FriendlyClearMap: an optimized toolkit for mouse brain mapping and analysis

Abstract

Background: Tissue clearing is currently revolutionizing neuroanatomy by enabling organ-level imaging with cellular resolution. However, currently available tools for data analysis require a significant time investment for training and adaptation to each laboratory's use case, which limits productivity. Here, we present FriendlyClearMap, an integrated toolset that makes ClearMap1 and ClearMap2's CellMap pipeline easier to use, extends its functions, and provides Docker Images from which it can be run with minimal time investment. We also provide detailed tutorials for each step of the pipeline.

Findings: For more precise alignment, we add a landmark-based atlas registration to ClearMap's functions as well as include young mouse reference atlases for developmental studies. We provide an alternative cell segmentation method besides ClearMap's threshold-based approach: Ilastik's Pixel Classification, importing segmentations from commercial image analysis packages and even manual annotations. Finally, we integrate BrainRender, a recently released visualization tool for advanced 3-dimensional visualization of the annotated cells.

Conclusions: As a proof of principle, we use FriendlyClearMap to quantify the distribution of the 3 main GABAergic interneuron subclasses (parvalbumin+ [PV+], somatostatin+, and vasoactive intestinal peptide+) in the mouse forebrain and midbrain. For PV+ neurons, we provide an additional dataset with adolescent vs. adult PV+ neuron density, showcasing the use for developmental studies. When combined with the analysis pipeline outlined above, our toolkit improves on the state-of-the-art packages by extending their function and making them easier to deploy at scale.

Keywords: light-sheet; parvalbumin; somatostatin; tissue clearing; vasoactive intestinal peptide.

Figure 1:

FriendlyClearMap’s processing steps illustrated. Top (gray): Data generation. Mouse brain hemispheres are stained and cleared with iDISCO+ and imaged with a light-sheet microscope. The coordinates of labeled cells can be identified with an external program or marked manually and subsequently imported into the pipeline. Middle (light blue): Dockerized pipeline steps. Middle left: Landmark-based atlas alignment with BigWarp. Use the preprocessing function (inside the Docker) to generate a downsampled image stack. Subsequently, use the ImageJ plugin BigWarp to mark corresponding landmarks on both downsampled stack and on the atlas. Middle center: Use of ClearMap1 and 2’s built-in cell-finding pipeline for threshold-based segmentation, or use an Ilastik Pixel Classifier workflow to identify cells with machine learning. (All of those steps are described in Supplementary Protocols 1 and 2. In each case, the landmarks are then used to transform the pixel coordinates from image-stack space to atlas space and assigned to atlas regions. Bottom (dark blue): Quantification steps outside of the Docker container. Left: Density heatmap as generated by ClearMap1/2. Middle: We also provide optimized scripts for visualizing the output with BrainRender, described in detail in Supplementary Protocol 3. Right: Voxel-level P value maps and region quantifications as generated by ClearMap1/2.

Figure 2:

Quantitative assessment of changes in parvalbumin expressing neuron distribution during development. We used our pipeline to identify PV⁺ neurons expressing the TdTomato reporter throughout the adolescent (P14) vs. adult (P56) brain. PV⁺ neurons start expressing PV as part of their maturation, driven by sensory inputs. (a, d) Sagittal views of representative image stacks for P14 (a) and P56 (d). Note that the cerebellum (arrowhead) was so bright for TdTomato due to the local abundance of the PV⁺ Purkinje cell dendrites that it saturated our detector; consequently, we only segmented cells outside of the cerebellum for P14 and cut it off for P56. (b, e) PV⁺ neuron density per brain area, after processing with our FriendlyClearMap pipeline and visualization with BrainRender. PV⁺ neuron density in the P14 cortex (b) is higher in sensory areas than in extrasensory areas and lower in V1 than in S1. Density is much higher in all of those areas at P56 (e). (c, f) Relative PV⁺ neuron density in the cortex at P14 (c) and P56 (f). At P14, PV⁺ neuron density is high in the somatosensory and auditory areas, which already receive sensory input at this time. In contrast, the visual cortex only starts receiving input around P12 to P14; consequently, PV⁺ expression is low. At P56, in contrast, PV⁺ expression is high across the cortex, with especially dense clusters in S1 and V1. Note that the heatmaps are only scaled relative to each dataset; therefore, the densities are not directly comparable between the 2 time points.

