The evolution of flower color, especially petal pigmentation, has received substantial attention. Less understood is the evolutionary ecology of pollen pigmentation though it varies among and within species and its biochemical properties affect pollen viability.  

We characterize the distribution of pollen color across 24 populations of the North American herb Campanula americana, and assess the degree to which this variation is genetically based. We identify abiotic factors that covary with pollen color and test whether germination of light and dark pollen is differentially affected by variable temperature and UaV.

Pollen color vaulries from white to deep purple in C. americana and is genetically determined. There was a longitudinal cline whereby pollen was darkest in western populations. Accounting for latitudinal variation, western populations experience elevated temperature and UV irradiance. Germination of light-colored pollen was reduced by 60% under high temperature, but dark pollen was unaffected. Exposure to UV reduced germination of light and dark pollen similarly.

The cline in pollen color across the range may reflect adaptation to heat stress. This study supports thermal tolerance as a novel function of pollen pigmentation and contributes to growing evidence that abiotic factors can drive floral diversity.

Campanula americana plants from 24 populations were grown to flowering in the greenhouse. Pollen color was scored as indicated in the paper. Pollen color of parents and offspring were used to calculat of heritability. Geographic variation in pollen color and its association with abiotic factors was estimated. Finally different color pollen was exposed to a range of temperatures and to different UV levels to determine the influence of these factors on pollen viability (measured as germination proportion). Raw data is uploaded. Additional data can be found in Table S1 of the Supporting Information.

Metadata is included in each of the files describing each vbariable. The methods section of the paper describes how each variable was collected.
Missing values for numerical data are blanks, i.e. empty cells, for non-numerical data they are "."