How genome organisation and architecture respond to variations in selection regimes is indeed poorly known at the microevolutionary scale. Supergenes are tight clusters of genes which simultaneously control the variation of various components of a complex trait, involving elements with different ontology1-4. Supergenes typically evolve in response to strong epistatic selection between neighbouring genes. They are usually maintained by some form of balancing selection, but the modalities of their evolution are still obscure and require empirical investigation. I propose to take advantage of the remarkable supergene controlling wing-pattern mimicry balanced polymorphism in the tropical butterfly Heliconius numata to investigate the structure and evolution of supergenes at the molecular level. The supergene P is a positional homologue of a loose cluster of loci controlling wing pattern mimicry in a related species. Positional cloning revealed the supergene is situated in a local inversion. Recombination is largely suppressed around P and two groups of haplotypes segregate in perfect association with wing pattern in natural populations. I propose to investigate the detailed structure of the inversion breakpoints and gene shuffling to test the working model of supergenes arising by rearrangement of distant ancestral loci. Using population genetics I propose to survey genetic diversity and look for signatures of balancing selection on the genes composing the supergene inversion. We will test for the effect of balancing selection in a patchy habitat, which has received little empirical data. We are now very close to genetically characterise the major loci controlling spectacular adaptations, involved in speciation and radiations, and determining species coexistence.