A time tree for the evolution of insect, vertebrate, wind, and water pollination in the angiosperms

There is much circumstantial evidence that flowering plants were diverse by the Lower Cretaceous and were pollinated by insects (Arber & Parkin, 1907; Crepet & Friis, 1987). Arguments supporting this come from extant and fossil flower morphology, fossilized traces of interactions, and the pollination modes of surviving early lineages. First, some extinct gymnosperms had bisporangiate cones (with both micro- and megasporangia) surrounded by bracts (Fig. 1), and many such cones show traces of having been chewed by mandibulate insects (Peris et al., 2017). Fossils of flower-associated flies also provide evidence of the existence of strobilus–pollinator interactions from the Permian to the Jurassic (Ren, 1998; Ren et al., 2009; Khramov et al., 2023). Second, if flowers evolved from bisporangiate strobili, they were not well suited for wind pollination because simultaneous optimization for pollen export and pollen capture is structurally difficult. The angiosperms' defining enclosure of the megasporangium inside surrounding structures may also point to ancestral insect pollination, as argued by Arber & Parkin (1907: 73), ‘In the case of the angiosperms such primitive entomophily was preserved and rendered permanent by a transference of the pollen-collecting mechanism from the ovule itself to the carpel or megasporophyll and by the closure of this organ.’ Third, all angiosperms, but no living gymnosperm, produce pollenkitt, an oily substance on the surface of pollen that serves as a glue to attach pollen to animal vectors (Hesse, 1980). In wind-pollinated plants, pollenkitt abundance is secondarily reduced. Lastly, the oldest lineages of flowering plants that still survive today are pollinated by flies, moths, and beetles (Luo et al., 2018).

While insect pollination thus undoubtedly played a decisive role in the evolution of flowers, a phylogenetically informed analysis of pollination by insects, vertebrates, wind, and water across a full modern phylogeny of plants has been lacking. This is what Stephens et al. now provide in an article published in this issue of New Phytologist (2023; 880–891). Using a time-calibrated phylogeny with 1201 species representing the major lineages of flowering plants, together with geographic occurrence data, Stephens et al. quantified the timing and environmental associations of pollination shifts. Where possible, they scored pollination at the species level, either from published fieldwork (n = 432) or from the pollinator syndrome approach (n = 728). Where no information was available for a particular species, taxa were scored at genus (n = 131) or family (n = 4) level. In some analyses, 180 taxa with missing or polymorphic data were excluded from the analyses.

All major angiosperm clades (magnoliids, monocots, eudicots, asterids, and rosids) and 57 of 64 angiosperm orders were reconstructed as ancestrally insect pollinated. Only the Zingiberales are ancestrally vertebrate pollinated and the Fagales and Picramniales wind pollinated. Stochastic character mapping found 42–50 transitions from insect to wind pollination and 4–12 reversals from wind to animal pollination, while there were 39–56 transitions from insect to vertebrate pollination and 26–57 reversals from vertebrate back to insect pollination. Two angiosperm clades are primarily water pollinated, the Ceratophyllales and the seagrasses within Alismatales (Ruppiaceae, Cymodoceaceae, Posidoniaceae, Zosteraceae, and Potamogetonaceae). Water pollination evolved from wind pollination, with no reversals.

Associations between pollination modes and environments are surprisingly weak, except that the probability of wind pollination increases with habitat openness. When averaging the total branch lengths spent in each state across all stochastic character maps, a mean of 86% of angiosperm evolutionary time since the crown node is spent in insect pollination, 10% of evolutionary time in wind pollination, 4% of time in vertebrate pollination, and 1% of time in water pollination.

These findings are based on a single phylogeny that does not consider any topological uncertainty. Deep nodes where the study's topology might be erroneous include the position of the monocots relative to the eudicots and the position of Amborella relative to the remaining angiosperms. The position of the magnoliids as a sister lineage to monocots as in Stephens et al.'s phylogeny is ambiguous. Other studies instead found magnoliids to be sister to all eudicots, including magnoliids (Wickett et al., 2014; Zeng et al., 2014; One Thousand Plant Transcriptomes Initiative, 2019; Yang et al., 2020). There is also stronger support for a position of Amborella as sister to the Nymphaeales than there is for Amborella as sister to the remaining angiosperms (Xi et al., 2014).

Any possible effects from such topological changes, however, would not have modified the reconstruction of entomophily as ancestral in the flowering plants nor of the frequent reversals between insect and vertebrate pollination, which all occur more recently than 66 Ma. As pointed out by Stephens et al., future work might focus on the question of which environmental conditions accompany shifts between insect and vertebrate pollination. Factors favouring such switches presumably involved animal physiology and nutritional needs relative to plants' ability to fulfil these needs, while balancing other challenges, such as drought, wind exposure, and growing season length. Finer-grained studies could follow the approach developed by Stephens et al., but include occurrence on oceanic islands or sky islands, or also plant growth form and niche, such as tree, climber, or epiphyte. Conceivably, pollinators might be scored more finely, too, by separating birds, bats, and bees from flies and beetles. However, our knowledge of pollinators, especially in tropical trees and epiphytes, is scarce, and much field work remains to be done.