A dynamic network model predicts the phenotypes of multicellular clusters from cellular properties

Cell division without cell separation produces multicellular clusters in budding yeast.

Two fundamental characteristics of these clusters are their size (the number of cells

per cluster) and cellular composition: the fractions of cells with different phenotypes.

Using cells as nodes and links between mother and daughter cells as edges, we

model cluster growth and breakage by varying three parameters: the cell division rate,

the rate at which intercellular connections break, and the kissing number (the

maximum number of connections to one cell). We find that the kissing number sets the

maximum possible cluster size. Below this limit, the ratio of the cell division rate to the

connection breaking rate determines the cluster size. If links have a constant

probability of breaking per unit time, the probability that a link survives decreases

exponentially with its age. Modeling this behavior recapitulates experimental data. We

then use this framework to examine synthetic, differentiating clusters with two cell

types, faster-growing germ cells and their somatic derivatives. The fraction of clusters

that contain both cell types increases as either of two parameters increase: the kissing

number and difference between the growth rate of germ and somatic cells. In a

population of clusters, the variation in cellular composition is inversely correlated

(r2=0.87) with the average fraction of somatic cells in clusters. Our results show how a

small number of cellular features can control the phenotypes of multicellular clusters

that were potentially the ancestors of more complex forms of multicellular development,

organization, and reproduction