To create a new dish, a chef must select ingredients and mix them to achieve different tastes and textures. Similarly, the expression of various combinations of thousands of genes creates and maintains the many diverse ‘flavours’ of each type of cell in the brain. Writing in Nature, Hodge et al.1 report their analysis of gene expression in single cells from the brain to present a ‘cookbook’ of molecular recipes for the neuronal cell types in the human cerebral cortex, a part of the brain required for many cognitive processes. By comparing gene-expression profiles of human and mouse neurons, they reveal a striking conservation of molecularly defined cell types during evolution, but also highlight many key species-specific differences in gene expression within conserved cell types.
The classification of neuronal cell types in the brain represents a long-standing goal in neuroscience that goes back to the anatomists of the early twentieth century. In the cerebral cortex alone, billions of neurons are organized into six sheet-like layers and distributed across dozens of anatomically distinct regions. Given the many sources of variation between individual neurons, systematically annotating brain cells under a common taxonomy represents an enormous challenge2. Previous studies have gained insight into the molecular landscape of the brain by measuring gene expression in different brain regions3 and by using bioinformatic analyses to identify molecular signatures of the major cell classes4. These approaches have been extended by technologies that measure gene expression in single cells5, but isolating intact cells from human brain tissue is technically challenging and many cells are lost in the process.
Hodge et al. overcame this problem by measuring the levels of different RNA transcript molecules in single nuclei isolated from samples of human brains obtained either post-mortem or during surgery. The authors then carried out a statistical analysis of the transcriptomic data; this grouped individual data points (each representing a single cell) into clusters that corresponded to the types of RNA transcript expressed by the cells. The analysis revealed 75 distinct clusters, including 24 types of excitatory neuron (the main signal-generating cells in the brain), 45 types of inhibitory neuron (which suppress neural activity) and 6 non-neuronal cell types. Many of those clusters represent previously defined cell types, but others revealed previously unknown distinctions within broad neuronal classes. Each cell type was given a four-part name on the basis of its general function (excitatory, inhibitory or non-neuronal), its anatomical position, and its expression of genes characteristic of major classes and of specific cell types. This systematic nomenclature integrates multiple modes of information that have historically been used to classify neuronal cell types.
The layer of the cortex in which a neuron resides (also known as its laminar position) has conventionally been considered a fundamental feature of neuronal identity, and so molecular markers of laminar position have previously been used to study cortical organization6. Cortical layers are generated sequentially during development, such that each neuron’s birthdate predicts its eventual laminar position7. To determine how well the laminar position of a cell predicts its type, the authors recorded the layer of origin of each cell and examined whether this correlated with the cell-type taxonomy defined by gene expression.
In some cell types, the expression of certain genes correlated with precise laminar position. However, in striking contrast to recent findings in the mouse8, Hodge et al. found that, in humans, almost all types of excitatory neuron reside in more than one layer (Fig. 1a). This finding was validated using fluorescent labelling of cell-type-specific RNA transcripts in intact tissue samples. It is further supported by a gene-expression survey9 in the developing human cortex that found that molecular signatures of cell type do not correlate well with birthdate, and that most molecular layer markers are not expressed when new cells are generated during development. Together, these studies challenge long-held assumptions about the discrete laminar organization of neuronal cell types in the cerebral cortex….Read more>>