MADS boxes, flower development, and evolution


I’ve been writing a fair amount about early pattern formation in animals lately, so to do penance for my zoocentric bias, I thought I’d say a little bit about homeotic genes in plants. Homeotic genes are genes that, when mutated, can transform one body part into another—probably the best known example is antennapedia in Drosophila, which turns the fly’s antenna into a leg.

Plants also have homeotic genes, and here is a little review of flower anatomy to remind everyone of what ‘body parts’ we’re going to be talking about. The problem I’ll be pursuing is how four different, broadly defined regions of the flower develop, and what that tells us about their evolution.


This is a greatly simplified flower, and for this article we can safely ignore most of the details. At this point, all we care about is that the parts of a flower can be categorized into four broad types, all organized in concentric rings. The outermost layer, Whorl 1, are the sepals. Sepals tend to resemble leaves, and are often green and photosynthetic. The next layer in, Whorl 2, are the petals—typically the pretty, colored part of the flower, specialized to attract pollinators. Whorl 1 and whorl 2, which do not contribute directly to the sex organs of the flower, are called the perianth, and the remaining two whorls form the reproductive organs proper. Whorl 3 is the ‘male’ part of the flower, containing the stamens, where pollen forms. The central whorl, Whorl 4, consists of the carpels, where ovules form; carpels can fuse into a central organ, the pistil.

There is, of course, an immense amount of diversity in the details of each of these pieces of a flower, but for now all we need to know is that there are four parts to a flower, sepals, petals, stamens, and carpels, and those are the body parts that have to be formed in their correct positions in development. And, since I warned you at the beginning that I was going to talk about homeotic mutants, you know that there exist mutants that transform flower parts into other flower parts. Here are some examples in Arabidopsis:


The apetala2 mutant fails to make either sepals or petals—they are all transformed into stamens and carpels. That plant on the left has an excess of sex organs, and they are all hanging out there naked and unadorned. The middle plant, the apetala3 mutant, also lacks petals, and it also has no stamens, only sepals and carpels. And on the right is the agamous mutant, a vegetable eunuch with no sex organs at all, only sepals and petals.

Observations of many different mutants and combinations of mutations has supported a relatively simple model of pattern specification in the flower, called the ABC model, diagrammed below. There are 3 classes of genes expressed in overlapping, concentric rings. The A class is expressed in the outermost ring and C is expressed in the center; B is expressed at the boundary of A and C. If A is expressed in a cell, it goes on to form a sepal. If C is turned on, it forms a carpel. Petals are formed where both B + A are active. Stamens are formed with the combination B + C.


Now the patterns of the mutants I illustrated above should start to make sense. The apetala2 gene falls into the A class, and losing it means there is no A gene expression anywhere: all the plant has is B and C. All it can make are C products and B + C products, or carpels and stamens. Looking at the diagram below, you can also see that in the absence of A, C is expressed everywhere—from that, you’d expect that one thing A genes must also do is inhibit the expression of the C gene.


The agamous gene, on the other hand, is in the C class. Lack of the C gene means the plant can’t make carpels (C) or stamens (C + B), and all the flower parts are either sepals (A) or petals (A + B). As you can also guess from the diagram above, C must also work to suppress A expression, and in its absence A gets turned on all across the flower.

Earlier, I showed a picture of a third mutant, apetala3. By now you should be a master of the regulatory logic behind flowering plant organization, so I shouldn’t need to say what class apetala3 belongs to, and you should be able to figure out which plant parts form in each of the four whorls all by yourself.

In animals, the core homeotic genes that define body parts along the anterior-posterior axis are called the Hox genes. They all possess a common motif, a 180-base-pair sequence called the homeobox, which gets translated into a 60 amino acid sequence called the homeodomain. The homeodomain is the region of the protein that binds to DNA.

In plants, all of the homeotic genes I mentioned above are called MADS box genes. The MADS box is another common motif, one that is 174 nucleotides or 58 amino acids long. Like the homeobox, the MADS box is a DNA binding domain.

It’s name is an acronym for the four genes in which this sequence was first identified: MCM1 (found in yeast), Agamous (from Arabidopsis; you just met it above), Deficiens (a snapdragon gene), and SRF (a human gene). As is obvious, the MADS box is also found in animals, including us, just as homeobox genes are found in other organisms than animals.

Where we differ is in the frequency of these genes. The yeast, Saccharomyces cerevisiae has a grand total of four MADS box genes in its genome; fruit flies and nematodes have only two each. The Arabidopsis genome has 82 different MADS box genes. Other flowering plants have equivalent numbers.

What’s clear is that the evolution of flowering plants has been accompanied by an explosion of diversity in the MADS box genes that are responsible for patterning the flower. One thing evolutionary theory now allows us to do is superimpose the distribution of MADS box genes, which we can relate to flower morphology, on plant phylogeny, and come up with models for the evolution of flowers. Here, for instance, is a diagram summarizing the lineages of the flowering plants from Irish (2003), with her interpretation of how the acquisition of new MADS box genes by duplication during evolution allowed the proliferation of novel flower forms.

A simplified phylogeny of the angiosperms. Several recent studies have all arrived at a virtually identical hypothesis of angiosperm phylogeny. The presumed relationships of the major groups of angiosperms, including the eudicots, monocots and basal angiosperms, are shown. In addition, an example of a familiar species is indicated for each group.

