Biologists have long been preoccupied with classifying diverse objects into meaningful categories. Even at the beginning of genetics, Gregor Mendel classified peas as smooth or wrinkled, to infer the mechanisms underlying the inheritance of specific features. Neurons, like peas, come in different shapes and types, but how many neuronal cell types exist and how neuronal diversity emerges during development is unclear. Answering these questions is important, because the diversity of neurons determines the diversity of circuits that can be built and, by extension, the scope of an animal’s behavioural repertoire. Two groups (one writing in Nature1 and one in Science2) now investigate the origins of neuronal diversity in the developing neocortex, a brain region that is the source of mammals’ complex behavioural and cognitive abilities3.
Two opposing — although not mutually exclusive — scenarios account for the generation of distinct kinds of neuron across the nervous system, and in the cerebral cortex in particular3–6. In the first scenario, diverse neuronal subtypes are born from correspondingly diverse progenitor cells (Fig. 1a). This is known as the premitotic model, because diversity arises in progenitors, before neurons are actually generated (neuronal birth occurs through mitotic cell division). In this scenario, the diversity of progenitors might result from the cells’ locations in different parts of the brain, or from their generation at different times. In the second scenario, known as the postmitotic model, progenitors are homogeneous, and diversity arises in neurons as they develop, including through interactions with the surrounding environment, which sculpt these cells into more-refined adult cell types (Fig. 1b).
In the current studies, the groups investigated how diversity arises in the neocortex by using single-cell RNA sequencing to identify and define distinct cell types by their transcriptional signatures. In contrast to previous approaches, in which cells were lumped together to yield average transcriptional activities, this technology allows several thousand individual cells to be simultaneously singled out and sequenced, yielding a more fine-grained view of neurons’ molecular identities7.
In the first study, Mayer et al.1 focused on a population of neurons in the mouse neocortex that releases the inhibitory neurotransmitter molecule GABA. These neurons arise from progenitors located in three transient structures of the developing mammalian brain, called ganglionic eminences8. Using single-cell RNA sequencing, the authors revealed that transcriptional programs are largely conserved in progenitors across the ganglionic eminences. Only a select set of genes is differentially expressed between the three regions — a limited level of premitotic diversity that is consistent with the postmitotic model.
Next, the researchers used elaborate bioinformatics approaches to further reconstruct the developmental trajectory of each neuronal subtype. This analysis demonstrated a link between the initial diversity of immature neurons and the diversity in mature neuronal subtypes, consistent with the premitotic model. The group thus showed that mature neuronal properties are already distinguishable in a rough form in newborn neurons. Subtypes then become more crisply defined as neurons, poised in the genetic ground state dictated by progenitors; they then differentiate and interact with the environment. Altogether, these data support a mixed model of development in which diversification occurs in both pre- and postmitotic cells (Fig. 1c).
In the second study, Nowakowski et al. focused on excitatory neurons that produce the neurotransmitter glutamate, in two neocortical areas in human fetuses — the prefrontal cortex and the primary visual cortex, which are involved in behavioural planning and in vision, respectively. Neurons in these regions are organized into archetypal cortical columns, which are core functional units of circuits. The precise structure of these columns is thought to be tweaked across cortical areas to allow for different functions3–5. How these spatial differences emerge is poorly understood, but explanations involving pre- or postmitotic divergences have both been put forward4,5.
The authors explored how these spatial differences emerge using similar experimental and analytical tools to those of Mayer and colleagues. Nowakowski et al. found that only select sets of genes are differentially expressed between progenitors across the two cortical regions, and that differences increase as neurons mature. These findings thus support a ‘soft’ premitotic situation that leaves leeway for extrinsic factors to drive the developmental trajectories of postmitotic newborn neurons (Fig. 1c). Despite fundamental differences in the biology of the cell types studied, then, the groups reach similar conclusions.
An intriguing finding made by Nowakowski et al. is that glutamate-producing neurons share their initial transcriptional trajectories with GABA-producing neurons. Similarly, Mayer et al. found that newborn neurons from all ganglionic eminences initially transit through overlapping transcriptional ground states. Thus, neuronal subtypes seem to emerge from subtype-specific processes that are superimposed on more-generic developmental programs.
These two studies shed long-awaited light on the origin of neuronal subtypes. However, the quest to understand neuronal diversity is only just beginning. Mendel’s classification of peas was later revisited by Raphael Weldon, who argued that there were many more categories than originally described9,10. Similarly, other studies might reveal more diversity than is reported in the current papers because, inevitably, the current studies are constrained by the criteria used to define cell classes.
To explain, single-cell RNA sequencing, in principle, provides an objective and comprehensive criterion for cellular classification. However, the granularity of the transcriptional assessment — factors such as sequencing quality and which kinds of RNA are analysed — is a key parameter in delineating cell types. Thus, future studies might reveal more progenitor types than those found here, and tilt the balance towards a more premitotic view of diversity. In fact, one recently published study11 reported substantial diversity in immature neurons. Furthermore, sequencing approaches differentiate between cell types only on the basis of molecular differences; other criteria such as electrophysiological properties and anatomy are probably crucial, even at early developmental stages, for a complete definition of cell types.
How identity progresses with time in postmitotic neurons remains elusive. It will be particularly useful to dissect the extent to which external factors, including sensory signals from the peripheral nervous system, drive differentiation and further diversification of neurons within these genetically poised subtypes. Resolving these issues will be central not only to addressing the mechanisms that generate neuronal diversity, but also to characterizing and understanding inter-individual differences in circuit structure and function.
Finally, both studies identify disease-related genes among the cell-type-specific transcripts, including some associated with autism and schizophrenia. As such, they provide a valuable resource for those seeking to understand the mechanisms underlying these disorders.
Nature 555, 452-454 (2018)