Quantitative spatial transcriptomics of the developing brain

Change log

Developmental and cellular heterogeneity can be studied at the transcriptomic level using RNA sequencing methods, but techniques that quantitatively investigate the spatial dynamics of cell heterogeneity across large tissue areas at a single-cell level are lacking. Applying this capability to the developing brain would enable reconstruction of a full spatial transcriptomic map of neural heterogeneity and the discovery of novel regional neural subtypes with potential functional implications. Such capability would ultimately culminate in the creation of single cell-level tissue atlases with preserved cellular topology. My goal has been to combine single-molecule fluorescence in situ hybridisation (smFISH), automated confocal imaging and analytical tools to derive ‘Quantitative Spatial Transcriptomics (QST)’ for the spatiotemporal characterisation of gene expression across the developing mammalian forebrain.

My thesis research has focused in two main areas:

  1. Novel methodology for insights into mammalian cortex architecture: I have co-developed (with Dr Omer Bayraktar) an automated, multiplexed smFISH and imaging pipeline for screening brain-wide gene expression at cellular resolution, termed Large-area Spatial Transcriptomics (LaST). LaST allows “mapping back” of transcriptome data from single-cell and single-nuclei RNA sequencing for high-quality in situ validation in a regional qualitative and quantitative manner. In addition, this tool can uncover novel cell heterogeneity and discover unique cell identities based on precisely validated combinations of cellular gene expression. For example, the organisation of neurons into six distinct layers is a hallmark of the mammalian neocortex but it is not known if glial cells also possess any diversified laminar features. Using LaST, the single-neuron expression of layer markers was mapped across the mouse cerebral cortex and identified diversified glutamatergic neuron subclasses and their laminar distribution in the neocortex, including rare transcriptomic types. Moreover, applying LaST to cortical astrocytes, I identified molecular distinctions that were associated with three dorso-ventral astrocyte layers in the somatosensory cortex, which deviate from known layer organisation of glutamatergic neurons. These published results (Bayraktar, Bartels et al., 2020) identify a previously unrecognised spatial complexity to cortical architecture when considering combined patterns of neuron and astrocyte heterogeneity, and are likely to have functional implications.
  2. High-resolution transcriptomic developmental biology of Cdkl5: I further developed LaST into QST for cell type-specific quantifications of a single subject gene and analysed the expression of Cyclin-dependent kinase-like 5 (Cdkl5) during forebrain development. CDKL5 mutations cause a severe human neurodevelopmental disorder called CDKL5 deficiency disorder (CDD) that is currently incurable. The expression of CDKL5 in developing neural cells, especially glia, is unclear, resulting in an incomplete understanding of the pathogenesis of the disease. Using QST, I quantified the spatiotemporal expression pattern of Cdkl5 mRNA in developing mouse brain and uncovered novel dynamic patterns of Cdkl5 enrichment. I discovered that Cdkl5 is initially enriched in upper layer neurons perinatally, followed by deep layer neurons during late postnatal development. Using QST and data derived from single-cell RNA sequencing, I found evidence that Cdkl5/CDKL5 is expressed by macroglia during early postnatal development and is enriched in distinct oligodendrocyte transcriptomic types in the mature mouse and human neocortex. An investigation of CDKL5 protein expression in mouse brain development and in human post-mortem tissue is yet to be fully characterised. These data, combined with future investigation of the function of CDKL5 in these cell populations, will improve our understanding of CDD pathogenesis and may uncover novel cellular targets for therapeutic intervention in CDD. In addition, the QST screen provides a quantitative reference map of physiological Cdkl5 expression levels during postnatal development that is instructive for future development of gene and protein replacement therapies.

Combined, the novel spatial transcriptomics techniques developed in this thesis identified higher-order cerebral cortex organisation, and allowed characterisation of Cdkl5 gene expression, significantly advancing our understanding of mammalian neocortical development and Cdkl5 gene function therein.

Rowitch, David
Neurodevelopment, Cerebral cortex, Astrocyte diversity, Neocortical astrocyte layers, CDKL5 deficiency disorder, Spatial transcriptomics, Single-cell omics
Doctor of Philosophy (PhD)
Awarding Institution
University of Cambridge
Wellcome Trust (109142/Z/15/Z)
Wellcome (109142/Z/15/Z)
Is supplemented by: