For a long time, internode inhibition in simple inflorescences was considered a universal path leading to the evolution of pseudanthia. For instance, capitula of Asteraceae were usually interpreted as derived from simple (28) or compound racemes (29), umbels (30), or spikes (31). On the other hand, different forms of thyrses (inflorescences with primary racemose and secondary cymose branching) were suggested as the underlying architecture of the umbels in Apiaceae (32, 33). Only recently, the knowledge about genetic patterning of meristems and their detailed morphological reinvestigations lead to the recognition of floral units (18, 27, 34). Such multiflowered structures arise from meristems resembling those of single flowers and, thus, are hardly comparable to conventional model plants used in developmental and evolutionary-developmental studies. The ontogenetic prolongation of FMs, which we hypothesize as the possible pathway towards FUMs (35), may facilitate the co-option of various conserved developmental regulators (such as CYC/TB1 genes) from the level of individual flower to the level of entire floral unit and constitute an important preadaptation for the evolution of floral pseudanthia. As floral units are morphogenetically determinate, they cannot continuously segregate new submeristems/primordia due to apical growth. Their patterning is thus highly dependent on the space created by the initial and ongoing expansion of the FUM. Additionally, as apioid pseudanthia develop within the envelope of bracts and vegetative leaves that press them against the stem, their patterning is subjected to significant mechanical constraints.
Floral pseudanthia in Apiaceae: the unique interplay among spatial constraints and morphogenetic gradients
Meristem geometry and mechanical constraints during morphogenesis can vastly alter plant morphology (36), including the number of organs (37), their proportion (38) and arrangement (39, 40). Our understanding on how growth rate differences between cells influence gene regulatory networks during plant development is far from comprehensive and although gene expression can be induced solely by mechanical pressure (without secondary, auxin-mediated response), the mechanisms of such mechano-transduction remain elusive (Fal et al., 2016; Landrein et al., 2015).
The importance of ontogenetic collisions for architecture of complex floral units has been explored with procedural modelling (43) and is vastly exemplified by experimental studies on simple and secondary heads of Asteraceae. Capitula primordia respond to wounding with a change of their phyllotactic pattern and only recently it has been proven that these alterations result from the disruption of natural auxin gradients (17). In Gerbera, the release of spatial constraints from floral primordia adjacent to the wounding site allows for their repatterning into ray flowers. The bisection of a sunflower capitulum conducted by Marc & Palmer (1982) almost four decades earlier yielded a similar outcome, i.e. the formation of two pseudanthial units with enlarged peripheral florets induced at the place of cut. Further proof for space-dependent patterning of FUM-derived pseudanthia comes from natural anomalies of syncephalous Asteraceae (45). Contrary to the secondary umbels in Apiaceae, secondary heads show only ray flowers at the periphery of the entire multi-headed unit. However, single capitula in Oedera capensis and Dyssodia decipiens, whose primordia were physically separated from their neighbours, are able to develop ray flowers around their entire margin (Claßen-Bockhoff 1996).
Based on the aforementioned studies and results of our investigation, we hypothesize that patterns of ray flower formation in pseudanthia of Apiaceae result from the interaction between peripheral promotion and spatial constraints which increase towards the centre of the meristem. Secondary umbel shaping is based upon the relative size among the incipient FUM and umbellet meristems. Large FUMs give rise to umbel-centred units (Figs. 2 and 3). Spatial constraints are imposed by the numerous peripheral umbellets, which likely delay the fractionation of inner umbellet meristems and development of adaxial, ray flowers in the peripheral umbellet meristems (Figs. 2D and 3D). Thus, from the very beginning, the entire pseudanthium develops as a single entity, divided in a promoted peripheral and a retarded central part. On the contrary, when the FUM gives only rise to few umbellets, their physical separation caused by sequential elongation of stalks (raylets) releases mechanical pressure on the adaxial sides (Figs. 4, 5 and 6). In consequence, though all umbellets originate from the same FUM, each of them develops independently and forms ray flowers around its entire margin. The establishment of an intermediate promotion pattern in secondary umbels proceeds almost identically to that of umbel-centred pseudanthia. However, due to the smaller size of the FUM and the corresponding smaller number of umbellets (Fig. 7A), its expansion goes along with a formation of additional space between the peripheral umbellets (Fig. 7C) allowing for the development of subperipheral units with smaller, weakly zygomorphic ray flowers (Fig. 7D). Interestingly, in Apiaceae, the promotion pattern might change within the individual plant (Additional file 1). The higher-order umbels (those that develop later on the reproductive shoot) are usually smaller and produce fewer umbellets. Species with umbel-centered promotion in strong terminal and first order secondary umbels can thus produce units with intermediate promotion in higher order as can be seen in carrot (Additional file 1).
