Integrating molecular, morphological, and connectomic properties is critical for unbiased classification of neuronal cell types in the mammalian brain. Here we propose a novel approach to classify neuronal cell types by brainwide comprehensive profiling of the dendritic morphology of genetically-defined neurons in the mouse brain. We have developed an innovative mouse genetic tool, called Mosaicism with Repeat Frameshift (or MORF), which enables sparsely and stochastically labeling of genetically-defined neurons in mice. MORF reporter mice can label in exquisite detail single neurons from dendrite and spines to axons and axonal terminals at a labeling frequency of 1-5% of a given neuronal population. We propose to cross our new MORF lines with Cre mouse lines for striatal medium spiny neurons (MSNs) of direct- and indirect pathways, and for cortical pyramidal neurons of distinct cortical layers (i.e. L2/3/4, L5 and L6). Each MORF/Cre mouse will allow us to image the detailed dendritic morphology for thousands of genetically-defined striatal and cortical neurons (i.e. dendritome). We have also developed and streamlined imaging and computational tools to acquire and register brainwide single neuron morphological data onto a standard reference mouse brain atlas. We will digitally reconstruct hundreds of thousands of MORF-labeled neurons using our novel program called G-Cut. Reconstructed neurons will subsequently used for morphology based clustering to define new morphological subtypes, which in turn can be analyzed for the expression of novel molecular markers neuronal cell types (e.g. from single cell RNA-sequencing). Finally, we will disseminate the data to the Brain Cell Data Center (BCDC) for data integration with those from other BRAIN Initiative Cell Census Network (BICCN) and for data access by the broader neuroscience research community. In addition to dendritome data generation and analyses, we will further advance our MORF method by generating new MORF reporter mouse lines with logarithmic fold decrease in the Cre- dependent labeling frequencies, which will permit imaging of the complete, brainwide morphology of genetically-defined single neurons that include both dendritic and axonal arborization. Such tool should greatly facilitate the neuronal morphology based cell type classification. Finally, we will develop integrated computer hardware and software for domain-specific computing for automated image processing and neuronal reconstruction, a major bottleneck in analyzing large bioimage datasets. Altogether we will provide rich dendritome information to enable unbiased, morphology-based neuronal cell type classification, and novel mouse genetic tools and computer software and hardware to advance the field of large-scale neuronal morphological imaging and analyses for the comprehensive study of the mammalian brain.
The mouse brain is composed of about one hundred million highly interconnected neurons, however, the precise number of neuronal cell types remain to be defined. Here we propose to apply a novel mouse genetic tool to image and analyze at large-scale neuronal morphological data (i.e. dendrites) from genetically- defined neurons in the mouse brain. The clustering of neurons based on their morphological features, and then molecular signatures, could help to define novel neuronal cell types and hence contribute to the BRAIN Initiative's effort to generate a reference mouse brain cell atlas.
