Labellum–androecial identity
Abnormality is often observed in species of ex situ conservation and in vitro conservation, as the results of the accumulation of adaptive genetic variation and mutation caused by interspecific hybridization, tissue culture, chemical treatment, radiation exposure, parasitism and climatic change during acclimatization. On one hand, the reduction of heritable variation reduces the ability of small populations to adapt to ex situ changing conditions, and the probability of adverse factors accumulating. On the other hand, inbreeding makes some recessive mutation appear and reduce the fitness of small populations. Nevertheless, we observed two labella of Amomum. sp. in the wild population (unpublished data). Multiple male buds phenomenon in banana is a natural but rare occurrence, caused by the branching of inflorescence axis or the differentiation of male flowers into male buds [11]. The outer staminodes initiate and usually abort during early development in Canna species [39]. Many cases of homeosis are based on a phylogenetic hypothesis; environmentally induced homeosis (phenocopying) may occur through the same or similar mechanism as genetically induced homeosis [57]. The determination or circumscription of an organ is itself rather equivocal. Staminodes belonging to an androecial whorl may become secondarily petaloid in Zingiberales [5]. The labellum, a petaloid organ, blends characteristics of more petal than stamen and cannot be precisely classified as either. In comparison with other monocots, it is certain that the labellum and other petaloid appendages are homologous with stamens on the basis of their position. In the two-staminate flower, direct replacement of the fertile stamen with labellum (and vice versa) results in a 1:1 substitution. The homeotic conversion of stamen into labellum also suggests that the labellum has an androecial identity. There are about 25 to 30 flowers in one inflorescence. Apart from about 2 two-staminate flowers, we observed another two types of abnormal flowers with one fertile stamen in A. intermedia, 1 to 2 flowers with two labella and 1 to 2 flowers with three appendages. Song et al. [59] reported that the number of fertile stamen, labellum, appendage, and petal varied in different species of Alpinia genus flowers, and the relative position of fertile stamen, labellum, appendage, and petal was different from normal flowers, rotated from adaxial plane to lateral to abaxial. The continuous variation in the number of stamen may be the reappearance of the evolutionary history of the stamen in the ginger group and even in Zingiberales [59]. The number of fertile stamen from 2, 1.5, 1, 0.5 to 0, varies continuously to some extent, with intermediate phenotypes of one stamen being the most common. We speculated that the homeotic conversion occurred due to the development of common petal–stamen primordia and petal–labellum primordia were disordered during the differentiation of the floral organs. Compared with sepal and carpel, androecial members and petals show much wider variation in abnormal flower, which coincide with androecial petaloidy in Zingiberales. The bias of mutant frequency between stamen and petaloid staminodes hints that they suffered the selective pressures, which were the result of an evolutionary balance between producing pollen and attracting pollination. Several lineages of floral organ identity MADS-box genes have been isolated from Zingiberales to investigate their potential roles in the determination of androecial members [1, 3, 15, 16, 34, 58, 61, 71]. Compared with model plant systems, broad expression of B-, C-class genes, on the one hand, has expanded the regulatory potential beyond their initial framework, on the other hand, may not be specific in determining petal and androecial member identity. Previous studies imply that the AGL6 MADS-box gene is required to specify stamen development, and low expression of AGL6 may promote petaloidy in the androecial whorl [33, 34, 71]. Most, if not all, floral organ identity genes are dosage-dependent [6, 20, 67,68,69, 72], and the effects of gene dosage on a range of quantitative trait variation may have driven homeosis to occur through gene duplication and loss events. Floral organ identity genes may play central roles as transcript factors to determine the type and number of the organs formed and affect the floral architecture. In addition, the interaction between AGL6 and E-class protein may play a “bridge” or “glue” role in the multimeric complexes formation to regulate floral organ determination.
