Fig 1.
New workflow for the Blue Brain Mouse Cell Atlas (BBCAv2).
Steps of the pipeline are displayed in rows with the data consumed in the left column, the method used in the middle column, and the produced results on the right. Each step of this pipeline builds on the results of the previous steps.
Fig 2.
Local impact of the annotation atlas on the estimation of cell densities.
Panels (A) and (B) show the misalignment between the Nissl datasets in grey and their respective annotation atlases overlaid as colors (coronal slices). Boundaries between annotated regions, when available, are displayed in yellow. Different versions are shown from left to right: CCFbbp, CCFv2 from the AIBS, and CCFv3 from the AIBS. (C) and (D) compare the estimated neuron densities from the BBCAv2 pipeline (step 1 and 2 of Fig 1) based on each reference atlas version to literature (bars in brick red with black outline). Each reference atlas version appears in different colors: dark blue for CCFbbp, orange for CCFv2, light blue for CCFv3, and green for the manually realigned AV2 and Nissl2. (A) Focus on the isocortex L1/L2 boundary. On the right of each image, L1 and L2 annotations are shown respectively in blue and orange, on top of the Nissl expression. The density difference between L1 and L2 creates a visible boundary (see left part of the images). However, in each CCF version, parts of the denser L2 are annotated as L1, raising the estimated density of neurons for L1. The magnifications shown for CCFv3 highlight some examples of this problem. (B) Focus on the cerebellar cortex, granular/molecular boundary (Purkinje layer not represented). Each subregion of the cerebellar cortex in the annotation atlas is displayed on top of the Nissl dataset with a different color. Molecular (ML) and granular (GL) layers are shown in different shades of the same color, except for CCFv3 which does not distinguish these layers anymore. Here also, the annotation boundaries do not follow the visible density changes between these layers in the Nissl data. (C) Focus on various regions of the isocortex L1. For these regions, the neuron counts produced with the pipeline using each CCF version are greater than their literature counterpart. The errors linked to misalignment will increase with the boundary surface area. Using the CCFv3 version (light blue bars) yields the worst estimates. Regions shown: Primary somatosensory area, barrel field (BF); Primary visual area (VA); Primary somatosensory area (SS); Prelimbic area (PV). (D) Focus on lingula (Ling) and flocculus (Floc) regions of the cerebellum. Results of the manual alignment of the AV2 with Nissl2 are shown with green bars. Density estimates are normalized according to the densities obtained with the CCFbbp. We observe that after manual realignment, estimates tend to be closer to literature values than for the other CCF versions. CCFv3 estimates are not shown, because its AV lacks the separation between molecular and granular layers.
Fig 3.
AIBS GAD67 ISH experiment data.
(A) raw image, (B) filtered image and (C) realigned binarized image. This AIBS ISH experiment highlights inhibitory cells of the mouse brain (experiment No 79556706—slice 22). The raw slice image (A) contains artifacts and a minimum expression value offsets the dataset. The filtered image (B) is derived from the raw image (A) by the AIBS. Somas reacting to GAD67 are detected in the raw image and isolated from the background. The different colors of the cells in (B) represent the different levels of expression (AIBS 2015 white paper). (A) and (B) have been downloaded on the AIBS website [68] with permission of the AIBS. The filtered images are realigned and thresholded to obtain the binarized image shown in (C). (D) Positions of the manually realigned coronal stained slices, shown in blue on a sagittal slice of the CCFbbp brain volume.
Table 1.
Summary of papers reviewed by marker.
Fig 4.
Overview of published data on inhibitory neuron densities in the mouse brain.
(A) Illustration of regions for which information on densities or absolute numbers of GAD67+ neurons were published. The disk is divided into rings and sectors. Each ring represents a hierarchical level in the AV and sectors represent the contained regions. The center of the disk represents the entire brain, each surrounding ring then represents the next hierarchy level. Colored areas represent regions where at least one study reports cell numbers or densities. The size of the sectors corresponds to the number of neurons in the region relative to the number of neurons in the brain. The colors correspond to regions according to the AV, such as the cerebellar cortex in yellow and cortical areas in green. (B) Variability of published cell density estimates. Each panel shows the range of reported cell densities (log scale). For each region, data points are depicted in blue, the average in orange. The minimum and maximum values are indicated by the whiskers. Regions shown: Dentate Gyrus (DG); Field CA1 (CA1); Field CA3 (CA3); PreLimbic area (PL); Primary SomatoSensory area (SSp); Primary VISual area (VISp); STRiatum (STR); SUBiculum (SUB); Primary somatosensory area, barrel field (BF); Lateral Amygdalar nucleus (LA); CereBellar corteX, Purkinje layer (CBXpu).
Fig 5.
Linear fitting of marker intensity to cell density in cerebellum, isocortex, and the rest of the brain.
