Ovules are the developmental precursors of seeds. In angiosperms the ovules are enclosed within the central floral organs, the carpels. We have identified a homeotic mutation in Arabidopsis, "bell" (bell), which causes transormation of ovule integuments into carpels. In situ hybridization analysis shows that this mutation leads to increased expression of the carpel-determining homeotic gene AGAMOUS (AG) A typical angiosperm flower includes four major organ types-sepals, petals, stamens, and carpels. The carpels include an additional set of distinct internal structures, the ovules. As the precursors of seeds, ovules play an essential role in higher plant sexual reproduction. Ovules are also found in gymnosperms, which lack carpels, and fossil evidence of angiosperm ancestors demonstrates that ovules precede the evolution of carpels (1-3). Thus, ovules evolved first, with subsequent evolution ofthe enclosing carpel. From this perspective, ovules can be viewed as separate floral organs, which, in angiosperms, develop from and are enclosed within the carpels.Little is known concerning the genetic control of determination and differentiation of ovules. This contrasts with recent progress in elucidation of control of development of other floral organs, where analysis offloral homeotic mutants in Arabidopsis thaliana and Antirrhinum majus has led to the formulation of two similar models for genetic determination of the identity of the four major floral organs (4-8). In Arabidopsis, at least five genes-APETALA1 (API), AP-ETALA2 (AP2), APETALA3 (AP3), PISTILLATA (PI), and AGAMOUS (AG)-are responsible for this determination (5,(8)(9)(10)(11). Several of these genes have now been cloned and found to be homologous to known transcription factors identified in yeast and mammals (12)(13)(14)(15). One of these genes, AG, important for carpel identity (5,8,16), is also potentially involved in ovule development. The expression ofAG occurs both early and late in flower development (16 (20). Mapping relative to RFLP (restriction fragment length polymorphism) (21) was performed on F2 progeny (96 chromosomes analyzed) from a bell-l/bell-l Ler x Co-3 cross. RFLP probes were M247 (Bgl II digestions) and g4028 (HindIII digestions). To determine the frequency ofintegument-to-carpel transformation in different alleles, ovules in >150 pistils from at least three different plants for each allele were scored for carpelloid features.Microscopy and in Situ Hybridization. Scanning electron microscopy was performed as described (17). In situ hybridization was carried out as described earlier (22 (16) were treated in the same experiment. Only exposures providing subsaturated grains were counted. For wild-type ovules, the area delimited by the endothelium and the embryo sac was not counted. The mean grain surface density (+SD) before background correction in the integument area on sections of wild-type ovules probed with anti-AG was 0.029 (±0.007) (10 ovules counted), and that for bell-i ovules with the same probe was 0.076 (+0.046) (13 ...