Figure 3:

Somatostatin⁺ and VIP⁺ neurons are distributed in a cell type–specific pattern in the late-adolescent (P28) mouse brain. (a) Example image stack of a SST⁺ reporter brain. The cortical SST⁺ neurons express TdTomato reporter in their axons, whose dense branches in layer 1 are visible as a bright layer even after iDISCO+ treatment. SST⁺ neurons are visible in the granular (layer 4) and infragranular (5–6) layers. Subcortically, they appear especially dense in the globus pallidus (arrowhead, top-right image). (b) SST⁺ neuron distribution in the isocortex, hippocampus, thalamus, and a subset of sensory cortical regions, after processing with our FriendlyClearMap pipeline and visualization with BrainRender. (c) Cortical SST⁺ neuron density showing a reduced density in the somatosensory and auditory regions at the center of the cortex and an elevation toward the edges of the cortical plate. (d) Example image stack of VIP⁺ reporter neurons in a mouse brain. VIP⁺ neurons are found mostly in the supragranular layers 2–3 in the cortex and the hippocampus but barely elsewhere in the brain. (e) VIP⁺ neuron density in the isocortex, hippocampus, thalamus, and the 3 primary sensory cortical regions, after processing with our FriendlyClearMap pipeline and visualization with BrainRender. Note the high density of VIP⁺ neurons along the upper half of the cortex is caused by the overlap of both sensory cortical areas and the motor/retrosplenial cortex along the midline. Also note the complete absence of VIP⁺ neurons in the thalamus. (f) Average VIP⁺ neuron density in the cortical plate, showing a relatively high density in the somatosensory areas and along the posterior and ventral edges of the cortex.

Figure 4:

Cell distribution across cortical span and depth visualized by cortical flatmaps. (a) Overview of the cortical areas that are visible in the flatmap (Butterfly) projection, color-coded as in the CCF3. Scale bar: 500 px (note that this projection does not preserve the spatial relations of the standard atlas space, so there is no direct micrometer equivalent except for depth). Orientation compass: A = anterior, P = posterior, M = medial, L = lateral. (b) Flatmap projections of normalized cell density for PV⁺, SST⁺, and VIP⁺ cells, with cortical region boundaries outlined in white. The cell density is scaled to the same min/max values, enabling a direct comparison of densities across datasets. Top row: all cortex; rows below: computed cell densities for cortical layers 2/3–6. Note that cortical layer 4 (i.e., granular layer), which would be found approximately underneath the indicated “midline,” is only present in granular cortices, such as sensory cortices. PV⁺ cell density at P14 (left column, n = 7) is highest in the primary somatosensory cortex, with detectable expression in auditory, visual, retrosplenial, and interestingly orbitofrontal cortices (arrowhead, top row). At adulthood, PV⁺ cell density (right column, n = 7) is much higher across the entire cortex, however relatively still highest in the sensory cortices, but again with interestingly high densities in retrosplenial and orbitofrontal cortices (arrowhead). SST⁺ cell density (column 3, n = 3) is generally lower and mostly found at the frontal and posterior poles, respectively retrosplenial and OFC/PFC. (c) In contrast to both PV⁺ and VIP⁺ cells, SST⁺ cells (left column, n = 3) are overrepresented in the infragranular layers 5 and 6, consistent with a population largely consisting of Martinotti cells. Lastly, VIP⁺ cells (right column, n = 7) are mostly found at the periphery of the cortical sheet, similar to SST⁺ cells. However, their density is lower and they are primarily found in the supragranular layers 2/3. (d) Cortical depth plots enabled by the flatmap projection. The progression from PV⁺ P14 to adulthood is clear: both have the highest density in the approximate location of layer 4, but there are substantially more PV⁺ cells in adult mice. (e) In contrast, SST⁺ cells are mostly found in the infragranular layers 5–6, whereas VIP⁺ cells are mostly found in the supragranular layers 2/3. (d, e) Normalized cell density (n/µm depth), starting from the pial surface (mapped to CCF3 atlas space for comparability). As this projection preserves cortical thickness differences between different areas, we have only roughly indicated the position of the layers: S.G. = supragranular layers 1, 2/3; G = granular layer 4, if present; I.G. = infragranular layers 5–6; W.M. = white matter. (b–e) Averages of n = 7 (PV P14), n = 7 (PV P56), n = 3 (SST), and n = 7 (VIP) independent mice.