Evidence for selectively constrained 3D flower shape evolution in a Late Miocene clade of Malagasy Bulbophyllum orchids

Abstract

Summary Questions concerning the evolution of complex biological structures are central to the field of evolutionary biology. Yet, still little information is known about the modes and temporal dynamics of three‐dimensional (3D) flower shape evolution across the history of clades. Here, we combined high‐resolution X‐ray computed tomography with 3D geometric morphometrics and phylogenetic comparative methods to test models of whole‐flower shape evolution in the orchid family, using an early Late Miocene clade (c. 50 spp.) of Malagasy Bulbophyllum as model system. Based on landmark data of 38 species, our high‐dimensional model fitting decisively rejects a purely neutral mode of evolution, suggesting instead that flower shapes evolved towards a primary adaptive optimum. Only a small number of recently evolved species/lineages attained alternative shape optima, resulting in an increased rate of phenotypic evolution. Our findings provide evidence of constrained 3D flower shape evolution in a small‐sized clade of tropical orchids, resulting in low rates of phenotypic evolution and uncoupled trait–diversification rates. We hypothesise that this deep imprint of evolutionary constraint on highly complex floral structures might reflect long‐term (directional and/or stabilizing) selection exerted by the group’s main pollinators (flies).

Evidence for an evo‐devo‐derived hypothesis on three‐dimensional flower shape modularity in a tropical orchid clade

“…In fact, the floral primordial initiation of the labellum usually follows immediately after the development of the lateral petals (e.g., in Phalaenopsis: 4). Notably, in a previous study by Artuso et al (2021), the OU model was found to best explain whole-flower shape variation among Bulbophyllum Clade A species and interpreted to reflect pollinator-driven (directional and/or stabilizing) selection (see also Serrano-Serrano et al 2015;Joly et al 2018;Ibañez et al 2019;Reich et al 2020). In consequence, based on the present results, the same selective pressures might be invoked for sepals, lateral petals, and labellum + column-foot.…”

“…S1). Phylogenetic half-lives are usually interpreted as the average time it takes a species to evolve halfway from its ancestral state (here: module shape) toward the optimum (see also Artuso et al 2021), thereby indicating the strength of the OU process acting on the species. When α is zero and/or t 1/2 is larger relative to tree height, the OU model collapses to the BM model (Beaulieu et al 2012;Cooper et al 2016).…”

“…2 and Table S2). Accuracy of LM placement and intraobserver error assessment, along with a pilot study of intra-vs. interspecific shape variation in the sister species Bulbophyllum bicoloratum/Bulbophyllum occultum, can be found in the Supporting Information of Artuso et al (2021).…”

“…In a recent study, Artuso et al (2021) combined highresolution X-ray computed tomography (HRX-CT) scanning (Staedler et al 2018) with landmark (LM)-based threedimensional (3D) geometric morphometrics (GM; Bookstein 1997; Klingenberg 2010) (see Fig. 2) and multivariate phylogenetic comparative methods (PCMs; Garamszegi 2014) to study whole-flower shape evolution across 38 species of Clade A, using a time-calibrated phylogeny (Gamisch et al 2021).…”

“…However, as for any OU-driven phenotypic evolution, intrinsic (e.g., developmental, genetic) factors remain viable alternative explanations (e.g., Hansen 2014; see also Futuyma 2010;Blomberg et al 2020). Nevertheless, there is evidence that the whole-flower shapes of some Clade A species evolved independently toward different phenotypic (OU) optima (Artuso et al 2021), suggesting that developmental/genetic factors are not the primary cause of constrained flower shape evolution in these orchids (see also Davis et al 2014). Unfortunately, testing multi-peak OU optima for high-dimensional morphometric data, such as our 3D-module dataset, is not yet achievable.…”