The social shape of sperm: using an integrative machine-learning approach to examine sperm ultrastructure and collective motility.

Sperm is one of the most morphologically diverse cell types in nature, yet they also exhibit remarkable behavioural variation, including the formation of collective groups of cells that swim together for motility or transport through the female reproductive tract. Here, we take advantage of natural variation in sperm traits observed across Peromyscus mice to test the hypothesis that the morphology of the sperm head influences their sperm aggregation behaviour. Using both manual and automated morphometric approaches to quantify their complex shapes, and then statistical modelling and machine learning to analyse their features, we show that the aspect ratio of the sperm head is the most distinguishing morphological trait and statistically associates with collective sperm movements obtained from in vitro observations. We then successfully use neural network analysis to predict the size of sperm aggregates from sperm head morphology and show that species with relatively wider sperm heads form larger aggregates, which is consistent with the theoretical prediction that an adhesive region around the equatorial region of the sperm head mediates these unique gametic interactions. Together these findings advance our understanding of how even subtle variation in sperm design can drive differences in sperm function and performance.