An Atlas of Cortical Arealization Identifies Dynamic Molecular Signatures

Aparna Bhaduri, Carmen Sandoval-Espinosa, Marcos Otero-Garcia, Irene Oh, Raymund Yin, Ugomma C. Eze, Tomasz J. Nowakowski, Arnold R. Kriegstein

Posted on: 14 July 2021

Preprint posted on 18 May 2021

Article now published in Nature at http://dx.doi.org/10.1038/s41586-021-03910-8

How is the human brain patterned into different functional areas during development?

Selected by Juan Moriano

Categories: bioinformatics, cell biology, developmental biology, genetics, neuroscience

Background

During early neurodevelopment, the neural tube consists of neuroepithelial cells arranged as a pseudostratified layer that expand through symmetrical divisions. Later, neuroepithelial cells lose some of their epithelial features and become radial glial cells. Also known as ventricular radial glia, these are bipolar stem cells that span the cerebral cortical wall with an apical process in contact with the ventricular surface, and a long basal process that reaches the pial surface. Radial glia can self-renew and differentiate into basal progenitors, namely intermediate progenitors and outer radial glia that occupy the subventricular zone. These progenitors cells also self-renew and eventually differentiate into neurons, which will in turn migrate radially -using the aformentioned radial glia processes- and position into the growing 6-layered neocortex. Early born neurons fill the deeper layers first, and later born neurons form the more superficial layers in an inside-out arrangement.

While this general scheme holds for the generation of excitatory neurons present throughout the neocortex, neocortical areas differ in cell density, morphology or connectivity, and are functionally specialized (for motor, sensory or higher order cognitive processing). For example, a portion of the neocortex at the caudal telencephalon is devoted to processing visual information (visual cortex), whereas the most rostral part (prefrontal cortex) is thought to be the seat of ‘higher level’ cognitive functions involving a broad set of processes such as perception, emotions or goal-directed behaviors. How the brain is patterned and such differences arise during neurodevelopment is a central topic in the field of developmental biology. Around three decades ago, two main hypotheses were proposed to account for the specification of neocortical areas. Rakic (1988) proposed that progenitor cells are predetermined to acquire area-specific identities and, thus, the neuroepithelium contains a protomap of the future neocortex. On the other hand, O’Leary (1989) supported the view of a initially rather uniform protocortex generated by progenitor cells that, upon environmental influences such as thalamic afferences, acquires area-specific identities. In this preprint, Bhaduri, Sandoval-Espinosa et al. performed single cell transcriptome sequencing and single-molecule in situ fluorescence to examine the dynamics of region and area-specific gene expression signatures in the developing human brain at around midgestation, covering several brain regions and neocortical areas. As a result, this study provides a novel and integrative view of these opposing hypotheses.

Main findings

Bhaduri, Sandoval-Espinosa et al. sequenced the transcriptome of around 700,000 cells from human brain samples aged between 14 to 25 gestational weeks (second trimester of pregnancy), covering ten major brain regions (such as the cortex, the thalamus, the striatum or the cerebellum) and several neocortical areas from the four major subdivisions: frontal, parietal, temporal and occipital regions (see Figure 1 below). Using an iterative hierarchical clustering strategy to group cells by transcriptome similarity and applying differential gene expression tests, the authors identified marker genes for each sampled region and for each main cell type (such as progenitor cells, neurons, astrocytes or oligodendrocytes), as well as inferred their dynamics through the differentiation process. Additionally, the authors used single molecule fluorescence in situ hybridization (smFISH), a technology that makes use of short, fluorescent DNA probes that hybridize selected gene transcripts, to evaluate the expression level and laminar localization in four cortical areas (prefrontal, somatosensory, temporal and visual cortex) of 31 genes of interest. The single cell RNA-seq and spatial transcriptomic data can be explored at https://kriegsteinlab.ucsf.edu/datasets/arealization

**Figure 1.** Brain regions and neocortical areas profiled with single-cell transcriptomics. Figure taken from Figure 1 of the preprint.

Performing hierarchical clustering on the single cell transcriptomic data, authors found that, while cells present in the developing human brain primarily group by type (radial glia, neurons, astrocytes, etc), within each cluster there is a segregation of cells coming from a particular region (subclusters for either the cortex, the striatum, etc). Thus, regional signatures are already well-established at the stages examined (from 10 to 25 gestational weeks) to discriminate cells by their place of origin. Indeed, even closely related regions such as the neocortex, the cingulate cortex and the hippocampus can be discriminated by their regional gene signatures, as shown in the following heatmaps (see Figure 2 below).

**Figure 2.** Marker gene signatures discriminate closely-related cortical regions. Each column represents a marker gene, whose expression is colored by a score from low (blue) to high (red) expression. Each cortical structure is color-coded as depicted at the right margin. Figure taken from Supplementary Figure 2 of the preprint.

