Cortical activity emerges in region-specific patterns during early brain development

Abstract

The development of precise neural circuits in the brain requires spontaneous patterns of neural activity prior to functional maturation. In the rodent cerebral cortex, patchwork and wave patterns of activity develop in somatosensory and visual regions, respectively, and are present at birth. However, whether such activity patterns occur in noneutherian mammals, as well as when and how they arise during development, remain open questions relevant for understanding brain formation in health and disease. Since the onset of patterned cortical activity is challenging to study prenatally in eutherians, here we offer an approach in a minimally invasive manner using marsupial dunnarts, whose cortex forms postnatally. We discovered similar patchwork and travelling waves in the dunnart somatosensory and visual cortices at stage 27 (equivalent to newborn mice) and examined earlier stages of development to determine the onset of these patterns and how they first emerge. We observed that these patterns of activity emerge in a region-specific and sequential manner, becoming evident as early as stage 24 in somatosensory and stage 25 in visual cortices (equivalent to embryonic day 16 and 17, respectively, in mice), as cortical layers establish and thalamic axons innervate the cortex. In addition to sculpting synaptic connections of existing circuits, evolutionarily conserved patterns of neural activity could therefore help regulate other early events in cortical development.

Keywords: GCaMP6s; activity pattern; cortical development; marsupial; spontaneous activity.

Fig. 1.

Conserved area-specific patterns of neural activity in the developing mammalian neocortex. (A) Experimental design. Developing dunnart joeys were electroporated in-pouch with PB-GcaMP6s at P15, during the peak of pyramidal neuron generation, which extends from around P10 to P20 (stages 20 to 23, equivalent to mouse E12-15; grey area). Animals were then imaged using live two-photon (2P) microscopy between P21 and P40 (stages 23 to 27). (B) Baseline fluorescence was readily apparent through joeys’ transparent skulls after opening a skin window. The square area is magnified below (Inset) and indicates approximate regions of the somatosensory (SS) and visual (VIS) cortices imaged in this study. (C) Coronal section through SS showing layered distribution of GcaMP6s-expressing pyramidal neurons (green) across layers (L) and white matter (wm) of an animal collected after live imaging at stage 27. (D) High-magnification section of a stage 27 somatosensory layers 1 and 2/3 stained against glial fibrillary acidic protein (GFAP, magenta). (E) Representative calcium traces (ΔF/F, significant events in red) of spatially contiguous medio-lateral regions (red bar) of SS (Top) and VIS (Bottom) of a dunnart at stage 27. Similar to equivalent-stage perinatal mice, activity in the dunnart neocortex is highly area-dependent. Patchwork-type activity is evident in SS, and travelling waves in VIS. (F) Representative snapshots of recordings in SS showing patchwork activity (color-coded on right). (G) Continuous time series of a VIS recording showing a travelling wave (time-color projection map on the right). In F and G, rostral is to the right and lateral to the top. (Scale bars: 500 μm in B; 100 μm in C; 10 μm in D; 2 mm in E; 200 μm in F and G.)

Fig. 2.

The onset and maturation of patchwork activity in the dunnart somatosensory cortex occurs between stages 23 and 25. (A) Representative ΔF/F traces of contiguous regions of interest (ROIs, Top Rows), network nodes mapped to their spatial arrangement within the SS imaged region [Middle Rows, rostral is to the Right and lateral to the Top; (scale bar, 200 µm.)], and functional connectivity networks estimated by Gaussian graphical modelling (Bottom Rows, color-coded nodes represent ROIs and edges represent nonzero partial correlations between pairs of ROIs), from three individual animals at stages 23, 24, and 25, respectively. Colors indicate distinct neural ensembles estimated by Louvain community detection. Grey nodes have zero partial correlation with all other ROIs. A single connected component (giant component) emerges during development, which encompasses the imaged region by stage 25. (B) Pairwise Pearson correlation between ROIs as a function of distance increases over development (Left) which is significant when averaged over distances less than 200 µm (Right). Curves show mean correlation with respect to distance as first averaged for each animal then over all animals, shaded regions are the SEM, and individual data points in the bar plots are values for each animal. (C) Properties of functional connectivity networks change over development, reflecting the emergence of patchwork-type activity. *P < 0.05, **P < 0.01, ***P < 0.001.

Fig. 3.

