Mari Sepp is a postdoctoral scientist at the Center for Molecular Biology of the University of Heidelberg, Germany. She grew up in Estonia, graduated from the University of Tartu and received her doctoral degree from Tallinn University of Technology.

Sepp’s primary research interest is elucidating the molecular mechanisms that connect genotypes to phenotypes, with the ultimate goal of understanding their implications in evolution and disease. During her doctoral research, she identified TCF4 as an activity-regulated transcription factor in neurons and demonstrated how autism-associated mutations in TCF4 impair its function. In her postdoctoral research, she compared cerebellum development across different mammalian species, revealing both conserved and human-unique aspects of its cellular dynamics and gene expression programs.

Sepp has been awarded the Otto Schmeil Prize from the Academy of Sciences of the State of Baden-Württemberg. She was selected as a SFARI Bridge to Independence fellow and EMBO Installation grantee in 2024.

Tracing cerebellum development and evolution with single-cell genomics

Cerebellum expanded in parallel to the neocortex during human evolution and is increasingly recognised to play important roles in the evolution of cognition. Such phenotypic innovations are thought to be largely driven by evolutionary shifts in time- and cell-type-specific gene regulation. I will present studies where we used single-nucleus measurements of gene expression and chromatin accessibility to characterise mammalian cerebellum development from early neurogenesis to adulthood. Our data from six species (human, bonobo, macaque, marmoset, mouse, opossum) include approximately 758,000 single-nucleus profiles. We found largely conserved developmental dynamics of cell-type generation, except for Purkinje cells, which showed an expansion of early-born subtypes in humans compared to marmosets, mice and opossums, suggesting this shift occurred within the last 40 million years after the human-marmoset split. Gene expression analyses revealed that cerebellar cell type-defining programs have been preserved for at least 160 million years of mammalian evolution. However, we also observed widespread gene repurposing at the cell-type level, identifying numerous genes that gained or lost expression during evolution. To investigate cis-regulatory elements (CREs) driving interspecies expression differences, we trained a sequence-based deep learning model using chromatin accessibility data. This showed that the regulatory code in cerebellar cell types is conserved, enabling us to apply the model to genomic sequences from over 200 mammalian species. We identified sets of CREs that are shared between different phylogenetic groups, as well as 3000 human-specific CREs that show signs of positive selection. We performed enhancer reporter assays for a selection of CREs in ex vivo cultures of cerebellar granule cells, confirming that our model’s predictions align with the evolution of enhancer activity. Altogether, our work unveils shared and lineage-specific programs governing cerebellum development, and expands our understanding of mammalian brain evolution.