Labellum–staminode fusion
The labellum is a major morphological structure in Lowiaceae, Zingiberaceae, Costaceae, and Cannaceae, but is not homologous among families. The labellum in Lowiaceae, which is regarded as the conspicuous petal or also receives the bundle of the adaxial, outer absent androecial member, is different from the labella in the other three families, which are only derived from staminodes [36]. In Zingiberaceae, the fate of androecial member varies, develops into fertile stamen, a highly specialized structure, or complete loss. The results of developmental work on unlobed or bilobed or trilobed labellum (Fig. 1) [7, 16, 24, 27, 58, 60, 73] generally share in ontogenic aspects and present a similar pattern: the two abaxial inner androecial members are joined through intercalary growth, fusing basally, to produce (the center of) the labellum. The two adaxial outer androecial members form two lateral staminodes, developing into small teeth at the base of the labellum, or petaloid free from the labellum, or lateral lobe fused with the central lobe of labellum to form the trilobed structure. It remains ambiguous whether the anterior, abaxial outer androecial member aborts early or never initiates since the cells in this region are indistinguishable from surrounding cells and fusion of the two segments of labellum derived from common primordia may almost be congenital. Recent transcriptome on Zingiber zerumbet, a trilobed labellum species, has investigated the implicated molecular mechanism that boundary genes may be involved in floral organ fusion by down-regulation, thus resulting in loss of boundary between the androecial primordia [73]. The silence of B- and C-class MADS-box genes in Nigella damascena not only led to homeotic conversion, but also boundary shift between floral whorls [68]. This suggests that the regulatory network of boundary genes and identity genes may contribute to the fusion between androecial members in a series of abnormal flowers.
In Alpinia calcarata, abnormal flowers showed two stamens, a median sterile appendage and a single gland in the postero-lateral positions [45]. The explanation was that one of the antero-lateral glands has become staminiferous, the additional fertile stamen occupies the missing gland in the abnormal flower. Therefore, the labellum which is a single organ, together with the fertile stamen, forms the outer androecial whorl. Another example of abnormal flower A. vittata had two lateral fertile stamens and a median sterile appendage as well as a single gland [47]. The explanation was that the two fertile stamens correspond to the two lateral staminodes, and the median sterile appendage corresponded to the posterior fertile stamen of the normal flower. Then the labellum was believed as the fusion of two antero-lateral androecial members. In this study, we observed that the two-staminate flower has two stamens, two lateral appendages, a median sterile appendage, and a single gland. The gland should not belong to the androecial whorl (see below nectary), so the labellum seems to be composed of three androecial members, as the presence of a third appendage. This result supports Gregory’s view that the labellum is a triple structure. From the above, how many androecial members composing labellum depends on largely the development of staminode. The existence of contradictory data raises the question of whether the profound integration of staminode in labellum organs could have resulted in a high degree of plasticity and dynamism. From the results of anatomy, the mid-anterior bundle in the floral tube is considered significant in the morphological nature of the labellum [41]. The mid-anterior bundle, which is derived from the vascular plexus, continues for a considerable length in the labellum, or shows an early division into two strands. Both of the above two cases are recorded in the different unlobed labellum species Elettaria flowers of the same plant, and the latter case also exists in some other species, such as bilobed labellum species Kaempferia scaposa and Curcuma amada [41, 51]. There is also a third case that the marginal bundles instead of mid-anterior bundle, laterally on either side of the mid-anterior line, keep upwards into the two segments of the labellum, e.g., trilobed labellum species Zingiber macrostachyum and bilobed labellum species Curcuma decipiens [51]. Pai suggested that when the labellum is emarginate, the mid-anterior bundle quickly divides into two and then run into the two components. In this study, two-staminate flowers show a median apical split in the labellum (Fig. 2c), whereas the normal ones do not, but they both possess a mid-anterior bundle. According to traditional opinion, the mid-anterior bundle may expediently be interpreted either as a composite bundle which is the fusion of the marginal bundles of the two-component members of the inner androecial whorl [41], or a vestige of the abaxial outer androecial member [17]. However, the existence of the mid-anterior bundle is independent of the labellum unlobed, bilobed or trilobed irrespective of whether the abaxial outer androecial member constitutes labellum. It seems that the development of a mid-anterior bundle and its further performance in the labellum is relevant to the degree of connation of its two components [25]. The result does not refute Payer’s view that the labellum is derived from the congenital fusion of two lateral staminodes. Recently, a study in Globba revealed that a “fifth staminode” develops as part of the labellum, supporting the view of reduction and loss of the outer abaxial staminode in Zingiberaceae [21]. Overall, the labellum comprises 2–5 androecial members, 2 or 3 androecial members in unlobed or bilobed labellum, while 4 or 5 androecial members in trilobed labellum. Anatomy of abnormal flowers may not provide enough evidence for elucidating relationships of the androecial members; it has been used to make inferences about the evolutionary homologies of different plant parts and the floral structure. However, this type of reasoning should be employed cautiously, especially the reasoning is based on the single example.