Onset and maturation of travelling waves in the dunnart visual cortex occurs between stages 25 and 27. (A) Example of automated detection and tracking of travelling wave trajectories in VIS by connected components analysis and Kalman smoothing (Methods and SI Appendix, Fig. S2). (B, Top Row) Trajectories for individual waves in VIS at stages 25, 26 and 27, pooled over all animals at each stage and color-coded by relative time. Waves with characteristic trajectory patterns emerge at stage 26 and become more complex by stage 27. (B, Bottom Row) Wave velocity vector fields over the VIS imaged region across development, corresponding to the wave trajectories above. The size and direction of each arrow corresponds to the average speed and direction, respectively, of waves travelling through each element of an 8 × 6 grid over the imaged region. Arrow color represents the number of waves that passed through each element of the grid. Directional structure appears at stage 26. Patterns at stage 27 show anticlockwise rotation over large sections of VIS. (C) Event duration, distance, size (total number of ROI activations in the event) and area (number of ROIs spanned by the event) increased significantly from stage 25 to stage 26. Each data point is the average value over all events for an individual animal. (D) Stacked histograms of event direction and distance. Direction is defined as the average along an event trajectory. At stage 25, detected wave events are few, only propagate over short distances (less than 200 µm, dark purple) and have no significant dominant direction. At stage 26, waves travel further and with statistically significant directional bias towards the upper-left quadrant of the imaged region (e.g., from rostro-medial to caudo-lateral). (Rayleigh test: P < 0.001; Rayleigh bimodal test: P = 0.03; Omnibus test: P = 0.001; Rao spacing test: P = 0.05; Hermans–Rasson test: P < 0.001). Wave distance continues to increase by stage 27, and there is no longer a single clear average direction of travel (no significant deviation from a uniform circular distribution). In A and B, rostral is to the right and lateral to the top. (Scale bars: 200 µm in A and B).

Fig. 4.

The emergence and specification of early neural activity in SS precedes that of VIS. (A) Representative traces of activity within the somatosensory (SS, Top) and visual (VIS, Bottom) cortices across stages 23 to 26, including mostly silent pre-onset, activity onset, and clear patchwork or wave features becoming evident one stage later in VIS than SS. Schematics showing approximate imaging locations of a stage 24 brain. (B) Representative raster plots of activity across ROIs from recordings of SS and VIS in one individual at stage 24 (Left) and another at stage 25 (Right) showing the broader delay in activity onset and maturation. Raster plots are masked to show only significant deviations from baseline fluorescence. (C) Low-dimensional representation of pre-onset and onset activity in both areas (t-SNE, t-distributed stochastic neighbor embedding; each point represents an individual calcium transient per ROI) reveal differences from their onset (SI Appendix, Fig. S5). (D) Violin plots of mean ΔF/F for individual events comparing SS and VIS for pre-onset and onset activity (mean ± SEM; Wilcoxon rank sum test: ***P < 0.001).

Fig. 5.

Developmental progression of correlated activity between somatosensory and visual cortex. (A) Linear discriminant analysis (LDA) on a reduced feature set at late stages 26 and 27, when patchwork and waves have established, clearly separates data by region (MANOVA test: P < 0.001; each point represents one animal; see Methods and SI Appendix, Fig. S6). (B) Principal component (PC) analysis of activity features from functional connectivity network analysis and wave event analysis for paired recordings of SS and VIS in single animals. Variables were scaled to zero mean and unit variance prior to computation of the PCs. Main axes show the projection onto the first two PCs which explain 93% of the variance in the data. Dashed grey lines between SS and VIS markers denote paired recordings from one animal. Inset panels show data separately for early stages (24 and 25) and later stages (26 and 27). Arrows show the average direction and magnitude of the difference between SS and VIS features for early and later stage recordings. The change in direction of the average feature difference over development suggests that patchwork and wave-like activity patterns in SS and VIS respectively do not develop in strict parallel. (C) Feature difference profiles between activity in SS and VIS highlight distinctive properties of activity across developmental stages (mean ± SEM). Activity features were computed by network analysis and wave analysis for paired recordings of SS and VIS in single animals and normalized for comparison. At early stages (24 and 25), all features have higher values in SS than VIS (i.e., positive values) with greater elevation for network features that characterize patchwork activity. At later stages (26 and 27) the difference between regions is dominated by the wave-like properties of activity in VIS.

Fig. 6.

Differences in cytoarchitecture and thalamic innervation between somatosensory and visual cortices of dunnarts during Stages 23 to 27. (A) Representative coronal sections of somatosensory (SS, Left) and visual (VIS, Right) cortices stained with the thalamocortical axon marker VGluT2 (red), DAPI (blue), and either the upper-layer marker Satb2 or the deep-layer marker Ctip2 (green). Insets show approximate rostrocaudal extent (Top) and a coronal schematic of the regions imaged in a red box (Bottom) of a stage (S) 23 brain. Note the size and cytoarchitectural differences between SS and VIS across stages, taken from single animals per stage. (B) Cortical thickness quantified as paired analyses showing significantly larger columns in SS than VIS at all stages. (C) Paired analysis of maturation index, quantified as a log₂ ratio of upper layers (L) 2 to 4 and ventricular zone (VZ). (D) Plot-profile quantifications of normalized VGluT2 fluorescence across layers (mean ± SEM, n > 3 per condition) in SS (red) and VIS (blue). Dotted lines represent the extent of layers as depicted in A averaged between SS and VIS for comparison, with asterisks representing significant differences between regions quantified within the middle 40% of each layer. (B–D) Paired analyses using repeated measures two-way ANOVAs with Sidak’s multiple comparisons correction between areas per stage. *P < 0.05, **P < 0.01, ***P < 0.001. IZ, intermediate zone; MZ, marginal zone; SP, subplate; UL, upper layers; wm, white matter. (Scale bars in A: 2,000 µm Insets, and 25 µm in Bottom sections.)