Imagine an art curator preparing for an exhibition. Paintings in the gallery can be arranged into various ensembles — for example by artist, medium, style or theme. Similarly, neurons can be categorized according to a variety of features, such as their size, shape or location in the brain. Writing in Nature, Tasic et al.1 and Economo et al.2 delve deep into the gallery of neuronal types in the cortex of the mouse brain, and use cutting-edge technologies to uncover previously unknown facets of these cells.
Around the turn of the twentieth century, Spanish neuroscientist Santiago Ramón y Cajal created a ‘portrait gallery’ of neurons by carefully examining slices of brain tissue to produce detailed drawings of the cells that captured their diverse shapes. Since then, neurons have been further characterized using measurements of shape, physiology or function. Now, technologies to analyse the gene-expression profile of single cells enable unbiased exploration of cell types.
The brain’s cerebral cortex is responsible for cognition and memory, and contains regions involved in sensory and motor functions. Tasic and colleagues used single-cell sequencing to profile the gene-expression landscapes of more than 20,000 cells, mostly neurons, from two anatomically distinct cortical areas in adult mice — the visual cortex, which processes visual sensory information from the eye, and the anterior lateral motor cortex, which is involved in movement. By doing so, they could compare cells of the same type located in regions with different functions (Fig. 1).
Broadly speaking, neurons of the cortex can be classified as excitatory or inhibitory, depending on the type of neurotransmitter molecule they produce and whether their activation leads to increased or decreased activity of neural circuits. The authors identified more than 100 different cell types, including 61 types of inhibitory neuron and 56 types of excitatory neuron. They found that most cell types were present in both cortical areas — with the exception of excitatory neurons.
These cells are the primary activity-generating units of cortical circuits, and have long been hypothesized to be identical across all cortical areas3. But Tasic et al. found that nearly every subtype of excitatory neuron was specific to either the visual or the anterior lateral motor cortex. The authors injected the cells with fluorescent tracers to track their neuronal projections into distant brain areas. Surprisingly, neurons with different gene-expression profiles also showed different patterns of long-range projection, suggesting that molecular definitions of cell types based on gene expression can provide information about multiple properties of an excitatory neuron.
A study published last year4 also found area-specific excitatory neurons in the developing human cortex, even before the cortical circuits begin to process sensory information. This, together with Tasic and colleagues’ observations, suggests the need to revise our framework for understanding how cortical areas process diverse types of information. In particular, these findings suggest that the functional specialization of cortical areas might rely not only on differences in microcircuits and connectivity patterns, but also on the use of different cell types to process information. Future work is needed to sample cells from more cortical areas to establish how many area-specific excitatory-neuron types exist and how their distribution affects cortical function.
Whereas Tasic and colleagues present an entire gallery of cortical neurons, Economo et al.2 zoom in to look at nuanced differences between neurons in one layer of the cortex — much like studying paintings of the same style on one gallery floor. Excitatory pyramidal tract neurons, which reside in a region called layer 5 in the anterior lateral motor cortex, communicate with other neurons located many thousands of cell diameters away by establishing physical contacts. Pyramidal tract neurons were presumed to all have similar functions5. However, Tasic and colleagues’ analysis revealed that these cells fall into different subtypes on the basis of their gene-expression profiles. Economo and colleagues sought to dissect the differences between subgroups.
Pyramidal tract neurons located primarily in the upper part of layer 5 send signals to a brain region called the thalamus that sends projections back to the cortex, forming a loop involved in preparing for motor activity. Tasic et al. demonstrated that these neurons are molecularly distinct from those located in a lower portion of layer 5 that project to the medulla, which is associated with the execution of movement. Economo et al. engineered each subpopulation of neurons to express the protein channelrhodopsin — a light-sensitive ion channel. This enables neuronal activity to be precisely controlled using light (a method known as optogenetics), and so allowed the authors to dissect the roles of the upper and the lower layer-5 neurons in different types of motor function.
Economo and colleagues used light to independently activate the pyramidal tract populations in mice, and simultaneously monitored both the activity patterns of the cells and the behaviour of the animals as they engaged in a motor-learning exercise. These experiments confirmed that the two populations of pyramidal tract neurons have separate roles: one in preparing for motor activity and the other in initiating movement. The authors’ results also provide a compelling demonstration of how understanding the molecular taxonomy of the brain can lead us to an understanding of how neurons connect and function.
Together, the two studies highlight the transformative potential of atlas-scale data sets in modern neuroscience6,7. They make a strong case for conducting similar studies of more cell types and of the brains of animals of different species, including humans, at various ages. In support of the need for data from different species, a recent single-cell sequencing study8 has reported a greater diversity of neurons in a cognition-associated region of the human cortex than has been described for mice — this might explain our ability for higher-order cognition. Further characterization of both neuronal and non-neuronal cell classes could also yield fresh insights into their selective vulnerabilities to disease states, and instruct the development of protocols to generate these cell types from stem cells in vitro, for use as disease models and for drug testing.
In the future, researchers will undoubtedly make use of the genetic markers of specific neuronal populations identified by Tasic and colleagues’ cell atlas. For example, these markers could be used to design more optogenetic experiments that target specific neuronal populations; to investigate whether ‘area-specific’ cell types can be found in other cortical regions; and to isolate populations of cells for further functional characterization.
However, translating the cellular composition of the brain into biologically meaningful insights will require new strategies for interrogating neuronal function. Technologies to manipulate cell types currently being developed through the support of the US National Institutes of Health BRAIN Initiative9 might enable these analyses. In doing so, they could allow us to fully appreciate the portrait gallery of cells that control brain function.
Nature 563, 38-39 (2018)
Tasic, B. et al. Nature 563, 72–78 (2018).
Economo, M. N. et al. Nature 563, 79–84 (2018).
Creutzfeldt, O. D. Naturwissenschaften 64, 507–517 (1977).
Nowakowski, T. J. et al. Science 358, 1318–1323 (2017).
Harris, K. D. & Shepherd, G. M. G. Nature Neurosci. 18, 170–181 (2015).
Zeisel, A. et al. Cell 174, 999–1014 (2018).
Saunders, A. et al. Cell 174, 1015–1030 (2018).
Hodge, R. D. et al. Preprint at bioRxiv https://doi.org/10.1101/384826 (2018).
Ecker, J. R. et al. Neuron 96, 542–557 (2017).