The floral organ identities have been intensively studied, and are well explained by the interaction of three kinds of genes (ABC model) in dicot species. However, stages earlier than flower formation such as the establishment of inflorescence architecture need to be studied in both dicots and monocots. Grass species have very different architecture of inflorescence and flower from dicot species. In grasses, reproductive induction converts the shoot apical meristem to rachis meristem, which produces primary rachis branches instead of leaves. Immediately after generating several primary branches, the rachis meristem of rice becomes aborted and remains as a vestigial organ. The apical meristem of the primary branch produces secondary branches and spikelets. The spikelet meristem is transformed to one (in rice) or more (in barley and maize) floral meristems. Thus, in grasses, reproductive development is driven by the four kinds of apical meristems: rachis meristem, branch meristem, spikelet meristem and floral meristem. This means that the architecture of panicle (inflorescence) is regulated by the activities of the four kinds of reproductive meristems. For understanding how panicle shape is genetically regulated, it will be important to identify genes associated with meristem identities in panicle. Here, we describe recessive aberrant panicle organization 1 (apo1) mutations affecting both panicle branches and flower organization.

We have identified three allelic mutations, apo1-1 and apo1-2 identified in M2 population of cv. Taichung 65, and apo1-3 in M2 population of cv. Kinmaze mutagenized with methyl-nitroso-urea (MNU). The three mutants exhibited common abnormalities in panicle architecture and abnormal floral organs. The developmental course of apo1 panicles was examined in detail. The abnormality of apo1 was first observed in the phyllotaxy of primary branches. The primary branches of apo1 mutants showed distichous phyllotaxy, in contrast to the 2/5 spiral in the wild type (Fig. 1A, E). The rachis meristem of apo1 was smaller than that of the wild type, which may cause the change of phyllotaxy. The abnormality of rachis meristem was also recognized in the number of primary branches. The apo1-1 and apo1-2 mutants showed the reduced number of primary branches without affecting each internode length of rachis, suggesting that apo1 lost the identity of the rachis meristem earlier than the wild type due to the precocious abortion or the conversion of identity. The number of primary branches of apo1-3 was comparable to the wild type. Detailed observation of rachis meristem revealed that in apo1-1 and apo1-2, most of the rachis meristems were converted to spikelet meristems before abortion, and formed a terminal flower (Fig. 1F). On the other hand, in apo1-3, about 40% of the panicles had the vestige of the rachis meristem, but in the others, rachis meristem was partially converted to spikelet meristem aborted after producing one or two glumes (Fig. 1G), or converted to complete spikelet. The above results indicate that the apo1 rachis meristem tends to be transformed to spikelet meristem before degeneration.

Similar tendency was observed in the primary branches. Each primary branch was shortened in the three apo1 mutants due to the precocious transformation of the apical branch meristem to spikelet. Thus, the total number of spikelets and secondary branches in apo1 was reduced to less than half that in the wild. The shortening of primary branches was acropetally enhanced. In the several apical branches, the branch meristem was converted to spikelet meristem immediately after initiation, resulting in the branch consisting of only one terminal spikelet lacking lateral organs. The phenotypes of rachis and branches

show that the meristem identities of rachis and branches are precociously converted to those of the advanced stages in apo1 mutants, suggesting that APO1 regulates the temporal maturation of reproductive meristem.

The apo1 mutants also showed abnormal floral organization. The rudimentary glumes, empty glumes, lemma and palea were not affected. The abnormalities were mainly detected in the number of floral organs. In the wild-type flower, two lodicules, six stamens and one pistil are produced in whorls. In the apo1, the lodicules were increased to three or four at the expense of the stamens (Fig. 1H), conserving the total number of organs in lodicule and stamen whorls to be about eight. In the stamen whorl, mixed organs of lodicule and stamen in which anther was formed on the lodicule were also produced. Thus, in apo1, the meristem at lodicule-differentiation stage is temporally extended to the stamen- differentiation stage. The carpels were increased to two in apo1-3 or to more than 5 in apo1-1 and apo1-2, showing the loss of determinacy of the floral meristem. (Fig. 1H). In apo1, one or two organs in which stigmatic papilae was formed on glume-like tissue were produced inside the stamen whorl. The modification of each floral organ number was more pronounced in apo1-1 and apo1-2 than in apo1-3. Thus, apo1-3 is weaker than apo1-1 and apo1-2.

In conclusion, APO1 gene is considered to temporally regulate the conversion of meristem identity in the reproductive phase.