The influence of bracts on the polarity and growth of floral meristems is frequently overlooked in developmental studies (Chandler, 2014; Kwiatkowska, 2008; Ronse De Craene, 2018). It is widely acknowledged that size and position of bracts might be a source of spatial constraints that influence the shape of floral meristems and floral organ initiation patterns (49–53). In Apioideae, the presence of bracts is highly variable but all species with floral pseudanthia have well-developed involucels (9). They originate from FUMs as common primordia with ray florets and are thus not a part of the plant’s foliage (Figs. 2A and 3A). The peripheral promotion stimulus acts on those common primordia and causes their simultaneous outgrowth. In the result, similarly to ray flowers, apioid involucels are subjected to spatial constrains that depend on their position within the secondary umbel and overall promotion pattern. In umbellet-centered pseudanthia, the influence of spatial constraints is noticeable in the developmentally retarded adaxial involucels which develop in direct contact with neighbouring umbellet meristems (Figs. 4, 5 and 6). In carrot, the high-order secondary umbels with intermediately-promoted ray flowers show a similar intermediate-promotion in bracts (Additional file 1). The involucels of peripheral umbellets are large and distinctly pinnatisect, while those of central umbellets are noticeably smaller and needle-like. In subperipheral umbellets, enlarged bracts occur under weak ray florets and are also asymmetrical and less developed in comparison to those found in peripheral umbellets. Besides being subjects of spatial constraints, bracts in Apioideae can also create mechanical pressure themselves. In Artedia squamata, involucre-derived spatial constraints (Fig. 3B) on peripheral umbellet meristems cause retardation of one of the abaxial FMs, as well as changes in the geometry of its neighbours (Fig. 3C). The mechanical forces acting on the sides of ray flower meristems inhibit the enlargement in one of their lateral and dorsal petals, leading to the establishment of Artedia-type zygomorphy. In Tordylium brachytaenium, a similar effect is achieved by the proximity of adjacent FMs (Figs. 7C and 7E) which press against each other. The pressure-dependent shift in ray flower meristem symmetry (from radial to zygomorphic) is also apparent in Echinophora trichophylla (Fig. 2H) however, its occurrence at late developmental stages – after petals, sepals and stamens are well-developed – does not cause a distortion in the patterning of Coriandrum-type zygomorphy. This observation implies that ontogenetic spatial constraints may have different effects on ray flower morphogenesis, depending on the timing of their occurrence (early or late shift in symmetry sensu Naghiloo, 2020).
Floral unit meristems – an important preadaptation for pseudanthia?
The negative feedback loop of CLAVATA3 (CLV3) and WUSCHEL (WUS) constitutes a key genetic component of stem-cell activity, accounting for self-perpetuation of IM and formation of cellular pool necessary for proper development of flowers (55). While WUS promotes cytokinin activity in the central zone of IM and incipient FMs, CLV3 restricts meristems’ size by preventing the build-up of excess cells (56). The disruption of WUS-CLV3 loop is necessary for cell differentiation and organogenesis. In Arabidopsis thaliana, when certain size of floral meristem is reached, WUS acts with LEAFY (LFY) to activate its own repressor – AGAMOUS (AG) (57). This process establishes the determinacy of future flowers. Other transcription factors that confer floral fate include UFO which works in combination with LFY to specify petals and stamens by activation of B-class MADS box genes (58, 59).
FUMs differ from inflorescence meristems and instead resemble flower meristems. Similar to FMs, FUMs are characterized by the early determinacy and lack of apical growth which in Asteraceae coincides with uniform expression of LFY (Zhao et al., 2016) homologue in the naked (i.e. undifferentiated) head meristem. According to our results, secomdary umbels in Apiaceae share this pattern (Fig. 9B-1). In carrot, the transcripts of DcLFY can be detected at the secondary umbel meristem and throughout the process of repeated fractionation in the umbellet meristems and flower primordia (Figs. 9B-2 and 9B-3). This indicates that flowerheads and secondary umbels, as well as various other floral units, are not condensed inflorescences but ontogenetic ‘matryoshka dolls’ that can be best described as flowers within flower. Following Claßen-Bockhoff & Frankenhäuser (2020), we hypothesize that FUMs might arise from floral FMs due to disruption of WUS-CLV3 signalling pathway, including the loss of size-restricting function of CLV3 and/or lack of negative feedback from other direct/indirect repressors of WUS. Such change would result in the expansion of FM and creation of space for additional fractionation, however, this hypothesis requires further studies to be ultimately confirmed or refuted.
The evolution of morphological novelties is frequently based on already existing gene regulatory networks (GRNs). As components of a single GRN are interconnected, differences in spatio-temporal expression of particular regulatory gene can potentially affect its downstream effectors, allowing for their redeployment in the novel context. The examples of such evolutionary co-options are widespread in both animals (61–63) and plants (64–66). As FUMs might evolve through heterochronic changes in FM genetic patterning, the co-option of multiple components of flower-specific GRN can tentatively explain the origin of some novelties associated with floral units, including formation of pseudanthia. In all hitherto published studies, including our own (Fig. 10), expression patterns that lead to the establishment of ray flowers were recovered for genes that normally participate either in the specification of identity or symmetry of floral organs (67–71). Most of those transcription factors, especially CYC/TB1 genes are known to have independently undergone major expansion in several plant lineages with FUM-derived floral pseudanthia (72–76). The expression of CYC-like genes is also documented in shoot-derived pseudocorollas of Actinodium cunninghamii (77). Interestingly, petaloid bracts found in several plant lineages with floral units, such as Nyssaceae (78), Cornaceae (79), are patterned by the expression of B-class MADS box genes (80, 81) which normally specify the identity of true petals. Moreover, in dove tree (Davidia involucrata), the origin of petaloid bracts can be traced to duplication and neofunctionalization of SUPRESSOR OF OVEREXPRESSION OF CONSTANS1 (82) – a universal flowering pathway regulator (83). Bract-derived pseudcorollas are also known to have evolved independently in numerous clades of Apioideae (9), and in some earlier-diverging umbellifer subfamilies (i.e, species of Alepidea and Astrantia from Saniculoideae, Pozoa coriacea from Azorelloideae and species of Actintous and Xanthosia from Mackinlayoideae). Although developmental data are scarce in Apiales, based on the occurrence of the secondary umbel as the basic architectural model in reproductive shoots (84–86), we expect floral units to have evolved before the divergence of the clade uniting Apiaceae, Araliaceae and Myodocarpaceae (87, 88). Later, the whole-genome duplication in the common ancestor of Apiaceae lead to the expansion of TCP gene family (89) which might have allowed for repeated co-option of newly-acquired paralogues into the patterning of floral pseudanthia.