Based on this rooting of the angiosperms, it seems likely that the primitive angiosperm flower was small, with few floral organs and lacking a well developed perianth, similar to that of Amborella. The fossil record, although sparse, is consistent with this possibility in that some of the oldest described fossil flowers are small and few-parted, similar to flowers of many of the most basal extant angiosperm taxa. Together, these data suggest that the primitive angiosperm flower contained reproductive organs and a few sterile subtending organs, but that the acquisition of distinct petals and sepals occurred during the angiosperm radiation. In fact, based on morphological evidence, petals appear to have evolved multiple times independently. Petals are thought to have evolved either as modifications of stamens (for instance in the Ranunculales, core eudicots and monocots) or as modifications of bracts or other leaflike structures (in the Magnoliales, Illiciales, and Laurales). The elaboration of these perianth organs resulted in larger, showier flowers, which has been proposed to be an evolutionary novelty that facilitated outcrossing by attracting pollinators. Based on both the phylogenetic data and fossil evidence, a number of other inferences can be made about the probable character traits in the earliest angiosperm flowers. Phyllotaxy, or the pattern in which organs arise around the floral axis, is generally spiral in basal angiosperm species and it is thought that this condition is ancestral. Variability in the number of particular organ types is associated with spiral phyllotaxy, and thus is also considered ancestral. The primitive angiosperm flower was also likely to have been bisexual; unisexual flowers, in which either stamen or carpel development is arrested, are thought to have been derived many times.

Isn’t that cool? I do appreciate it when molecular biology, development, evolution, and something as pretty and diverse as flowers all converge on a beautiful, elegant answer.

There is much more detail in the Irish paper. For example, the homologies of specific MADS box genes can be used to map out a pattern of gene duplications during evolution, as shown below. Although they fall into those 3 ABC classes, remember that there are many more than just 3 genes that all work together to form the intricacies of a flower.

MADS box gene duplications. Genes corresponding to the different MADS box gene lineages are indicated in the clades where they have been identified. FUL-like genes have been identified in the Magnoliales, monocots and Ranunculales, while the derivative euAP1 and euFUL lineage genes have been identified in members of the core eudicot clade. Similarly, the paleoAP3 and AG-like lineage genes have been found in the basal angiosperms, monocots and basal eudicots, while the derivative AP3 lineage (euAP3 and TM6) and derivative AG (euAG and PLE) lineage genes have only been identified in core eudicot species.

There is still much more to learn. I mentioned before that the models of early pattern formation in Drosophila were almost certainly unique to long germ-band insects, and that although the same molecules get used in other species, there are more complications and elaborations that need to be worked out. Irish says something similar of flower evolution: the ABC model works well for the core eudicots, but there is evidence to suggest that other mechanisms are at play in other angiosperm groups.

Irish VF (2003) The evolution of floral homeotic gene function. BioEssays 25:637–646.

Wolpert L, Beddington R, Jessel T, Lawrence P, Meyerowitz E, Smith J (2002) Principles of Development. Oxford University Press.


  1. says

    Ooo, am I the first one to comment? Cool.

    Thanks for the phytocentric post. And thanks for being so gracious as to deliberately leave out one of the few things I can add, which is that the ABC model of flower development has its roots (hoho) in work done in the late 18th century by Goethe (yes, that Goethe). Enrico Coen’s written about it a lot in his ‘Art of Genes’ book and this article. I’ll leave people to follow the links, rather than attempt to summarise his work…

  2. G. Tingey says


    I grow a few things, but … how does this relate to the “conifers”, or to the non-flowering plants without “cones” and the cryptogametes????

    Come to that, what about those flowering plants that have single fused petals.
    The Orchids and the Mint/Labiate famililes come to mind.

  3. llewelly says

    Interesting article. Did you choose the apetela[2,3] mutants, and the phrase ‘hanging out there naked’ to bring a smile to the lips of those of us who noted the similarity between ‘homoeotic’, and ‘homoerotic’ ?

  4. amph says

    Both plant and animal homeotic genes are apparently regulated by Polycomb-group genes. Isn’t that amazing? What do we have to think of that in terms of molecular evolution? MADS and Hox genes are absolutely non-homologous/orthologous.

  5. Steviepinhead says

    Well, I expect most online dictionaries could provide a pronunciation for “whorl” which is related to “whirl.”

    But on the off-chance that you really care, you won’t go to far wrong rhyming it with “quarrel.” Though there’s not quite so much of the second syllable in “whorl,” by gorl.

  6. Steviepinhead says

    Uh, I meant “whorl” and “whirl” are etmylogically related.

    Not that the vowel sounds are the same.

  7. says

    What I find most interesting is that animals have MADS box genes and non-animals have homeobox genes. I had naively assumed these things would have arisen after their lineages diverged.

  8. Cinclus says

    Does anyone know what function(s) the MADS genes perform in animals, and the role(s) of HOX genes in plants?

  9. says

    Oh, no — these are transcription factors. Transcription factors are found in bacteria, too. The ability to modulate gene expression is a property inherited from our single-celled ancestors.

  10. amph says

    Aesmael & Cinclus: Plants have homeobox genes, but no Hox clusters. Hox is a particular flavor of homeobox, not a synonym (although historically an acronym). MADS factors have various functions in animals, but these are not related to regional specification, as those of homeotic genes are. SRF (Serum Response Factor) is an example of an animal MADS factor.

  11. says

    Well, I sit corrected. I will not claim I understand now but I think I see where my outsiders perspective caused me to make bad assumptions. Thank you.

  12. says

    I’ve been poking around on homeotic genes and there’s
    a small bit in here that confuses me: the
    entry says that plants have *Hox* genes but they
    don’t serve homeotic functions.

    From what I’ve been able to dope out (I am not
    a scientist) *Hox* genes are homeotic homeobox genes.
    So is that just a minor slip of wordcraft, meaning
    plants have homeobox genes?

    Not nitpicking, honestly puzzled. The whole
    developmental gene field has some confusing