PREMISE

Floral shape, i.e. the relative arrangement and position of floral organs, is critical in mediating fit with pollinators and maximizing conspecific pollen transfer. This seems particularly true for functionally specialized systems. To date, however, few studies have attempted to quantify flowers as the inherently three-dimensional structures that they are, and determine the effect of intraspecific shape variation on pollen transfer. We here address this research gap using a functionally specialized system, buzz pollination, where bees extract pollen through vibrations, as a model. Our study species, Meriania hernandoi (Melastomataceae), undergoes a natural floral shape change from pseudo-campanulate corollas with more actinomorphically-arranged stamens (first day) to open corollas with more zygomorphic stamens (second day) over anthesis, providing a natural experiment to test how variation in floral shape affects male and female fitness.

METHODS

In one population of M. hernandoi, we bagged 51 pre-anthetic flowers and exposed half of them to bee pollinators when they were in either stage of their shape transition. We then collected flowers, obtained 3D flower models through X-ray Computed Tomography for 3D geometric morphometrics, and counted the amount of pollen grains remaining per stamen (male fitness) and stigmatic pollen loads (female fitness).

KEY RESULTS

We found significantly higher male fitness in open flowers with zygomorphic androecia than in pseudo-campanulate flowers. Female fitness did not differ among floral shapes.

CONCLUSIONS

These results suggest that there is an ‘optimal’ shape for male fitness, while the movement of bees around the flower when buzzing the spread-out stamens results in sufficient pollen deposition regardless of floral shape.

Experimental field data on pollen removal from flowers and pollen deposition collected in Cosanga, Ecuador; flowers were exposed to pollinators for two hours, then flowers were collected in 70% ethanol, and subsequently micro-CT scanned in the laboratories of the University of Vienna. After CT scanning, stamens were removed from flowers for pollen counting using a liquid particle counter, and stigmas were sliced off flowers for counting stigmatic pollen loads under a fluorescence microscope. Flower shapes were captured through the placement of landmarks on 3D models of the respective flowers, using the software AMIRA, and then analyzed in R. See details in our manuscript.