — is also lost, so the meristem does not stop producing whorls). Loss-of-function mutations in the B-function genes apetala3 (ap3) or pistillata (pi) convert whorl 2 petals into sepals and whorl 3 stamens into carpels, giving sepal–sepal–carpel–carpel. Independently, in snapdragon, mutations in deficiens (def) and globosa (glo) — the B-function orthologs — produced the identical sepal–sepal–carpel–carpel transformation, and a C-function-like mutant called plena (ple) gave the AG-like repeating-flower phenotype. The fact that mutations in two independently evolved lineages, studied by two different groups (Meyerowitz's lab on Arabidopsis; Coen's lab on Antirrhinum), produced matching combinatorial transformations was the strongest early argument that a shared, conserved genetic logic — not species-specific idiosyncrasy — was at work.

2. Molecular cloning and in situ expression mapping confirmed the predicted domains. Once AP1, AP2, AP3, PI, AG and the snapdragon orthologs DEF, GLO, PLE were cloned in the late 1980s and early 1990s, in situ hybridization directly visualized their mRNA expression in developing floral meristems. The observed domains matched the model's predictions almost exactly: AP1/AP2 transcripts (or protein activity, in AP2's case) restricted to whorls 1–2, AP3/PI to whorls 2–3, and AG to whorls 3–4 — with the AG domain visibly expanding into whorls 1–2 in ap2 mutants, and the AP2-regulated domain expanding into whorls 3–4 in ag mutants, exactly as the mutual-repression hypothesis predicted.

3. Ectopic (misexpression) experiments showed the genes are sufficient, not just necessary, for organ identity — the decisive "gain of function" test. The most rigorous test of a model in which genes specify identity is to express them where they normally are not, and ask whether they impose their organ identity there. When AG (C-function) was placed under a constitutive promoter and expressed throughout the flower (and even in leaves in some transgenic lines), whorls 1 and 2 organs were converted toward stamen- and carpel-like identities — carpel/stamen identity followed AG expression wherever it was forced to occur. Reciprocally, constitutive co-expression of AP3 and PI (B-function must act as a heterodimer) converted sepals into petals and carpels into stamen-like organs, again showing that B-function is sufficient to impose whorl 2/3 identity outside its normal domain. These gain-of-function experiments were essential because loss-of-function mutants alone can only show a gene is necessary; misexpression shows it is sufficient — together, necessity and sufficiency constitute the classic double test of causal gene function used throughout developmental genetics. Later, the discovery of the sepallata quadruple mutant (sep1 sep2 sep3 sep4, since single and triple mutants show near-normal flowers due to redundancy) showed that in the complete absence of E-function, all four whorls revert to leaf-like organs even though A, B, and C genes are still expressed — proving that A, B, and C activity alone is not sufficient without an E-function partner, and completing the case for the modern ABCE model. This is itself a clean necessity/sufficiency result: E-function genes are necessary for organ identity in every whorl, and adding SEP3 to an A+B or B+C combination is sufficient to convert vegetative leaves into petal-like or stamen-like organs in overexpression experiments.

Example 1 — predicting the phenotype of a double mutant. Suppose you cross a plant homozygous for a loss-of-function ap2 allele (A-function absent, so C-function has expanded into all four whorls) with a plant homozygous for a loss-of-function ag allele (C-function absent, A-function expanded into all four whorls), and examine the ap2 ag double homozygote in the F2. Work through it whorl by whorl using the model's logic, not memorized phenotypes:

• With both A and C absent everywhere, only B-function and E-function remain as spatial information (B is still restricted to whorls 2–3; E is present everywhere).
• Whorl 1: no A, no C, no B → only E present → organ reverts toward a leaf-like structure.
• Whorl 2: no A, no C, B present → B + E → petal.
• Whorl 3: no A, no C, B present → B + E → petal.
• Whorl 4: no A, no C, no B → only E present → leaf-like structure.