Specifically for a region considered fundamental for higher cognition, the neocortex, most cell type clusters contain cells from several neocortical areas. Area-specific markers are already observed in progenitor cells – suggesting some degree of early prepatterning- and, more prominently, a marked transcriptomic difference is detected between radia glial cells from two distinct areas: the prefrontal and visual cortical areas. For example, the known arealization gene NR2F1 and the nuclear factors NFIA, NFIB and NFIX are enriched in radial glia from the visual cortex, whereas the outer radial glia-marker HOPX and HES4 are enriched in radial glia from the prefrontal cortex.
- Some of the area marker genes identified are implicated in neurodevelopmental disorders, as it is the case of the aforementioned NFIA, NFIB and NFIX genes, linked to macrocephaly and cognitive impairment, or the prefontal cortex-enriched gene, HMGB3, in microcephaly.
The gene expression of areal identity markers is highly dynamic, with only a small fraction present across cell types at all time points examined (between 10 to 25 gestational weeks). In addition, the area-specific gene signatures observed in neurons at the stages evaluated differ from those present in neurons of the adult human brain, which suggests that the gene expression programs need to be further refined.
Some areal identity markers are mutually exclusive in their expression patterns across neocortical areas, and even present a distinctive laminar distribution as observed analyzing set of 31 marker genes using smFISH. For example, subplate markers NR4A2, NEFL and SERPINI1 show co-expression in prefrontal cortex, but differential laminar distribution in the three other regions examined (somatosensory, temporal and visual cortex) (see Figure 3 below). Overall, the networks of co-expressed genes substantially change throughout neocortical areas, and those strongly associated to specific area identities are mutually exclusive (not co-expressed in the same cell).

**Figure 3.** Spatial and laminar distribution of subplate markers *NEFL, NR4A2* and SERPINI1. Figure taken from Figure 4 of the preprint.

Why I chose this preprint

How do neurons acquire brain region and cortical area-specific identities? Bhaduri, Sandoval-Espinosa et al. used the power of single-cell transcriptomics in combination with single-molecule in situ fluorescent hybridization to offer new insights into this key question in the field of developmental biology. This outstanding effort led the authors to present new evidence for a model that combines, and extends, two previous hypotheses: Early during neurodevelopment, neural progenitor cells are primed toward either two fates, to frontal or occipital identities, in line with a partial protomap scenario. Later on, these transcriptional differences become more pronounced and specific areal identities in between the rostrocaudal axis gradually emerged, in consonance with the protocortex hypothesis.
Advancing our understanding of cortical area specification will likely help in the investigation of certain neurodevelopmental disorders. Indeed, novel areal markers identified in this study are implicated in disorders such as micro- and macrocephaly. Additionally, the comprehensive dataset generated, encompassing several weeks of neurogenesis and a wide variety of brain regions and neocortical areas, will be of great value as a reference dataset for future studies aiming at modelling brain development and evolution, for example with the use of brain organoid technologies.

Questions to authors

1. The maturation rates of brain regions is known to differ along the rostrocaudal axis. Could the authors discriminate this source of variation in the analysis to compare the dynamics of gene expression across regions?

2. Is there any functional category over-represented among the gene markers for region and area-specific signatures?

3. Differences in areal gene signatures are already detected in radial glia from the prefrontal and visual cortex. Transcriptional differences become then stronger in neurons. At a stage in between radial glia and neurons, intermediate progenitor cells, however, do not show clear area-specific gene signatures. Can the authors find any explanation for this transient ‘dilution’ of areal signatures in this cell-type?

Tags: areal patterning, brain development, cortical development, single cell rna seq, smfish

doi: https://doi.org/10.1242/prelights.30020

Read preprint

(No Ratings Yet)

Author's response

The author team shared

1. The maturation rates of brain regions is known to differ along the rostrocaudal axis. Could the authors discriminate this source of variation in the analysis to compare the dynamics of gene expression across regions?

The author team:
Maturation across brain regions is known to vary across the rostrocaudal axis. In this study we did not focus on this feature, but it may explain subsets of the differentially expressed genes. It is hard to control for because the maturation axis differences are not constant. We have previously used gene expression network analysis to show genes that are correlated with maturation signatures (Nowakowski, Bhaduri, Pollen et al Science 2017), and when we look across our larger list of differentially expressed genes, we do validate that the maturation genes are a subset. Additionally, although maturation may be influenced by certain genes, we do not have evidence for whether these may also play a role in patterning the areal identities of the cortex as well.

2. Is there any functional category over-represented among the gene markers for region and area-specific signatures?

The author team:
We have not found functional over-representation yet, but much of this may be because the functions of genes we are seeing during developmental timepoints are not well described or annotated in this context. This is an exciting area for future work.

3. Differences in areal gene signatures are already detected in radial glia from the prefrontal and visual cortex. Transcriptional differences become then stronger in neurons. At a stage in between radial glia and neurons, intermediate progenitor cells, however, do not show clear area-specific gene signatures. Can the authors find any explanation for this transient ‘dilution’ of areal signatures in this cell-type?

The author team:
Actually, in this study we do see evidence for some areal specific genes in intermediate progenitor cells (IPCs). However, we find that the signatures are dynamic, as we observe in the radial glia (RG) and the neurons, meaning they change both across developmental stages and across the differentiation trajectory. We were surprised by how much more similar these IPC transcriptional identities were to neurons than to their RG progenitor counterparts. However, it is true that the intensity of the transcriptomic signature is more pronounced in neurons that in the IPCs.