Scatter plots of the PV+, SST+, VIP+ and GAD+ densities reported in literature (y-axis) according to the region mean intensity (x-axis) grouped by the main regions of the brain, respectively cerebellum, isocortex, and the rest of the brain. Each point represents a single literature density value and is color-coded according to the different levels of confidence (ratio of standard deviation over mean value) from literature data. The linear fit is represented with a black diagonal line. Each panel shows 95% of the available points. Regions that are purely inhibitory (i.e., based solely on the neuron counts obtained at step 2 of the pipeline), with no inhibitory neurons and regions with null region mean intensity are not displayed and were not considered for the fitting.
Table 2.
Results of the linear fitting.
Fig 6.
Results of the density pipeline.
(A) Sagittal and (B) coronal view of the Nissl reference atlas used for BBCAv2, showing cells of the mouse brain. Regions with high cell density appear in dark grey. Regions that required significant corrections (see Section 2.7) for their neuron subtype density estimates are colorized (Neu = neuron). The coronal slice chosen is displayed as an orange line on the sagittal slices. Arrow in black shows the location of the lateral ventricle of the mouse brain. The blue line behind the sagittal slice highlights the drop of Nissl expression coming from the original Nissl experiment from Dong [18]. (C). Sagittal and (D). coronal view of the BBCAv2, showing the positions of the different types of neurons. PV+, SST+ and VIP+ cells appear respectively in yellow, dark orange and blue. The rest of the GABAergic population is color-coded in green and the remaining neurons in gray. The variation of the distribution of neurons is following the original distribution of Nissl expression.
Fig 7.
Distribution of the GABAergic neurons in the mouse brain.
(A) Ratios of PV+, SST+, VIP+, GAD67+, and remaining neurons of the mouse brain. InhR cells appear in green, PV+ in yellow, SST+ in orange, and VIP+ in blue. The remaining neurons appear in grey. (B), (C) Circular distribution of GAD67+ neurons in different regions of the mouse brain according to different parcellation schemes ranging from coarse at the center to fine-grained at the periphery (colors and disposition similar to Fig 4A). (B) displays results of BBCAv1 and (C) results of BBCAv2.
Table 3.
Estimates of inhibitory neurons in the mouse brain in millions and percentage (%) subtypes.
Fig 8.
Cortical excitatory/inhibitory ratios.
(A) Relative cortical layer cell densities from the primary somatosensory cortex lower limb (SSp-ll), primary auditory cortex (AUDp) and temporal association areas (TEa). Blue lines correspond to excitatory neuron densities (Exc) and the orange lines inhibitory neuron densities (Inh) in cells/mm3. Final confidence intervals are also displayed for each layer as vertical lines. (B) Ratios of inhibitory/excitatory neurons (Inh/Exc) across anatomical regions of the isocortex, expressed as percentage and arranged in five subnetworks based on Kim et al. [7]. Acronyms correspond to the AV naming convention [60] (see Table A in S1 Document). The portion of inhibitory neurons that belongs to L1 is shown with a black contour. (C) Somatosensory barrel field neuron distributions according to pia. Each layer is represented by a different color and the total density according to distance to pia appears in black. (D) Somatosensory barrel field excitatory/Inhibitory neuron distributions (Exc/Inh) according to pia compared to the results extracted from figures 2EFG from Meyer et al. [95]. AV mean layer limits are represented as dash lines and aligned to the ones provided by Meyer et al.
Fig 9.
Impact of the amount of literature on the BBCAv2.
(A) Evolution of the mean standard deviation of the BBCAv2 generated densities for each genetic marker (shown in different colors) using different percentages of the total amount of literature values available. (B) Comparison of the generated density values of the BBCAv2 with their literature counterparts using 90% of the total amount of literature values available (20 trials). The color encodes the brain regions according to the AV, while the shapes of the points encode for cell types. The middle line delimits equal quantities, while the dashed line shows the average deviation of 2.7-fold between literature values reporting on the same region for PV, SST and VIP neurons. (C) Evolution of the mean coefficient of determination R2 of the fitting, for each genetic marker (shown in different colors), performed using different percentages of the total amount of literature values available.
Fig 10.
Qualitative validation of the BBCAv2 and comparison with BBCAv1.
GAD67 sagittal brain slices comparison between (A) the filtered image from an ISH experiment by the AIBS (experiment No 480—slice 13) at the top and (B) the BBCAv2 predicted GABAergic neurons positions at the bottom. Cells reacting to the marker are displayed in purple. Blank regions in our model correspond to fiber tracts regions which host no neuron. This slice does not take regional soma sizes into account. Most of the density features of the real experiment can be found in the BBCAv2 slices. Black arrows point to the thalamus (left arrow), molecular layer of the dentate gyrus (center arrow) and granular layer of the cerebellum (right arrow) of each brain. (C) Circular distributions of the densities of PV+, SST+ and VIP+ neurons computed according to the Erö et al. method on the left [17] and our method on the right (colors and disposition as in Fig 4A). Distribution of PV+, SST+ and VIP+ neuron density values reported in literature against generated densities, using the BBCAv1 method (D) and the BBCAv2 method (E), for similar regions. The color encodes the brain regions according to the AV, while the shapes of the points encode for cell types. The middle line delimits equal quantities. In (E) the model is built on the regional literature references used in the figure which explain the numerous points on the middle line.