The predicted (and experimentally observed) ap2 ag double mutant phenotype is therefore leaf–petal–petal–leaf, an indeterminate, repeating structure — a striking confirmation that removing both A and C does not simply give "no organs" but reveals the residual B+E pattern directly, exactly as the combinatorial arithmetic predicts. This kind of epistasis analysis — reasoning out a double mutant from single-mutant logic before consulting the literature — is the core analytical skill the ABC model is meant to teach.

Example 2 — reading an unusual natural double flower. A gardener notices that a cultivated variety of a normally single-flowered ornamental plant produces "double flowers": far more petal-like organs than normal, extending inward to occupy almost the entire whorl 3 and whorl 4 territory, with drastically reduced or absent stamens and carpels (and hence reduced fertility, since seed set requires functional carpels). Using the ABC logic, what is the most parsimonious genetic explanation? Since petal identity requires A + B, and the expansion of petal identity into whorls 3 and 4 (normally B+C and C-only, respectively) implies that C-function has been lost or reduced there while A and B remain active — reproducing the classic agamous-type loss-of-function transformation (sepal–petal–petal–sepal, repeating). This is in fact the correct historical explanation for many popular ornamental doubles, including cultivated roses and carnations: they typically carry loss-of-function or reduced-function mutations in the AGAMOUS ortholog, and the resulting loss of carpel-forming, meristem-terminating C-function both converts stamens to petals and prevents the flower from ever "finishing," so it keeps initiating new petal-bearing whorls, producing the many-layered "double" flower prized horticulturally — at the direct cost of the plant's own seed fertility, since it has, in effect, lost its reproductive organs to attractive ones.

The ABC model emerged from the near-simultaneous, complementary work of several researchers, but is most closely associated with Enrico Coen (John Innes Centre, Norwich, working on Antirrhinum majus) and Elliot M. Meyerowitz (Caltech, working on Arabidopsis thaliana). Coen's group had access to a rich collection of classical and transposon-induced snapdragon mutants — snapdragon transposons made cloning genes by transposon-tagging tractable years before genome sequencing existed — while Meyerowitz's group exploited Arabidopsis's small genome, fast generation time, and ease of transformation to clone and characterize the corresponding genes and perform the decisive ectopic-expression experiments. Coen and Meyerowitz co-authored the 1991 synthesis in Nature ("The war of the whorls: genetic interactions controlling flower development") that gave the ABC model its now-standard name and its clean logical formulation. The subsequent completion of the model with E-function was principally the work of Martin Yanofsky's group (UC San Diego), who identified the SEPALLATA genes and demonstrated the quadruple-mutant leaf-reversion phenotype, and of Günter Theißen, who proposed the floral quartet model explaining the ABC(E) logic at the level of MADS-protein complex assembly.

The ABC(E) model is not a museum piece; it underpins active work in agriculture, horticulture, and evolutionary developmental biology. Plant breeders manipulating the AGAMOUS ortholog (or its regulators) deliberately engineer or select for "double flower" ornamentals — many commercial roses, carnations, and camellias owe their extra petals to naturally occurring or induced C-function loss, exactly as worked through in the example above. Conversely, understanding B- and C-function genes has informed efforts to engineer male-sterile lines for hybrid seed production in crops, since converting stamens to another organ identity is a direct route to preventing self-pollination without hand-emasculation. In cereal biology, ABC-gene orthologs help explain floral variation among grasses, including differences relevant to grain yield architecture in wheat, rice, and maize. In evolutionary developmental biology ("evo-devo"), comparing ABC(E) gene expression across angiosperm lineages — including basal lineages like Amborella and even gymnosperm MADS-box relatives — has illuminated how floral novelty (double perianths, unusual symmetry, sex-specific flowers in dioecious species) can arise from relatively modest shifts in the boundaries of these ancient, deeply conserved expression domains, making the ABC model a foundational case study in how small changes to a combinatorial regulatory code can generate large-scale morphological diversity.

Ask the tutor: How would the ABC model predict the phenotype of a plant with a gain-of-function AG mutation that is expressed in every whorl?