(548c) Single-Cell Transcriptomic Analysis of Retinal Neurons in Nine Species Facilitates Comparison of Cell Type Diversity across Evolution

Vision begins in the retina, the neural tissue that lines the back of the eye. Neurons in the retina can be divided into five retinal neuronal classes based on an intersection of morphology, synaptic connections, physiology and molecular expression. These retinal classes and their anatomical organization into three nuclear layers and two synaptic layers are conserved in all extant vertebrate species, from lampreys to humans. This suggests that the neuronal classes and their basic organization arose 500 million years ago, making the vertebrate retina truly ancient. However, each of these neuronal classes can be further subdivided into discrete types based on morphological, physiological and molecular criteria. Recent studies employing high throughput gene-expression profiling have estimated 80-130 distinct molecularly defined retinal neuronal cell types in vertebrate species such as mouse, chick, zebrafish and primates. Furthermore, evidence in mice suggests that these estimates align with cell type inventories based on morphological and physiological criteria.

However, a major unanswered question is the extent to which cell types within the major neuronal classes are conserved among different species. Equally important is to connect differences in cell types to the rewiring of gene expression programs, and connect these changes to species-specific retinal adaptations. Understanding these patterns of conservation and divergence among cell types across species will be useful to evaluate preclinical animal models of human retinal diseases.

To further understand the differences between retinal cell types across species and the gene expression patterns that underlie them, we collected single cell RNA sequencing (scRNA-seq) data from 9 vertebrate species to catalog gene expression in two heterogeneous cell classes that we have studied extensively using scRNA-seq: bipolar cells (BCs) and retinal ganglion cells (RGCs). The variation in the average transcriptomic profile of BCs among species mirrors their phylogenetic structure based on DNA-sequence, while the variation in RGC transcriptomes is more divergent. To embed these cell types in a shared latent space of gene expression, we use autoencoders to learn a phylogeny-guided low dimensional representation of the data, allowing us to draw comparisons across species. Finally, we identify cell type specific genes under evolutionary selection based upon variation in gene expression.

Together, this dataset facilitates a systematic investigation of the conservation and divergence of retinal cell types at various evolutionary scales and reveals new molecular information about cell type variation in the retina across the vertebrates.