Nectary
Nectaries are universally present in Zingiberales, except in Lowiaceae where they are aborted (only left nectary slits) and in Zingiberaceae and Costaceae where they are highly transformed [22, 26, 46]. In Musaceae family, nectaries are limited to the upper part of the ovaries above the locules in female flowers, while nectaries entirely occupied the aborted ovaries in male flowers. Carpel margins fuse incompletely to develop the septal nectarines which occur in the septa of the ovary, such as in Strelitzia flower and Heliconia flower. In contrast, the carpels fuse so completely that the structure of septal nectaries is replaced by epigynous glands in many Zingiberaceous flowers. The glands of Zingiberaceae are typically two in number and are above the septa on the antero-lateral of the flower. In some, they are basally connate on the anterior side. In rare cases, the antero-lateral glands may extend to the posterior side. The gland on the posterior side becomes reduced and absent in the evolution of the Zingiberaceous flower [47]. The two basal glands were interpreted as staminodes, epidermal appendages or stylodes. Gregory regarded the glands as epidermal appendages of the ovary, and Pai suggested that glands are more deeply associated with organs of the ovary through comparative observations [17, 41]. It is very questionable that the gland is equivalent to a staminode as it is derived from septal nectaries, which still present in other families of Zingiberales. From a positional homology point of view, stamen regressive is pervasive almost across the whole Zingiberales, in which fertile stamen evolved into staminode or totally absent. If the nectary belongs to an androecial member, then it would conflict with the establishment of monocots architecture, which is based on the trimerous plan of floral construction. Therefore, the origins of glands in Zingiberaceae could not be the androecial members. In Arabidopsis, the nectary is not regulated by ABC functional genes, but is independent of any floral organogenesis gene and is position-determined [5]. It supported the independence of nectary development from the specification of floral organ identity.
Characteristics of vasculature in Zingiberaceae
In Zingiberales, for the common MAD petal initiation, two primary patterns of floral zygomorphy are observed, which differ mainly in the configuration of the androecium [55]. In pattern 1, the suppression of the adaxial median stamen is often associated with the presence of a relatively well-differentiated MAD petal, such as in Musaceae, Lowiaceae, and Strelitziaceae. In pattern 2, one or more abaxial stamens are reduced or modified, such as in more derived Heliconiaceae and ginger families. Interestingly, pattern 2 normally occurs in taxa that are embedded within pattern 1 clades [9, 55]. The two-staminate flower seems to accord with pattern 1 and pattern 2 on account of suppression in both the adaxial and abaxial stamen. In contrast to the normal flower, variation in the two-staminate flower occurred not only in the number and arrangement of androecium, but also in the distribution of petals and ovary. It displays putative developmental rotation in floral orientation with the median-abaxial petal by torsion of the ovary. Floral organ differentiation appears to be regulated along the adaxial–abaxial radius, with stamens developing on the adaxial side and the labellum on the abaxial side. This seems to be genetically fixed, optimizing effectiveness of the labellum as a landing platform for insect pollinators. However, whether resupination occurs needs a direct evidence, especially in developmental investigation.
Based on the previous studies, the outer androecial whorl receives its vascular supply from the CDBs and the inner androecial whorl from the PBs in both the basal banana group grade and the more derived ginger clade [35, 36, 64]. The plexus, which contributes to the petals, gland(s), and androecium, is a significant anatomical development and has been universally observed in Zingiberaceae [41]. The formation of vascular plexus is a derived character after the ginger group divergence and loss in Cannaceae. Although the stamen has degraded or lost in the process of long-term evolutionary adaptation, there is still the corresponding vascular bundle system that persists in the origin position.
Although the arrangement of the organs of the normal and abnormal flowers seems very different, the vascular bundles tend to be more consistent in supplying floral organs: (1) the bundles of the outer ring in the pedicel travel out exhibiting extensive branching while they pass upward through the flower, the sepals are supplied by these small traces derived from the outer ring and the outer branches of CDBs; (2) the large bundles from the outer ring in the pedicel, along with the branches of CDBs and the PBs, form an anastomosing vascular plexus; (3) small traces left from plexus and outer branches of PBs go into the petals; (4) the fertile stamen(s) incorporate PBs; (5) appendages are supported by inner branches of CDBs. The difference is comparing to the labellum in normal flower receives abaxial (antero-lateral) PBs and inner branches of CDBs, the smaller labellum in the two-staminate flower only receives abaxial PBs.
Transcriptome and expression data from representative species in Zingiberales have provided information for studying molecular mechanisms of floral development and diversification, but what is the specific gene and how the gene regulates coordinates the relationships of the androecial members is still unclear. Precise positional transcriptome combined with microdissection and RNA-seq techniques would facilitate the understanding of the underlying mechanism of androecial primordia development.